senteur

DC microgridEnergy managementPredictionSmart grid

gridalazatimos

supervision design with predicted powers ow optimization for DC microgrid based on photovoltaic

ationsmust bwith

information exchange. Smart grid is electric networks that employinnovative and intelligent monitoring, control communication, andself-healing technologies to deliver better services for power pro-ducers and distributors, exible choices for end-users, reliabilityand security of power supply [4,5]. Smart grid is expected mainlyfor the following aspects: bidirectional power distribution; bidirec-

tives of providinger into the utilityach, severanergy pene

level, facilitating the smart grid implementation, bettersupply for remote areas, power balancing at local level wisupplying possibility, and maintaining load supply duringding operation or off-grid mode [12]. Thus, the microgrid controllerbecomes essential for balancing power and energy management,and facilitates the sources pooling during islanding.

Depending on the usage of AC or DC bus for coupling differentelements within microgrid, AC microgrid, DC microgrid and hybridAC/DC microgrid structures exist [13]. At present, the DC grid is notubiquitous [14,15], but more HVDC transmission lines are beingbuilt in MW level, while low voltage DC grid is being adopted,

Corresponding author. Tel.: +33 344234964; fax: +33 344235262.

Electrical Power and Energy Systems 58 (2014) 140149

Contents lists availab

Electrical Power an

.e lE-mail address: manuela.sechilariu@utc.fr (M. Sechilariu).ment are being developed, as well as information on grid needsand availability [3], which could assist in power balancing byavoiding undesired injection and performing load shaving duringpeak hours. For this, the smart grid is being created to facilitate

generation and power demand to meet the objeclocal power, ancillary services, and injecting powgrid if required [8]. Concerning microgrid approadvantages can be given: improving renewable e0142-0615/$ - see front matter 2014 Elsevier Ltd. All rights reserved.http://dx.doi.org/10.1016/j.ijepes.2014.01.018l maintrationenergyth self-islan-The grid capacity is built to satisfy the peak consumption. If thepeak consumption can be shifted during a day, referred to as peakshaving, the power adjustment, often ensured by excess capaci-ties working in stand-by mode, could be largely reduced. To builda more robust utility grid, strategies and means of power manage-

reduces energy cost [610]. Microgrids are considered as one ofthe possible approaches helping to develop the smart grid [11].By aggregating loads and multi-source, renewable and traditional,microgrid can operate in both off-grid and grid-connected congu-ration. It is generally considered that microgrid controls on-siteSimulationSupervision

1. Introduction

Aiming to avoid grid voltage uctuat any time instant, the electric gridthe production and the consumptionsources, storage, grid connection and DC load. The supervision control, designed as four-layer structure,takes into account forecast of power production and load power demand, storage capability, grid powerlimitations, grid time-of-use tariffs, optimizes energy cost, and handles instantaneous power balancing inthe microgrid. Optimization aims to reduce the microgrid energy cost while meeting all constraints and iscarried out by mixed integer linear programming. Simulation results, show that the proposed control isable to implement optimization in real-time power balancing with resistance to uncertainties. Thedesigned supervision can be a solution concerning the communication between loads and smart grid.

2014 Elsevier Ltd. All rights reserved.

[1,2], or even blackout,alance power betweena small margin of error.

tional communication, and reduction mismatching betweensupply and demand.

The concept of microgrid is proposed for better renewable en-ergy penetration into the utility grid and helps energy manage-ment to respond to some grid issues, such as peak shaving, andKeywords:

ing optimization in real-time operation has not been reported. The difculty is to offer resistance to opti-mization uncertainties in real-time power balancing. To cover the research gap, this paper presents theSupervision control for optimal energy coin DC microgrid: Design and simulation

Manuela Sechilariu , Bao Chao Wang, Fabrice LocmUniversit de Technologie de Compigne, AVENUES-GSU EA 7284, BP 60203, rue du Doc

a r t i c l e i n f o

Article history:Received 30 January 2013Received in revised form 7 January 2014Accepted 18 January 2014

a b s t r a c t

The development of microinstantaneous grid power bas power balancing, optimiaspects have been studied

journal homepage: wwwt management

tSchweitzer, 60203 Compigne, France

s could facilitate the smart grid feasibility which is conceived to improvencing as well as demand response. It requires microgrid control functionson, prediction, and smart grid and end-user interaction. In literature, thesetly separately. However, combining them together, especially implement-

le at ScienceDirect

d Energy Systems

sevier .com/locate / i jepes

andNomenclature

CG grid energy cost ()CLS load shedding cost ()CPVL PV production limitation cost ()CS storage energy cost ()Ctotal microgrid energy cost ()cG grid energy tariff (/kW h)cNH grid energy tariff for normal hours (/kW h)cPH grid energy tariff for peak hours (/kW h)cLS load shedding tariff (/kW h)cPVL PV production limitation tariff (/kW h)cS storage energy tariff (/kW h)CP proportional gainCREF storage nominal capacity (Ah)iPV PV current (A)iPV PV current reference (A)KD distribution coefcientKL load shedding coefcientKL_lim load shedding limit coefcientp* power reference (W)pG grid power (W)pG grid power reference (W)pG_I grid injection power (W)pG_S grid supply power (W)p grid injection power limit (W)

M. Sechilariu et al. / Electrical Powerstarting with data centers, for the reason of more efciency, lesscost, less occupied space, lower lifetime cost and more reliability[1618].

Paper [13] presents a three-levels hierarchical control accordingto ISA-95 and applied to AC or DCmicrogrids. This general approachof hierarchical control for microgrids is conceived for a large-scalepower system, upstream in the utility grid hierarchy. Imitating thebehavior of a grid synchronous generator control, the proposedhier-archical control strategy aims at balancing power between multiinverters coupled on the same bus without communication, whilecontrolling the power at the point of common coupling (PCC) atthe same time. The proposed hierarchical control is considered asa part of the central control and does not take into account the pre-diction of the power generation and the energy optimization.

In [18], support for autonomous DC microgrid applications isproposed by integrating the device-level service oriented architec-ture paradigm into the international standard IEC 61850 applica-tions. In order to create self-manageable microgrid withsemantic-enabled plug-and-play process for distributed energy re-sources, this solution provides generic middleware platform re-quired for vertical communication. However, the proposedsolution applied to the real microgrid power systems requiresadditional control and regulation policy.

A high-level energy management supervision, by means of mul-ti-agent systems, is presented in [19]. In this work, the authors fo-cus on two-level architecture for multiple interconnectedmicrogrids aiming to manage distributed energy resources in orderto match the buyers and sellers in the energy market.

G_I_lim

pG_S_lim grid supply power limit (W)pG_I_prediction grid injection power prediction (W)pG_S_prediction grid supply power prediction (W)pL load power (W)pL_D load power demand (W)pL_lim load power limit (W)pL_max load maximum power (W)pL_prediction load power prediction (W)pPV PV power (W)pPV_lim PV limited power (W)pPV lim PV limited power reference (W)pPV_MPPT PV MPPT power (W)pPV_prediction PV power prediction (W)pS storage power (W)pS storage power reference (W)pS_C storage charging power (W)pS_D storage discharging power (W)soc storage state of charge (%)SOCmax SOC upper limit (%)SOCmin SOC lower limit (%)SOC0 initial soc (%)v DC bus voltage (V)v* DC bus voltage reference (V)vPV PV voltage (V)vPV PV voltage reference (V)vPV lim PV limited voltage reference (V)vPV MPPT PV MPPT voltage reference (V)vS storage voltage (V)

AbbreviationACR automatic current regulatorAVR automatic voltage regulator

Energy Systems 58 (2014) 140149 141A generalized formulation for intelligent energy management ofamicrogrid is proposed in [20] usingmultiobjective optimization tominimize the operation cost and the environmental impact. An arti-cial neural network ensemble is developed to predict renewableenergy generation and load demand. In addition, a battery schedul-ing is proposed as a part of an optimal online energy management,seen as a decision-making process. However, smart grid data ex-changes online or dynamic energy pricing are not considered.

To increase penetration of small PV production into the grid, alocal hierarchical control with energy management is proposedin [22]. The system is presented as multi-layer control structure,each layer with a different function, and is based on an optimalpower ow management with predictions, which considers batter-ies ageing and day-ahead approach into the optimization process.However, the exchange data with the smart grid, such as limita-tions of the grid capacity, is not taken into account. Moreover,due to uncertainty of prediction and lack of grid information, thegrid power could be out of control.

Concerning the energy management two main approaches areconsidered: rule-based and optimization-based approaches. Rule-based approach manages the system according to prexed rules,such as simple rule base, multi-agent system [19] and fuzzy logicapproaches [20,21]. Optimization based approach manages thesystem by mathematical optimization, carried out with objectivefunction and constraints. The optimization methods include thearticial intelligence joint with linear programming [20], linearprogramming [21] or dynamic programming [22,23], and geneticalgorithms [24].

HMI humanmachine interfaceMPPT maximum power point trackingNH normal hoursPH peak hoursPI proportionalintegralPV photovoltaicP&O Perturb & ObservePWM Pulse Width Modulation

To sum up the results presented in these works, the rule basedsystem is simple and robust, but not guarantee the optimal perfor-mance with given operating conditions. Moreover, rules becomecomplex when facing different scenarios. The optimization givesan optimal solution within given constraints and operation condi-tion. However, optimization is usually treated as separate problemfrom the power balancing strategy, and optimization is not imple-mented in real-time operation. Since optimization are preformedbased on prediction, errors between prediction and real conditioncould result in degraded real-time operation by violating certainconstraints, or even result in failure. Thus, optimization and powerbalancing should be designed together and coupled with adequateinterface.

Hence, in contrast of the above cited works, in this study thegoal is to design and develop a microgrid local supervision control

to implement optimization in real-time operation withresistance to uncertainties;

which feed directly a DC load through their dedicated converters.Concerning the DC bus, following considerations are assumed:for the demand side, 90% of a tertiary buildings electrical load ispossible to be DC fed in efciency manner; for the grid side, thepower factor can be controlled at 1 [5].

Aiming to interact with the load environment (user, metadata,smart grid), a supervision system is added, whose main role is tobalance instantaneous power in power system following an energymanagement algorithm. The system operation must keep powerbalance while respecting constraints of certain elements. Fig. 2shows the powers ow in the microgrid, using unidirectional pow-ers whose sign convention is always positive.

The system operates respecting the available storage level and

reference is taken and the MPPT algorithm is stopped. By con-

142 M. Sechilariu et al. / Electrical Power and Energy Systems 58 (2014) 140149 to process prediction data used for optimization; to exchange data with the smart grid and with the end-user; to adapt various data in microgrid operation, such as real-timemarket pricing and user demand.

Thus, this paper focuses on supervisory control of DC microgridwhich is supposed to manage the power ow in microgrid andpower ow exchanged with the utility grid, with the objective ofmaking full use of each source while respecting their constraintson capacity and power. The microgrid based on PV sources, storage,grid connection and DC load is presented in Section 2. The supervi-sion control, designed as four-layer structure, which is supposed toexchange data with the smart grid, deal with end-user demand,forecast of PV production and load consumption is described inSection 3. Taking into account forecasting data, storage capability,grid power limitations, grid time-of-use tariffs, the predicted pow-ers ow is optimized by mixed integer linear programming andsolved by CPLEX solver. The DC microgrid control is simulatedfor two different cases, based on optimized powers ow and with-out optimization; the results are presented in Section 4. Conclud-ing remarks are given in Section 5.

2. Microgrid overview

The microgrid presented in Fig. 1 is suggested for local PVpower generation combined with storage and grid connection,

DCDCPV

Control ACDC

DCDC

*PVv

*Gp

System states

POWER SYSTEM

*Sp LK

v

SUPERVISION SYSTEM

DC Load

PV Grid which can respond to all following requirements:

to design optimization and power balancing together andcouple them through adequate interface;StorageSources connection

Fig. 1. DC microgrid overview.strained power closed loop control, the PI controller controls thePV power at the limited level. In case of low solar irradiance, thePV output power ability is less than the limited power reference,so the PI controller will decrease vPV lim until the lower limit, andvPV MPPT is taken to control the PV system. So, the limited power

Grid

PV Sources

Storage

DC Load

_S Cp _S Dp

_G Ip _G Sp

PV power through loadtaking into account the grid connection. In case of insufcient PVenergy toward the load, the system security is ensured thanks tothe grid connection and by storage, as well as the load sheddingprogram. If any excess PV power, the storage could be chargedand the grid connection gives the possibility to trade it back.

2.1. PV sources control

PV system is controlled either by a maximum power pointtracking (MPPT) method or by an algorithm to output a limitedpower pPV lim to protect the storage from overcharging or to main-tain grid injection power within imposed constraint. In this study,the chosen MPPT method is the very well known Perturb & Ob-serve (P&O) [25,26], according to which PV sources produce aMPPT power pPV MPPT . The PV production could be limited thanksto the limited power reference pPV lim calculated by the supervisionwith the algorithm described in Section 3.4.

The PV power pPV control strategy is shown in Fig. 3. The P&Oalgorithm and the PV production limiting algorithm give at thesame time corresponding voltage reference vPV MPPT and vPV lim tooperate PV system. The maximum of these two references is takenas the PV voltage control reference vPV , which represents the min-imum power. Following vPV , the PV system is operated by voltageand current double closed-loop control via automatic voltage reg-ulator (AVR) and automatic current regulator (ACR).

The output of control, duty cycle of Pulse Width Modulation(PWM) signal, is given to the power electronic devices to controlthe PV system. During the MPPT operation, if a limited power with-in the MPPT ability is given, the proportionalintegral (PI) control-ler would increase the vPV lim. For vPV lim > vPV MPPT the vPV limconnection

Fig. 2. Powers ow representation.

control does not affect MPPT algorithm and MPPT power is

*

_ limPVp

andproduced. Experiment validation of the control is provided in [27].

2.2. Storage control

Lead-acid batteries are selected as storage for the DC microgridapplied to building, because of relatively low cost and mature tech-nology [28]. The storage is operated by current closed-loop control,and the storage power is controlled by supervision system whichcalculates the corresponding power reference. The storage stateof charge soc must be respected to its upper and lower limitations,SOCmax and SOCmin respectively, to protect the storage from over-charging and over discharging, as described by Eq. (1). The soc iscalculated by Eq. (2), with SOC0 as initial soc at t0, CREF as storagenominal capacity (Ah) and vS as storage voltage. When the soc limitis not reached, the PV production should not be limited, as in Eq.(3).

SOCmin 6 soct 6 SOCmax 1

soct SOC0 13600 vS CREF

Z tt0

pS Ct pS Dtdt 2

pPV limt pPV MPPTt if soct < SOCmax 3+-

maxP&O

PI

STOP

PVv

PVi

*

_ limPVp

*

_ limPVv

*

_PV MPPTv

*

PVv

*

PVi*

PVv

PWMACRAVR+-

+-

PVp

Fig. 3. PV system control strategy.Supervision algorithm

M. Sechilariu et al. / Electrical Power2.3. Grid connection control

The grid connection is controlled by current closed-loop control.The grid power is controlled by supervision system which calcu-lates the corresponding power reference. Furthermore, as mes-sages transmitted by the smart grid via supervision system,limits for grid supply power pG_S_lim and grid injection powerpG_I_lim could be imposed. By these two limitations, grid problems,such as performing peak shaving, avoiding undesired injection ordownscaling injection uctuations caused by intermittent PV pro-ductions, can be improved. During microgrid operation, the gridpower should be controlled to satisfy Eqs. (4) and (5).

0 6 pG It 6 pG I lim 4

0 6 pG St 6 pG S lim 5

2.4. DC load control

The load power demand pL_D should be satised; nevertheless,in case of insufcient storage and grid access limits, pL_D cannotbe fully met, and the load must be partially shed, forming loadpower pL, to keep power balance. The load limited power pL_lim iscontrolled by the load coefcient KL dened by Eq. (6):

KL pL lim=pL max 6where pL_max is a constant as contractual subscribed maximumpower. It is supposed that the load could not be supplied with apower that exceeds this limit. Given the constant denominatorpL_max, therefore KL e [0, 1] changes according to the availablepower for supplying the load. If pL_D > pL_lim, the load would be shedwithin the limit level. These operations should be controlled by thesupervision system. Temporarily load partial shedding could be asolution to reduce utility grid mismatching, or to obtain less energyconsumption, within the limit agreed by end-user.

2.5. Power balancing principle

According to the powers sign convention (powers alwayspositive), the physical law of power balancing is described by:

pLt pG It pS Ct pG St pS Dt pPV t 7with pL(t) = min (pL_D(t), pL_lim(t)) and pPV(t) = min (pPV_MPPT(t),pPV_lim(t)).

Fluctuations in the DC bus voltage, which is noted v, are causedby the difference between load consumption and PV generation.The required power reference p* for power balancing is calculatedby regulating v with a proportional controller as in Eq. (8):

pt pPV t pLt CPvt vt 8where v* is the DC bus voltage control reference and CP is the pro-portional gain. For stabilizing the DC bus voltage, power balancein the system is performed by adjusting storage and grid power.Thus, p* is shared by the storage and the grid as in Eq. (9)

pt pGt pSt 9with pG(t) = pG_I(t) pG_S(t) and pS(t) = pS_C(t) pS_D(t).

After calculating p*, the grid power reference pG and storagepower reference pS are calculated according to a distribution coef-cient KD(t), as described in Eqs. (10) and (11).

pSt KDtpt 10

pGt pt pSt 11The power balancing of studied DC microgrid operation can workwith any KD value. To improve efciency and reduce energy cost,optimization calculations using specically metadata (power pre-dictions, power limits, real-time grid energy tariffs) are done bythe supervision which outputs the distribution coefcient, KD(t),whose time depending values represent the predicted optimizedpowers ow.

3. Supervision control design

The supervision system, proposed in Fig. 4, is designed in four-layer structure, which consists of humanmachine interface (HMI),prediction layer that predicts load consumption and PV produc-tion, energy management layer that optimizes the predicted pow-ers ow, and operation layer that balance instantaneous power,based on unique interface parameter KD(t), in power system.

The supervision control takes into consideration prediction dataand various constraints, such as grid power limits and storagecapacity, which can be fully respected for both optimization andreal operation. Each layer provides an independent function andthus, the structure is exible and can be implemented in several

Energy Systems 58 (2014) 140149 143microcontrollers or computers so that real-time power balancingcontrol and complicated optimization can be executed at the sametime without affecting each other. The multi-layer structure

mize the microgrid energy cost through predictive data and thus

Prediction layer

Human-machine interfaceUser demand

Metadata

Supervision

{ __

L prediction

PV prediction

pp

Load schedulingWeather forecast

Wheather

Load management data

_ limLK

Fig. 6. Prediction layer design.

144 M. Sechilariu et al. / Electrical Power and Energy Systems 58 (2014) 1401493.1. Humanmachine interface layer

The HMI design is presented in Fig. 5. User can specify the low-est limit of the load coefcient, KL_lim. Aiming to maintain powerbalance, if the system requires KL < KL_lim, the operation is assumedwith the necessary KL value, but the user will be notied.

3.2. Prediction layer

The prediction layer design is presented in Fig. 6. It calculatespossible PV power and load power evolutions for the next day.obtain the predicted optimized power ow which is then trans-lated into KD(t) sequence. On the other hand, the power balancingcontrol in operation layer is an independent function that can workwith any KD(t) value. The distribution coefcient KD(t) is a singleinterface parameter yet represents the power ow from differentsources. Hence, the communication of KD(t) does not need highspeed communication between layers.simplies implementation of such complex control strategy. Thus,the control of the power balancing is separated from energy man-agement layer, yet they are linked through one interface parameterKD(t). On one hand, the energy management layer is able to opti-

Power system states

Operation layer

Energy management layerSmart grid messages

Power system

Fig. 4. Design principle of supervision.Based on solar irradiance and temperature forecasting, and on PVmodel, built with parameters identication or PV solar irradiancemapping [29,30], the PV predictable power pPV_prediction could becalculated with as less error as possible.

The load power supply can be predicted by statistical data and/or by information of load scheduling from building managementsystem and with respect of user demand [31,32]. The load schedul-ing, that represents the operating program of building facilities,building energy needs linked to weather, and the partial and fullload shedding program are supposed known, and they are imple-mented and updated continuously by the building managementsystem. According to the value of KL_lim chosen by the user, the loadprediction power pL_prediction can be calculated by this layer.

Load shedding limit User demand

_limLK

Fig. 5. HMI layer design.3.3. Energy management layer

Energy management layer (Fig. 7) interacts with the predictionlayer and controls the operation layer by calculating the distribu-tion coefcient KD following an optimization method. The optimi-zation goal is to obtain the best power distribution between thegrid and the storage, so to reduce energy cost, grid power peak con-sumption, load shedding and limiting PV production at the sametime. In this study, the smart grid message is supposed to providereal-time grid energy tariffs and grid power limits, which assist inreducing peak supply and avoid undesired injection. Furthermore,thanks to the smart grid connection, this layer is able to inform thegrid operator of the grid supply power prediction pG_S_prediction andgrid injection power prediction pG_I_prediction.

The energy cost of system Ctotal consists of grid energy cost CG,storage energy cost CS, PV production limitation cost CPVL and loadshedding cost CLS, as in Eq. (12).

Ctotal CG CS CPVL CLS 12By calculating the energy cost for each time duration Dt, CG is

dened by Eq. (13). According to this denition, the grid powercould be bought or sold.

CG 13:6 106

XtFtit0

cGti Dt pG Iti pG Sti

with ti ft0; t0 Dt; t0 2Dt; . . . ; tFg13

This study takes into account the same price for energy pur-chased or sold, and the grid energy tariff is dened by Eq. (14).

cGt cNH 0:1=kW h for t 2 normal hoursNHcPH 0:7=kW h for t 2 peak hoursPH

14

Storage aging should be considered to give an energy tariff ofstorage using. For this study, the storage energy cost CS and an arbi-trary storage energy tariff are dened in Eq. (15).

CS 13:6 106

XtFtit0

cSti Dt pS Cti pS Dti

with ti ft0; t0 Dt; t0 2Dt; . . . ; tFg and cst 0:05=kW h15nvi: network variable inputnvo: network variable output

{ __

L prediction

PV prediction

pp

_ _ lim _ _ lim, G S G Ip pGrid time-of-use tariffs

}_ __ _

G I prediction

G S prediction

pp

_ _ lim

_ _ lim

G S

G I

D

ppK

Smart grid messagesnvo

nvi

Fig. 7. Energy management layer design.

pL lim pPV pG S lim 22

tion layer for a constant value of KD is provided in [15]. As the pre-diction layer requires available and reliable forecast data, for thePV sources geolocation and for few hours ahead, in this study,the optimized DC microgrid is tested by simulation.

4. Simulation results

The microgrid control is simulated for an operation on 23rd ofApril 2011 in Compiegne, France. The objective of this study ismore to validate a comprehensive approach rather than purelynumerical results. For this reason we do not give the numerical val-ues of various system components studied, yet powers values are

andThe cost of PV system shedding and an arbitrary PV sheddingtariff are dened in Eq. (16).

CPVL 13:6 106

XtFtit0

cPVLti Dt pPV MPPTti pPV limti

with ti ft0; t0 Dt; t0 2Dt; . . . ; tFg and cPVLt 1=kW h16

The cost of load shedding and an arbitrary load shedding tariffare dened in Eq. (17).

CLS 13:6 106

XtFtit0

cLSti Dt pL Dti pL limti

with ti ft0; t0 Dt; t0 2Dt; . . . ; tFg and cLSt 1=kW h17

Aiming to limit the power grid uctuations, grid power chang-ing rate limits are introduced as:

pGti pGti1 6 LimitpGti pGti1P Limit

with ti ft0; t0 Dt; t0 2Dt; . . . ; tFg18

As PV energy grid injection benets incentive tariffs, energy gridinjection by power grid charged storage is forbidden. Thus, the fol-lowing limits are imposed in order to ensure storage energy chargeand grid injection only from PV production.

pGtiP 0;pStiP 0 if pPV ti pL DtiP 0pGti 6 0; pSti 6 0 if pPV ti pL Dti < 0with ti ft0; t0 Dt; t0 2Dt; . . . ; tFg

19

Finally, by considering the discrete time instant ti, from t0 to tF,with the time interval Dt, the optimization problem can be com-pletely mathematically expressed by (20):

Minimize Ctotal CG CS CPVL CLSwith respect to :

pLti pG Iti pS Cti pG Sti pS Dti pPV tiSOCmin 6 socti 6 SOCmax

socti SOC0 13600vS CREFXtFtit0

pS Cti pS DtiDt

0 6 pG Iti 6 pG I lim0 6 pG Sti 6 pG S limpGti pGti1 6 LimitpGti pGti1P LimitpGtiP 0; pStiP 0 if pPV ti pL DtiP 0pGti < 0;pSti < 0 if pPV ti pL Dti < 0pPV MPPTt pPV limt 0 if soct < SOCmax

8>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>>:ti ft0; t0 Dt; t0 2Dt; . . . ; tFg

20

Energy optimization are usually solved by linear programming [21]or dynamic programming [22,33] technique. Dynamic program-ming can solve non-linear problem, while linear programmingsolves problems satisfying linear forms. Linear programming canbe more efciently solved with less time and memories. It can beveried that, mathematical formulation given by Eq. (20) followsstandard linear programming form, except for the last constraintwhich does not take a linear form. However, this constraint canbe easily linearized by introducing for each time point ti, a variablearray which is integer; this is why the optimization formulation is a

M. Sechilariu et al. / Electrical Powermixed integer linear program [34].In this study, the optimization problem is solved using the IBM

ILOG CPLEX solver [35], which is a powerful tool for solving differ-The system compares the actual load power pL with the loadpower limit pL_lim; if pL > pL_lim the load would be shed within thelimit level. When the storage is full and the grid injection limitdoes not permit absorbing all excess of PV production, the PV lim-ited production is performed by calculating pPV lim:

pPV lim pL pG I lim 23To sum up the above described power balancing, and taking

into consideration the limits of storage and grid given in Eqs. (1),(3), (4), and (5), the overall algorithm of the operation layer isshown in Fig. 9.

Experiment validation of the technical feasibility of the opera-ent types of optimization problems. However, any other mixedinteger linear programming solver also can be used (LP_SOLVE,GUROBI, . . . ). In addition, in order to express our problem in thesyntax of the solver and to call the solving algorithm of the solver,a procedure written in C++ is used. This procedure output to a lethe optimal powers ow, which is the evolution of pS_D(t), pS_C(t),pG_S(t), pG_I(t). The estimated optimum powers ow is then trans-lated into a control parameter that is the optimum distributioncoefcient KD(t). Taking into account Eqs. (9) and (10), the opti-mum KD values are calculated by Eq. (21):

KDt pS Ct pS DtpS Ct pS Dt pG It pG St21

3.4. Operation layer

According to the energy management layer output (KD(t) andgrid power limits), the operation layer presented in Fig. 8 aims atbalancing power in the power system while meeting allconstraints.

If there is not sufcient power to supply the load (insufcientPV production when the grid power is limited and the storage isempty), the load shedding is performed. In such cases, the loadpower has to be limited to pL_lim, which is calculated as in Eq. (22).

,, , , PV L Gp p soc p v

{ _ _ lim _ _ lim,G S G ID

p pK

min max, SOC SOC

* * *

_ lim, , ,G S L PVp p K p

Power system

_ limL LK K alert user<

Fig. 8. Operation layer design.

Energy Systems 58 (2014) 140149 145based on our experimental multisource system platform[12,15,36]. The simulation results of the DC microgrid were ob-tained under the MATLAB-Simulink. All the implemented auto-

*( )PV L Pp C v v

*

_ _ _ limG I G Ip pYes No

* * * * *

_

with G S G I Gp p p p p= =

* 0p

maxsoc SOCYes No

* *

S Dp K p= * 0Sp =

1LK =

No

*

_ _ limG G Ip p=

*

_ lim _ _ limPV L G Ip p p= +

min max

agemem: v, , , , , ,G PV Lp p p soc SOC SOC

Update PV limited power reference

minsoc SOC

maxsoc SOC

No

No

Yes

_ lim _V PV MPPTp=

1LK =

operation layer control.

and*p p=

minsoc SOCYes

* *

S Dp K p= * 0Sp =

Yes

*

_ _ _ limG S G Sp pYes No

* * * * *

_

with G S G S Gp p p p p= =

No

_ lim _ _ limL PV G Sp p p= +

_ lim _ maxL L LK p p=

Update load shedding control

*

_ _ _ limG S G Sp p=

Read from the energy manRead from the power syst

Yes

*

_ lim _PV PV MPPTp p=

*

Pp

Fig. 9. Flowchart of

146 M. Sechilariu et al. / Electrical Powermatic controls, described earlier, are working satisfactorily. Theproportional controller, whose proportional gain CP is used in Eq.(8), provides a wide control bandwidth adapting to simulationstep. However, synthesis for parameter tuning is not realized inthis work.

4.1. Optimization results

Nowadays services can be found providing worldwide PV powerforecast according to location and weather information. However,public data of solar irradiance (W/m2) forecasting are not yet pre-cise enough for a specied location. This is why is very difcult tocalculate the PV power prediction following real solar irradianceforecasting. The measured PV MPPT power, whose values havebeen recorded by our experimental device, is shown in Fig. 10a:the green curve shows the real time PV power evolution, whilethe gray bars are the hourly average PV power.

As this paper aims at demonstrating feasibility of the designedsupervision, to overcome the lack of real solar irradiance forecast-ing, the PV power prediction data are calculated from the real mea-surement data. PV power hourly prediction data given in Fig. 10bare assumed having random 10% error with the hourly averageof the measurement data.

Load prediction data are supposed to be given by load manage-ment system, which implies additional uncertainties. In this study,a simple arbitrary load power evolution is considered. The differ-ence of load power and load power prediction is shown inFig. 11. In order to perform an optimized operation for the nextday, prediction layer is supposed to give to the energy manage-ment layer the forecasts of the PV power and the load powerhourly evolutions.

Peak hours during the day are assumed 11:0013:00 and16:0018:00. Grid and storage constraints, as arbitrary values,_ _ lim _ _ liment layer: , ,D G S G IK p p

Energy Systems 58 (2014) 140149are imposed for the system operation. To mitigate grid powerstrong uctuations, grid power changing limit is imposed as20W/s. Arbitrary soc limits are considered as 45% and 55%, whileSOC0 = 50% and CREF = 130 Ah. Considering the storage capacity of

9:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:000

500

1000

1500

P PV

(W

)

Time

PPV_MPPT_measurement PPV_hourly average_

9:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:000

500

1000

1500

p PV

(W

)

Measure Prediction

Time

PPV_measurement PPV_prediction

(a)

(b)

Fig. 10. PV MPPT power evolution (a), hourly PV power measurement andprediction (b).

our experimental platform, whose parameters values are used inthe simulation study, these soc limits are selected to show the sys-tem behavior with relevant storage events (full, empty) in a dayrun. Grid power injection limit is imposed as 700W, and supplylimit as 600W. Based on the prediction information, the energymanagement layer calculates the optimization problem by CPLEXand gives the optimum powers ow evolution, as presented inFig. 12a. Corresponding KD(t) sequence is calculated by Eq. (21)from the optimum powers ow evolution, as presented inFig. 12b. For performing a day optimization during 9 h, the data

age has to be proposed to supply the load. During the peak hoursthe soc decreases continually, as shown in Fig. 12b. Hence, the stor-age is used to supply the load as much as possible with respect toits soc lower limit. However, as the storage energy is not enough,the grid power is also used to supply the load. This is why, duringpeak hours, storage power and grid power are proposed to sharethe necessary power to supply the load in an optimized mannerwhile respecting all constraints. This sharing power is proposedin an intermittent manner by the used solver.

4.2. Powers ow simulation controlled by KD(t) optimum evolution

The calculated KD(t) optimum evolution is given to the opera-tion layer to run the power system following the conditions givenby 23rd of April 2011 (meteorological and load). The operationpowers ow, as real situation, simulated by MATLAB, is shown inFig. 13a. During this day operation, grid and storage share powerfor supplying energy or for receiving energy at the same time. Inthe rst off-peak hours (9:0011:00), grid mainly supplies the loadfor reserving storage for peak hour supply.

During the rst peak hours (11:0013:00), the load is suppliedby storage and grid, the sharing proportion is determined by opti-mization calculation which aims also to reserve storage for supply-ing during second period of peak hours.

Just before 13:00 the surplus of the PV production is injectedinto the grid in order to make the maximum prot. Aiming to re-

9:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:000

500

1000

1500

(W)

Measure Prediction

Time

PL_measurement PL_prediction

Fig. 11. Load power measurement and day-ahead prediction.

M. Sechilariu et al. / Electrical Power and Energy Systems 58 (2014) 140149 147resolution is chosen at 10 s/point, i.e. 3240 points each powercurve. The optimization program execution time is within 10 sfor a computer with CORE i5 processor. The optimum cost esti-mated by CPLEX is 0.517 .

Fig. 12 shows the optimization calculation results. It can be ob-served that for peak hours the PV production is not sufcient tosupply the load; so, the storage and the grid have to supply theload for the remain part of power. The considered optimizationproblem is formulated to minimize the global energy cost for thewhole period from 9:00 to 18:00, while respecting all constraints.So, as for the peak hours the grid energy tariff is very high and lar-gely superior to storage energy tariff, it seems normal that the stor-

500

1000

1500

2000

Pow

er (W

)

pL_D pPV pG_I pG_S pS_C pS_D pPV_MPPT pL

Peak hours

(a)9:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:000

Time

00.10.20.30.40.50.60.70.80.9

1

Time9:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:00

KD soc

44454647484950515253545556

(b)

%

Fig. 12. Optimized powers ow (a), optimized KD(t) and soc evolution (b).duce the energy cost by avoiding grid to supply during peak hours,in the second peak hour period (16:0018:00), the storage ismainly used for supplying the load. During 13:0015:00 with theexcess PV production, the storage is charged for supplying in thesecond peak hour. Grid power injection limit and supply limitare respected. Short time load shedding can be seen in the opera-tion after 17:00, when the battery is empty. The load shedding isperformed based on instantaneous power information. To avoidload shedding uctuations in PV power uctuating circumstances,it is also possible to impose duration for load shedding in CPLEXoptimization, and optimized load shedding information could be

9:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:000

500

1000

1500

2000

Time

Pow

er (W

)

pL_D pPV_MPPT pG_I pG_S pS_C pS_D pPV pL

370375380385390395400405410415420425430

(V)

Time18:009:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00

DC bus voltage soc

44454647484950515253545556

%

Peak hours

(b)

(a) Fig. 13. Simulated powers ow (a), DC bus voltage and soc evolution (b) foroptimum KD(t).

given to operation layer to override the operation layer load shed-ding control. The energy cost is 0.512 , which is close to the opti-mization cost. The soc evolution with the optimum KD(t) and theDC bus voltage are illustrated in Fig. 13b. The DC bus voltage uc-tuations are negligible compared to the value of 400 V, signifyingthe power is well balanced.

By comparing the results presented in Figs. 12 and 13, it can beseen that the simulated powers ow is slightly different from theoptimization due to the uncertainties of solar irradiance predictionand load power prediction. During the solar uctuations between16:00 and 17:00, the storage provided more powers. However, inthis case, the storage is still able to be the main load supply duringthe second period of peak hours, but a slight load shedding occurswhen the soc reaches its low limit.

4.3. Powers ow simulation controlled by constant KD

In order to further analyze, a simulation case for a constant KD ispresented in Fig. 14. It is chosen as constant value KD = 0.5885which is the average value of optimum KD(t) evolution shown inFig. 12b. In this case, the obtained energy cost, 0.652 , the differ-ence with optimization is larger compared with using optimalKD(t), and longer load shedding can be seen during this operation.The soc evolution, illustrated in Fig. 14b, is very different from theoptimum soc evolution shown in Fig. 13b. Even optimization effect

and updating the K (t) sequence during the actual operation with-

through simple interface. It handles also constraints such as stor-age capability, grid power limitations, energy grid tariffs, grid peakhours. The optimization is based on mixed integer linear program-ming, solved by CPLEX. Simulation results, even with uncertaintiesof prediction and arbitrary energy tariffs, taken as assumptions inthis study, show that the proposed supervision design is able toperform efciency and cost effective powers ow in real-timeoperation with respect to constraints such as grid power limitsand storage capacity. Load shedding and PV power limiting ensurespower balancing in any case. The simulation shows that the opti-

148 M. Sechilariu et al. / Electrical Power andis affected, the power balancing is robust. Regarding the DC busvoltage illustrated in Fig. 14b, it can be seen that the DC bus voltageremains stable with very slight uctuations, signifying the power iswell balanced.

4.4. Simulation results: comparison and discussion

Table 1 shows the energy cost of the microgrid Ctotal given by Eq.(12) and occurrences of load shedding for these three cases: opti-mized operation by energy management layer with 10% uncer-tainties prediction data, simulated operation in case of a real PVproduction with the calculated optimum KD(t), and simulated

9:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00 18:000

500

1000

1500

2000

Time

Pow

er (W

)

pL_D pPV_MPPT pG_I pG_S pS_C pS_D pPV pL

370375380385390395400405410415420425430

(V)

Time18:009:00 10:00 11:00 12:00 13:00 14:00 15:00 16:00 17:00

DC bus voltage soc

44454647484950515253545556

%

Peak hours

(b)

(a) Fig. 14. Simulated powers ow (a), DC bus voltage and soc evolution (b) forconstant KD = 0.5885.D

out interrupting the power balancing. Thus, hourly or morefrequent optimization that updates KD(t) sequence, with latest pre-diction and power system status, is expected to give better energyperformance of the supervision control.

However, the limit of the supervision control is that optimizingeffectiveness is affected on the prediction precision. Predictionsuncertainties do not inuence power balancing but the optimal en-ergy cost is affected. Future research should focus on enhancingoptimization performance, especially facing low prediction preci-sion. An optimization technique that able to optimize power owwith consideration on uncertainties of the prediction combinedwith a rule based algorithm in operation layer that corrects KD(t)in real time with respect of power system status can be developedas one solution. Besides, a second storage can be installed as back-up for correcting the errors between optimized power and realoperation.

5. Conclusions

Facing the advent of smart grid context, a microgrid controlcombining power balancing, optimization and smart grid interac-tion is proposed through a multi-layer supervision control struc-ture. The research issue of implementing optimization in real-time operation is particularly addressed.

Based on PV sources, storage, power grid connection and DCload, the microgrid aims at self supply with limited access to grid.Taking into account forecast of PV production and load power de-mand, the four-layer supervision system performs optimizationand implements optimization in instantaneous power balancingoperation with constant KD = 0.5885. It can be seen that simulatedcost for the optimum KD(t) is close to the prediction cost, and theerror is within 1%, which is from the uncertainties of prediction.

For the constant KD case, the cost is 26% more than the predic-tion case; moreover, longer load shedding can be seen in this case.Even with some uncertainties, the optimized KD(t) operates themicrogrid with respect of the utility grid requirements and storagecapacity. This comparison validates the presented simulation casefor the proposed supervision of the proposed DC microgrid.

As previously mentioned, the communication of KD(t) does notneed high speed communication between layers. So, the supervi-sion control provides the possibility of re-performing optimization

Table 1Comparison of different cases.

Caseoperation

KD Ctotal cost()

Load sheddingduration

Optimization Optimum KD(t) as given inFig. 12b

0.517 No shedding

Simulation Optimum KD(t) as given inFig. 12b

0.512 7 min

Simulation KD = 0.5885 0.652 37 min

Energy Systems 58 (2014) 140149mization gives better energy performance while minimizing loadshedding and PV production limitation and the operation layer re-spects all constrains of power system elements.

On the other hand, the optimization efciency is based on pre-diction precision, which may limit the nal performance. Thedesigned operation layer can work with any KD value, so theprediction errors and non-optimum KD does not affect the powerbalance. However, if uncertainties are higher than 10%, the energycost and the load shedding duration could be severely affected. Thedeveloped simulation will permit in further work to design anadditional supervision layer aiming to mitigate the differencesbetween the optimized powers ow and the real one.

To sum up, the feasibility of the proposed DC microgrid super-vision control structure, that combines grid interaction and energymanagement with power balancing, is proved by simulation re-

[13] Guerrero JM, Vasquez JC, Matas J, de Vicuna LG, Castilla M. Hierarchical controlof droop-controlled AC and DC microgridsa general approach towardstandardization. IEEE Trans Ind Electron 2011;58(1):15872.

[14] Shenai K, Shah K. Smart DC micro-grid for efcient utilization of distributedrenewable energy. In: Proc of IEEE Energytech; 2011. p. 16.

[15] Wang BC, Sechilariu M, Locment F. Intelligent DC microgrid with smart gridcommunications: control strategy consideration and design. IEEE Trans SmartGrid 2012;3(4):214856.

[16] AlLee G, Tschudi W. Edison Redux: 380 Vdc brings reliability and efciency tosustainable data centers. IEEE Power Energy Mag 2012;10:509.

[17] Patterson BT. DC, come home: DC microgrids and the birth of the Enernet.IEEE Power Energy Mag 2012;10:609.

[18] Sucic S, Havelka JG, Dragicevic T. A device-level service-oriented middlewareplatform for self-manageable DC microgrid applications utilizing semantic-enabled distributed energy resources. Int J Electr Power Energy Syst

M. Sechilariu et al. / Electrical Power and Energy Systems 58 (2014) 140149 149sults. Experimental test is going be carried out in real conditionsonce the PV power prediction service is ready. Although the micro-grid only refers to a building scale and involves only a few sources,the idea of parameterize power balancing and interfacing withoptimization, as well as smart grid interaction, can be generalizedand thus can be used as solution for advanced energy managementfor other microgrids to optimize local power ow and improvefuture PV penetration.

References

[1] Lee TL, Hu SH, Chan YH. D-STATCOM with positive-sequence admittance andnegative-sequence conductance to mitigate voltage uctuations in high-levelpenetration of distributed-generation systems. IEEE Trans Ind Electron2013;60(4):141728.

[2] Wei-Lin H, Chia-Hung L, Chao-Shun C, Hsu CT, Te-Tien K, Cheng-Ta T, et al.Impact of PV generation to voltage variation and power losses of distributionsystems. In: Proc 4th international conference on electric utility deregulationand restructuring and power technologies; 2011. p. 14748.

[3] Liserre M, Sauter T, Hung JY. Future energy systems, integrating renewableenergy sources into the smart power grid through industrial electronics. IEEEInd Electron Mag 2010;4(1):1837.

[4] Peeters E, Belhomme R, Batlle C, Bouffard F, Karkkainen S, Six D, et al.ADDRESS: scenarios and architecture for active demand development in thesmart grid of the future. In: Proc of CIRED 20th international conference onelectricity distribution; 2009. p. 14.

[5] Sechilariu M, Wang BC, Locment F. Building-integrated microgrid: advancedlocal energy management for forthcoming smart power grid communication.Energy Build 2013;59(1):23643.

[6] Lasseter RH, Eto JH, Schenkman B, Stevens J, Vollkommer H, Klapp D, et al.CERTS microgrid laboratory test bed. IEEE Trans Power Delivery2010;26(1):253140.

[7] Hatziargyriou N, Asano H, Iravani R, Marnay C. Microgrids. IEEE Power EnergyMag 2007:7894.

[8] Guerrero JM, Chandorkar M, Lee TL, Loh PC. Advanced control architectures forintelligent microgridsPart I: decentralized and hierarchical control. IEEETrans Ind Electron 2013;60(4):160718.

[9] Georgilakis PS. Integration of Distributed Generation in the Power System, M.Bollen, F. Hassan. WileyIEEE Press, New Jersey (2011). Int J Electr PowerEnergy Syst 2013;48:6970.

[10] Alvarez E, Campos AM, Arboleya P, Gutirrez AJ. Microgrid management with aquick response optimization algorithm for active power. Int J Electr PowerEnergy Syst 2012;43(1):46573.

[11] Lasseter RH. Smart distribution: coupled microgrids. Proc IEEE2011;99:107482.

[12] Sechilariu M, Wang BC, Locment F. Building integrated photovoltaic systemwith energy storage and smart grid communication. IEEE Trans Ind Electron2013;60(4):160718.2014;54:57688.[19] Kumar Nunna HSVS, Doolla S. Multiagent-based distributed-energy-resource

management for intelligent microgrids. IEEE Trans Ind Electron2013;60(1):167887.

[20] Chaouachi A, Kamel RM, Andoulsi R, Nagasaka K. Multiobjective intelligentenergy management for a microgrid. IEEE Trans Ind Electron2013;60(1):168899.

[21] Chakraborty S, Weis MD, Simoes MG. Distributed intelligent energymanagement system for a single-phase high-frequency AC microgrid. IEEETrans Ind Electron 2007;54:97109.

[22] Riffonneau Y, Bacha S, Barruel F, Ploix S. Optimal power ow management forgrid connected PV systems with batteries. IEEE Trans Sustain Energy2011;2(3):32532.

[23] Bo G, Mills JK, Dong S. Energy management control of microturbine-poweredplug-in hybrid electric vehicles using the telemetry equivalent consumptionminimization strategy. IEEE Trans Vehicular Technol 2011;60:423848.

[24] Bustos C, Watts D, Ren H. MicroGrid operation and design optimization withsynthetic wind and solar resources. IEEE Trans Latin America2012;10:155062.

[25] Houssamo I, Locment F, Sechilariu M. Maximum power tracking forphotovoltaic power system: development and experimental comparison oftwo algorithms. Renew Energy 2010;35(10):23817.

[26] Houssamo I, Locment F, Sechilariu M. Experimental analysis of impact of MPPTmethods on energy efciency for photovoltaic power systems. Int J ElectrPower Energy Syst 2013;46:98107.

[27] Wang BC, Houssamo I, Sechilariu M, Locment F. A simple PV constrainedproduction control strategy. In: Proc of IEEE international symposium onindustrial electronics; 2012. p. 96974.

[28] Tan X, Li Q, Wang H. Advances and trends of energy storage technology inmicrogrid. Int J Electr Power Energy Syst 2013;44(1):17991.

[29] Houssamo I, Wang BC, Sechilariu M, Locment F, Friedrich G. A simpleexperimental prediction model of photovoltaic power for DC microgrid. In:Proc of IEEE international symposium on industrial electronics; 2012. p. 9638.

[30] Lorenz E, Hurka J, Heinemann D, Beyer HG. Irradiance forecasting for thepower prediction of grid-connected photovoltaic systems. IEEE J Select TopicsAppl Earth Observ Remote Sens 2009;2:210.

[31] Amjady N, Keynia F, Zareipour H. Short-term load forecast of microgrids by anew bilevel prediction strategy. IEEE Trans Smart Grid 2010;1:28694.

[32] Ren P, Xiang Z, Qiu Z. Intelligent domestic electricity management systembased on analog-distributed hierarchy. Int J Electr Power Energy Syst2013;46:4004.

[33] Fuselli D, De Angelis F, Boaro M, Squartini S, Wei Q, Liu D, et al. Actiondependent heuristic dynamic programming for home energy resourcescheduling. Int J Electr Power Energy Syst 2013;48:14860.

[34] Franco JF, Rider MJ, Lavorato M, Romero R. A mixed-integer LP model for theoptimal allocation of voltage regulators and capacitors in radial distributionsystems. Int J Electr Power Energy Syst 2013;48:12330.

[35] IBM.com. IBM ILOG CPLEX Optimizer. .[36] Locment F, Sechilariu M, Houssamo I. DC load and batteries control limitations

for photovoltaic systems. Experimental validation. IEEE Trans Power Electron2012;27(9):40308.

Supervision control for optimal energy cost management in DC microgrid: Design and simulation1 Introduction2 Microgrid overview2.1 PV sources control2.2 Storage control2.3 Grid connection control2.4 DC load control2.5 Power balancing principle

3 Supervision control design3.1 Humanmachine interface layer3.2 Prediction layer3.3 Energy management layer3.4 Operation layer

4 Simulation results4.1 Optimization results4.2 Powers flow simulation controlled by KD(t) optimum evolution4.3 Powers flow simulation controlled by constant KD4.4 Simulation results: comparison and discussion

5 ConclusionsReferences