ieee communications surveys and tutorials 1 a survey … author's copy.pdf · ieee...

IEEE COMMUNICATIONS SURVEYS AND TUTORIALS 1

A Survey on Demand Response Programsin Smart Grids: Pricing Methods

and Optimization AlgorithmsJohn S. Vardakas, Member, IEEE, Nizar Zorba, Member, IEEE, and Christos V. Verikoukis, Senior Member, IEEE

Abstract—The smart grid concept continues to evolve andvarious methods have been developed in order to enhancethe energy efficiency of the electricity infrastructure. DemandResponse (DR) is considered as the most cost-effective andreliable solution for the smoothing of the demand curve, whenthe system is under stress. DR refers to a procedure that isapplied to motivate changes in the customers’ power consumptionhabits, in response to incentives regarding the electricity prices.In this paper, we provide a comprehensive review of various DRschemes and programs, based on the motivations offered to theconsumers in order to participate in the program. We classify theproposed DR schemes according to their control mechanism, tothe motivations offered to reduce the power consumption and tothe DR decision variable. We also present various optimizationmodels for the optimal control of the DR strategies that havebeen proposed so far. These models are also categorized, basedon the target of the optimization procedure. The key aspectsthat should be considered in the optimization problem are thesystem’s constraints and the computational complexity of theapplied optimization algorithm.

Index Terms—Smart grid, demand response, pricing methods,optimization algorithms.

I. INTRODUCTION

SMART grid uses new technologies, such as intelligentand autonomous controllers, advanced software for data

management, and two-way communications between powerutilities and consumers, in order to create an automated anddistributed advanced energy delivery network [1]. In the nextgeneration power systems, these intelligent technologies areincorporated across the entire system, from power generation,transmission and distribution, to electricity consumption at thecustomers’ premises, with the aim of improving the efficiency,reliability, and safety of the system [2].

One of the key objectives in smart grid is the transition to anenergy-efficient power grid [3]. Energy efficiency is obtainedwhenever volatile demands and renewable energy are man-aged, through the utilization of scalable information process-ing architectures. The concept of Demand Side Management(DSM) includes all activities which target to the alteration ofthe consumer’s demand profile, in time and/or shape, to makeit match the supply, while aiming at the efficient incorporationof renewable energy resources [4]. Furthermore, DSM canalso be employed to facilitate the integration of distributed

J. S. Vardakas is with Iquadrat, Barcelona, Spain e-mail: ([email protected]).

N. Zorba is with Qatar University, Doha, Qatar, (e-mail: [email protected])C. V. Verikoukis is with the Telecommunications Technological Centre of

Catalonia (CTTC), Barcelona, Spain, (e-mail: [email protected]).

generation that can yield significant savings both in the energygeneration and transmission [5]. Other advantages of DSMinclude the blackouts elimination, the reduction of operationalcosts and decreased CO2 emissions [6].

Currently, one of the main DSM activities is DemandResponse (DR), since DR is considered as a subset of thebroader category of DSM, together with energy-efficiency andconservation programs [5], [7], [8]. The US Department ofEnergy defined DR as “a tariff or program established tomotivate changes in electric use by end-use customers, inresponse to changes in the price of electricity over time, or togive incentive payments designed to induce lower electricityuse at times of high market prices or when grid reliabilityis jeopardized” [9]. Based on this definition, the idea is tomake DR attractive to consumers, in order to manage theirpower usage preferences in a way that will benefit not onlythemselves, but also the power grid [10]. This customer-enabled power consumption management is the key smart-gridfeature that enables the adaptation of power demands to timepricing or incentives, while it also improves the efficiency andthe reliability of the power grid [11], [12]. It should be notedthat some researchers and practitioners assume that DSM andDR are interchangeable [4], [6].

The design of efficient DR programs is a crucial componentfor the smart grid deployment [13]. To this end, the study ofDR is an important issue and the various types of DR schemesand actual programs should be identified, in order to extractthe advantages and limitations of these schemes. In this paper,we present various DR schemes that have been proposed inthe literature. Specifically, we organize the DR schemes intothree basic categories, as shown in Fig. 1, while the researchworks for each category are presented in Tables I, II and III.In the first category, DR schemes are classified according tothe control mechanism into centralized and distributed [14].In the centralized mode, consumers communicate directly tothe power utility, without interacting with each other; whilein the distributed mode interactions between users provideinformation to the utility about the total consumption [14].

In the second category, DR schemes are classified accordingto the motivations offered to consumers in order to reduce theirpower consumption [15]. In general, these motivations breakdown into time-based DR and incentive-based DR. In the time-based DR (also known as price-based DR [9], [16]), consumersare granted time-varying prices that are defined based on theelectricity cost in different time periods. On the other hand,customers in incentive-based DR schemes are offered fixed or

2 IEEE COMMUNICATIONS SURVEYS AND TUTORIALS

Fig. 1. Classification of DR programs.

time-varying payments, in order to motivate the reduction oftheir electricity usage during periods of system stress [17], butthey are also under specific constraints or they are penalizedfor not participating in the program.

Finally in the third category, DR schemes use the decisionvariable to identify task-scheduling and energy-management-based DR schemes (also known as energy or power schedulingDR schemes) [18]. In task scheduling DR, the key function isthe control on the activation time of the requested load, whichcan be shifted to peak-demand periods [19]. Different powerconsumption in peak-demand hours is achieved by the energy-management-based DR schemes through reducing the powerconsumption of specific loads [18].

In recent years, there has been an extensive research efforton the optimization and control of the smart grid. Fig. 2presents the increased trend on the development of opti-mization models based on DR programs (indexed by GoogleScholar) within the last 15 years. These efforts are compre-hensively reviewed in this paper, and their key characteristicsare presented, such as the objective function, the appliedoptimization technique and the constraints that are used toformulate the optimization problem. Furthermore, we organizethese optimization models into 5 groups according to the targetof the proposed optimization model. These categories are: a)minimization of electricity cost, b) maximization of socialwelfare, c) minimization of aggregated power consumption,d) minimization of both electricity cost and aggregated powerconsumption, and e) both the maximization of social welfareand minimization of aggregated power consumption. We alsopresent game-theoretic methods that have been proposed forthe solution of the demand-response optimization problem.Moreover, we highlight the optimization methods for two newsmart grid paradigms: Vehicle-to-Grid (V2G) systems [20] andmicrogrids [21]. A V2G system is able to provide energy andancillary services from an electric vehicle to the grid. Thisfunction is achieved through the utilization of bidirectionalpower flows that transfer the discharging energy back to thegrid, or through unidirectional power flows by changing ratemodulation [20], [22], [23]. On the other hand, microgrids are

entities that coordinate distributed energy resources, energystorage devices and electric loads in a decentralized way[21]. In a microgrid environment, the controller facilitatessupply side management, demand side management, as wellas voltage and frequency control [24]. In a grid-connectedoperational mode, these parameters follow the same policythat is used in the main grid. However, in an islanded mode,microgrids are independently controlled and therefore theyefficiently deal with events like faults and voltage sags [25].

Other surveys on DR can be found in the literature [26]-[29]. In [26], the authors study the state of DR technologyin electricity markets, and the magnitude of energy savingsunder DR and other efficiency standards that have been used inelectricity markets. A description of existing DR architecturesis presented in [27] with a report on their requirements, bene-fits and costs, and also a brief review of DR implementationsin USA, Europe and China. Furthermore, authors in [28]perform a bibliographic survey on pricing signals in electricitydistribution systems, while briefly reviewing some demand-side programs. Recently, a survey on DR programs is presentedin [29], where authors present the enabling technologies andsystems, such as smart meters, energy controllers, and com-munication systems that are required for the application of DRin smart grids.

To position our contribution, we are motivated to presentthis survey on DR to be used in future research efforts on moresophisticated and realistic DR optimization models. As theresearch and development of DR programs are evolutionary,this survey provides a summary and a detailed taxonomy of thecurrent status. Moreover, our contribution complements the ex-isting surveys by presenting: a) an overall look of DR schemesand actual DR programs, and recent research approaches thatapply these schemes, b) a classification of DR programsbased on the control mechanism, the motivations offered toconsumers and the decision variable. c) optimization methodsfor the minimization of electricity cost/power consumption aswell as the profits maximization, d) a detailed classification ofthe optimization models based on the target of the optimizationprocedure, the solution methods that have been considered

VARDAKAS et al.: A SURVEY ON DEMAND RESPONSE PROGRAMS IN SMART GRIDS: PRICING METHODS AND OPTIMIZATION ALGORITHMS 3

Fig. 2. Increased trend of number of optimization models on DR programs.

for each case, the ability to include uncertainties, scalability,responsiveness, communications requirements, and support ofmultiple load types, and e) the application of optimizationmethods in V2G systems and microgrids.

The rest of this work is organized as follows. Section IIprovides the DR background by discussing the objectivesof DR schemes, management issues, types of consumersthat participate in DR programs, communication requirementsand adversative conditions in DR implementation. SectionIII presents different categories of DR schemes that havebeen presented in the literature. In Section IV we tackle anextensive survey of the major optimization models that havebeen proposed for smart grid environments. Finally, SectionV provides concluding remarks for this work, together withthe lessons learned and future directions. Furthermore, a listof abbreviations used in this survey is provided in AppendixA.

II. BACKGROUND: DR CONCEPTS

A. Main objectives of DR

The main objectives of the application of a DR scheme aresummarized as follows:

• Reduction of the total power consumption, so that mutualprofit for the power utility and the consumers is achieved.This reduction should occur not only in the consumer’sdemand, but also in the losses of the transmission anddistribution systems [7].

• Reduction of the total needed power generation, which isthe main result of the aforementioned objective. Underthe successful implementation of a DR scheme, theneed of activating expensive-to-run power plants to meetpeak demands is eliminated, while it enables the energyproviders to meet their pollution obligations [9].

• Change of the demand in order to follow the availablesupply, especially in regions with high penetration ofrenewable energy sources, such as solar panels and windturbines, in order to maximize the overall power-system’sreliability [30].

• Reduction or even elimination of overloads in the distri-bution system. This objective is met by the operation ofa Distribution Management System (DMS) that monitorsthe operation of the distribution system, and takes near

real-time decisions that enhance the reliability of thesystem [31].

A DR scheme should also consider security mechanisms,for the protection of personally identifiable energy usage infor-mation that is collected by smart meters for the DR provision[32]. Furthermore, DR schemes should target the reduction ofloads in the distribution system to unload transmission lines,in order to prevent emergency conditions [33]. Moreover, aDR scheme should be designed in such a way that attracts theinterest of consumers to participate in the program, throughthe provision of incentives to change their power consumptionhabits, while at the same time minimizing the consumers’discomfort [34].

B. DR Management

The implementation of a DR method targets the control ofthe customer’s power consuming behavior, in order to meetthe aforementioned objectives that are presented in SectionII.A [35]. The adjustment of the customers’ electric usageis realized as a response to changes in electricity price overtime or when system reliability is threatened. This functionis executed through the cooperation of four main participants[36], as illustrated in Fig. 3: a) energy consumers that takepart in the DR program and they can be either residential,commercial or industrial consumers, b) a DR aggregator thatis connected to the customers and executes the DR program,c) a Distribution System Operator (DSO) that controls thedistribution grid and d) an Independent System Operator(ISO) or Regional Transmission Operator (RTO). In general,the process of a DR program starts by the ISO/RTO thatdetermines the preferred demand volume and the time durationthat it is offered. This information is submitted to the DRaggregators, who then select the participating customers basedon their availability. By taking into account the number ofcustomers that agree with the proposed DR, the aggregatorcalculates the total demand and reports back to the ISO/RTO.In order to evade uncertainty problems in the distributionsystem, the aggregators may initially report the total DR tothe DSO, who then informs the most available substationsabout the total power demand [35]. In this case, the DRcalculations are performed at the DR aggregators, and they arethen used by the DSO for executing optimization proceduresor for discovering problems in the distribution grid.

The aforementioned model of the four participants is gen-eral, and it may also involve a number of agents that interactin competition or in cooperation. In these multi-agent sys-tems, distributed decision making is implemented either in alocal domain or inside the entire system [37]. The decisionmaking process is a result of negotiations and trading onan electronic market and involves procedures like DR anddistributed generation [38]. There are several examples ofmulti-agent energy management systems that are proposed anddescribed in the literature. PowerMatcher [39] is a hierarchicalmarket based algorithm, in which multiple agents that controlelectronic devices can bid for energy, by considering theirown bidding strategy. A similar multi-agent system, knownas the Dezent project, is proposed in [40], where scalability is


Fig. 3. Main participants in a DR program.

improved by incorporating balancing group managers insteadof a central market place. A more scalable system is presentedin [41], where agents not only use price information butalso information about the environment and current status fortheir bidding strategies. Due to the nature of the multi-agentsystems, they are well-fit in microgrid environments [42], [43].

C. DR Applicability

A DR program could increase its effectiveness by takinginto consideration the types of consumers that are appliedto. Typically, four different sectors are the main electricityconsumers, as illustrated in Fig. 4: transportation, residen-tial, commercial and industrial sectors [44]. However, DRprograms are mostly applied to residential, commercial andindustrial consumers.

1) Residential consumers: The design of an efficient DRprogram for residential users is far more complicated, com-pared to industrial customers, mainly due to their near-randomconsumption patterns that require vigilant modeling. This taskcan be achieved by designing residential load managementprograms that either reduce or shift power consumption [45].The reduction of power consumption is realized through theencouragement of energy-aware consumption patterns and theconstruction of buildings with high energy efficiency [8].However, by shifting consumption from peak demand to off-peak hours, a significant reduction of the peak-to average ratiocan be achieved. Consequently, there are possibly abundantopportunities for the DR application in domestic areas. Nev-ertheless, the applied DR program should not assume thatall customers have the same power consuming behavior. Asreported in [35], residential consumers can be grouped intodifferent categories: a) short range consumers, who are onlyconcerned about the power price at the current time instant,b) real-world advancing customers, with consumer perceptionin current and past periods only, c) real world-postponingconsumers, whose perception depends on current and futureprices only, d) real-world mixed consumers, who are a mixtureof postponing and advancing customers, and e) long rangeconsumers, who are able to shift their consumption over awide range of time.

In addition to the consumer’s response to the DR program,other factors should be taken into account during the designof DR programs for residential areas. The advent of Plug-in Hybrid Electric Vehicles (PHEVs) is expected to place a

significant load on the power grid [46]. A smart scheduling ofPHEVs charging hours (e. g. during the night) can reduce theirimpact on the grid. Another issue that should be addressed ina DR program is the proliferation of locally generated powerat the residential level. This local generation provides thecustomers with the opportunity to supply their excess electricalpower back to the grid. Finally, DR programs should considerthat each residence is equipped with appliances with diverseenergy requirements, operational times and arrival rates ofpower requests

2) Commercial consumers: Typically, commercial build-ings are identical in terms of energy consumption patterns,which are determined by weather conditions, design styles, andoperational behaviors. Furthermore, these types of consumerscan be assumed autonomous, regarding the way they respondto electricity prices [47]. In such environments, the mainpower consuming processes are Heating, Ventilation and Air-Conditioning (HVAC), lighting systems and electronic equip-ment. The reduction of the power consumption on these heavyloads can be achieved by either the adoption of energy-efficientbuilding technologies, and/or by the control of the build-ings’ energy consumption behavior through price elasticity ofpower demand. By applying a DR program in a commercialenvironment, building operators or their automated controlsystems make modifications to building operations, with theaim of reducing the building’s total electric load during peakelectric usage times. These modifications vary, depending onthe consuming process. HVAC systems usually use automatedoperational DR functions that are based on temperature and/orair distribution adjustments, in order to achieve power con-sumption reductions [47]. Lighting DR strategies depend onthe season of the year and the time of day. For instance, ona summer day the demand reduction in over-lit buildings canprovide savings, which can be further increased by coolingsavings, since lighting produces heat [47]. For the results of theapplicability of DR methods on various case studies regardingcommercial buildings, the interested reader may refer to [47]-[53].

3) Industrial consumers: Industrial plants are high energyconsumers, with typical peak loads of hundreds of MWsat high voltage levels. In such cases, power and voltageefficiency are extremely vital. Besides, many manufacturingprocesses have critical temporal dependencies, which mustbe scheduled with high timing precision. In contrast to the


Fig. 4. Share of electricity consumed in 2013 in USA [44].

residential consumers, where it is sufficient to control loadsbased on near-real time data, in many industrial environ-ments millisecond-scale monitor and control is essential [54].Furthermore, security issues are particularly important in theindustry. The access to information regarding the load profilesor load shapes is highly confidential and competition-sensitive,since it may indicate the type of equipment that is active andin what time periods. These requirements are vital for theinnocuous operation of the industry equipment. Hence, formany years various industrial facilities participate in actionsthat are very familiar to the smart grid applications, frequentlythrough the application of dynamic pricing models. A numberof examples of such activities are listed in [54]. However,intelligent DR methods may increase the reliability of theindustrial system and the economic efficiency of the electricityinfrastructure [55], [56]. The application of these methods mayalso benefit the power utilities, through the collaboration withthe industrial partners, to analyze DR and determine optimalsolutions for the response at peak-demand hours, and toprovide downstream benefits through end-use monitoring [57].The results of the application of DR in different sections ofindustry can be found in [58] (industrial refrigerated systems),[59] (meat industry), [60] (cement industry) and [61] (foodindustry).

D. DR communication requirements

A communication infrastructure that provides connectivityamong systems, devices and applications is essential for theefficient and reliable operation of the smart grid [1]. Thegeneral communication requirement for the implementationof a DR program refers to the provision of a two-way flowof information between the various entities that participatein the program, through the communication infrastructure.However, smart grids should also satisfy other requirementsfor the effective and reliable communication between thevarious elements of the grid, which are necessary for theimplementation of a DR program [62]:Quality of Service - (QoS): the provision of QoS guaranteesfor the communication technologies used in the smart grid isessential for the smooth implementation of a DR program. Thebidirectional networking of the smart grid should ensure thatcontrol commands, emergency response and pricing signals

are reliably transfered without being affected by the number ofthe connected consumers. For real-time sensing and meteringpurposes that are used in various pricing schemes, latencyvalues of a few milliseconds should be achieved [63]. Onthe other hand, bandwidth requirements of DR programs arein the range of a few kilobits per second per consumer,and depend on the communication frequency between utilitiesand consumers [64], [65]. However, as the number of theinvolving smart-grid elements increase, the communicationinfrastructure should be able to provide ample bandwidthfor the transport of the controlling messages, with minimumfailure rates and latencies.Interoperability: the cooperation of different systems is vi-tal, in order to realize data exchange between the differentcomponents of the smart grid. To provide interoperability andseamless data exchange between interconnected elements ofthe smart grid, the adoption of standards across the communi-cation infrastructure is essential [65]. The provision of inter-operability through open standards is an important operationalobjective of the DR infrastructure, which guarantees that theoverall system is insensitive to changes or modifications in anyone of its underlying components [66]. The interoperabilityfeature can be added to a DR management system, mainlythrough the consideration of a layered architecture that ensureshigh flexibility, together with extensibility and composability[67].Scalability and flexibility: DR becomes more effective whena large number of consumers participate in the DR program,since more adjustable loads are available for regulating thedemand [68]. Thus, a highly scalable communication infras-tructure is essential for the accommodation of a large numberof devices and services, through the evolutionary implemen-tation of the infrastructure on a broader scale. On the otherhand, flexibility allows for the provision of multiple redundantalternate routes for the data flows, as well as the support ofthe mobility feature for the end devices [63]. Cloud-basedarchitectures for DR implementation can be considered as aneffective solution that leverages data-centric communicationfor scalable and flexible communication between the utilityand the consumers [69].Security: network security is an important factor in operationof smart grids, since it provides the means to maintain dataintegrity, confidentiality and authentication, while it facilitatesnon-repudiation [70]. Several security issues may compromisethe effectiveness of a DR scheme; tampering of information ofa pricing program may trigger financial and legal problems,while malwares that may infect the grid can cause a severedamage to the power delivery system [71]. It is thereforeessential to implement secured DR programs that protect theprivate data of consumers, avert unauthorized access that at-tempt to delay, block or even corrupt information transmission,and provide authentication, authorization, auditability and trustcomponents to the communication infrastructure [72].

For a more comprehensive study on the communication re-quirements, challenges and solutions for DR, the interestedreader may refer to [6], [62], [63], [71]-[73].


Fig. 5. Centralized and distributed control mechanisms.

E. Adversative conditions in DR implementation

The execution of a DR program may have unpredictedresults when specific conditions are encountered. An importantdrawback occurs when the end users consume more powerthan their original level of reported consumption. This increasein power consumption is usually caused by loads, such aswater heaters, air conditions or electric stoves, and is calledCold Load PickUp (CLPU) [74]. On the other hand, voltageviolations may occur when Voltage/Var control is applied inthe distribution system. A possible solution to this deficiencyis the integration of Voltage/Var control applications and DR,which may significantly increase the productiveness of thedistribution management system [33]. Furthermore, violationsassociated to the DR activation may be the result of asym-metric DR balancing between phases. This may lead to theincrease of neutral wire current and three-phase bus voltageimbalance [36].

The installation of smart meters in the consumers’ premisesallows the implementation of more dynamic pricing schemes,for the triggering of the peak-demand reduction. However, theapplication of such pricing schemes to a wide range of con-sumers may result in several important implications, regardingthe consumers’ reaction to these rates (especially low-incomeconsumers) and the possible volatility in electricity prices.This is mainly due to low-income consumers, who have asignificantly lower level of price-elasticity than higher-incomeconsumers. An interesting article that considers these issuescan be found in [75].

III. CLASSIFICATION OF DR MODELS

The design of DR programs in a smart grid environment hasdrawn much research attention in recent years. Fig. 1 illustratesthe classification of these research efforts; this classificationis based on the control mechanism of the DR procedure,on the motivations offered to customers to reduce or shifttheir demands, or on the DR decision variable. The followingsubsections present the main characteristics of each one ofthese DR schemes.

A. DR methods based on the control mechanism

This class of DR schemes can be further classified intocentralized and distributed programs [76], according to wherethe decisions for the execution of the DR program are made. Incentralized programs, response decisions for load activation orload scheduling are only tackled by the power utility, throughconsidering that a number of users form a group. In thisway, each consumer contributes to the program individually,without requiring the knowledge of the involvement of theother consumers in the group [77]. However, the operation andcontrol of the grid in a centralized manner is highly difficultin complex and large grids. As an alternative, in large gridsthe communication between energy suppliers and consumerscan be distributed [76]. In such distributed schemes, the powerutility’s main contribution is the transmission of price signals,which are dependent on the overall system load; users can co-ordinate directly with each other, in order to achieve an aggre-gated load reduction. This decentralized control assures scala-bility, while it is also a means of consumer privacy protection,by preventing central authorities from collecting informationfor decision making. In the following subsections we presentmore details on centralized and distributed DR schemes. Basedon the aforementioned description, in centralized schemescommunication connections are only necessary between theutility and the consumers (Fig. 5a), while distributed schemesrequire the additional consumer interconnection (Fig. 5b).

1) Centralized schemes: In a centralized scheme, the DRprocedure is monitored and coordinated by a central controller,who collects demand information from consumers, and DRdecisions are then made for the demand scheduling. Forexample, in [77] and [78] an aggregator is used in order toderive scheduling decisions. Centralized management of loadsis an effective solution for controlling thermostatically con-trolled appliances [79], buildings [80], and charging stationsfor PHEVs [81]- [84]. For instance, a centralized method forPHEV charging is proposed in [81] that schedules PHEVcharging times based on weights, which define critical andnon-critical demand periods.

Central controllers are used in islanded microgrids that havebeen introduced as a coordinated approach. They are used tosimplify the penetration of distributed generation units intothe utility network [85]. In microgrids, different power micro-sources operate as a single system that provides energy to acluster of loads in a local area [86]. The main function of themicrogrid is the conservation of power balance independentlyof the main grid. The application of DR in microgrids hasbeen studied in several articles ( [87]- [89]). For example, in


[88], an active control load strategy is applied in a microgridenvironment, which is triggered by the voltage level in themicrogrid and it enables the full exploitation of the installedrenewable energy sources.

2) Distributed schemes: Distributed DR control programsconsider that demand information is not centrally collectedand consumers can directly access indicators of the grid’sstate [90]. By using this information, consumers are able toreact, if the system’s state is critical. Many researchers havebeen inspired by the distributive nature of the Internet, inorder to provide efficient control mechanisms in smart gridenvironments [91]- [93]. In [92], authors present a distributedscheme that is based on congestion pricing in Internet trafficcontrol. Consumers receive only pricing information from theutility; this information is a function of the current aggregatedload, while it is used by the consumers to adapt their loads.This is achieved by using a mechanism that is based ondecentralized congestion control mechanism for IP networks.This scheme is also used in [91] and it is applied to adistributed charging system for PHEVs. In this scheme, theuser’s preference is modeled through a parameter for thewillingness to pay, which can be seen as an indicator ofdifferential quality of service. Similar studies are presentedin [90], [93].

Distributed schemes are also used in cooperation withother mechanisms that target the control of crucial systemparameters. In [94], a distributed DR method is integrated witha Voltage/Var control scheme that provides improved voltagecontrol and reduction of power consumption. Additionally,in [95] the problem of frequency control is discussed andauthors argue that frequency-sensitive demand-response couldbe achieved in such power systems. Furthermore, the presenceof distributed energy resources is considered in [96]; authorsprovide distributed algorithms for the coordination and controlof both DR resources and distributed energy resources.

The research articles from literature presented in this sub-section are summarized in Table I, for both centralized anddistributed schemes. Section IV presents several other researchefforts that have been made on both these schemes that useoptimization methods.

B. DR methods based on offered motivations

Another way to distinguish DR schemes is by consideringthe motivation method that is offered to customers for theirefforts to reduce or shift their power demands. There exist twomain classes within this group: time-based DR and incentive-based DR. The former group is usually more suited for resi-dential customers, while incentive-based programs are moresuited for industrial consumers [35]. For example, a studythat is based on real data from several industrial and large-scale commercial customers proved that incentive-based DRprograms are more suitable, through the provision of explicitbill credits or payments for pre-contracted or measured loadreductions [7]. Furthermore, the application of a time-basedDR program and an incentive-based DR-program to a 24-bus IEEE Reliability Test System showed that the offeredincentives have key impact on customer habit formation in

TABLE ICLASSIFICATION OF DR SCHEMES BASED ON CONTROL MECHANISM.

response to DR programs [97]. The research articles fromliterature that are presented in the following two subsectionsare summarized in Table II and Table III, respectively.

1) Time-based DR: These programs offer customers time-varying prices that are defined based on the cost of electricityin different time periods [27]. Customers receive this infor-mation and have a propensity for consuming less electricalpower in time periods when prices are high. There are variouspricing schemes that have been proposed for DSM, which areeither retail price structures or DR-based programs [98]- [99].In the former case, either fixed prices or consumption-basedelectricity rates are offered to consumers in order to reducetheir electricity usage. However, customers do not participatein the determination of the prices, while no economic incen-tives are offered to the consumers to respond to hourly changesin electricity prices [98]. On the other hand, in DR-basedprograms the reduction of the electricity usage is achievedwith the contribution of customers, who respond to motivationsignals, being sent from the energy provider [107]. In thefollowing paragraphs we present the different pricing schemes,by firstly introducing the retail pricing schemes and then theDR-based programs.

Flat pricing has been used in traditional energy systemsand has been ingrained in the users’ mind. Under this scheme,customers know that the only way to reduce their electricitybills is by simply using less electricity throughout the durationof the day. In some cases, seasonal flat pricing can be applied,where prices are fixed within a season but they can changefrom one season to another [98].

Time-Of-Use (TOU) pricing is the application of flat pricingin different time periods. Under a TOU pricing scheme, pricesare retained fixed within different pricing periods, which canbe different hours within a day or different days within a week[27]. For example, in California, USA, TOU is used for largecommercial customers, who are charged different rates forthe energy they consume in three different periods: off-peak,mid-peak and peak. During the off-peak period the customersare charged $0.05/KWh, during the mid-peak $0.078/KWhand during the peak $0.099/KWh [99]. TOU tariffs are alsoused as incentives in a household simulation model thatgenerates realistic load profiles in [100]. Based on this model,bill savings are estimated, when the household invest in thesmart appliance technology. However, the effectiveness of suchschemes to the reduction of the total power consumption islimited, since customers do not receive any practical incentivesto reduce their demands. This customers’ response to TOUschemes is also triggered by the fact that they receive attractiveoff-peak prices, but relatively high prices in peak-demand


hours [101]. A study in [102] showed that TOU programsoffer the smallest reduction in the peak demand among alltested programs.

Critical Peak Pricing (CPP) has similarities with TOUpricing, regarding the fixed prices in different time periods.However, the price for at least one period can change, eitherregularly or in most cases, due to occasions of system stress[103]. The participant consumers receive notification of thenew energy price, usually a day ahead. As in the case ofTOU, CPP is not economically efficient for the consumers,due to the preset prices. Furthermore, the ratio of on-peakto off-peak price is higher on CPP event days than in aTOU program [104]. On the other hand, from the energyprovider point of view, significant load reductions duringcritical periods can be achieved under this pricing scheme[105], but with high probability of negative net benefits [106].The implementation of a CPP program in California, USA,demonstrated respectable responses to the state’s announcedcritical events [107]. The basic CPP program is called fixedperiod CPP (CPP-F), where power utilities maximize theirsavings by selecting a single critical peak price and choosingthe event window with the highest wholesale market [108].A number of variations of CPP have been proposed, whereadditional considerations are made for the maximization ofthe utilities’ profit. In variable peak pricing (VPP) the peakprice may vary, since the utility may select from a number ofprice levels on an event period. For example, the OklahomaGas and Electric Company offers a VPP program, with anoff-peak electricity rate of 0.045 $/ kWh and three peak ratelevels: a standard peak rate of 0.113 $/kWh, a high peak rateof 0.23 $/kWh, and a critical peak rate of 0.46 $/ kWh [109].In variable-period CPP (CPP-V) the event start time and theduration of the event period are controlled by the utility, whichimposes a maximum number of event hours. For example,Dominion Virginia Power may trigger a CPP event 25 timesper year, for a maximum of 5 hours per event, and a maximumof 125 hours per year [108]. Finally, Extreme Day CPP isanother CPP variant, in which a critical peak price is appliedto critical peak hours, but there is no variable tariff on otherdays [110].

In Peak Load Pricing (PLP) the day is divided into a numberof periods and different prices are determined for each period.These prices are announced to the customers ahead of each day[111]. The price value for each time period is calculated basedon the average power consumption of the consumers in eachtime period, in order to maximize the payoff of the energyprovider [112]. In addition, the price calculation targets thedemand shift away from peak-demand periods, by expectinga reaction from the customers’ side according to the highprice. A study on the effectiveness of PLP in Auckland, NewZealand showed that the participation of the consumers in thePLP program is highly affected by the high-peak prices [113].A similar to the PLP pricing method is the adaptive pricing,where prices are not announced to customers at the beginningof day. Instead, based on the power consumptions on previoustime periods, the energy provider calculates prices in real-timeand announces them to customers at the beginning of each timeperiod [114].

TABLE IICLASSIFICATION OF TIME-BASED DR SCHEMES.

Under the Peak Day Rebates (PDR) pricing scheme (alsoknown as Peak Time Rebates (PTR)), customers decidewhether they respond to a critical event. Specifically, cus-tomers are under their standard tariff, but they have theopportunity to receive a rebate payment for any load reductionthey can achieve below an estimated baseline load threshold[98]. The results of a pilot study conducted in Connecticut,USA, showed that PDR is more advantageous compared toTOU, in terms of power reduction and consumer’s satisfaction[115]. On the other hand, the same study showed that CPPis more beneficial that PDR. Furthermore, due to the factthat the baseline load threshold must be calculated for eachcustomer and for every critical event, additional resources areneeded. Besides, it is possible that some customers will receiverebate for the reduction of their power consumption that theywould have made, regardless of the critical event [116]. Anexperiment involving 123 residential consumers of the cityof Anaheim, CA, USA, showed that the rebate rewarded toconsumers is pre-determined to be very high, which does notreveal the actual supply-demand balance at different operatingconditions [117].

Another pricing scheme that is based on voluntary participa-tion of customers is the Vickrey-Clarke-Groves (VCG) scheme.Customers are requested to provide their power demand in-formation, which is then used by a centralized mechanism forthe price calculation, for each time period [112]. Payments areprovided to the customers in a way that they have motivationsto provide their demand information truthfully. The VCGpricing scheme has been used in order to reduce the totalpower consumption [112], or to shift it to off-peak time periods[118].

Real-Time Pricing (RTP) requires the maximum customerparticipation. Under an RTP scheme, the energy providerannounces electricity prices on a rolling basis; these prices aredetermined and announced before the start of each time period(e.g., 15 minutes beforehand) [119]. Therefore, the successfulimplementation of an RTP scheme relies on the two-waycommunication capabilities of the smart grid, which togetherwith an Energy Management Controller (EMC) installed atthe customer’s premises, significantly increases the decisiontaking velocity [120]. EMCs support continuous flow of dataand are based on the consumer’s preferences. The consumersmake smart decisions to modify the energy usage across thebuilding, which will guarantee higher reductions in the elec-tricity bill. The energy provider also makes decisions to definethe prices for the upcoming time period. These decisions are


influenced by random events, the total power consumption andthe response of the consumers to the previous prices [121].RTP mechanisms have already been applied to large industrialand commercial customers [122]. However, in the residentialdomain, RTP schemes have small implementation success,since most consumers are risk-averse and see the necessityof taking systematic electricity decisions as an importantdrawback [123], [124]. In addition, in some cases the costsavings resulting from the participation in an RTP programwill exceed the costs imposed on customers to follow theprogram [75], [125].

One of the main challenges for the implementation of anRTP scheme is that it requires continuous real-time communi-cation between the energy provider and the customers, whichis not attractive from the user perspective. [126]. Furthermore,the mass flow of data that is exchanged between the energyprovider and the EMCs, the lack of efficient smart meteringand the high complexity could limit the effectiveness of sucha scheme. The Day-Ahead RTP (DA-RTP) is an alternativeRTP-based solution, wherein the next day’s predicted real timeprices are announced to the customers beforehand and theyare billed for their consumption based on the price of thisday-ahead [98], [127]. A test system with 320 customers inOntario, Canada showed that the DA-RTP scheme achievedflatter demand curve, lower losses, lower peak-to-peak dis-tance and higher load factor [98]. Also, the integration of aDR program based on DA-RTP and Voltage/Var control isproposed in [128], where the effects of demand reductionon system voltage are studied and results show significantimprovement in system voltage under the proposed scheme.

The results of the application of the aforementioned time-based DR programs in various markets across the USA arepresented in [115]. Several case studies are tackled, with awide range of consumer sizes, program durations and energymanagement services. In addition, the authors in [129] outlinethe benefits of the applied RTP programs by the AmericanElectric Power (AEP) in Columbus, Ohio. Similar studies arepresented in [99], [105] and [130], where time-based schemesin North America, in California and in the city of Chicago,IL, are respectively presented and analyzed.

2) Incentive-based DR: They consist of programs thatoffer fixed or time-varying incentives (payments) to customersthat reduce their electricity usage during periods of systemstress [31]. Customer enrollment and response are voluntary,although some of these programs penalize customers thatfail the contractual response when events are declared. Theincentive-based DR programs can be further sub-categorizedinto classical programs and market-based programs, whilethey can be offered in both retail and wholesale markets[131]. Consumers that participate in classical incentive-basedprograms receive participation payments, usually as bill cred-its or discount rates. In market-based programs, participantsare rewarded with money for their performance, dependingon the amount reduction of electricity usage during criticalconditions.

The Direct Load Control (DLC) is a classical program,and it enables the power utility to remotely cycle or turn offconsumers’ electrical equipment [5]. These loads (typically

TABLE IIICLASSIFICATION OF INCENTIVE-BASED DR PROGRAMS.

appliances such as air-conditions and water heaters) may bedirectly dispatched by the power utility, based on the balancebetween consumption and generation. The load control isfeasible through the installation of switches at the customer’spremises that communicate with the power utility. In somecases, the power entity can also send control signals tothe customer in order to influence the control action. DLCprograms are mainly offered to residential or small commercialcustomers and they can be normally deployed within a rela-tively short notice [114]. Consumers that participate in a DLCprogram receive incentive payments in advance, in order toreduce their consumption below predefined thresholds. DLC isa common solution for the residential sector in the U.S.A. [54],where utilities offer incentives for the installation of remotecontrol switching systems for air conditioners or other directlycontrolled loads. However, DLC programs are rarely useful inthe industry section, due to the specific characteristics of theindustry loads that prevent any load adjustments [54]. DLChas been applied also in various DR programs [132]- [136].For instance, a DLC program is considered in [132], whichtargets the reduction of the power consumption in an in-homeenvironment, where both real-time and scheduled appliancesare considered.

Another classical program subclass is the Interrupt-ible/Curtailable (I/C) load, where upfront incentives are alsoprovided to participant consumers. An I/C program considerscurtailment options, e.g. curtail a specific part of electricload or curtail the total consumption to a predefined level.Furthermore, they provide a rate discount or bill credit byagreeing to reduce load during system emergencies [28].Customers that do not respond to these options receive anumber of penalties that are defined in the program’s termsand conditions [137]. These programs are traditionally offeredto larger customers with power consumption that range from200 KW, for the baseline interruptible program, up to 3 MW[138]. These customers must respond within 30 to 60 minuteswhen being notified by the utility, while the total amount oftime that a utility can call interruption is capped (not morethan 200 hours per year) [138].

Emergency DR Programs (EDRPs) are a kind of Market-based programs, but can be also considered as a combina-tion of DLC and I/C programs, since they provide incentivepayments to consumers for reducing their power consump-tion during reliability triggered events [139]. Consumers can


choose not to curtail and therefore forgo the payments, whichare usually specified beforehand. EDRP has been included inthe programs offered by the New York ISO, which managesNew York’s electricity grid [140]: participants that are ableto reduce consumption are subscribed to the EDRP and theyare called during emergency conditions. The EDRP pays forenergy during times of emergency, but does not pay forcapacity [139]. In [141], the authors propose an event-drivenEDRP scheme that prevents a power system from experiencingvoltage collapse. The avoidance of a critical event is ensuredby triggering DR, based on a table of DR actions that containsinformation regarding locations and the amount of electricitythat are needed in these locations. However, several real casesof EDRP operation have pointed out that an excessive sheddingof the EDRP could lead to unpredicted power oscillations,which complicate the sequential generation control [142].

Another Market-based program is the Capacity MarketProgram (CMP) that is offered to consumers who are ableto provide predefined load reductions to replace conventionalgeneration or delivery resources [143]. For the reduction ofpower consumption, customers that participate in a CMPusually receive a day-ahead notification and they are penalizedif they do not contribute to the load reduction [28]. Further-more, participants are obliged to demonstrate that a minimumload curtailment is achievable, while they receive guaranteedpayments, even if they are not called to curtail. Customerswho participate in a CMP offered by New York ISO receivepayments when specific requirements are covered: 100 kWminimum load reduction, minimum four hour reduction periodwith a two hour notification, while customers are subject toone test or audit per market period [138]. An economic modelbased on both CMP and I/C programs has been developed in[138], and a simulation study has been conducted that revealthe strong relationship of the incentive-based program and thecorresponding penalties, with the satisfaction level of bothconsumers and electricity suppliers.

Demand Bidding (DB) is another market-based program(also known as negawatt program) and it is usually appliedto large consumers, who offer curtailment capacity bids in theelectricity wholesale market [139]. A bid is accepted if it isless than the market price, where the consumer must curtailhis load by the amount specified in the bid, otherwise he facespenalties [144]. The Federal Energy Regulatory Commission(FERC) assumes that demand bidding “will be less costlythan a program where an end user receives payments greaterthan the market-clearing price to reduce its demand”, [4].Authors in [4] argue that it is a debate whether this idea isacceptable, since the amount of the energy reduction is mainlybased on the consumption history of the bidder; due to thisfact, market designing problems may arise. For example, theElectric Reliability Council of Texas (ERCOT) market hasapplied a balancing up load program, where demand biddingis permitted through the submission of formal bids from theconsumers [145]. However, the capacity payment has notbeen sufficient enough to encourage consumers into submittingformal offers. A possible solution to this problem could bethe DR-aided DB [146], which is a combination of DR (byconsumers) and demand biding that allows energy retailers to

Fig. 6. Effectiveness of DR schemes based on offered motivations in reducingthe peak demand.

request from consumers to curtail loads, in order to reducepeak demands.

As in the case of demand bidding programs, the consumersfollowing the Ancillary Service Market (ASM) subfamily areallowed to bid on load curtailments, but the bids refer to theancillary service market. If the bids are accepted, participatingconsumers are paid for committing to be on standby. In casethe load curtailments are needed, the participants are notifiedby the power operator and they are paid for the energyprovision [139]. For example, the Midwest ISO selected $3500per MWh as “the average cost to consumers of an interruptionof firm demand”and the highest price on its ancillary servicesdemand curve [147]. In addition, California ISO acceptsbids for ancillary services through a demand responsivenessprogram and under specific terms and conditions [140]. Fur-thermore, New York ISO includes ancillary services througha demand side ancillary services program and it providesthree different services: 10-Minute Spinning (Synchronous)Reserve, 10-Minute Total (includes 10-Minute Synchronousand 10-Minute Non- Synchronous Reserve), and 30-MinuteReserve Total (Synchronous and Non-Synchronous) [140].

Apart from their classification to classical and market-basedprograms, the aforementioned incentive-based programs canbe also classified as voluntary, mandatory and market clearingprograms [148]. DLC and EDRP voluntary programs andtherefore participants are not penalized for not contributingto the program. I/C and CMP mandatory programs, whereparticipants that do not curtail loads receive penalties. Demandbidding and ASM are market clearing programs, where usuallylarge customers inform the utility operator for the amount ofload that they are willing to curtail at posted prices.

Incentive-based programs not belonging to the aforemen-tioned categories have been also proposed as parts of DRprograms in other research efforts. An incentive-based modelimplemented at end-user’s premises to curtail/shift electricloads to the right time of the day has been presented in[149]. This model targets on spreading out the demand profileand allowing the utilization of renewable energy sources. In


Fig. 7. Examples of task scheduling and energy management DR methods.

[150], a coupon incentive-based DR program is presentedthat exploits the capabilities of mobile communication andsmart grid technologies. This scheme preserves a simpleflat retail rate structure, while it also provides a voluntaryincentive-based structure to trigger demand reduction, insteadof paying the high wholesale price. In [151], authors proposean incentive-based DR scheme that covers the in-home loadmanagement with the objective to control appliances, suchas air conditions, water heaters, clothes-dryers and electricvehicles. This method assumes that appliances have differentpriorities, depending on the degree of necessity of use.

By utilizing the results of the aforementioned studies onboth time-based and incentive-based DR programs, in Fig. 6we illustrate the effectiveness of these DR programs in reduc-ing the peak demand for the cases of residential, commercialand industrial consumers. Finally, the interested reader mayrefer to [152] for an overview of DR programs (both price- andincentive-based) that have been implemented in competitiveelectricity markets.

C. DR methods based on the decision variable

DR methods can be also sub-classified according to thedecision variable into two main groups: the first group refers toDR programs that decide when to activate the requested loads,while DR programs in the second group decide the amount ofenergy can be allocated to each consumer (or appliance) duringeach time period [28]. Examples of both DR methods areillustrated in Fig. 7. The research articles from literature thatare presented in the following two subsections are summarizedin Table IV.

1) Task scheduling DR methods: The key function of theseDR programs is the control of the activation time of therequested loads. Two types of loads are considered: must-run(or non-schedulable) loads that cannot tolerate any activationdelay (e.g. illumination or refrigerators) and schedulable loadsthat can be stopped, adjusted, or shifted to other time slots(e.g. water-heaters or PHEVs) [153]. Other parameters thatare taken into account are related to the available energy,

predefined deadlines and operating times of the loads [154].The main target of these DR programs is to reduce thepower consumption in peak-demand hours by shifting loadsto off-peak hours. This is typically realized by using a targetpower-level that should not be reached at peak-demand hours.Such a target power-level is used in [132], as a methodbased on communication protocols in order to achieve uniformoverall power consumption. Furthermore, the current powerconsumption is used in order to decide the scheduling of powerrequests in [19], where two power demand control policies areproposed and corresponding analytical models are presented.The first policy assumes that a power controller activatesimmediately or postpones power requests, based on the currentpower consumption. In the second policy, a new request isactivated immediately, if the total power consumption is lowerthan a threshold, else it is queued.

The shifting of the activation time should be followed byelectricity bill reductions or incentive provisions. In [77] adirect load scheduling algorithm is proposed that mediatesbetween the central control model of DLC and RTP. In thisscheme, costumers voluntarily release the control of theirloads to a central controller, so that their energy use ismanipulated to follow a desired demand profile closely. ADR formulation that takes into account both must-run andscheduled services is presented in [155], by considering theDA-RTP pricing scheme. The problem of optimally schedulinga set of appliances is mapped to the multiple knapsack methodand uniformity in the home energy profile is achieved. Inaddition, a TOU pricing scheme is used in [156], whichdefines the energy price in two different scenarios: the Power-Constrained, Minimum-Cost Scheduling with Fixed Pricesscenario assumes boundaries for the total power consumptionof the consumer, while the Minimum-Cost Scheduling withPower-Dependent Variable Prices scenario assumes that theelectrical energy price is also a function of the total powerusage of the consumer.

In the literature, there is a large number of research articlesthat target the optimization of the task scheduling procedurein order to minimize the total power consumption and/or tomaximize the social welfare. These methods are presented inSection IV.

2) Energy-management-based DR methods: The main ob-jective of energy-management-based DR programs is to reducethe power consumption of specific loads, so that the totalpower consumption in peak-demand hours is reduced [18].This is realized by controlling the appliance’s operation toconsume less power during system stress. For example, ina summer day an air condition could be adjusted to 25oCinstead of 22oC, thus less power is consumed and peoplestill feel comfortable. To motivate consumers to control thepower consumption of their appliances, bill reductions or otherincentives are provided by the power utility. The satisfactionlevel to the results of energy management is the subject of thestudy in [76], where a power-scheduling scheme is proposed.This scheme is driven by the Quality of Experience (QoE)factor that describes the consumer’s satisfaction degree onthe grid’s performance and defines the social welfare of thesystem.


TABLE IVCLASSIFICATION DR PROGRAMS BASED ON THE DECISION VARIABLE.

The consideration of schedulable and non-schedulable loadsin an energy-management scheme is taken into account in[76], [157]-[159]. In these studies a task-scheduling schemeis considered for appliances that consume power in adjustabletime-slots and an energy-management scheme for appliancesthat have flexible demands but are non-schedulable. Therefore,these DR programs decide when to activate a specific setof appliances, while they also decide how much energy toallocate to another set of appliances during each time-slot. In[157], a water-filling based scheduling algorithm is proposedthat allows consumers to shift part of their loads to off-peak hours in a probabilistic way. This algorithm has lowcomplexity, since it utilizes statistical information on powerconsumption, which is available from the power utility. On thecontrary, a cooperative scheduling approach in [158] requiresdetailed and continuously updated information between theutility company and the consumers. Apart from schedulableand non-schedulable loads, this scheme also considers loadsthat must consume a certain amount of power (e.g. recharge-able batteries and PHEVs). Finally in [159], a distributedincentive-based algorithm for scheduling power consumptionis presented. The optimal energy consumption schedule foreach consumer is derived based on an optimization algorithm,while game theory is used to derive a pricing model that offersa motivation for consumers to reduce their loads. Optimizationmethods and game theoretic analysis have been also used ina number of energy-management-based DR programs. Theseprograms are presented in Section IV.

IV. OPTIMIZATION: METHODS IN DR PROGRAMS

In this section, we review the work on the optimization ofvarious DR programs and schemes. Given that optimizationis defined as the process of finding the conditions that givethe maximum benefit or minimum cost of a process ( [160]),published studies on optimization of DR programs targetthe minimization of the total power consumption and/or themaximization of the social welfare. The latter term refersto the utility’s profit (total consumer willingness to pay)minus the total cost experienced by all the generators andwastage cost caused by transmission losses [18]. Therefore, themaximization of the social welfare is achieved by maximizingthe difference of the utility’s profit by total electricity cost.

The target of an optimization problem is to find a set ofvariables that minimizes (or maximizes) a function (or a setof functions) of this set of variables, while these variablesare subject to a set of constraints. The set of variables isknown as the design vector, while the function is termed asthe objective function [161]. The design vector is defined by

the variables of the specific DR problem. For example, in atask-scheduling scheme the design vector can be determinedby the demand request start time, the operational time of theload, the type of load (e.g. the type of an appliance in theconsumer’s residence) and the priority of the request. Also, inan energy-management based scheme, the design vector can bedefined by the load type, the amount of power that is reducedand the load operational duration under the reduced load. Theobjective function is defined based on the desired characteristicthat is optimized, e.g. the total power consumption or the socialwelfare. Finally, the constraints are determined based on theconditions of the DR scheme under study. Typical parametersthat are constraints in an optimization problem refer to theoperation of the system, such as capacity constraints, energystorage constraints and appliance constraints (e.g. the totalenergy required for the operation of an appliance).

Optimization problems can be classified based on the natureof the design vector, the objective function and the con-straint functions. For example, if at least one of the objectiveand constraint functions is non-linear and all the variablesof the design vector are integers, then the problem is aninteger, non-linear programming problem, while if some ofthe variables are integers, then the problem is an mixedinteger non-linear programming problem [160]. Furthermore,based on the deterministic or the stochastic nature of thevariables involved, optimization problems can be classifiedinto deterministic and stochastic programming problems; thelatter case defines optimization problems that deal with re-newable energy sources, due to the stochastic nature of thesesources, or other uncertainties and correlations. Based onthe type of the optimization problem, a technique is definedfor the derivation of the solution; therefore for an integerlinear programming problem, an integer linear technique isapplied for the optimal solution derivation. It should be notedthat optimization does not necessarily mean that an optimumsolution is reachable. There are optimization problems (e.g.NP-hard problems), where a solution is not feasible to find[162], or the computational times are too high. In such cases,classical optimization procedures, such as linear programming(which have been widely deployed when large problemsare modeled), or quadratic programming cannot be applied;therefore, if the complexity of a solution technique is high,heuristic approaches can be used, since they provide fast andnear optimal solutions.

In this section we firstly categorize the optimization ap-proaches in DR problems based on the target of the opti-mization procedure. These categories are: a) minimization ofelectricity cost, b) maximization of social welfare, c) mini-mization of aggregated power consumption, d) joint minimiza-tion of electricity cost and aggregated power consumption,and e) joint maximization of social welfare and minimizationof aggregated power consumption. Therefore, the first threecategories deal with a single objective function, while thetarget of the last two categories is the optimization of twoobjective functions. Furthermore, we present game-theoreticmethods for the determination of the optimal solution, whicheither refer to the minimization of electricity cost or themaximization of the social welfare. Finally, we highlight the


optimization models for two important smart grid paradigms;V2G systems and microgrids. We also categorize the opti-mization problems based on the solution method that is usedin order to derive the optimal solution. Table V presents theDR optimization models for each one of the aforementionedcategories, together with the solution methods that have beenconsidered for each case. Furthermore, in Table VI we classifythese optimization models according to the control mechanism,the decision variable and the pricing scheme. Finally, inTable VII the presented optimization models are classifiedaccording to the ability to include uncertainties, scalability,responsiveness, communications requirements, and support ofmultiple load types.

A. Minimization of electricity cost

The main objective of an optimization algorithm aims tobring the final load curve close to the objective load curve,such that the desired objective of the DR strategy is achieved.The objectives of a DR strategy could be to maximize theuse of renewable energy resources, to maximize the economicbenefit for the power utility, to minimize the electricity orgeneration cost, and/or to reduce the peak load demand.The formulation of the cost minimization problem is basedon the derivation of an optimal load scheduling procedurethrough the definition of an objective function that is basedon the application of an appropriate pricing scheme. The mainchallenge for the determination of the optimization problemis the design of an optimal load scheduling scheme thatconsiders the characteristics of each load and the specialneeds of the consumers (e.g. the temperature bounds specifiedby users for thermostatically control appliances, to reflecttheir comfort). The latter features define the design vectorof the optimization problem, while the restrictions of thesefeatures determine the nature of the optimization problem(e.g. as a nonlinear optimization problem). Furthermore, thecharacteristics of the objective function and the design vectordetermine the computational complexity of the solution, whichis also affected by the number of consumers considered in theoptimization problem.

There are several research articles that aim to the costminimization objective. These articles can be further sub-categorized based on the applied optimization procedure andthe technique that is used for the derivation of the solution.

A number of cost minimization problems are solved byusing complex, well-known optimization procedures. An In-teger Linear Programming (ILP) method is used in [163] toderive the minimum electricity cost for local (single house)or global (multiple houses) applications. In both cases, theproposed methods decide when appliances are switched on oroff, as well as when generators are activated and deactivated.The ability of the proposed method to deal with predictionerrors allows the preservation of the consumers comfort,even in conflict situations. However, the proposed approachprovides optimal solutions over a specific scheduling window,without accounting for time periods beyond the set window,which could lead to sub-optimal solutions. Furthermore, aMixed Integer Programming (MIP) method is used in [164]

to minimize the total cost (operation, reserve and expectedload-not-supplied cost), while it demonstrates that DR canbe efficiently used as both a reserve supplying and a peakclipping resources. The calculation of operation and reservecosts is based on two utilization patterns of DR methods.The Peak Clipping DR activates a task-scheduling procedurewhen the system is at stress, while the Reserve SupplyingDR is an I/C-based DR method that is applied as a reserveand it is activated in emergency conditions. A Mixed IntegerLinear Programming (MILP) method is used in [165] forthe minimization of the household’s electricity payment byoptimally scheduling the operation and energy consumptionfor each appliance, while considering the waiting time as acomfort setting for the operation of each appliance, and anRTP scheme. In addition, an appliance commitment algorithmis presented in [166] and solved with a multiple-loopingalgorithm and enhanced by a linear sequential optimizationprocess. This algorithm schedules Thermostatically ControlledAppliances (TCAs) based on price and consumption forecasts.Customer comfort constraints are incorporated in the proposedapproach, by specifying time-varying temperature ranges foreach TCA.

The optimization model of [167] considers two types ofappliances: the first type consists of task-scheduling appliances(their consumption can be adjusted across time), while thesecond type consists of energy-management appliances (theirconsumption can be reduced but cannot be shifted to nexttime-slots). Together with the electricity cost minimization,the formulated problem deals with the minimization of theconsumers’ discomfort, due to the adjustments in the ap-pliances’ operation. The resulting optimization problem isnon-convex. In general, non-convex optimization models aredifficult to solve, since they are computationally intractableand convex relaxation techniques should be applied, in order toconvert the problem to a convex optimization problem [168].However, in [167] under the consideration of a continuoustime horizon, the non-convex problem has a zero duality gap(difference between the primal and the dual problem, whichis usually observed in convex optimization problems), and itcan be solved by using Lagrangian algorithms. Similarly, themodel in [169] considers multiple appliance’s types; the dualoptimization problem of minimizing the electricity cost andmaximizing the consumer’s satisfaction is balanced into theoptimization problem of maximizing its payoff. This problemis also non-convex and a simulated-annealing-based pricecontrol algorithm is developed to provide the optimal solution.

The aforementioned optimization approaches that are basedon linear or convex programming provide efficient solutions,with polynomial time complexity, while the optimality of thesolution is definite [170]. However, the main problem of theseapproaches is their high complexity, especially when a largenumber of consumers is considered. For such cases, heuristicapproaches can provide fast and near optimal solutions. Forexample, a heuristic-based evolutionary algorithm is proposedin [171], with primary objective to reduce the utility bills ofconsumers in residential, commercial and industrial areas. Thisis realized through a load shifting technique, for the support ofa large number of loads of various types. The Binary Particle


TABLE VCLASSIFICATION OF DR OPTIMIZATION PROBLEMS AND THE SOLUTION METHODS FOR EACH CASE.

Swarm Optimization (BPSO) heuristic algorithm is used in[172] for the optimization of the demand management anda Particle Swarm Optimization (PSO) [173], for the optimalresource management. A near optimal solution is also providedin [174], which is based on a greedy search heuristic. Theapplication of this heuristic method results in an effectivelyflattened demand curve, even though the consumer’s dailyelectricity bill is not minimized; however, this method focuseson managing individual household demand. The approach in[175] also focuses on individual household demands, but itconsiders various types of appliances, by taking into accountboth inelastic (must-run) and elastic (shift-able) power de-mands from various residential appliances, renewable energysources and energy storage devices, for the formulation ofthe optimization problem. The proposed cost minimizationproblem in [175] is transformed to a relaxed problem withlower complexity, which is solved by a modified Lyapunovoptimization technique. In the same way, the convex pro-gramming problem of cost minimization is relaxed in [176],so that it can be applied to a large number of consumers.

This is realized by relaxing the binary decision variables,associated with the consumer-appliances’ status, from integerto continues values, in order the use of a complicated ILPscheme. Results show that in general, the proposed modelprovides a solution that falls within 1% deviation of theoptimal solution.

Several research efforts appear in the literature that deal withthe optimization problem in V2G systems. V2G technologyallows PHEVs to feed energy directly back into the powergrid [177]. This energy transaction is realized though theexchange of a lot of information between the vehicle, thecharging station, and the utility. This data exchange betweenthe three parties requires a robust two-way communicationinfrastructure, in order to offer reliable services regarding theregistration of the PHEV to the V2G program, the estab-lishment of the charging session, the delivery of chargingstatus data, and to correctly bill the customers, based ontheir selected rates. However, the formulation of optimizationproblems should also consider a number of challenges in bothunidirectional and bidirectional V2G solutions. In the bidirec-


tional power flow option, additional hardware is required in thevehicles to pump the energy back to the grid, while the powerutility must convince consumers to allow discharging theirbatteries. On the other hand, the limitations in unidirectionalsystems refer to the reduced participation times, due to batterycharging and the lower power levels. This is mainly due to theinability to pump the energy stored in the batteries back to thegrid, and the lower overall performance of the unidirectionaloption, compared to the bidirectional solution [83], [178].Nevertheless, it is acknowledged that both V2G systems havethe potential to provide financial benefits to utilities [179], CO2

emission reductions [180], and power consumption increaseduring off-peak hours [181].

It should be noted that the optimal resource schedulingproblem that leads to the minimization of the cost may havesignificant computational complexity, especially if the V2Gsystem supports a large number of vehicles. To overcome thisproblem, artificial-intelligence-based techniques can be usedin order to reduce the execution time. PSO is a techniquethat exploits simple analogues of social interaction, instead ofpurely individual cognitive abilities. This technique is used in[23] and [181] for the cost minimization in a V2G system,while considering renewable energy sources. The optimalsolution in [23] achieves low operational costs and CO2

emissions; however, it was pursued not for each vehicle, butonly from the perspective of efficient grid operation, while thepricing mechanism of regulation was based on the availablepower capacity, not the generation cost [182]. On the otherhand, the PSO approach of [181] is applied on dynamic datafor the generation of intelligent control of the vehicles andachieves intelligent scheduling of vehicles, as well as thermalunits; however, the aimed operational cost and CO2 emissionsreduction can only be achieved through the utilization ofrenewable energy sources.

The effectiveness of the PSO approach is verified throughits comparison with a reference methodology based on Mixed-Integer Non-Linear Programming (MINLP) in [183]. Thiscomparison reveals that the PSO approach achieves signifi-cantly shorter execution times, but slightly higher total cost(production cost plus V2G discharge cost), compared to theMINLP solution. A similar conclusion is presented in [184]for the performance of another technique that reduces thealgorithm execution time. This technique is called simulatedannealing technique [185], and its application to a V2G systemis based on several constrains regarding the electric vehicles,such as battery capacity, charging/discharging rates, startingtravel times and minimum travel distances, while it considersthat vehicles are distributed through the distribution networkand there is no centralized parking lot. The evaluation of thistechnique under a scenario that considers 1000 vehicles ledto a faster solution, compared to both mixed-integer nonlinearand linear programming approaches, but its solution is slightlymore expensive, compared to the optimal solution of the linearapproach.

B. Maximization of social welfareThe maximization of social welfare is achieved by defining

the objective function as the difference of the total utility’s

profit minus the total cost of energy generators and transmis-sion networks. In this case, the design vector is defined notonly by the specific characteristics of the loads, but also bythe generation and transmission line capacities. However, ifmultiple generators are considered (e.g. in a renewable energyresources environment), additional constraints should be de-fined, while distributed algorithms are required for the deriva-tion of optimal solutions. Therefore, the various solutions thathave been proposed for the maximization of the social welfarevary depending on the implemented optimization technique,the constraints that are taken into consideration in the problemformulation, as well as on the applied pricing scheme.

The maximization of the social welfare has been formulatedas a convex optimization problem in a number of researchapproaches. The model in [186] assumes that specific paymentrules should be applied to consumers that are unwillingto reveal their real power demands. The proposed convexoptimization model is solved based on the knowledge of theconsumers’ demands. This solution determines the optimalpower allocation to each user and provides the maximum so-cial welfare, although the significant computational complexitylimits the applicability of this model to small residentialnetworks. Convex optimization problems are also formulatedin [187], [188], that affect the energy procurement decisionsof the consumers for the amount of the purchased balancingpower needed to meet the aggregated demand. In this case, thesocial welfare maximization is the result of a joint optimizationprocedure. The first part of the overall problem refers tothe power utility decisions for the day-ahead procurement ofelectricity on the wholesale electricity markets, and the secondpart refers to the real-time decisions of consumers for the loadschedule. In addition, the model captures the uncertainty of theelectricity supply from renewable energy sources, but it adjustsall appliance’s power consumptions; the later considerationmay not be valid for some non-interruptible appliances. Fur-thermore, the convex optimization problem that is presentedin [18] is solved by a decentralized Lagrange-Newton method,by considering a single power demand per consumer. Forthe determination of the maximum social welfare, the modelconsiders the energy demand and the generation decisionsthat reduce transmission losses. The interior point method isused for solving the convex problem of maximization of thesocial welfare in [112]. This approach considers both must-run and controllable loads, while it suggests that users areindependent decision makers and schedule their loads basedon an energy-management scheme, so that predefined powerconsumption levels are met. The energy-based scheduling ismotivated by a VCG pricing method, which is proved to bea more efficient solution than a PLP method, by providing anumerical evaluation.

A mixed discrete-continuous optimization nonlinear prob-lem with a single integer variable is presented in [189], whichtackles the optimal integration of renewable energy systems.A number of constraints, such as voltage level, active andreactive power constraints for generators and consumers andflow constrains for lines and transformers are all included inthe proposed analysis. The presented results evidenced thatthe combined operations of renewable energy systems and


price responsive demands alleviate network constraints, whilesatisfying greater demand levels and reducing energy costs. ANon-Linear Programming (NLP) problem is formulated in [98]for the maximization of the power utility’s profit and is basedon the assumption of a single power demand per household,while also considering other parameters, such as consumer’sbenefits and reactions to energy prices, minimum daily powerconsumption and constraints in the distribution network. Theproposed scheme is based on a DA-RTP pricing method.The optimization problem is solved by using commercialsoftware, while the Benders decomposition method is appliedto make the model applicable to a large number of consumers.Furthermore, variations of a Coevolutionary PSO (CPSO) areused in [190], in order to incorporate the coordination of dis-tributed energy resources to the social welfare maximization.The model proposed in [190] considers multiple appliancesper consumer residence, but its main disadvantages are lowconvergence rates and large communication requests.

The target of the optimization model presented in [191]is the maximization of the utility’s profit, together with lowcosts to the consumers in a unidirectional V2G system. Inboth cases, the problem is formulated as a linear program.It should be noted that unidirectional V2G systems requiresimpler control and can support a larger number of vehicles,which are likely to be connected for relatively short intervals,since vehicle owners may not need to connect a fully chargedvehicle [46]. Therefore, the same authors study the equivalentproblem, but for a bidirectional V2G system in [192], in orderto higher benefits for the utilities, although bidirectional V2Gsystems introduce higher risks due to the extra capital costs forthe implementation of a bidirectional system. Furthermore, theprofit maximization problem is also tackled in the bidirectionalV2G system of [82], where a real-time pricing scheme isconsidered in order to deal with the price uncertainty in V2Gsystems. The scheduling problem is formulated as a MarkovDecision Problem (MDC), where the decisions on the pricesare made by considering the future profits. This problem isthen solved by using a Q-learning algorithm, which is aniterative method that learns from experience and updates ineach step. The main limitation of this algorithm is that thelearning process is time consuming.

Heuristic approaches have been also proposed for the de-termination of the maximum social welfare. In [193], twomodels are developed for the minimization of the consumer’selectricity cost and the minimization of the generating costfor the utility companies. The former problem is solved by agreedy algorithm that is applied to each time period for theoptimal load scheduling, while the latter problem is solved byusing a filling method. PSO has been used in an optimizationmodel for V2G systems in [193] for the maximization ofthe utility’s profit, which satisfies both the system’s and thevehicle’s owners constraints. However, the model of [194] doesnot consider the effects of a fully occupied parking lot on thedistribution system of the grid, while its centralized algorithmlimits the applicability of the model to V2G systems with alarge vehicle population and dynamic arrivals.

TABLE VICLASSIFICATION OF DR OPTIMIZATION METHODS BASED ON CONTROL

MECHANISM, DECISION VARIABLE AND PRICING SCHEME.

C. Minimization of aggregated power consumption

The minimization of the power consumption is usuallyachieved by finding the optimal scheduling solution that con-sumers can use to schedule their loads to off-peak hours. Theoptimal scheduling determination is based on the incentivesoffered to consumers for reducing their power consump-tion. Therefore, in these power-consumption minimizationproblems, the scheduling plans should be carefully definedin the design vector, especially when different applianceswith diverse power demands are considered, in order toachieve reduced cumulated power consumption. Moreover,the scheduling decisions should take into account the con-sumer preferences, the applied pricing procedure, the presenceof scheduling-flexible and inflexible loads and the require-ments/constraints of the different loads. On the other hand,the minimization of the aggregated power consumption canbe also achieved by encouraging consumers to optimize theirown loads.

Various optimization techniques have been applied for thedetermination of the optimal load scheduling. A MINLP isproposed in [195], that involves resource scheduling with dif-ferent anticipation times: day-ahead, hour-ahead and 5 minutesahead, while generators, storage units and intra-day marketare considered by the hour-ahead management. Additionally,the proposed model considers the intensive penetration ofdistributed generation and the load curtailment opportunitiesthat are enabled by the DR program, while it supports non-shiftable, time-shiftable, and power-shiftable appliances. Com-mercial optimization software is used to solve the presentedproblem based on a genetic algorithm, while the obtainedsolutions are validated by using a power system simulation.However, the utilization of genetic algorithms leads to slowconvergence and long execution times, although they generallylead to the global optimum.

Optimal daily load scheduling is also achieved in [196]through an ILP technique, which is solved by using the Branchand Bound method. The resulting solution is used to derive


the optimal power consumption, by taking into account loadsthat can shift their operation in succeeding time periods andloads that can adjust their power consumption in the currentperiod, but non-interruptible and uninterruptible loads. Theproposed mechanism can be applied either in the residentialenvironment, where a home energy management unit makesoptimal scheduling decisions for all appliances, or in a localarea, where scheduling decisions are made by a central controlunit. Similarly, a linear programming model is also usedin [197] for the optimization of the peak load reduction,through customer direct load control programs for schedulingcommercial, industrial and residential loads; however thisapproach does not consider the underlying physical process,but only the consumed power of loads.

The convex optimization model that is proposed in [126]is easily scalable to large number of users. This is due to thefact that the analysis is based on a task scheduling schemewith heterogeneity in delay tolerances. This fact allows theenergy provider to estimate these delay tolerances based onlyon the aggregated data and not on the parameters of eachconsumer. Similarly, another convex optimization problem isformulated in [198], that takes into account distributed powergenerators and electricity prices that are analogous to thetotal power consumption. This problem consider users withand without renewable energy sources and considers that eachuser optimizes his own load, so that their electricity bills arereduced. A parallel distributed optimization approach is usedfor the solution of this problem, which significantly reducesthe time complexity and communication costs. Moreover, aheuristic approach that is proposed in [199] can be appliedto a large number of consumers. This methodology is calledSignaled PSO (SPSO) and it is used in order to addressthe energy resources management problem. The comparisonof the SPSO method with other methodologies revealed thesuperiority of the proposed scheme in terms of convergence,cost and time execution and absolute error. The considerationof a large number of thermostatic loads in [200] convertsthe optimization problem into a non-convex problem, witha significantly high complexity due to the on-off controlof such devices. To this end, the authors of [200] proposea distribution structure for the optimal solution derivation,with a supervisor center (that broadcasts coordination signals)and local controllers for the consumer’s appliances. Eventhough the distributed approach of the problem requires thecommunication between the thermostatic devices, the authorsclaim that the communication requirements of their proposedscheme are low.

Significant work has been conducted for the developmentof optimization models in microgrids [201]- [43]. In bothgrid-connected and islanded microgrid schemes, the optimalcontrol is directly affected by several factors, such as theavailability of renewable energy sources, the load distributionvariations among users and the changes electricity prices.These parameters are key factors that should be consideredfor the formulation of optimal scheduling mechanisms ofelectricity supply and demand in microgrids. The incorporationof renewable energy sources and storage devices in a microgridcase is modeled as a mixed integer programming model in

[201], which determines the optimal operation schedule ofloads. The impact of renewable generators and power storagedevices on the optimal schedule of loads is studied and anestimation of the storage capacity according to the size of themicrogrid is derived. The optimal scheduling of residentialpower consumption is also studied in [204], by introducinga mixed integer programming model. This model targetsthe minimization of the total one day-ahead expense of theresidential power consumption, which is achieved through thedetermination of an optimal scheduling for both the operationof the consumer’s appliances and for the distributed energygenerators. In the latter case, the derivation of the optimalscheduling is based on renewable energy output forecast (e.g.weather forecast is used for the case of wind generators).

The combination of DR and distributed generators is chal-lenging, mainly due to the high computational complexity ofcentralized optimization process. To overcome this problem,a 3-step optimization procedure is proposed in [43], for theminimization of the total power consumption of a distributedmicrogrid that services a residential community. The first stepconsiders the dynamic DR based on day-ahead time-varyingpricing, for the reduction of the energy consumption cost of theconsumer, while the second step considers the reduction of theelectricity cost in the entire microgrid. The third step involvesthe management of the storage of surplus wind energy and thedischarge of this energy in high-demand hours. The electricitydemands that are taken into account in this model are not onlyschedulable tasks, but also tasks that can be dropped by theagent to prevent system stress. The optimization problem forthe first two steps is solved with a PSO algorithm, which isconsidered as an advantageous method over other evolutionaryalgorithms, while the optimal solution for the third step isderived by applying a Q-learning algorithm.

D. Minimization of electricity cost and aggregated powerconsumption

In this section we target on both the minimization of theelectricity cost and the total power consumption. Typically, op-timization problems that consider multiple objective functionsare solved either by considering decomposing algorithms, orutilizing Pareto-based optimization methods [205]. The maintarget of a decomposing algorithm is the uncoupling of thepower system into subsystems with reduced complexity; thesesubsystems are optimized separately by utilizing methods thatare compatible with the optimum of the entire system. InPareto-based methods, relations are created among solutionsbased on a Pareto-dominance concept and the set of optimalsolutions is extracted based on these relationships. A third so-lution is the utilization of aggregated weight functions, wherethe objective functions are combined into a single function,while weights are considered to depict the importance ofeach original objective [206]. For example, the dual-objectiveproblem of minimizing the cost in a microgrid and at the sametime minimizing the amount of pollutants released into the airby thermal units (which are functions of the power generation)of [207] is solved by generating the non-dominated set ofPareto-optimal solutions.


The scheduling energy consumption problem is formulatedas a linear programming problem in [208], for a deterministicapproach of the scheduling problem. The proposed robustoptimization model explicitly addresses the problem of cor-related price data, while the principal component analysisand the minimum power decomposition methods are used forthe solution of the robust problem. The model considers thatcorrelation may exist among the uncertain electricity price dataof successive time periods over which consumption is to bescheduled. For the evaluation of the proposed model, two casestudies are presented that use prices from the Brazilian market.They tackle the determination of the appropriate consumptionscheduling algorithms that could be used in the specific marketand achieving the highest possible energy transfer from high-to low-demand periods.

The dual minimization problem of both the operational costand the load power are solved by using a MILP model in[209], for the proposed home energy management system withdistributed energy resources. The presented case study refersonly to a single household, which cannot be an indicator forthe complexity level of the model. On the other hand, the casestudy for the home energy management system proposed in[210] considers a scenario with 60 residential users with threecontrollable loads in each residence. The proposed analysistargets the minimization of the real-time market cost and thetotal energy consumption. Customers are encouraged to par-ticipate in the program by using the motivation that they willnot pay additional money compared to the cost they optimizedby participating in the program. The optimization problemis formulated as an approximate certainty equivalent controldynamic programming problem, while its computational com-plexity is highly affected by the number of residential usersand the number of controllable appliances. To increase themodel scalability to higher numbers of users and appliances,a decentralized algorithm decomposes the problem to parallelsubproblems, where each residence computes its schedulingsolution locally. However, due to multiple iterations requiredby the decentralized algorithm, the model requires real-timeexchange messages between the users, thus causing possibleoverheads in the communication network.

E. Maximization of social welfare and minimization of aggre-gated power consumption

The dual optimization problem of social welfare maximiza-tion and optimal power consumption can be solved by usingvarious optimization schemes. Convex optimization modelshave been used in [114], [158]. In the latter case, the dualproblem can be solved under a cooperative multi-residencescheduling approach, which assumes that each consumer hastwo classes of adjustable loads. Loads that belong to the firstclass must consume a specified total amount of energy overthe scheduling horizon, but the consumption can be adjustedacross different time periods. Loads that belong to the secondtype have adjustable power consumption patterns, without atotal energy requirement, but the load operation at reducedpower results in discomfort at the end-user side. The resultingconvex optimization problem is solved through a distributed

TABLE VIICLASSIFICATION OF DR OPTIMIZATION METHODS BASED ON VARIOUS

PERFORMANCE METRICS.

subgradient algorithm. Similar communication requirements,as the ones in [114], are needed in the analysis presented in[158], but in the latter case a single type of adjustable loadis assumed. The proposed energy and welfare optimizationmodels are based on a real-time pricing scheme. This problemcan also be solved by using a convex programming technique,such as the centrally-fashioned interior-point method.

Authors in [211] argue that the social welfare maximizationand the optimal resource allocation introduce communicationnetwork externalities, caused by the uncertainty in messagesignaling due to transmission network constraints. Therefore,


the proposed framework jointly optimizes the data networkcomponent of the smart grid, so that the uncertainty on thecommunication (e.g. delay) is reduced. The optimal schedulingprocedure is derived by a distributed algorithm that considersthe fact that a consumer may be affected by the action of otherconsumers.

The aforementioned optimization models consider a groupof consumers that are served by a single load-serving entityor that each consumer has only two classes of appliances.The key assumption of the analysis presented in [212] is thateach residence is equipped with different appliances of diversepower demands. In addition, a consumer is also equipped witha battery that provides further flexibility for optimization ofthe residence’s power consumption across time. The optimalpower scheduling and maximization of social welfare prob-lems are solved by a distributed algorithm, which considersthat the power utility and the consumers jointly computean equilibrium based on a gradient algorithm. In the latteralgorithm, the power utility sets the prices to be the marginalcosts of electricity and each consumer solves its own netbenefit maximization problem in response.

F. Application of Game Theory to DR Programs

Game theory can be considered as a collection of analyticaltools that provides the understanding of phenomena that areobserved when decision-makers interact [213]. Consequently,game theory is suitable to address demand response man-agement, where the players are the consumers, the actionsare the strategies that players follow to optimize a utilityfunction, while the solution (the outcome of the game) isthe optimal utility function [214]. Based on the target of thesolution, game-theoretic methods can be classified by usingthe categorization of the optimization procedures that arepresented in the previous subsections.

Game theory has been used for the formulation of theappropriate energy consumption procedure that results in min-imum electricity cost. The scheduling game that is proposedin [8] considers consumers as the players and the daily loadschedules as their strategies. The target of this approach is tominimize the electricity cost, which is achieved at the Nashequilibrium, and also to minimize the Peak-to-Average Ratio(PAR). Users interact with each other via message exchange,in order to coordinate their electricity usage, so that reductionof the PAR is achieved. The resulting problems are solved byusing the interior point method.

A game theoretical method can also be used to capturethe conflicting economic interests of the retailer and theirconsumers. Authors in [215] propose optimization models forthe maximization of the expected market profits for the retailerand the minimization of the electricity cost for the consumer.The proposed approach considers real-time prices for must-runloads. Solutions to these two separated problems are providedthrough the formulation of one bilevel problem as an MILPmodel. The MILP solution is provided by using commercialoptimization software. However, the computational complexityof the proposed approach is significant and the authors providean example with only three consumers. In addition, auctioning

games have been used in [216] to allocate load demandsamong customers, while maximizing the social welfare. Theserepeated Vickrey auctions use the optimal demand schedulingproblem, which is solved by applied a water filling heuristicmethod. Furthermore, the utility cost minimization problem isformulated as a convex optimization, with a solution that isderived under the generation capacity constraint. The solutionsof the two problems are based on the assumption of a single-used type scenario and are used to maximize customers’ socialwelfare. Similarly, the target of the proposed market modelsin [217] is the maximization of the social welfare, as wellas the matching of supply with demand in competitive andoligopolistic markets. However, the proposed models considera single type of load per consumer, while the models are basedon convenient forms of objective functions for the electricitycost and utility functions.

A problem that can significantly affect the performance of aDR program is the unwillingness of consumers to reveal theirreal power demands. This problem is the subject of the studyin [218], where a cheat-proof game theoretic DR method isproposed. The participation in the program is motivated bya simple RTP scheme, where consumers calculate their ownoptimal demand and report it to the utility. A similar problemmotivated the game-theoretic approach in [219], where theutility uses a TOU pricing scheme that is announced aheadof time. In this way, consumers are not obliged to respondto the complex procedure of a time-varying pricing process.The designed utility function (profit minus the cost of theusers) is optimized with linear constraints and solved for Nashequilibrium.

Game theoretical models have also been employed for theoptimal solution to the dual minimization problem of theelectricity cost and the aggregated power consumption. In[220], a two-level optimization framework is presented, witha game-theoretic framework at the upper level and a staticconvex optimization problem at the lower level. On the upperlevel, a nonzero-sum differential game is used to capturethe interaction among different players, who seek to find anoptimal demand policy to maximize their long term payoff. Atthe lower level, a static convex optimization problem leads toan optimum solution, which is found for the scheduling of eachconsumer, while considering different appliances per house-hold. Similarly, a Stackelberg game model is used in [119]for the optimal scheduling of loads. In this game, the utilityplays the leader level game by setting the real-time price andthe consumer plays the follower level game and schedules theappliances’ operation. An RTP-based pricing scheme is used,where prices are determined by the utility according to thecurrent load level, while optimal power consumption occursby shifting power consumption in forwarding time periods.Furthermore, a non-cooperative game theoretical framework isapplied, to model the DR problem in a distributed smart gridenvironment, which is equipped with energy storage devices,connected in a decentralized fashion. Based on the gametheoretical approach, the distributed algorithm is determinedand minimization of both the aggregated power consumptionand the total energy cost is achieved. A Stackelberg gamemodel has also been used for the interaction between operators


and consumers in [221]. This game allows operators to de-termine optimal electricity price and consumers their optimalpower consumption in a network of multiple utility companiesand consumers. The latter assumption increases the model’scomputational complexity and limits its applicability to largesmart grid networks.

Game theory may also be applied for power scheduling andcontrol, especially for the case of distributed microgrids. Theinterested reader may refer to [214] for an overview of gametheoretic models in distributed microgrids. Furthermore, gametheory has also been applied for the interaction of PHEVs andthe retailer, where players are the PHEVs who are involved in anon-cooperative game, while their actions are the load demandvalues. The solution of the game is obtained for the maximumprofit from the retailer’s perspective, and the optimal tradeoffbetween the benefit from battery charging and the associatedcost, from the PHEV’s perspective. This target is achieved bythe generalized Stackelberg game presented in [222], wherePHEVs select their strategies in order to optimize their benefit,while the retailer consider these strategies to maximize itsprofit. Apart from the optimal profit of either the retaileror the PHEV owners, other characteristics of the PHEV-retailer system can be optimized. The authors in [223] presenta four-stage nested game, where players are not only thePHEV owners, but also residential users. The objective of theretailer is to maximize its profit, but also to perform frequencyregulation through matching the power supply with demand.The optimal performance of frequency regulation in a real-timepricing scenario is also the objective of the game-theoreticmodel presented in [224]. However, this model does notincorporate the dynamics of the regulation signals, the energyrestrictions of the EVs’ batteries and the battery degradationdue to frequent charging/discharging. These characteristics ofthe PHEV-aggregator system are considered in [225], wherea stochastic optimization problem is proposed, based on theLyapunov optimization technique. It should be noted that theconstraints in the aforementioned models refer to the aggre-gated demand of the PHEVs and to the charging/dischargingprocedures (e.g. charging and discharging cannot be performedsimultaneously).

V. CONCLUSION: LESSONS LEARNED AND FUTUREDIRECTIONS

In this paper we presented the background and key char-acteristics of DR programs that have been proposed for theefficient operation of the smart grid. We provided an extensivereview on DR methods and we classified them into categoriesbased on the control mechanism of the DR program, on themotivations offered to customers to reduce or shift their powerdemands, and on the DR decision variable. We also reviewed awide range of optimization algorithms that have been proposedfor the optimal operation of the smart grid and we provided adetailed classification of these optimization models based onthe target of the optimization procedure, the solution methodsthat have been considered for each case, the ability to includeuncertainties, scalability, responsiveness, communications re-quirements, and support of multiple load types, with specialattention given to V2G systems and microgrids.

A. Lessons learned from DR programs

From the study of the various DR schemes and programs anumber of useful lessons can be derived. Firstly, the successfulimplementation of a DR program relies on the participation ofthe consumers that contribute to reduce of the overall powerconsumption in peak-demand hours. This involvement doesnot imply that consumers have to pull the plug on majorappliances, or to compromise their lifestyle and intimidatetheir comfort. It is up to them to decide the amount of their par-ticipation in a DR program. The study of 70 tests performed ondynamic pricing-based DR pilots carried out in Europe, NorthAmerica and Australia showed that in general, consumersmarginally respond to these programs, while others do notrespond at all [227]. Furthermore, the result of a DR programmay not be beneficial for all customers. For example, theimplementation of DR programs in Victoria, Australia [228]and in Illinois, U.S.A. [75], resulted in higher electricity billsfor the low income earners, since they usually do not use muchenergy and so, the amount of demand reduction is limited. Thelatter fact shows that a number of DR programs do not assurethat all participants are rewarded based on their contributionin achieving the program’s objectives. Additionally, severalDR programs penalize their participants, who usually pay highprices for electricity consumed during peak load hours, evenif the usage is for the base load [194], while non-participantusers are unfairly benefit from the participants’ effort, withouthaving any contribution to the program’s objectives [229].

Based on above-mentioned facts, a DR program should alsooffer information tools to the consumers, regarding the partic-ipation benefits and the optimal use of their appliances, inorder to increase the consumer participation rate. In addition,increased consumers’ participation can be achieved throughthe provision of a variety of DR contract types that reflect thedifferent power consumption preferences. A study in [230]showed that the diversity of the contract types offered to con-sumers is highly important for DR programs to be appealing toa variety of participants, who have various consumption needs.The diversity of the offered contract types can be enhancedthrough market forces, if diversified DR intermediaries enterthe market. Furthermore, financial incentives should be offeredto retail customers to invest on smart metering infrastructurethat enables the switch from a fixed retail rate to a moredynamic pricing [231]. Finally, DR programs should not onlytarget on minimizing the total generation cost or minimizingthe peak-to-average ratio in the load demand, but they shouldalso indicate how consumers are charged to assure fairness[232]. The incorporation of fairness in DR program could bebeneficial not only to the participants, who will be rewardedfor their efforts on achieving the program’s objectives, butalso to the power utilities who will motivate a larger set ofconsumers to participate in the program. A study on the incor-poration of fairness within DR programs [233], observes thatthe basic criteria for the provision of a fair charging schemeto the consumers (fixed charges for must-run loads and multi-dimensional differential pricing based on the user-type) arepartially addressed by the known DR schemes. Therefore, theeffectiveness of a DR program can be significantly improved


by incorporating pricing schemes that consider a number offairness criteria for both participant and non-participant users,while achieving optimal resource management.

Secondly, the majority of DR programs are based oncustomer baselines that are used to determine the demandreduction, while they also define the price formation. Thesetting of the customer baselines is a complex procedure,since it is based on varying load patterns. Therefore, thechallenge for the DR provider is to establish an efficientbaseline that avoids producing significant problems in theprice formation in wholesale markets, while also urges theconsumers to realize the potentials and advantages of demandcontrol [234]. However, a successful DR program is notonly founded on the proper definition of customer baselines;other uncertainties may have a penetrating influence on thesuccess of a DR program. Government policies, fuel prices,technology breakthroughs, demand fluctuations and capitalcosts for the infrastructure renovations are highly related tothe price formation and are expected to play a significantrole in the elaboration of efficient DR programs [235]. Apowerful process that enables the uncertainty resolution isthe implementation of pilot programs; in this way, valuableinformation and credible results may improve the effectivenessof a DR program [106].

Thirdly, useful information can be derived from the resultsof already implemented pilots. A survey on various casestudies of several dynamic pricing programs is presented in[115]. The general conclusion from these case studies is thatthe greatest motivation for participation of the consumers isthe bill reduction. However, only a few consumers respondedto critical peak events, either by reducing or shifting theirsloads, unless significant incentives are offered. For example,Pepco’s customers in Washington, DC, that utilize a CPPprogram, reduced their peak summer usage by an additional20%, compared to those who do not participate in the program.Similar results are obtained from the case studies presented in[236]. What is interesting is that in the case of the HydroOne’s pilot in Ontario, Canada, an average 6.5% of powerconsumption reduction is achieved without offering any priceincentives, but by simply providing the customer with real-time feedback, through the use of in-home displays. Thegeneral conclusion of these studies, as well as of other casestudies in Europe and China (e.g. [237]- [239]), is that thesuccessfulness of a DR program is not only based on theoffered incentives; DR providers should also consider theprovision of information tools for the efficient participation,as well as the supply of complimentary smart equipment tothe consumers.

B. Future research directions

Based on the above survey, we can focus on the challenges,advantages and also limitations that arise from the implemen-tation of these DR methods, which should be addressed inorder to make more efficient and cost effective decisions forthe implementation of the future smart grid. The employmentof DR programs relies on a robust, secure and reliable commu-nication infrastructure, which is an indispensible component of

the future power grid . Efficiency, QoS support, secure routing,interoperability and scalability are critical to enabling a smartgrid communication infrastructure and should be considered asdirections for future research. Significant attention should bealso paid on privacy issues that arise from the managementof metering data, which contains private information andactivities or choices of the consumers. It is therefore essentialto implement a secured communication infrastructure for thesmart grid, which should overrun the security vulnerabilitiesand the unauthorized smart metering data access in the futurepower grid.

The key factor that aims at changing the demand in orderto follow the available supply is the efficient and reliableutilization and control of various energy sources. This effortwill be even more challenging in the future, mainly dueto the high penetration of renewable energy sources. Thestochastic nature of these sources and the large variations ofthe renewable energy are triggering the need for the efficientmodeling of there sources. Stochastic game and stochasticinventory theory are techniques that can be used for thederivation of optimal decisions for the interactions of multiplerenewable energy sources with the grid, and for the chargingof energy storage devices.

Even more challenging is the implementation of real-timepower generation, and demand forecasting methods that can beused by the power utility to adjust its forthcoming operationand properly schedule the consumers’ demands. A numberof efforts have been made in order to design an accurateforecasting model [8], [219], [240]-[242]. These efforts includealgorithms that are based on fuzzy logic [240], neural networks[241], [242] and game theory [8], [219]. Also, attention hasbeen given on using machine learning for forecast decisionsin smart grid environments [243]- [245]. Nevertheless, theresearch in the field of demand forecasting is still at itsinfancy and there are many fundamental issues that still needto be addressed, in order to jointly consider the randomdistribution of the various smart grid components and thestochastic behavior of the renewable energy sources. A keyissue on the derivation of effective forecasting methods isthe availability of statistics. In the absence of reliable data,techniques that jointly consider statistical learning and optimaldecision-making should be implemented for the training offorecasting methods. These techniques can be applied to theaforementioned artificial intelligence techniques, or to othertechniques, such as sum of products, regression fusion andmeta-classification. These artificial intelligence techniques canprovide practical in advance results on the definition of themargins for reliable operation of the smart grid system.Finally, it is highly desirable to generate simpler dynamicpricing schemes, based on the application of these forecastingtechniques, and more efficient automated procedures for theDR implementation that consider the probabilistic behavior inthe usage of appliances.

ACKNOWLEDGMENT

This work has been funded by the E2SG project, an ENIACJoint Undertaking under grant agreement No. 296131, and bythe Smart-NRG, No. 612254.


APPENDIX ALIST OF ABBREVIATIONS

AEP American Electric PowerASM Ancillary Services MarketBPSO Binary Particle Swarm OptimizationCMP Capacity Market ProgramsCPSO Coevolutionary Particle Swarm OptimizationCLPU Cold Load PickUpCPP Critical Peak PricingDA-RTP Day-Ahead Real-Time PricingDB Demand BiddingDR Demand ResponseDSM Demand Side ManagementDLC Direct Load ControlDMS Distribution Management SystemDSO Distribution System OperatorERCOT Electric Reliability Council of TexasEDRP Emergency Demand Response ProgramsEMC Energy Management ControllerFERC Federal Energy Regulatory CommissionHVAC Heating, Ventilation and Air-ConditioningISO Independent System OperatorI/C Interruptible/CurtailableILP Integer Linear ProgrammingMDC Markov Decision ProblemMILP Mixed Integer Linear ProgrammingMINLP Mixed-Integer Non-Linear ProgrammingMIP Mixed Integer ProgrammingNLP Non-Linear ProgrammingPSO Particle Swarm OptimizationPDR Peak Day RebatesPLP Peak Load PricingPTR Peak Time RebatesPAR Peak-to-Average RatioPHEVs Plug-in Hybrid Electric VehiclesQoE Quality of ExperienceRTP Real-Time PricingRTO Regional Transmission OperatorSPSO Signaled Particle Swarm OptimizationTCAs Thermostatically Controlled AppliancesTOU Time-Of-UseVCG Vickrey-Clarke-GrovesV2G Vehicle-to-Grid

REFERENCES

[1] X. Fang, S. Misra, G. Xue, and D. Yang, “Smart grid-The new andimproved power grid: A survey”, IEEE Commun. Surveys Tuts., vol.14, no. 4, pp. 944-980, 4th Quarter 2012.

[2] J. Gao, X. Xiao, J. Liu, W. Liang, and C. L. Chen, “A survey ofcommunication/networking in Smart Grids”, Fut. Generat. Comp. Syst.,vol. 28, no. 2, pp. 391-404, Feb. 2012.

[3] G. W. Arnold, “Challenges and opportunities in smart grid: A positionarticle”, Proceed. IEEE, vol. 99, no. 6, pp. 922-927, Jun. 2011.

[4] M. Alizadeh, X. Li, Z. Wang, A. Scaglione, and R. Melton, “Demand-Side Management in the Smart Grid: Information Processing for thePower Switch”, IEEE Signal Process. Mag., vol. 29, no. 5, pp. 55-67,Sept. 2012.

[5] P. Palensky and D. Dietmar, “Demand side management: Demandresponse, intelligent energy systems, and smart loads”, IEEE Trans.Ind. Informat., vol. 7, no. 3, pp. 381-388, Aug. 2011.

[6] F. Zhong, P. Kulkarni, S. Gormus, C. Efthymiou, G. Kalogridis,M. Sooriyabandara, Ziming Zhu, S. Lambotharan, and Woon HauChin, “Smart Grid Communications: Overview of Research Challenges,Solutions, and Standardization Activities”, IEEE Commun. SurveysTuts., vol. 15, no. 1, pp. 21-38, 1st Quarter 2013.

[7] M. M. Eissa, “Demand side management program evaluation based onindustrial and commercial field data”, Energy Policy, vol. 39, no. 10,pp. 5961-5969, Oct. 2011.

[8] A.-H. Mohsenian-Rad, V.W.S. Wong, J. Jatskevich, R. Schober, andA. Leon-Garcia, “Autonomous Demand-Side Management Based onGame-Theoretic Energy Consumption Scheduling for the Future SmartGrid”, IEEE Trans. Smart Grid, vol. 1, no. 3, pp. 320-331, Dec. 2010.

[9] US Department of Energy, Benefits of Demand Response in ElectricityMarkets and Recommendations for Achieving them, report to the UnitedStates Congress, Feb. 2006.

[10] S. Mohagheghi, F. Yang, F. and B. Falahati, “Impact of demandresponse on distribution system reliability”, in Proc. IEEE PES Gen.Meet. 2011, Detroit, MI, USA, 24-28 Jul. 2011.

[11] A. Brooks, E. Lu, D. Reicher, C. Spirakis, and B. Weihl, “Demanddispatch”, IEEE Power Energy Mag., vol. 8, no. 3, pp. 20-29, May-Jun. 2010.

[12] L. A. Greening, “Demand response resources: Who is responsible forimplementation in a deregulated market?”, Energy, vol. 35, no. 4, pp.1518-1525, Apr. 2010.

[13] P. P. Varaiya, F. F. Wu, and J. W. Bialek, “Smart operation of smartgrid: Risk-limiting dispatch”, Proceed. IEEE, vol. 99, no. 1, pp. 40-57,Jan. 2011.

[14] B. B. Alagoz, A. Kaygusuz, and A. Karabiber, “A user-mode distributedenergy management architecture for smart grid applications”, Energy,vol. 44, no. 1, pp. 167-177, Aug. 2012.

[15] M. Wissner, “The Smart Grid-A saucerful of secrets?”, Appl. Energy,vol. 88, no. 7, pp. 2509-2518, Jul. 2011.

[16] D. T. Nguyen, M. Negnevitsky, and M. de Groot, “Pool-based demandresponse exchange–concept and modeling”, IEEE Trans. Power Syst.,vol. 26, no. 3, pp. 1677-1685, Aug. 2011.

[17] A. Iwayemi, P. Yi, X. Dong, and C. Zhou, “Knowing when to act:An optimal stopping method for smart grid demand response”, IEEENetw., vol. 25, no. 5, pp. 44-49, Sept.-Oct. 2011.

[18] Q. Dong, L. Yu, W. Z. Song, L. Tong, and S. Tang, “DistributedDemand and Response Algorithm for Optimizing Social-Welfare inSmart Grid”, in Proc. 26th IEEE IPDPS, pp. 1228-1239, Shanghai,China, 21-25 May 2012.

[19] I. Koutsopoulos, and L. Tassiulas, “Optimal Control Policies for PowerDemand Scheduling in the Smart Grid”, IEEE J. Sel. Areas Commun.,vol. 30, no. 6, pp. 1049-1060, Jul. 2012.

[20] Y. Ota, H. Taniguchi, T. Nakajima, K. M. Liyanage, J. Baba, and A.Yokoyama, “Autonomous distributed V2G (vehicle-to-grid) satisfyingscheduled charging”, IEEE Trans. Smart Grid, vol. 3, no. 1, pp. 559-564, Mar. 2012.

[21] H. Jiayi, J. Chuanwen, and X. Rong, “A review on distributed energyresources and MicroGrid”, Renew. Sustain. Energy Rev., vol. 12, no. 9,pp. 2472-2483, Dec. 2008.

[22] C. Liu, K. T. Chau, W. Diyun, and S. “Opportunities and Challengesof Vehicle-to-Home, Vehicle-to-Vehicle, and Vehicle-to-Grid Technolo-gies”, Proceed. IEEE, vol. 101, no. 11, pp. 2409-2427, Nov. 2013.

[23] A. Saber and G. Venayagamoorthy, “Intelligent unit commitment withvehicle-to-grid - A cost-emission optimization”, J. Power Sources, vol.195, no. 3, pp. 898-891, Feb. 2010.

[24] C. Gouveia, J. Moreira, C. L. Moreira, and J. A. Pecas Lopes, “Co-ordinating Storage and Demand Response for Microgrid EmergencyOperation”, IEEE Trans. Smart Grid, vol. 4, no. 4, pp. 1898-1908,Dec. 2013.

[25] R. H. Lasseter, “Smart distribution: Coupled microgrids”, Proc. IEEE,vol. 99, no. 6, pp. 1074-1082, Jun. 2011.

[26] K. Spees and L. B. Lave, “Demand response and electricity marketefficiency”, Electricity J., vol. 20, no. 3, pp. 69-85, Apr. 2007.

[27] J. Aghaei and M.-I. Alizadeh, “Demand response in smart electricitygrids equipped with renewable energy sources: A review”, Renew.Sustain. Energy. Rev., vol. 18, pp. 64-72, Feb. 2013.

[28] M. H. Albadi, E.F. El-Saadany, “A summary of demand response inelectricity markets”, Elect. Power Syst. Res., vol. 78, no. 11, pp. 1989-1996, Nov. 2008.

[29] P. Siano, “Demand response and smart grids - A survey”, Renew.Sustain. Energy. Rev., vol. 30, pp. 461-478, Feb. 2014.


[30] E. Santacana, G. Rackliffe, L. Tang, and X. Feng, “Getting smart”,IEEE Power Energy Mag., vol. 8, no. 2, pp. 41-48, Mar.-Apr. 2010.

[31] S. Mohagheghi, J. Stoupis, Z. Wang, Z. Li, and H. Kazemzadeh,“Demand response architecture: Integration into the distribution man-agement system”, in Proc. IEEE SmartGridComm 2010, pp. 501-506,Gaithersurg, MA, USA, 4-6 Oct. 2010.

[32] M. A. Lisovich, D. K. Mulligan, and S. B. Wicker, “Inferring personalinformation from demand-response systems”, IEEE Security Privacy,vol. 8, no. 1, pp. 11-20, Jan.-Feb. 2010.

[33] N. Markushevich and E. Chan, “Integrated voltage, var control anddemand response in distribution systems”, in Proc. IEEE PES PowerSyst. Conf. Expo., Seattle, WA, USA, 15-18 Mar. 2009.

[34] J. M. Yusta, H. M. Khodr, and A. J. Urdaneta, “Optimal pricing ofdefault customers in electrical distribution systems: Effect behaviorperformance of demand response models”, Electr. Power Syst. Re., vol.77, no. 5, pp. 548-558, Apr. 2007.

[35] N. Venkatesan, S. Jignesh, and S. K. Solanki. “Residential DemandResponse model and impact on voltage profile and losses of an electricdistribution network”, Appl. Energy, vol. 96, pp. 84-91, Aug. 2012.

[36] J. Medina, N. Muller, and I. Roytelman, “Demand Response andDistribution Grid Operators: Opportunities and Challenges”, IEEETrans. Smart Grid, vol. 1, no. 2, pp. 193-198, Sept. 2010.

[37] L. Hernandez, C. Baladron, J. M. Aguiar, B. Carro, A. Sanchez-Esguevillas, J. Lloret, D. Chinarro, J. J. Gomez-Sanz, and D. Cook,“A Multi-Agent System Architecture for Smart Grid Managementand Forecasting of Energy Demand in Virtual Power Plants”, IEEECommun. Mag., vol. 51, no. 1, pp. 106-113, Jan. 2013.

[38] J. Warrington, S Mariethoz, C. N. Jones, and M. Morari, “Predictivepower dispatch through negotiated locational pricing”, in Proc. IEEEPES ISGT Europe 2010 Gothenburg, Sweden, 11-13 Oct. 2010.

[39] K. Kok, C. J. Warmer, and I. G. Kamphuis, “Powermatcher: multiagentcontrol in the electricity infrastructure”, in Proc. AAMAS 2005, pp. 75-82, New York, USA, 2005.

[40] H. F. Wedde, S. Lehnhoff, E. Handschin, and O. Krause, “Real-timemulti-agent support for decentralized management of electric power”,in Proc. 18th ECRTS, pp. 43-51, Dresden, Germany, 5-7 Jul. 2006.

[41] S. Lamparter, S. Becher, and J. G. Fischer, “An agent-based marketplatform for smart grids”, in Proc. AAMAS 2010, pp. 1689-1696,Toronto, Canada, 10-14 May 2010.

[42] Z. Jiang, “Agent-based control framework for distributed energy re-sources microgrids”, in Proc. IEEE/WIC/ACM IAT 2006, pp. 646-652,Hong-Kong, 18-22 Dec. 2006.

[43] B. Jiang, and Y. Fei, “Dynamic residential demand response anddistributed generation management in smart microgrid with hierarchicalagents”, Energy Procedia, vol. 12, pp. 76-90, 2011.

[44] U.S. Energy Information Administration, Annual Energy Outlook 2014,Sept. Apr. 2014.

[45] Energy Conservation Committee Report and Recommendations, “Re-ducing electricity consumption in houses”, Ontario Home Builders’Assoc., May 2006.

[46] M. Yilmaz and P. T. Krein, “Review of the impact of vehicle-to-grid technologies on distribution systems and utility interfaces”, IEEETrans. Power Elect., vol. 28, no. 12, pp. 5673-5689, Dec. 2013.

[47] N. Motegi, M. A. Piette, D. S. Watson, S. Kiliccote, and P. Xu, “In-troduction to commercial building control strategies and techniques fordemand response”, Lawrence Berkeley National Laboratory, Berkeley,CA, May 2007.

[48] Z. Zhou, F. Zhao, J. Wang, “Agent-Based Electricity Market SimulationWith Demand Response From Commercial Buildings”, IEEE Trans.Smart Grid, vol. 2, no. 4, pp. 580-588, Dec. 2011.

[49] C. A. Bel, M. A. Ortega, G. E. Escriva, and A. G. Marin, “Technicaland economical tools to assess customer demand response in thecommercial sector”, Energy Convers. Managem., vol. 50, pp. 2605-2612, Jul. 2009.

[50] J. L. Mathieu, P. N. Price, S. Kiliccote, and M. A. Piette, “Quanti-fying changes in building electricity use, with application to demandresponse”, IEEE Trans. Smart Grid, vol. 2, no. 3, pp. 507-518, Sept.2011.

[51] S. Kiliccote, M. A. Piette, E. Koch, and D. Hennage, “UtilizingAutomated Demand Response in commercial buildings as non-spinningreserve product for ancillary services markets”, in Proc. 50th IEEECDC-ECC, pp. 4354-4360, Orlando FL, USA, 12-15 Dec. 2011.

[52] S. Liliccote, M. A. Pietre, J. Mathieu, and K. Parrish, “Findingsfrom Seven Years of Field Performance Data for Automated DemandResponse in Commercial Buildings”, in Proc. 2010 ACEEE, PacificGrove, CA, 15-20 Aug. 2010.

[53] M. Alcazar-Ortega, G. Escriva-Escriva, and I. Segura-Heras, “Method-ology for validating technical tools to assess customer Demand Re-sponse: Application to a commercial customer”, Energy Convers.Managem., vol. 52, no. 2, pp. 1507-1511, Feb. 2011.

[54] T. Samad and S. Kiliccote. “Smart Grid Technologies and Applicationsfor the Industrial Sector”, Comp. & Chem. Eng., vol. 47, pp. 76-84,Dec. 2012.

[55] M. Zimmerman, “The industry demands better demand response”, inProc. IEEE PES ISGT 2012, Washington, DC, USA, 16-20 Jan. 2012.

[56] S. Mohagheghi and N. Raji, “Intelligent demand response scheme forenergy management of industrial systems”, in Proc. 2012 IEEE IASAnn. Meet., pp.1-9, Las Vegas NV, USA, 7-11 Oct. 2012.

[57] A. T. McKane, M. A. Piette, D. Faulkner, G. Ghatikar, A. RadspielerJr., B. Adesola, S. Murtishaw, and S. Kiliccote, “Opportunities, Barriersand Actions for Industrial Demand Response in California”, PIERIndustrial/Agricultural/Water End Use Energy Efficiency Program, Jan.2008.

[58] A. Grein and M. Pehnt, “Load management for refrigeration systems:Potentials and barriers”, Energy Policy, vol. 39, no. 9, pp. 5598-5608,Sept. 2011.

[59] M. Alcazar-Ortega, C. Alvarez-Bel, G. Escriva-Escriva, and A. Domi-jan, “Evaluation and assessment of demand response potential appliedto the meat industry”, Appl. Energy, vol. 92, pp. 84-91, Apr. 2012.

[60] D. Olsen, S. Goli, D. Faulkner and A. McKane, “Opportunities forEnergy Efficiency and Demand Response in the California Cement In-dustry”, PIER Industrial/Agricultural/Water End Use Energy EfficiencyProgram, Dec. 2010.

[61] M. Alcazar-Ortega, Evaluation and assessment of new demand re-sponse programs based on the use of flexibility in industrial processes:Application to the food industry, Ph.D. Dissertation, University ofSouth Florida and Universidad Politecnica de Valencia, Feb. 2011.

[62] Y. Yan, Y. Qian, H. Sharif, and D. Tipper, “A Survey on SmartGrid Communication Infrastructures: Motivations, Requirements andChallenges”, IEEE Commun. Surveys Tuts., vol. 15, no. 1, pp. 5-20,1st Quarter 2013.

[63] C.-H. Lo and N. Ansari, “The Progressive Smart Grid System fromBoth Power and Communications Aspects”, IEEE Commun. SurveysTuts., vol. 14, no. 3, pp. 799-821, 3rd Quarter 2012.

[64] Q.-D. Ho, Y. Gao, and T. Le-Ngoc, “Challenges and research oppor-tunities in wireless communication networks for smart grid”, IEEEWireless Commun., vol. 20, no. 3, pp. 89-95, Jun. 2013.

[65] V. C. Gungor, D. Sahin, T. Kocak, S. Ergut, C. Buccella, C. Cecati,and G. P. Hancke, “A Survey on Smart Grid Potential Applications andCommunication Requirements”, IEEE Trans. Ind. Informat., vol. 9, no.1, pp. 28-42, Feb. 2013.

[66] S. Mohagheghi, “Communication services and data model for demandresponse”, in Proc. IEEE GreenCom 2012, pp. 80-85, 25-28 Sept. 2012.

[67] G. T. Costanzo, G. Zhu, M. F. Anjos, and G. Savard, “A SystemArchitecture for Autonomous Demand Side Load Management inSmart Buildings”, IEEE Trans. Smart Grid, vol. 3, no. 4, pp. 2157-2165, Dec. 2012.

[68] S.-J. Kim and G. B. Giannakis, “Scalable and Robust Demand Re-sponse With Mixed-Integer Constraints”, IEEE Trans. Smart Grid, vol.4, no. 4, pp. 2089-2099, Dec. 2013.

[69] X. Fang, S. Misra, G. Xue, and D. Yang, “Managing smart gridinformation in the cloud: opportunities, model, and applications”, IEEENetw., vol. 26, no. 4, pp. 32-38, Jul.-Aug. 2012.

[70] K. Moslehi and R.Kumar, “A Reliability Perspective of the SmartGrid”, IEEE Trans. Smart Grid, vol. 1, no. 1, pp. 57-64, Jun. 2010.

[71] J. Liu, Y. Xiao, S. Li, W. Liang, and C. L. Philip Chen, “Cyber Securityand Privacy Issues in Smart Grids”, IEEE Commun. Surveys Tuts., vol.14, no. 4, pp. 981-997, 4th Quarter 2012.

[72] Y. Yan, Y. Qian, H. Sharif, and D. Tipper, “A Survey on Cyber Securityfor Smart Grid Communications”, IEEE Commun. Surveys Tuts., vol.14, no. 4, pp. 998-1010, 4th Quarter 2012.

[73] T. Khalifa, K. Naik, and A. Nayak, “A Survey of CommunicationProtocols for Automatic Meter Reading Applications”, IEEE Commun.Surveys Tuts., vol. 13, no. 2, pp.168-182, 2nd Quarter 2011.

[74] B. Ramanathan and V. Vittal, “A framework for evaluation of advanceddirect load control with minimum disruption”, IEEE Trans. Power Syst.,vol. 23, no. 4, pp. 1681-1688, Nov. 2008.

[75] B. R. Alexander, “Dynamic pricing? Not so fast! A residential con-sumer perspective”, Electricity J., vol. 23, no. 6, pp. 39-49, Jul. 2010.

[76] L. Zhou, J. J. Rodrigues, and L. M. Oliveira, “QoE-driven powerscheduling in smart grid: architecture, strategy, and methodology”.IEEE Commun. Mag., vol. 50, no. 5, pp. 136-141, May 2012.


[77] M. Alizadeh, A. Scaglione, and R. J. Thomas, “From packet topower switching: Digital direct load scheduling”, IEEE J. Sel. Areas.Commun., vol. 30, no. 6, pp. 1027-1036, Jul. 2012.

[78] H. Hildmann, and F. Saffre, “Influence of variable supply and loadflexibility on demand-side management”, in Proc. 8th IEEE EEM, pp.63-68, Zagreb, Croatia, 25-27 May 2011.

[79] N. Lu and Y. Zhang, “Design Considerations of a Centralized LoadController Using Thermostatically Controlled Appliances for Continu-ous Regulation Reserves”, IEEE Trans. Smart Grid, vol. 4, no. 2, pp.914-921, June 2013.

[80] P. Palensky, G. Zucker, F. Judex, R. Braun, F. Kupzog, T. Gamauf,and J. Haase, “Demand response with functional buildings usingsimplified process models”, in Proc. 37th IEEE IECON, pp. 3230-3235, Melbourne, Australia, 7-10 Nov. 2011.

[81] H. Yano, K. Kudo, T. Ikegami, H. Iguchi, K. Kataoka, and K. Ogimoto,“A novel charging-time control method for numerous EVs basedon a period weighted prescheduling for power supply and demandbalancing”, in Proc. IEEE PES ISGT 2012, Washington, DC, USA,16-20 Jan. 2012.

[82] W. Shi and V. W. Wong, “Real-time vehicle-to-grid control algorithmunder price uncertainty”, in Proc. IEEE SmartGridComm 2011, pp.261-266, Brussels, Belgium, 17-20 Oct. 2011.

[83] S. Han and K. Sezaki, “Development of an optimal vehicle-to-gridaggregator for frequency regulation”, IEEE Trans. Smart Grid, vol. 1,no. 1, pp. 65-72, Jun. 2010.

[84] J. Lee, H. J. Kim, G. L. Park, and H. Jeon, “Genetic algorithm-basedcharging task scheduler for electric vehicles in smart transportation”,LNCS 7196, pp. 208-217, 2012.

[85] A. Mehrizi-Sani and R. Iravani, “Potential-Function Based Control of aMicrogrid in Islanded and Grid-Connected Modes”, IEEE Trans PowerSyst., vol. 25, no. 4, pp. 1883-1891, Nov. 2010.

[86] N. Hatziargyriou, H. Asano, R. Iravani, and C. Marnay, “Microgrids:An overview of ongoing research, development, and demonstrationprojects”, IEEE Power Energy Mag., vol. 5, no. 4, pp. 78-94, Jul.2007.

[87] S. A. Pourmousavi and M. H. Nehrir, “Demand response for smartmicrogrid: Initial results”, in Proc. IEEE PES ISGT 2011, Manchester,UK, 5-7 Dec. 2011.

[88] T. L. Vandoorn, B. Renders, L. Degroote, B. Meersman, and L.Vandevelde, “Active load control in islanded microgrids based on thegrid voltage”, IEEE Trans. Smart Grid, vol. 2, no. 1, pp. 139-151, Mar.2011.

[89] A. A. Zaidi, F. Kupzog, T. Zia, and P. Palensky, “Load recognitionfor automated demand response in microgrids”, in Proc. 36th IEEEIECON, pp. 2442-2447, Glendale, AZ, USA, 7-10 Nov. 2010.

[90] Z. Tan, P. Yang and A. Nehorai, “An Optimal and Distributed DemandResponse Strategy With Electric Vehicles in the Smart Grid”, IEEETrans. Smart Grid, vol. 5, no. 2, pp. 861-869, Mar. 2014.

[91] H. Hermanns and H. Wiechmann. “Demand-Response Management forDependable Power Grids”, Embedded Syst. Smart Appliances EnergyManagem, vol. 3, pp. 1-22, 2013.

[92] Z. Fan, “Distributed demand response and user adaptation in smartgrids”, in Proc. IFIP/IEEE IM 2011, pp. 726-729, Dublin, Ireland, 23-27 May 2011.

[93] Z. Fan, “A distributed demand response algorithm and its applicationto PHEV charging in smart grids”, IEEE Trans. Smart Grid, vol. 3,no. 3, pp. 1280-1290, Sept. 2012.

[94] K. Anderson, and A. Narayan, “Simulating integrated volt/var controland distributed demand response using GridSpice”, in Proc. IEEESGMS 2011, pp. 84-89, Brussels, Belgium, 17 Oct. 2011.

[95] A. Molina-Garcia, F. Bouffard, and D. S. Kirschen, “DecentralizedDemand-Side Contribution to Primary Frequency Control”, IEEETrans. Power Syst., vol. 26, no. 1, pp. 411-419, Feb. 2011.

[96] A. D. Dominguez-Garcia, and C. N. Hadjicostis, “Distributed algo-rithms for control of demand response and distributed energy re-sources”, in Proc. 50th IEEE CDC-ECC, pp. 27-32, Orlando, FL, USA,12-15 Dec. 2011.

[97] P. T. Baboli, M. Eghbal, M. P. Moghaddam, and H. Aalami, “Customerbehavior based demand response model”, in Proc. IEEE PES Gen.Meet., pp. 1-7, San Diego, CA, USA, 22-26 Jul. 2012.

[98] M. Doostizadeh and H. Ghasemi, “A day-ahead electricity pricingmodel based on smart metering and demand-side management”, En-ergy, vol. 46, no. 1, pp. 221-230, Oct. 2012.

[99] A. Faruqui, R. Hledik, and J. Tsoukalis. “The power of dynamicpricing”, Electricity J., vol. 22, no. 3, pp. 42-56, Apr. 2009.

[100] S. Gottwalt, W. Ketter, C. Block, J. Collins, and C. Weinhardt,“Demand side management - A simulation of household behavior undervariable prices”, Energy Policy, vol. 39, no. 12, pp. 8163-8174, Dec.2011.

[101] S. Shao, T. Zhang, M. Pipattanasomporn, and S. Rahman, “Impact ofTOU rates on distribution load shapes in a smart grid with PHEVpenetration”, in Proc. IEEE PES 2010, New Orleans, LA, USA, 19-22Apr. 2010.

[102] A. Faruqui, S. Sergici, and A. Sharif, “The impact of informationalfeedback on energy consumption - A survey of the experimentalevidence”, Energy, vol. 35, no. 4, pp. 1598-1608, Apr. 2010.

[103] Q. Zhou, W. Guan, and W. Sun. “Impact of demand response contractson load forecasting in a smart grid environment”, in Proc. PES Gen.Meet., pp.1-4, San Diego, CA, USA, 22-26 Jul. 2012.

[104] K. Herter, “Residential implementation of critical-peak pricing ofelectricity”, Energy Policy, vol. 35, no. 4, pp. 2121-2130, Apr. 2007.

[105] G. R. Newsham, and B. G. Bowker. “The effect of utility time-varying pricing and load control strategies on residential summer peakelectricity use: A review”, Energy Policy, vol. 38, no. 7, pp. 3289-3296,Mar. 2010.

[106] A. Faruqui, R. Hledik, and S. Sergici. “Piloting the smart grid”,Electricity J., vol. 22, no. 7, pp. 55-69, Aug.-Sept. 2009.

[107] K. Herter and S. Wayland, “Residential response to critical-peak pricingof electricity: California evidence”, Energy, vol. 35, no. 4, pp. 1561-1567, Apr. 2010.

[108] C. Weiwei, X. Wang, J. Petersen, R. Tyagi, and J. Black, “OptimalScheduling of Demand Response Events for Electric Utilities”, IEEETrans. Smart Grid, vol.4, no.4, pp. 2309-2319, Dec. 2013.

[109] Standard Pricing Schedule: R-VPP Residential Variable Peak PricingProgram, Oklahoma Gas and Electric Company May 2010 [Online].Available: http://www.oge.com/Documents/OK/3.50%20R-VPP.pdf

[110] A. Faruqui and S. S. George, “The value of Dynamic Pricing in MassMarkets”, Electricity J., vol. 15, no. 6, pp. 45-55, Jul. 2002.

[111] Y. Liang, L. He, X. Cao, and Z.-J. Shen, “Stochastic Control for SmartGrid Users With Flexible Demand”, IEEE Trans. Smart Grid, vol. 4,no. 4, pp. 2296-2308, Dec. 2013.

[112] P. Samadi, H. Mohsenian-Rad, R. Schober, and V. W. Wong, “Advanceddemand side management for the future smart grid using mechanismdesign”. IEEE Trans. Smart Grid, vol 3, no. 3, pp. 1170-1180, Sept.2012.

[113] E. Ragnarsson, Price it Right: Household Response to a Time-of-Use Electricity Pricing Experiment in Auckland, New Zealand. PhDdissertation, University of Otago, 2012.

[114] P. Samadi, A. Mohsenian-Rad, R. Schober, V. Wong, and J. Jatskevich,“Optimal real-time pricing algorithm based on utility maximizationfor smart grid”, in Proc. IEEE SmartGridComm 2010, pp. 415-420,Gaithersburg, Maryland, USA, 4-6 Oct. 2010.

[115] J. Wang, M. A. Biviji, and W. Maria Wang, “Lessons learned fromsmart grid enabled pricing programs”, in Proc. IEEE PECI 2011,Urbana, IL, USA, 25-26 Feb. 2011.

[116] S. Braithwait, “Behavior modification”, IEEE Power Energy Mag., vol.8, no. 3 pp. 36-45, May-Jun 2010.

[117] F. Wolak, “Residential Customer Response to Real-time Pricing: TheAnaheim Critical Peak Pricing Experiment”, Center for the Study ofEnergy Markets Working Paper 151, May 2006.

[118] P. Samadi, R. Schober, and V. W. Wong, “Optimal energy consumptionscheduling using mechanism design for the future smart grid”, in Proc.IEEE SmartGridComm 2011, Brussels, Belgium, 17-20 Oct. 2011.

[119] C. Chen, S. Kishore, and L. V. Snyder, “An innovative RTP-basedresidential power scheduling scheme for smart grids”, in Proc. IEEEICASSP, pp. 5956-5959, Prague, Czech Republic, 22-27 May 2011.

[120] G. Webber, J. Warrington, S. Mariethoz, and M. Morari, M. Commu-nication limitations in iterative real time pricing for power systems, inProc. IEEE SmartGridComm 2011, pp. 445-450, Brussels, Belgium,17-20 Oct. 2011.

[121] P. Luh, Y. Ho, and R. Muralidharan, “Load adaptive pricing: Anemerging tool for electric utilities,”IEEE Trans. Autom. Control, vol.AC-27, no. 2, pp. 320-329, Apr. 1982.

[122] P. Cappers, C. Goldman, and D. Kathan, “Demand response in USelectricity markets: Empirical evidence”, Energy, vol. 35, no. 4, pp.1526-1535, Apr. 2010.

[123] H.Zhong, L. Xie, and Q. Xia, “Coupon Incentive-Based DemandResponse: Theory and Case Study”, IEEE Trans. Power Syst., vol. 28,no. 2, pp. 1266-1276, May 2013.

[124] H. Allcott, “Real time pricing and electricity markets”, Working Paper,Harvard Univ., Feb. 2009.


[125] H. Allcott, “Real Time Pricing and Electricity Market Design”, NewYork University Working Paper, May 2012.

[126] C. Joe-Wong, S. Sen, S. Ha, and M. Chiang, “Optimized Day-AheadPricing for Smart Grids with Device-Specific Scheduling Flexibility”,IEEE J. Sel. Areas Commun. vol. 30, no. 6, pp. 1075-1085, Jul. 2012.

[127] E. Shayesteh, A. Yousefi, and M. Parsa Moghaddam, “A probabilisticrisk-based approach for spinning reserve provision using day-aheaddemand response program”, Energy, vol. 35, no. 5, pp. 1908-1915,May 2010.

[128] N. Venkatesan, J. Solanki, and S. K. Solanki, “Demand response modeland its effects on voltage profile of a distribution system”, in Proc.IEEE PES Gen. Meet. 2011, Detroit, MI, USA, 24-28 Jul. 2011.

[129] S. Widergren, C. Marinovici, T. Berliner, and A. Graves, “Real-timepricing demand response in operations”, in Proc. IEEE PES Gen. Meet.2012, San Diego, CA, USA, 22-26 Jul. 2012.

[130] H. Allcott, “Rethinking real-time electricity pricing”, Resour. EnergyEcon., vol. 33, no. 4, pp. 820-842, Nov. 2011.

[131] M. Parvania, and M. Fotuhi-Firuzabad, “Demand response Schedulingby Stochastic SCUC”, IEEE Trans. Smart Grid, vol. 1, no. 1, pp. 89-98,Jun. 2010.

[132] G. Xiong, C. Chen, S. Kishore,and A. Yener, “Smart (in-home) powerscheduling for demand response on the smart grid”, in Proc. IEEE PESISGT 2011, Manchester, UK, 5-7 Dec. 2011.

[133] B. Ramanathan and V. Vittal, “A framework for evaluation of advanceddirect load control with minimum disruption”, IEEE Trans. Power Syst.,vol. 23, no. 4, pp. 1681-1688, Nov. 2008.

[134] A. Gomes, C. H. Antunes, and A. G. Martins, “A multiple objectiveapproach to direct load control using an interactive evolutionary algo-rithm,”IEEE Trans. Power Syst., vol. 22, no. 3, pp. 1004-1011, Aug.2007.

[135] N. Ruiz, I. Cobelo, and J. Oyarzabal, “A direct load control model forvirtual power plant management”, IEEE Trans. Power Syst., vol. 24,no. 2, pp. 959-966, May 2009.

[136] Q. Wu, P. Wang, and L. Goel, “Direct Load Control (DLC) ConsideringNodal Interrupted Energy Assessment Rate (NIEAR) in RestructuredPower Systems”, IEEE Trans. Power Syst., vol. 25, no. 3, pp. 1449-1456, Aug. 2010.

[137] P. Faria, and Z. Vale, “Demand response in electrical energy supply:An optimal real time pricing approach”, Energy, vol. 36, no. 8, pp.5374-5384, Aug. 2011.

[138] H. A. Aalami, M. P. Moghaddam, and G. R. Yousefi, “Demandresponse modeling considering Interruptible/Curtailable loads and ca-pacity market programs”, Appl. Energy, vol. 87, no. 1, pp. 243-250,Jan. 2010.

[139] H. A. Aalami, M. P. Moghaddam, G. R. Yousefi, “Modeling andprioritizing demand response programs in power markets”, ElectricPower Syst. Res., vol. 80, no. 4, pp. 426-435, Apr. 2010.

[140] W. J. Lee, F. L. Quilumba, J. Shi, and S. H. Huang, “Demand response-An assessment of load participation in the ERCOT nodal market”, inProc. IEEE PES Gen. Meet. 2012, San Diego, CA, USA, 22-26 Jul.2012.

[141] Y. Wang, I. R. Pordanjani, and W. Xu, “An event-driven demandresponse scheme for power system security enhancement”, IEEE Trans.Smart Grid, vol. 2, no. 1, pp. 23-29, Mar. 2011.

[142] J.Mickey, “Using load resources to meet ancillary service requirementsin the ERCOT market: A case study,”in Proc. IEEE PES Gen. Meet.2010, Minneapolis, MN, 25-29 Jul. 2010.

[143] P. Khajavi, H. Abniki, and A. B. Arani, “The role of incentive basedDemand Response programs in smart grid”, in Proc. IEEE EEEIC2011, Rome, Italy, 8-11 May 2011.

[144] K. Kostkova, L’ Omelina, P. Kycina, and P. Jamrich, “An introductionto load management”, Electric Power Syst. Res., vol. 95, pp. 184-191,Feb. 2013.

[145] J. W. Zarnikau, “Demand participation in the restructured ElectricReliability Council of Texas market”, Energy, vol. 35, no. 4, pp. 1536-1543, Apr. 2010.

[146] H. S. Oh and R. J. Thomas, “Demand-side bidding agents: Modelingand simulation”, IEEE Trans. Power Syst., vol. 24, no. 3, pp. 1050-1056, Aug. 2008.

[147] P. Centolella, “The integration of Price Responsive Demand intoRegional Transmission Organization (RTO) wholesale power marketsand system operations”, Energy, vol. 35, no. 4, pp. 1568-1574, Apr.2010.

[148] M. P. Moghaddam, A. Abdollahi, and M. Rashidinejad, “Flexible de-mand response programs modeling in competitive electricity markets”,Appl. Energy, vol. 88, no. 9, pp. 3257-3269, Sept. 2011.

[149] M. Marwan, F. Kamel, and W. Xiang, “A demand-side response smartgrid scheme to mitigate electrical peak demands and access renewableenergy sources”, in Proc. 48th AuSES Solar2010, Canberra, Australia,1-3 Dec. 2010.

[150] H. Zhong, L. Xie, and Q. Xia, “Coupon incentive-based demandresponse (CIDR) in smart grid”, in Proc. IEEE PES Gen. Meet., SanDiego, CA, USA, 22-26 Jul. 2012.

[151] M. Pipattanasomporn, M. Kuzlu, and S. Rahman. “Demand responseimplementation in a home area network: A conceptual hardwarearchitecture”, in Proc. IEEE PES ISGT 2012, Washington, DC, USA,16-20 Jan. 2012.

[152] P. T. Baboli, M. P. Moghaddam, and M. Eghbal, “Present status andfuture trends in enabling demand response programs”, in Proc. IEEEPES Gen. Meet., Detroit, MI, USA, 24-28 Jul. 2011.

[153] M. Rastegar, M. Fotuhi-Firuzabad, and F. Aminifar, “Load commitmentin a smart home”, Appl. Energy, vol. 96, pp. 45-54, Aug. 2012.

[154] J. Lee, G. L. Park, S. W. Kim, H. J. Kim, and C. O. Sung, “Powerconsumption scheduling for peak load reduction in smart grid homes”,in Proc. ACM SAC 2011, pp. 584-588, Taichung, Taiwan, 21-24 Mar.2011.

[155] N. Kumaraguruparan, H. Sivaramakrishnan, and S. S. Sapatnekar,“Residential task scheduling under dynamic pricing using the multipleknapsack method”, in Proc. IEEE PES ISGT 2012, Washington, DC,USA, 16-20 Jan. 2012.

[156] H. Goudarzi, S. Hatami, and M. Pedram, “Demand-side load schedul-ing incentivized by dynamic energy prices”, in Proc. IEEE SmartGrid-Comm 2011, pp. 351-356, Brussels, Belgium, 17-20 Oct. 2011.

[157] M. Shinwari, A. Youssef, and W. Hamouda, “A Water-Filling BasedScheduling Algorithm for the Smart Grid”, IEEE Trans. Smart Grid,vol.3, no. 2, pp. 710-719, Jun. 2012.

[158] N. Gatsis and G. Giannakis, “Cooperative multi-residence demand re-sponse scheduling”, in Proc. 45th Annu. Conf. Inf. Sci. Syst., Baltimore,MD, USA, 23-25 Mar. 2011.

[159] A.-H. Mohsenian-Rad, V. Wong, J. Jatskevich, and R. Schober, “Op-timal and autonomous incentive-based energy consumption schedulingalgorithm for smart grid”, in Proc. IEEE PES ISGT 2010, Gothenburg,Sweden, 11-13 Oct. 2010.

[160] S. S. Rao, Engineering optimization: theory and practice, John Wiley& Sons, 2009.

[161] J. Borwein and A. Lewis, Convex analysis and nonlinear optimization:theory and examples (Vol. 3). Springer, 2006.

[162] M. R. Garey and D. S. Johnson, Computers and intractability: a guideto the theory of NP-completeness, W. H. Freeman and Company, SanFrancisco, 1979.

[163] A. Molderink, V. Bakker, M. G. Bosman, J. L. Hurink, and G. J. Smit,“Domestic energy management methodology for optimizing efficiencyin smart grids”, in Proc. IEEE PowerTech 2009, Bucharest, Romania,28 Jun.-2 Jul. 2009.

[164] M. Behrangrad, H. Sugihara, and T. Funaki, “Analyzing the systemeffects of optimal demand response utilization for reserve procurementand peak clipping”, in Proc. IEEE PES Gen. Meet. 2010, Minneapolis,MN, 25-29 Jul. 2010.

[165] A.-H. Mohsenian-Rad and A. Leon-Garcia, “Optimal residential loadcontrol with price prediction in real-time electricity pricing environ-ments”, IEEE Trans. Smart Grid, vol. 1, no. 2, pp. 120-133, Sep. 2010.

[166] P. Du and N. Lu, “Appliance commitment for household load schedul-ing”, IEEE Trans. Smart Grid, vol. 2, no.2, pp. 411-419, Jun. 2011.

[167] N. Gatsis and G. B. Giannakis, “Residential Demand Response withInterruptible Tasks: Duality and Algorithms”, in Proc. 50th IEEECDIC-ECC, Orlando, FL, USA, 12-15 Dec. 2011.

[168] M. Kraning, E. Chu, J. Lavaei, and S. Boyd, “Dynamic network energymanagement via proximal message passing”, Optimization, vol. 1, no.2, pp. 70-122, 2013.

[169] L. P. Qian, Y. J. A. Zhang, J.Huang, and Y. Wu, “Demand responsemanagement via real-time electricity price control in smart grids”, IEEEJ. Sel. Areas Commun., vol. 31, no. 7, pp. 1268-1280, Jul. 2013.

[170] S. Boyd and L. Vandenbergher, Convex Optimization, Cambridge, U.K.:Cambridge Univ. Press, 2004.

[171] T. Logenthiran, D. Srinivasan, and T. Z. Shun, “Demand side manage-ment in smart grid using heuristic optimization”, IEEE Trans. SmartGrid, vol. 3, no. 3, pp. 1244-1252, Sept. 2012.

[172] N. Gudi, L. Wang, and V. Devabhaktuni, “A demand side managementbased simulation platform incorporating heuristic optimization formanagement of household appliances”, Int. J. Electr. Power EnergySyst., vol. 43, no. 1, pp. 185-193, Dec. 2012.

[173] R. Poli, J. Kennedy, and T. Blackwell, “Particle swarm optimization”,Swarm intelligence, vol. 1, no. 1, pp. 33-57, Jun. 2007.


[174] J. Xiao, J. Y. Chung, J. Li, R. Boutaba, and J. K. Hong, “Near optimaldemand-side energy management under real-time demand-responsepricing”, in Proc. IEEE CNSM 2010, Niagara Falls, Canada, 25-29Oct. 2010.

[175] Y. Guo, M. Pan, and Y. Fang, “Optimal power management ofresidential customers in the smart grid”, IEEE Trans. Parallel Distrib.Syst., vol. 23, no. 9, pp. 1593-1606, Sept. 2012.

[176] K. M. Tsui and S. C. Chan, “Demand Response Optimization for SmartHome Scheduling Under Real-Time Pricing”, IEEE Trans. Smart Grid,vol. 3, no. 4, pp. 1812-1821, Dec. 2012.

[177] W. Su, H. Rahimi-Eichi, W. Zeng, and M.-Y. Chow, “A Survey on theElectrification of Transportation in a Smart Grid Environment”, IEEETrans. Ind. Informat., vol. 8, no. 1, pp. 1-10, Feb. 2012.

[178] J. Tomic and W. Kempton, “Using fleets of electric-drive vehicles forgrid support,”, J. Power Sources, vol. 168, no. 2, pp. 459-468, Jun.2007.

[179] J. Mullan, D. Harries, T. Braunl, and S. Whitely, “The technical,economic and commercial viability of the vehicle-to-grid concept”,Energy Policy, vol. 48, pp. 394-406, Sept. 2012.

[180] R. Doucette and M. McCulloch, “Modeling the prospects of plug-inhybrid electric vehicles to reduce CO2 emissions”, Appl. Energy, vol.88, no. 7, pp. 2315-2323, Jul. 2011.

[181] A. Saber and G. Venayagamoorthy, “Efficient utilization of renewableenergy sources by gridable vehicles in cyber-physical energy systems”,IEEE Syst. J., vol. 4, no. 3, pp. 285-294, Sept. 2010.

[182] H. Khayyam, H. Ranjbarzadeh, and V. Marano, “Intelligent control ofvehicle to grid power”, J. Power Sources, vol 201, pp. 1-9, Mar. 2012.

[183] J. Soares, T. Sousa, H. Morais, Z. Vale, and P. Faria, “An optimalscheduling problem in distribution networks considering V2G”, inProc.IEEE CIASG 2011, Paris, France, 11-15 Apr. 2011.

[184] T. Sousa, H. Morais, Z. Vale, P. Faria, and J. Soares, “Intelligentenergy resource management considering vehicle-to-grid: A simulatedannealing approach”, IEEE Trans. Smart Grid, vol. 3, no. 1, pp. 535-542, Mar. 2012.

[185] E. Aarts and J. Korst, Simulated Annealing and Boltzmann Machines:A Stochastic Approach to Combinatorial Optimization and NeuralComputing, Wiley, New York, 1989.

[186] J. Cao, B. Yang, C. Chen, and X. Guan, “Optimal Demand ResponseUsing Mechanism design in the Smart Grid”, in Proc. 31th CCC, Hefei,China, 25-27 Jul. 2012.

[187] J. Jiang and S. H. Low, “Multi-period optimal procurement and demandresponses in the presence of uncertain supply”, in Proc. IEEE CDC2011, Orlando, FL, USA, 12-15 Dec. 2011.

[188] L. Jiang and S. H. Low, “Multi-period optimal energyprocurement and demand response in smart grid withuncertain supply”, Technical Report, Caltech, 2011. Available:http://rsrg.cms.caltech.edu/greenIT/papers/multiperiod.pdf.

[189] C. Cecati, C. Citro, and P. Siano, “Combined operations of renewableenergy systems and responsive demand in a smart grid”, IEEE Trans.Sustain. Energy, vol. 2, no. 4, pp. 468-476, Oct. 2011.

[190] M. A. A. Pedrasa, T. D. Spooner, and I. F. MacGill, “Coordinatedscheduling of residential distributed energy resources to optimize smarthome energy services”, IEEE Trans. Smart Grid, vol. 1, no. 2, pp. 134-143, Sept. 2010.

[191] E. Sortomme and M. A. El-Sharkawi, “Optimal charging strategies forunidirectional vehicle-to-grid”, IEEE Trans. Smart Grid, vol. 2, no. 1,pp. 131-138, Mar. 2011.

[192] E. Sortomme and M. A. El-Sharkawi, “Optimal scheduling of vehicle-to-grid energy and ancillary services”, IEEE Trans. Smart Grid, vol.3, no. 1, pp. 351-359, Mar. 2012.

[193] T. Cui, H. Goudarzi, S. Hatami, S. Nazarian, and M Pedram, “Concur-rent optimization of consumer’s electrical energy bill and producer’spower generation cost under a dynamic pricing model”, in Proc. IEEEPES ISGT 2012, Washington, DC, USA, 16-20 Jan. 2012.

[194] C. Hutson, G. K. Venayagamoorthy, and K. A. Corzine, “Intelligentscheduling of hybrid and electric vehicle storage capacity in a parkinglot for profit maximization in grid power transactions”, in Proc. IEEEENERGY 2008, Atlanta, GA, USA, 17-18 Nov. 2008.

[195] M. Silva, H. Morais, and Z. Vale, “An integrated approach for dis-tributed energy resource short-term scheduling in smart grids consid-ering realistic power system simulation”, Energy Convers. Managem.,vol. 64, pp. 273-288, Dec. 2012.

[196] Z. Zhu, J. Tang, S. Lambotharan, W. H. Chin, and Z. Fan, “Aninteger linear programming based optimization for home demand-side management in smart grid”, in Proc. IEEE PES ISGT 2012,Washington, DC, USA, 16-20 Jan. 2012.

[197] C. N. Kurucz, D. Brandt, and. S. Sim, “A linear programming modelfor reducing system peak through customer load control programs”,IEEE Trans. Power Syst., vol. 11, no. 4, pp. 1817-1824, Nov. 1996.

[198] P. Yang, P. Chavali, P. and A Nehorai, “Parallel autonomous optimiza-tion of demand response with renewable distributed generators”,in ProcSmartGridComm 2012, pp. 55-60, Tainan City, Taiwan, 5-8 Nov. 2012.

[199] J. Soares, M. Silva, T. Sousa, Z. Vale, and H. Morais, “Distributedenergy resource short-term scheduling using Signaled Particle SwarmOptimization”, Energy, vol. 42, no. 1, pp. 466-476, Jun. 2012.

[200] L. C. Totu, J. Leth, and R. Wisniewski, “Control for large scale demandresponse of thermostatic loads”, in Proc. IEEE ACC 2013, pp. 5023-5028, Washington, DC, USA, 17-19 Jun. 2013.

[201] S. Choi, S. Park, D. J. Kang, S. J. Han, and H. M. Kim, “A microgridenergy management system for inducing optimal demand response”,in Proc. IEEE SmartGridComm 2011, pp. 19-24, Brussels, Belgium,17-20 Oct. 2011.

[202] P. O. Kriett and M. Salani, “Optimal control of a residential microgrid”,Energy, vol. 82, no. 1, pp. 321-330, Jun. 2012.

[203] M. A. Sofla and R. King, “Control method for multi-microgrid systemsin smart grid environment-Stability, optimization and smart demandparticipation”, in Proc. IEEE PES ISGT 2012, Washington, DC, USA,16-20 Jan. 2012.

[204] D. Zhang, L. G. Papageorgiou, N. J. Samsatli, and N. Shah, “Optimalscheduling of smart homes energy consumption with microgrid”, inProc. ENERGY 2011, pp. 70-75, Venice, Italy, 22-27 May 2011.

[205] D. E. Goldberg, Genetic algorithms in search optimization and machinelearning, Addison-Wesley, 1989.

[206] O. Battaıa and A. Dolgui, “A taxonomy of line balancing problemsand their solution approaches”Int. J. Production Economics, vol. 142,no. 2, pp. 259-277, Apr. 2013.

[207] J. Aghaei and M. I. Alizadeh, “Multi-objective self-scheduling ofCHP (combined heat and power)-based microgrids considering demandresponse programs and ESSs (energy storage systems)”, Energy, vol.55, pp.1044-1054, Jun. 2013.

[208] R. S. Ferreira, L. A. Barroso, and M. M. Carvalho, “Demand responsemodels with correlated price data: A robust optimization approach”,Appl. Energy, vol. 96, pp. 133-149, Aug. 2012.

[209] J. Wang, Z. Sun, Y. Zhou, and J. Dai, “Optimal dispatching model ofSmart Home Energy Management System”, in Proc. IEEE PES ISGTASIA 2012, Tianjin, China, 21-24 May 2012.

[210] T. H. Chang, M. Alizadeh, and A. Scaglione, “Coordinated homeenergy management for real-time power balancing”, in Proc. IEEEPES Gen. Meet. 2012, San Diego, CA, USA, 22-26 Jul. 2012.

[211] M. G. Kallitsis, G. Michailidis, and M. Devetsikiotis, “Optimal powerallocation under communication network externalities”, IEEE Trans.Smart Grid, vol. 3, no. 1, pp. 162-173, Mar. 2012.

[212] N. Li, L. Chen, and S. H. Low, “Optimal demand response based onutility maximization in power networks”, in Proc. IEEE PES Gen.Meet. 2011, Detroit, MI, USA, 24-28 Jul. 2011.

[213] M. J. Osborne, A course in game theory. Cambridge, Mass.: MIT Press,1994.

[214] W. Saad, Z. Han, H. Poor, and T. Basar, “Game-theoretic methodsfor the smart grid: an overview of microgrid systems, demand-sidemanagement, and smart grid communications”, IEEE Signal Process.Mag., vol. 29, no. 5, pp. 86-105, Sept. 2012.

[215] M. Zugno, J. M. Morales, P. Pinson, and H. Madsen, “A bilevelmodel for electricity retailers participation in a demand response marketenvironment”, Energy Economics, vol. 36, pp. 182-197, Mar. 2013.

[216] D. Li, S. K. Jayaweera, and A. Naseri, “Auctioning game based demandresponse scheduling in smart grid”, in Proc. Online IEEE GreenComm2011, pp. 58-63, 26-29 Sept. 2011.

[217] L. Chen, N. Li, S. H. Low, and J. C. Doyle, “Two Market Modelsfor Demand Response in Power Networks”, in Proc. IEEE SmartGrid-Comm 2010, Gaithersburg, MA, USA, 4-6 Oct. 2010.

[218] Y. Chen, W. S. Lin, F. Han, Y. H. Yang, Z. Safar, and K. J. Liu, “Acheat-proof game theoretic demand response scheme for smart grids”,in Proc. IEEE ICC 2012, pp. 3362-3366, Ottawa, Canada, 10-15 Jun.2012.

[219] P. Yang, G. Tang, and A. Nehorai, “A Game-theoretic approach foroptimal time-of-use electricity pricing”, IEEE Trans. Power Syst., vol.28, no. 2, pp. 884-892, May 2013.

[220] Q. Zhu, Z. Han, and T. Basar, “A differential game approach todistributed demand side management in smart grid”, in Proc. IEEEICC 2012, pp. 3345-3350, Ottawa, Canada, 10-15 Jun. 2012.

[221] S. Maharjan, Q. Zhu, Y. Zhang, S. Gjessing, and T. Basar, “DependableDemand Response Management in the Smart Grid: A Stackelberg


Game Approach”, IEEE Trans. Smart Grid, vol. 4, no. 1, pp. 120-132,Mar. 2013..

[222] W. Tushar, W. Saad, H. V. Poor, and D. B. Smith, “Economics ofelectric vehicle charging: A game theoretic approach”, IEEE Trans.Smart Grid, vol. 3, no. 4, pp. 1767-1078, Dec. 2012.

[223] Y. Li, Y. Wang, S. Nazarain, and M. Pedram, “A Nested Game-BasedOptimization Framework for Electricity Retailers in the Smart Gridwith Residential Users and PEVs”, in Proc. IEEE Online Conf. onGreen Communications, 29-31 Oct. 2013.

[224] C. Wu, H. Mohsenian-Rad, and J. Huang, “Vehicle-to-aggregatorinteraction game”, IEEE Trans. Smart Grid, vol. 3, no. 1, pp. 434-442, Mar. 2012.

[225] S. Sun, M. Dong, and B. Liang, “Real-Time welfare-maximizing regu-lation allocation in aggregator-EVs systems”, in Proc. IEEE INFOCOMWKSHPS, pp.13-18, Turin, Italy, 14-19 Apr. 2013.

[226] W. Meng, and L. Kai, “Optimization of electric vehicle charging stationlocation based on game theory”, in Proc. TMEE 2011, pp. 809-812,Changchun, China, 16-18 Dec. 2011.

[227] A. Faruqui, R. Hledik, and S. Sergici. “Rethinking prices: The changingarchitecture of demand response in America”, Public Utilities Fort-nightly, vol. 148, no. 1, pp. 30-39, Jan. 2010.

[228] A. Faruqui, “The ethics of dynamic pricing”, Electricity J. vol. 23,no.6, pp. 13-27, Jul. 2010.

[229] Z. Baharlouei, H. Narimani, and H. Mohsenian-Rad, “Tackling co-existence and fairness challenges in autonomous Demand Side Man-agement”, in Proc. IEEE Globecom 2012, pp. 3159-3164, Anaheim,USA, 3-7 Dec. 2012.

[230] X. He, N. Keyaerts, I. Azevedo, L. Meeus, L. Hancher, and J. M.Glachant, “How to engage consumers in demand response: a contractperspective”, Utilities Policy, vol. 27, pp. 108-122, Dec. 2013.

[231] H. P. Chao, “Price-responsive demand management for a smart gridworld”, Electricity J., vol. 23, no. 1, pp. 7-20, Jan.-Feb. 2010.

[232] Z. Baharlouei, M. Hashemi, H. Narimani, and H. Mohsenian-Rad,“Achieving Optimality and Fairness in Autonomous Demand Response:Benchmarks and Billing Mechanisms”, IEEE Trans. Smart Grid, vol.4, no. 2, pp. 968-975, Jun. 2013.

[233] S. K. Vuppala, K. Padmanabh, S. K. Bose, and S. Paul, “Incorporatingfairness within demand response programs in smart grid”, in Proc.IEEE PES ISGT 2011,Anaheim, CA, USA, 17-19 Jan. 2011.

[234] A. Zibelman and E. N. Krapels, “Deployment of demand response asa real-time resource in organized markets”, Electricity J., vol. 21, no.5, pp. 51-56, Jun. 2008.

[235] M. Kezunovic, J. D. McCalley, and T. J. Overbye, “Smart grids andbeyond: achieving the full potential of electricity systems”, Proceed.IEEE, (Special Centennial Issue), vol. 100, pp. 1329-1341, May 2012.

[236] J. Wang, M. Biviji, and W. M. Wang. “Case studies of smart griddemand response programs in North America”, in Proc. IEEE PESISGT 2011, Anaheim, CA, USA, 17-19 Jan. 2011.

[237] J. Torriti, M. G. Hassan, and M. Leach, “Demand response experiencein Europe: Policies, programmes and implementation”, Energy, vol. 35,no. 4 pp. 1575-1583, Jun. 2009.

[238] E. ten Heuvelhof, and M. Weijnen, “Investing in smart grids withinthe market paradigm: The case of the Netherlands and its relevancefor China”, Policy and Society, vol. 32, no. 2, pp. 163-174, May 2013.

[239] H. Sæle and O. S. Grande, “Demand response from household cus-tomers: experiences from a pilot study in Norway”, IEEE Trans. SmartGrid, vol. 2, no. 1, pp. 102-109, Mar. 2011.

[240] H. Hagras, V. Callaghan, M. Colley, G. Clarke, A. Pounds-Cornish,and H. Duman, “Creating an ambient-intelligence environment usingembedded agents”, IEEE Intell. Syst., vol. 19, no. 6, pp. 12-20, Nov.Dec. 2004.

[241] H. Zhang, F. Xu, and L. Zhou, “Artificial neural network for loadforecasting in smart grid”, in Proc. of ICMLC 2010, pp. 3200-3205,Qingdao, China, 11-14 Jul. 2010.

[242] L. J. Ricalde, B. Cruz, G. Catzin, A. Y. Alanis, and E. N. Sanchez,“Forecasting for smart grid applications with Higher Order NeuralNetworks”, in Proc. IEEE WAC 2012, pp. 1-6, Puerto Vallarta, Mexico,24-28 Jun. 2012.

[243] B. Li, S. Gangadhar, S. Cheng, and P. K. Verma, “Predicting usercomfort level using machine learning for Smart Grid environments”,in Proc. IEEE PES ISGT 2011, Anaheim, CA, USA, 17-19 Jan. 2011.

[244] T. Hong, M. Gui, M. E. Baran, and H. L. Willis, “Modeling andforecasting hourly electric load by multiple linear regression withinteractions”, in Proc. IEEE PES Gen. Meet. 2010, Minneapolis, MN,25-29 Jul. 2010.

[245] O. Kramer, B. Satzger, and J. Lassig, “Power Prediction in Smart Gridswith Evolutionary Local Kernel Regression”, LNCS 6076, pp. 262-269,2010.

John S. Vardakas received the Dipl.-Eng. in Elec-trical & Computer Engineering from the DemocritusUniversity of Thrace, Greece, in 2004 and his Ph.Dfrom the Electrical & Computer Engineering Dept.,University of Patras, Greece. He is currently a seniorresearcher at Iquadrat Informatica. His research in-terests include performance analysis and simulationof communication networks and smart grids. He is amember of the IEEE, the Optical Society of America(OSA) and the Technical Chamber of Greece (TEE).

Nizar Zorba holds a BSc in Electrical Engi-neering from JUST University (Jordan, 2002), aMSc in Data Communications by the University ofZaragoza (Spain, 2004), an MBA by the Universityof Zaragoza (Spain, 2005) and a Ph.D. in SignalProcessing for Communications by UPC-Barcelona(Spain, 2007). He led and participated in more than20 research projects (European, Qatari and Spanishfunded). He is author of six patents, two books,five book chapters and more than 100 peer-reviewedjournals and international conferences. His research

interests are QoS/QoE in wireless systems, Energy efficiency and resourceoptimization in WLAN and LTE systems.

Christos Verikoukis received his Ph.D. from theTechnical University of Catalonia in 2000. He iscurrently a Senior Researcher at CTTC and anadjunct professor at UB. His area of expertise is inthe design of energy efficient layer two protocolsand RRM algorithms, for short range wireless co-operative and network coded communications. Dr.Verikoukis has participated and coordinated severalnational and European projects. He has published40 journal papers and over 100 conference papers.He is also co-author in two books, 12 chapters in

different books and in two patents. Dr. Verikoukis has participated more than20 competitive projects (IST, ICT, CELTIC, MEDEA+, CATRENE, Marie-Curie, COST) while he has served as the Principal investigator in threenational projects in Greece and Spain as well as the technical manager in nineEC funded projects. He has served as co-editor in five special issues while hehas participated in the organization of several international conferences. He isalso a regular reviewer in a number of international journals. He has appointedto serve as a reviewer in FP7 projects and as an evaluator in ARTEMIS-JUand for research funded projects in Greece and in Spain. He has supervised15 Ph.D. students and five Post Docs researchers since 2004.