Tampere University of Technology

Optimising design and power management in energy-efficient marine vessel powersystems

CitationJaurola, M., Hedin, A., Tikkanen, S., & Huhtala, K. (2019). Optimising design and power management in energy-efficient marine vessel power systems: a literature review. Journal of Marine Engineering and Technology, 18(2).https://doi.org/10.1080/20464177.2018.1505584Year2019

VersionPublisher's PDF (version of record)

Link to publicationTUTCRIS Portal (http://www.tut.fi/tutcris)

Published inJournal of Marine Engineering and Technology

DOI10.1080/20464177.2018.1505584

LicenseCC BY-NC-ND

Take down policyIf you believe that this document breaches copyright, please contact cris.tau@tuni.fi, and we will remove accessto the work immediately and investigate your claim.

Download date:05.06.2020

JOURNAL OF MARINE ENGINEERING & TECHNOLOGYhttps://doi.org/10.1080/20464177.2018.1505584

Optimising design and power management in energy-efficient marine vesselpower systems: a literature review

Miikka Jaurola a, Anders Hedinb, Seppo Tikkanen a and Kalevi Huhtala a

aLaboratory of Automation and Hydraulic Engineering (AUT), Tampere University of Technology (TUT), Tampere, Finland; bWärtsilä Norway AS,Rubbestadneset, Norway

ABSTRACTThe conventional marine vessel power systems typically have the potential to improve their fuelconsumption and their emissions. This can be done by redesigning the system configuration, themachinery and the power management strategy. The addition of options in power managementallows for the running of individual power sources closer to their optimal operating point. However,this immediately raises questions about how to redesign the system and how to operate it to max-imise the benefits. The information needed to answer these questions is often scattered aroundseparate sectors of the marine industry. The system integrator needs to be able to combine thecomplex dependencies of these individual sectors to formulate the big picture that describes thewhole power system. Numerical optimisation algorithms provide solutions to develop methodolo-gies to solvemulti-variable and potentially multi-objective problems. This literature review presentsthe authors’ findings of design and power management optimisation cases in marine vessel powersystems.

ARTICLE HISTORYReceived 8 May 2017Accepted 19 July 2018

1. Introduction

Marine engineers have been motivated to replace theconventional power systems in marine vessels with com-plex power systems to adjust for tightened emission leg-islation and to reduce operational costs. Here, the powersystem includes the propulsion, the electric and the auxil-iary systems. This type of trend is already familiar in theautomotive industry, electric power production and theheating and cooling industry to name a few. Hybridisa-tion has the greatest potential with vessel types that have avariant operational profile, are powered by a single primemover, such as an internal combustion engine (ICE) andthat have, by conventional design, a simplified powertrainconfiguration, referred to as topology, to enable optionsto be added in the power management with reasonablejustification. The potential also exists, although in lesserquantity, with more complex power systems of modernday.

A system integration designer who is responsible forthe vessel power system must ask the question if thehybrid power system is the correct solution. The addi-tional investment cost of a complex system architectureshould be paid back with the lower operational costs in areasonable time. The following defining questions arise:

CONTACT Miikka Jaurola miikka.jaurola@tut.fi

• Can the system be retrofitted or should the wholepower system be redesigned?

• Which system topology should be used?• How big should the battery pack be?• Can a smaller main engine be chosen?• What is the lifecycle of the power system?• When should the battery be charged/discharged?• Is it possible to charge the battery from the land grid?• What if the propulsion load is 5% higher?

These questions are often answered by choosing asystem topology, sizing either the retrofitted compo-nents or the whole machinery system and using a rule-based power management strategy to find the potentialimprovements in the system’s performance. However, ifthe same system output can be generated with multipledifferent control combinations, then which combinationwill best achieve the defined goals? The different lev-els in the system design are choosing the system topol-ogy, sizing the components and controlling the compo-nents and they all depend on each other, as explained inGuzzella and Sciarretta (2007). Thus, the design problemshould be solved as a nested problem as suggested alsoin Silvas (2015). The solution to this multi-level problem

© 2018 The Author(s). Published by Informa UK Limited, trading as Taylor & Francis GroupThis is an Open Access article distributed under the terms of the Creative Commons Attribution-NonCommercial-NoDerivatives License (http://creativecommons.org/licenses/by-nc-nd/4.0/), which permits non-commercial re-use, distribution, and reproduction in any medium, provided the original work is properly cited, and is not altered, transformed, or built upon inany way.

2 M. JAUROLA ET AL.

depends on the intended operation of the vessel. Theoperation of the vessel in a time frame is referred to asa duty cycle.

In numerical optimisation problems, the interestingsystem outputs are described with an objective function.The objective function value is dependent on optimisa-tion variables. The optimisation algorithm observes thebehaviour of the objective function and tries to find a setof variable values that minimise or maximise the func-tion value. In colloquial language, the system is optimisedwith a parametric variation study using a subset of inter-esting parameters of the system.While this kind of brute-force search (BF) is useful in certain cases, for hybridpower systems, the high number of design variables andthe fineness of the search grid makes BF computation-ally exhaustive. To find the minimum/maximum faster,a numerical optimisation algorithm follows either thegiven objective function gradients or tries to estimate thegradients. Additional definitions for the problem, calledconstraint functions, can also be defined. In the case ofa hybrid power system, the objective function is mostoften formulated to minimise fuel consumption. Some-times additional targets, such as emissions and invest-ment costs, are also included. The constraints typicallyset a demand for the total delivered power whichmust behigher than the load. In addition, operational limitationsof individual components are described with constraintfunctions. Although numerical optimisation is a usefultool for system design, Edgar et al. (2001) emphasisethe importance of choosing the correct algorithm for theproblem at hand.Otherwise, the problemmust be formu-lated to make it suitable for the available algorithm. Thedemand for finding the solution online and in real-timealso affects the choice of methodology.

This article presents a review of studies where eitherthe power management or the power system designwas optimised using numerical optimisation algorithms.Current system solutions are described in Section 2.Section 3 presents the optimisation cases in which onlythe system power management was studied. This is fol-lowed by Section 4 in which the component sizing isincluded in the optimisation problem. A selection ofexisting power system design tools are presented inSection 5. Projections for future work and the conclu-sions are discussed in Sections 6 and 7, respectively.

2. Increasing options for power management ina vessel power system

In conventional diesel-mechanical (DM) propulsion sys-tems, each propeller is mechanically linked to a propul-sion engine, and the engine is designed according to amaximumpropulsion load.Many vessel types, such as the

tugboat, spend the majority of their time on part load asshown by Cavalier and Caughlan (2009). This means thatthe propulsion engines run close to the optimum pointfor only a short time.

In the 1990s, the diesel-electric (DE) propulsionsystems emerged from various system designers. Themain difference in the system architecture was that themechanical power was converted into electric power andthen back to mechanical propulsion power. One largediesel engine could be replaced with a number of smallergensets, which not only provided power to the propul-sion system, but also to the auxiliary system. Part of thegensets can be shut off under low propulsion loads, mak-ing the active gensets run closer to their rated power.Although this additional degree of freedom in powermanagement enables running the gensets at more fuel-efficient operating point, the conversion losses in thegenerators and the variable-speed drive (VSD) lower theoverall system efficiency (Ådnanes 2003; Molland 2011;Babicz 2015).

By definition – a hybrid powertrain includes morethan one type of energy storage system (ESS) and powerplant for meeting the needs of a power consumer. Thebasic technologies for ESS are presented in Baseleyet al. (2007), Ibrahim et al. (2008), Lukic et al. (2008). Theapplication determines whether a good specific energydensity of modern batteries is needed or if high powerdensity storages, such as a hydraulic accumulator anda supercapacitor, are more beneficial. For example, thechosen technology in electric machines (EM) and bat-teries affects the investment cost and profitability as dis-cussed in Reed (2015). Currently, the combination ofcombustion engine and electric battery is the most famil-iar hybrid technology because of the automotive industry.In the marine industry, the definition of a hybrid powersystem is different. Power systems that utilise differentfuels in a single engine or combine the DM and DEconfigurations are sometimes called hybrids.

The common goal in a hybrid drivetrain is to oper-ate the primary power source at the optimum pointas much as possible. This is done using the secondaryenergy source to cover the difference between the oper-ating point of the primary power source and the powerconsumer as explained inHawksey (2014). For a combus-tion engine, even balancing the oscillating load comingfrom the waves is beneficial, even though the optimumoperating point may not be reached.

Figure 1 shows the typical topologies in a hybridpropulsion system. The functionalities and characteris-tics of the series and parallel topologies are explainedin NSBA (2015). From a design point, the main dif-ference is that the EM in a series topology need tobe able to cover the maximum power demand, while

JOURNAL OF MARINE ENGINEERING & TECHNOLOGY 3

Figure 1. Hybrid topologies using a combustion engine and abattery.

in parallel topologies the EM can be sized for lowerpower levels. In parallel topology, the engine can feedthe propulsion loads without additional electric conver-sion losses, which is beneficial under high propulsionloads. The series–parallel topology combines the featuresof both the series and parallel topologies. While enablingan asymmetric sizing of the EM and having a redundantEM for electric propulsive power, the increased numberof components increases the investment cost.

The design problem in marine hybrid power systemsis similar to the system in a hybrid electric vehicle (HEV).The main difference is the environment that the marineapplications work in. On-land vehicles tend to operatein environments in which the interaction with the sur-rounding traffic is more frequent. In general, marinevessels operate in open spaces with less traffic, whichmeans that the machinery controls are less transient. Thehydrodynamic interactions between the water, the pro-peller and the vessel hull affect the propeller operatingas described in Ghose and Gokarn (2004). In practicaldesigns, these interactions lead to fairly low dynamicsbetween the machinery operating point and the actualvessel motion. Therefore, the static propeller loads aresufficient in system modelling in most cases as stated inVu et al. (2015). Due to the hydrodynamics of the pro-peller, the ability to recuperate energy efficiently usingregenerative braking is a feature that is typically missingin marine vessels. Although use of an ESS would allowregenerative braking, it is not a highly interesting optiondue to the poor efficiency of the propeller when work-ing as turbine. Unlike cars, marine vessels typically havemore than one combustion engine in the power system.Although the conventional DM topologies can gener-ate electric power with the propulsion engines, there are

typically diesel fuel-powered gensets onboard as well.By linking the multiple power sources to form a hybridpower system, a designer has more flexibility to optimisethe power management in the system.

The feasibility of using a hybrid power system for dif-ferent vessel types was studied by Dedes et al. (2012),Völker (2013), Díaz-de Baldasano et al. (2014), Ander-sson and Logason (2015), NSBA (2015) and Yumet al. (2016). The service life of the battery, availabilityof shore power, vessel size and chartering commands anduse of generated heat were found to influence the feasi-bility of the hybridisation depending on the vessel type.

3. Power management optimisationmakes thebest use of existing components

The target when using parallel options for power man-agement is to know how to split the load between theparallel power sources. This target is eminent not onlyfor hybrid power systems but also for DE configurations.There is an additional dimension if a single componentcan deliver the same output with different system states;an example of this is if the controllable pitch propelleris capable of generating the same thrust with differentcombinations of rotational speed and pitch angle. Alter-natively, in a DC grid, the generators do not have to runat the same fixed speed to maintain the required voltagelevel in the grid. In addition, the optimisation problemin power management has one more degree of freedomwhen one of the energy sources has a limited capacitycompared with the other sources. For example, the bat-tery in a hybrid power system typically has a smallerenergy capacity compared with the capacity on the com-bustion side. The maximum and minimum limits of theSOC introduce time dependency in the power manage-ment optimisation which sets the question whether thebattery should be charged or discharged at the currentmoment of time? A recent literature review by Geertsmaet al. (2017) presented findings on different system con-trol strategies categorised by the vessel system topology.Before the actual system design is optimised, the powermanagement optimisation is studied with a pre-definedsystem topology consisting of a set of pre-chosen com-ponents.

Baldi et al. (2016) used the mixed integer non-linearprogramming method to optimise the load sharing ina hybrid power system without batteries. Combiningsequential quadratic programming for the continuousvariables and using the branch and boundmethod for theinteger variables, the problem was solved with Matlab. Acruise ship was used for the case study. The shaft gen-erator/motor was sized for the hybrid power system bytrying out different unit sizes. The hybrid system involved

4 M. JAUROLA ET AL.

lowered operational costs especially when the dissipatedheat from themain engines was utilised in the ship’s heat-ing system. The investment costs and payback period,however, were not considered.

The integrating element of a rechargeable energy stor-age with capacity limits brings a time aspect to the prob-lem. Vu et al. (2014a,2014b,2015) presented variations ofa powermanagement optimisationmethod using a serieshybrid tugboat. The penalty function included elementsfor fuel consumption, load demand tracking and changesin battery energy. A genetic algorithm (GA) was used forthe search. Studies compared algorithms when the dutycycle was known beforehand and if the duty cycle wasunknown, but could be predicted based on the statisticaland repetitive nature of the tugboat duty cycle. Com-pared with the rule-based power management strategydescribed in Sciberras and Norman (2012), 9% fuel sav-ings were obtained with the optimised machinery usage.The difference between the predicted duty cycle and theknown duty cycle was less than 0.5%.

Zahedi et al. (2014) studied the fuel savings in a serieshybrid off-shore support vessel with a DC grid. Theproposed power management strategy charges and dis-charges the battery between theminimumandmaximumcapacity in a repetitive cycle. If the pulsed cycle of thissquare wave-like usage becomes infeasible, a so-calledcontinuous mode is engaged, where the SOC oscillatesin a smaller range to even the load oscillation at theactive diesel generators. Using a simulation model fromZahedi and Norum (2013), they showed that with theproposed algorithm, the DC grid series hybrid leads to15% fuel savings compared with a conventional DE con-figuration with an AC grid and 7% fuel savings comparedwith a DC grid configuration without a battery. The BFalgorithm was used to find the optimum operating pointfor the generators under a dynamic load. Both onlineand offline algorithms were proposed; the online versionof the algorithm uses a search space with less points forfaster computation and a low-pass filter to estimate theaverage load and the load ripple.

Haseltalab et al. (2016) expanded the power manage-ment problem with a velocity tracking error, and useda model predictive controller (MPC) to minimise boththe velocity tracking error and the fuel cost in the load-sharing optimisation case. Using the simulation modelfrom Zahedi and Norum (2013) with a combination ofbattery and an ultra-capacitor as the energy storage, thefeasibility of theMPCunder surge conditions was shown.The actual algorithm for the numerical optimisation orperformance improvements to a reference systemwas notpresented.

A battery hybrid tugboat with a parallel topology wasstudied by Grimmelius et al. (2011). Two cost functions

were compared side by side to find the global minimumfor the problem: one for themodewhich the EMworks asa motor and one which EMworks as a generator. Despiteevaluating two separate cost functions, a good potentialfor real-time implementation was obtained with a lin-earised controller, and utilising linear programming andthe Simplex optimisation method in Matlab/Simulink.The cost function in the used equivalent cost minimisa-tion strategy (ECMS) includes the cost of energy storedin the battery; thus, the future use of the battery is con-sidered. The study did not present a comparison to thebaseline system.

Bassam et al. (2017) compared different energy man-agement strategies for a hybrid passenger vessel pow-ered by fuel cells and a battery. The authors’ multi-scheme strategy was a combination of the other strategiesreviewed in the comparison. A Simscape and Matlabenvironment was used for system modelling and opti-misation. The duty cycle of the studied passenger vesselwas fairly flat excluding the docking phase with higherload peaks and the acceleration phase with a higherand wider power peak. The size of the battery and theenergy management strategies that were used, led toa small variation in the battery SOC during the oper-ation excluding the charge-depleting–charge-sustaining(CDCS) strategy. The battery was charged to the initialSOC using shore power after the daily shift. Includingthe battery charging cost, the total operating costs of thedifferent energy management strategies were reportedto be similar, although most cases favoured the ECMSstrategy over the developed multi-scheme strategy. Thisoccurred even when the sensitivity towards energy priceswas considered. The search algorithm used and the base-line results of the vessel powered by the fuel cells alonewere not reported. The results imply that the chosen lead-gel battery might not be ideal for the duty cycle of thestudied vessel with high power and low energy transients.

As a bridge between Sections 3 and 4, theMaster’s the-sis by Kwasieckyj (2013) is briefly reviewed. Proposing adesign methodology built with MS Excel, four differentvessel types were studied that had the potential to applythe hybrid power system. However, in this case, the useof battery was not included in the hybridisation. For eachvessel type, the author handpicked a number of differentmachinery topologies that enabled load sharing betweenparallel power sources. The load share was optimisedusing the generalised reduced gradient search algorithm.Because this is an algorithm for finding a local optimum,a set of initial guesses was pre-selected for each topol-ogy for a more complete coverage in the search space.The objective function included the fuel consumptionof the combustion engines, while the constraint func-tion ensured that the load demand ismet withmachinery

JOURNAL OF MARINE ENGINEERING & TECHNOLOGY 5

operating within the specified limits. Although the actualpower system design was not optimised, this researchstudied a small subset of possible system topologies. Inaddition, it considered the sensitivity of the solutionswith changes in the operational profile, component effi-ciencies and control strategy. Although the investmentcost was not included in the formulation of the opti-misation problem, it was estimated for the comparedtopologies.

4. Increasing the frame size in the optimisationproblem: solving the sizing problem

On a general level, minimising the fuel consumptiontends to decrease the load share of the combustion sideand increase the load share of the electric side. However,this means that the components in the electric side wouldhave to be sized for a higher rated power and capacity,which would lead to larger, heavier and more expen-sive electric components. Thus, the two objectives of fuelefficiency and sizing conflict with each other. Two formu-lation approaches are typically used for a multi-objectivesizing optimisation: a penalty function that weighs thesetwo objectives and leads to a single optimum just as inthe previous power management optimisation cases, orthe solutions are ranked in groups based on their fitnesstowards the problem. The results of the latter approachare often presented in Pareto fronts, shown in Figure 2in which the best solutions get the smallest rank. Withineach rank, all solutions are equally as good as the othersolutions. Therefore, the final decision needs additionalinformation that is either not available or is not used forinput when the problem is formulated.

Component sizes affect the objective function and theconstraints set for the typical powermanagement optimi-sation problem. Reflecting cases presented in Section 3,the size of the combustion engine defines the fuel con-sumption rate, which is typically included in the objectivefunction to be minimised. The fuel consumption rate foran engine is specific for each size, but often simplifiedusing a generic specific fuel consumption (SFC) chart inwhich the fuel consumption rate is scaled with the ratedpower. The capacity of the battery typically affects theconstraints of the problem; since the battery is a time-integrator element, the capacity defines when chargingor discharging should be done and the magnitude ofcharging. Optimising the component sizing requires thatthe individual component sizes are set as optimisationvariables.

Solem et al. (2015) presented a design optimisationstudy for a vessel with a DE topology. Although this wasnot a hybrid system per se, it was a good example ofsimplifying the system characteristics to find a global

Figure 2. Pareto optimum of a generic multi-objective optimisa-tion problem.

solution for the multi-objective optimisation problem.The diesel generators of a small size anchor handlingvessel were sized using the branch and bound method,and the cost function, which included penalties for fuelcost, investment cost, NOx taxes and the area requiredfor the installation. The use of special ordered sets oftype 2 (SOS2) guaranteed that a global optimum couldbe found when using piecewise linear fuel consumptiondata to model the diesel engines. Using engine data ofmultiple models from multiple manufacturers, the studypointed out, that with the used penalties the objectivefunction typically becomes flat and the feasible enginemodel selection is large. This means that the solutionsbecome more sensitive for the stopping criteria for thealgorithm.

Del Pizzo et al. (2010) presented two different designmethods for sizing the main components in a seriestopology hybrid system. The methods were designedaccording to the primary power source operation type.Both methods were formulated to sustain the batterySOC; the final SOC was the same as the initial SOC overthe studied vessel duty cycle. An evaluation was carriedout for a hybrid waterbus using the exhaustive searchmethod. Discrete parametric variation of the ratio for themaximum and average generator power showed that forthe estimated duty cycle of the waterbus, a minimum fuelcost was obtained with a ratio of 1.15.

Dupriez-Robin et al. (2009) used the so-called powermodelling method for sizing the main components ina sailboat. The sailboat incorporated a series topologyin the electro-mechanical driveline, which was comple-mentedwith a parallel wind energy source (the sails). Theoptimum power management was then searched by for-mulating a discrete search space with dynamic program-ming (DP). Using DP, a global optimum was found for

6 M. JAUROLA ET AL.

the powermanagement problem. The authors recognisedthat uncertain wind conditions had a strong influenceon the optimum solution. Thus, test cycles with vary-ing wind conditions were studied. It was suggested thatthis optimisation method could be further refined bystudying the neighborhood of the found optimum withlocal optimisation routines. No comparison to a baselinesystem was reported.

Studying a series hybrid of a DC grid type, Soley-mani et al. (2015) optimised the sizing of the machin-ery. A particle swarm optimisation (PSO) algorithm wasused for rightsizing a combustion engine, generator, elec-tric motor and battery pack. The objective function onlyincluded the fuel costs over the studied duty cycle. Theconstraint function included penalties for the vessel oper-ational requirements (maximum acceleration, accelera-tion time and maximum velocity). A simple thermo-stat control strategy was used to switch the combustionengine on and off based on the battery SOC, in a similarway used by Zahedi et al. (2014). Continuous scaling fac-torswere used for the fuel consumption and the efficiencymaps of the main components leading to a continuousoptimisation space. An interesting result in the sizingwas that, although no penalty was used for the compo-nent space requirement or mass, the optimum solutiondid not maximise the battery capacity. With optimisedcomponent sizes, the authors reported a considerablefuel consumption saving compared with the baseline sys-tem. After finding optimum sizes for the components,the control strategy was further refined with a com-bined thermostat-fuzzy logic controller. The improvedPSO-fuzzy controller was only used to refine the powermanagement and it did not contribute in the sizing opti-misation routine.

Sciberras and Grech (2012) and Sciberras and Nor-man (2012) studied a motor yacht with parallel topologyto find an optimum solution between the two conflict-ing objectives of fuel consumption and physical size ofthe system. Instead of continuous scaling of the com-ponents as done by Soleymani et al. (2015), the authorsused a database of discrete components for a more real-life selection of available components. Using a rule-basedcontrol strategy, the control values for the componentswere set based on the velocity demand, battery SOCand the operating limits of the components. For thisnon-linear and discontinuous problem, non-dominatedsorting genetic algorithm (NSGA-II) and Pareto rank-ing were used. An interesting finding was that all feasiblesolutions had the same ICE as the original baseline sys-tem. This is likely because the duty cycle is fairly flat withone larger energy peak; the electric drive with higherrated power and larger battery capacity was penalized

heavily if this energy peak was to be covered in theelectric assist mode.

A similar study to the one by Sciberras and Nor-man (2012) was reported by Zhan et al. (2015). Thetest case involved retrofitting a battery pack in a trailingsuction hopper dredger in which the diesel engine, theVSD and the battery pack were sized using the NSGA-IIalgorithm. Retrofitting a parallel hybridwas supported bythe fact that the original system already included a shaftgenerator and that the duty cycle included power peaksfor each cycle during the operation. A database of discretemachinery units was used, but in addition to the objec-tive function used by Sciberras and Norman (2012), anadditional progressive punishment was used if the min-imum limit of the battery SOC was not respected. Theinteresting outcome in the results is that compared withthe fuel consumption and installation weight of the origi-nal system, bothmain objectives improvedwith solutionsin the Pareto front. Although a clear report on the originalsystem component characteristics, such as SFC, efficien-cies and rated powers was missing, the results hinted ona successful rightsizing of the system.

Skinner et al. (2009) used a multi-objective GA todesign the propulsion system of a nuclear-powered attacksubmarine. With four mission scenarios, the followingpropulsion drive topologieswere studied: a puremechan-ical drive, a direct electric drive, a geared electric drivewith multiple parallel electric motors and a parallelgeared combination of steam turbine and electric motor(referred to a hybrid). The sizes of the components rele-vant to each topology, including the propeller diameter,were set as the optimisation variables. Multiple objectivefunctions were formulated to maximise the componentefficiencies and to minimise the EM size and total sys-tem energy consumption. The results showed the tradeoffcharacteristics and, since multiple different mission sce-narios were studied, the designer was forced to choosewhich scenario to optimise the machinery. By using asupervisory controller to minimise the combined powerloss within the topology, the hybrid topology showed thebest energy consumption of the three alternatives. Thisresults from avoiding the poor efficiency of a steam tur-bine at low loads. The authors recognised the downsidesof the physical size of the complex hybrid system alongwith the fact that the single prime mover in the othertopologies could be replaced with multiple smaller onesto improve their energy efficiency.

Wen et al. (2016) studied battery size optimisation in alarge oil tanker powered by a combination of photovoltaicsolar panels, gensets and a battery pack. The battery bal-anced the oscillating output of solar panels as the shipmoves in the waves. The problem was formulated for the

JOURNAL OF MARINE ENGINEERING & TECHNOLOGY 7

interval optimisation method and INTLAB in coopera-tion with Matlab was used. The multi-objective problemwas solved using weighted penalties for fuel cost, bat-tery installation and replacement cost and the capitalrecovery factor. In addition to the load demand and com-ponent limitations, penalty for the amount of greenhousegas emissions was included. The article reported that theswinging motion of the ship influences the optimal bat-tery size significantly. Optimal sizing of the battery withthe proposed interval optimisation method reduced thenet present cost (NPC) of the system by 12% for the cho-sen operating route. The greenhouse gas emissions werereduced to almost one fourth with an optimum batterysize. Lan et al. (2015) studied the same case without theeffect of swinging motion, and used a combination ofthe PSO and NSGA-II algorithms reaching to 24% NPCreduction.

The characteristics of five optimisation algorithms forthe design optimisation of a HEV powertrain were sum-marised in Silvas et al. (2014). In all cases, DP was usedto optimise the power management level. The non-linearand potentially discontinuous nature of the design prob-lem requires the use of a global search method. Althoughthe PSO and GA are widely used in design optimisationof hybrid power systems, the fast computation and the re-usability without tuning favoured DIRECT algorithm indesign optimisation level according to Silvas.

Regarding the optimal sizing of hybrid machinery, avery important point was made by Wen et al. (2016).The ideal target would be to size the machinery foroptimum energy consumption. However, in reality, thesafety and usability aspects harshly override the load-sharing target. For instance, the diesel gensets in Wenet al. (2016) needed to be sized based on the maximumpower demand of the tanker, while only battery size wasoptimised for the application. Certain vessel types oper-ating in a harbour or near the coast might be more fea-sible for utilising a limp-home mode, which is not thecase with vessels operating far offshore. For leisure boats,it is also crucial to maintain a smooth driver experienceunder a faulty system condition; it is inconvenient for aboat captain to sail at a reduced speed when the batteryis drained or a fault occurs in the electric grid. Anotherreliability issue with hybrid systems in marine vessels isthat the crew needs to be trained to operate andmaintainthese more complex systems making it more challengingfor the boat owners to find skilled people for the crew.

Finally, a comment about the choice of time horizonand the power management strategy should be made.An offline design problem is formulated using a chosenpower management strategy and a duty cycle known apriori. Thus, the found optimum is valid only for thatparticular duty cycle if the chosen power management

Figure 3. Optimisation time domains.

strategy is used. For example, using a global searchmethod, such as DP, for the optimum power manage-ment over the duty cycle, a designer can find the utopiapotential of the system. In real life, the decision aboutthe power management is typically based on the presentsystem state or, in sophisticated approaches, the systemstate is predicted over a definite prediction window (seeFigure 3). Since the strategy is different, the real-life appli-cation is not capable of delivering the same performanceas the offline design even if the offline model would oth-erwise be an ideally accurate description of the system.Although offline strategies are not directly realisable inreal-life application, they are useful in providing a bench-mark solution and they can be modified to be used inonline strategies, as stated in Waschl et al. (2014).

5. Available power system design tools

There is a shortage of commercial simulation anddesign optimisation tools for hybrid power systems inthe marine sector. The RAptures tool in den Hertoget al. (2009) was developed to calculate the fuel con-sumption, emissions and investment cost of the powersystem, which is specified by the user. The input datafor the model include vessel and component specificdata, the duty cycle of the vessel and the topology ofthe machinery. However, power management optimisa-tion capabilities between parallel power sources were notreported.

A system engineering framework for marine energysystems called DNV COSSMOS was reported inDimopoulos et al. (2014). Themodelling, simulation andoptimisation software, developed by Det Norske Ver-itas (DNV), allows users to design the vessel powersystem from a library of components and study theenergy efficiency, emissions, reliability and cost of thesystem. The level of detail in the modelling can be cho-sen from a dynamic presentation of the system usingpartial differential equations to a simplified steady-statedescription formore high-level design purposes. Embed-ded optimisation features and the capability of using

8 M. JAUROLA ET AL.

Figure 4. Block diagram for author’s own optimisation methodology.

external optimisation routines arementioned, but are notreported in detail.

HOMER Pro, in Unknown (2016), is a microgriddesign tool which is suitable for various sectors in energyproduction. Although it seems to be mainly aimed fordesigning onland powergrids, it could well be used alsoin marine sector. It has a library with diverse rangeof component types for the user to compile and sim-ulate the microgrid. It optimises the microgrid energyproduction using a ‘modified grid search’ algorithmand allows the user to make a sensitivity analysis, forinstance, for different fuel prices. The algorithm itself isnot described in detail as it is one of themain assets of thesoftware.

6. Future work

The future target of the authors is to develop a toolfor a vessel system integrator. Focus is on assisting inthe decisions of power system design as early as possi-ble during the design process. To keep the tool genericfor different system topologies, the components and sys-tem configuration need to be easy to build and easy tomodify for the sake of comparing different system alter-natives. First, user specifies the components and vesseldata in the library of the tool. For example, diesel engines,gensets, batteries, gearboxes and propellers can be speci-fied. Using quasi-static component models, less param-eters are needed and the data needed to specify themis easier to access through sales brochures or manufac-turer measurement data. Also, a quasi-static duty cycle isspecified by defining either the propeller load or with acombination of vessel resistance and velocity. Similar topropulsion duty cycle, user also specifies the hotel andauxiliary load in the electric grid. Then user specifieswhich components are chosen for the case study and howthey are connected to each other and which modes theycan operate. The load-sharing between multiple powersources is optimised in a global scheme by investigating

the different modes the power system is capable of oper-ating in. Optimisation algorithms, such as gradient basedlocal algorithm, PSO and DP are already available for theuser to choose from.

As a short preview, Figure 4 shows a block diagramof the optimisation procedure. Essentially, the core of thetool will be split in two layers: the design layer and thepower management layer. The latter is an internal layerwhich finds the optimum usage for the machinery spec-ified in the higher design layer. The design optimisationlayer makes the decisions for sizing the machinery for achosen topology. The power management layer will bepresented next in the pipeline of future publications. Thechosen propeller type (fixed/controllable pitch) and howit affects powermanagement, is also brought in the frameof the design optimisation. For the future work, the mainfindings in this literature reviewwere the nested structureof the design optimisation procedure from Silvas (2015)and themode-wise local optimisationwith electricmotoras seen in Grimmelius et al. (2011). The approach usedin the latter article can be used in conjunction with cre-ating combinatory control modes seen in Linjama andVilenius (2005).

7. Conclusions

This literature review presented an outlook on designand power management optimisation studies for hybridpower systems inmarine vessels. The feasibility of using ahybrid power system inmarine vessels is strongly affectedby the duty cycle of the vessel. Numerical optimisationcan be a useful tool to assess the feasibility of hybridpower system. This requires that the system integrationdesigner knows enough about the whole power systembehaviour and not just small sectors of marine technol-ogy. Although a vessel might show potential in cost sav-ings using an energy-efficient system design and powermanagement, the tight safety and reliability regulationscoupled with more conservative mindset in the marine

JOURNAL OF MARINE ENGINEERING & TECHNOLOGY 9

sector tend to set harder constraints in system designcompared with the automotive industry.

The problem complexity and the required computa-tional effort depend on the choice of the problem framesize, the method used to the search for the solution andthe level of detail and dynamics of the system model.Whereas the power management problems for simplifiedconvex systems can be solved on a laptop within seconds,the topology and sizing optimisation problem of a non-linear and discontinuous system may take days to solveeven with parallel computing.

The authors of this article are developing a design toolfor optimising marine vessel power system integration.More detailed description of this tool will be published infuture and the next article will focus on the methodologyfor power management optimisation in this tool.

Acknowledgements

This work was conducted as a part of the first author’s PhDstudies at the Doctoral School of Industry Innovations (DSII).

Disclosure statement

No potential conflict of interest was reported by the authors.

Funding

This work was conducted in collaboration with Wärtsilä andTampere University of Technology.

ORCID

Miikka Jaurola http://orcid.org/0000-0002-6103-8868Seppo Tikkanen http://orcid.org/0000-0003-2973-094XKalevi Huhtala http://orcid.org/0000-0003-4055-0392

References

Ådnanes AK. 2003. Maritime electrical installations and dieselelectric propulsion. ABB. P. 7.

Andersson A, Logason K. 2015. Feasibility study of hybridpropulsion systems for long-liner fishing vessels [master’sthesis]. Chalmers University of Technology.

Babicz J. 2015. Wärtsilä encyclopedia of ship technology: Az.Helsinki: Wärtsilä Corporation; p. 175–176.

Baldi F, Ahlgren F, Melino F, Gabrielii C, Andersson K. 2016.Optimal load allocation of complex ship power plants.Energy Convers Manag. 124:344–356.

Baseley S, Ehret C, Greif E, Kliffken MG. 2007. Hydraulichybrid systems for commercial vehicles. SAE International.SAE Technical Paper; 10.

BassamAM, Phillips AB, Turnock SR,Wilson PA. 2017. Devel-opment of amulti-scheme energymanagement strategy for ahybrid fuel cell driven passenger ship. Int JHydrogenEnergy.42(1):623–635.

Cavalier D, Caughlan S. 2009. Design of a low emissions har-bour tug. In: International tug and salvation 2008, Singapore.[accessed 2018 Jan 10]. http://seaways.net.au/wp-content/uploads/2015/11/D2P7.pdf.

Dedes EK,HudsonDA,Turnock SR. 2012. Assessing the poten-tial of hybrid energy technology to reduce exhaust emissionsfrom global shipping. Energy Policy. 40:204–218.

Del Pizzo A, Polito RM, Rizzo R, Tricoli P. 2010. Designcriteria of on-board propulsion for hybrid electric boats.In: 19th International Conference on Electrical Machines;ICEM.

den Hertog V, Harford K, Stapleton R. 2009. RAptures: resolv-ing the tugboat energy equation. In: Tugnology ’09, Ams-terdam. [accessed 2018 Jan 10]. http://ral.ca/wp-content/uploads/2016/05/D1P3-RAptures-Resolving-the-Tugboat -Energy-Equation.pdf.

Díaz-de Baldasano MC, Mateos FJ, Núñez-Rivas LR, Leo TJ.2014. Conceptual design of offshore platform supply vesselbased on hybrid diesel generator-fuel cell power plant. ApplEnergy. 116:91–100.

Dimopoulos GG, Georgopoulou CA, Stefanatos IC, ZymarisAS, Kakalis NM. 2014. A general-purpose process mod-elling framework for marine energy systems. Energy Con-vers Manag. 86:325–339.

Dupriez-Robin F, Loron L, Claveau F, Chevrel P. 2009. Designand optimization of an hybrid sailboat by a power modelingapproach. In: Electric Ship Technologies Symposium, 2009;IEEE. p. 270–277.

Edgar T, Himmelblau D, Lasdon L. 2001. Optimizationof chemical processes. McGraw-Hill chemical engineer-ing series. [accessed 2018 Jan 10]. https://books.google.fi/books?id=PqBTAAAAMAAJ.

Geertsma R, Negenborn R, Visser K, Hopman J. 2017.Design and control of hybrid power and propulsion sys-tems for smart ships: a review of developments. Appl Energy.194:30–54.

Ghose J, Gokarn R. 2004. Basic ship propulsion. New Delhi:Allied Publishers; p. 28–93.

Grimmelius H, de Vos P, Krijgsman M, van Deursen E. 2011.Control of hybrid ship drive systems. In: 10th Internationalconference on computer and IT applications in the maritimeindustries. p. 1–15.

Guzzella L, Sciarretta A. 2007. Vehicle propulsion systems.Berlin: Springer; vol. 1, p. 37–38.

Haseltalab A, Negenborn RR, Lodewijks G. 2016. Multi-levelpredictive control for energy management of hybrid ships inthe presence of uncertainty and environmental disturbances.IFAC-PapersOnLine. 49(3):90–95.

Hawksey G. 2014 Oct. Hybrid options. Electric Hybrid MarineTechnol Int. 2014:30–34.

Ibrahim H, Ilinca A, Perron J. 2008. Energy storage systems– characteristics and comparisons. Renew Sust Energ Rev.12(5):1221–1250.

Kwasieckyj B. 2013. Hybrid propulsion systems [master’s the-sis]. Delft University.

Lan H, Wen S, Hong YY, David CY, Zhang L. 2015. Optimalsizing of hybrid pv/diesel/battery in ship power system. ApplEnergy. 158:26–34.

Linjama M, Vilenius M. 2005. Digital hydraulic tracking con-trol of mobile machine joint actuator mockup. In: Pro-ceedings of the Ninth Scandinavian International Confer-ence on Fluid Power, SICFP’05, June 1-3, 2005, Linköping,Sweden.

Lukic SM, Cao J, Bansal RC, Rodriguez F, Emadi A. 2008.Energy storage systems for automotive applications. IEEETrans Industr Electron. 55(6):2258–2267.

10 M. JAUROLA ET AL.

Molland AF. 2011. The maritime engineering reference book:a guide to ship design, construction and operation. Oxford:Elsevier; p. 375–380.

[Nova Scotia Boatbuilders Association] NSBA. 2015. Reviewof all-electric and hybrid-electric propulsion technology forsmall vessels. Nova Scotia Boatbuilders Association. ReportNo: 902.

Reed G. 2015 Oct. Architecture and motor selection. ElectricHybrid Marine Technol Int. 2015:84–87.

Sciberras E, Grech A. 2012. Optimization of hybrid propul-sion systems. Int J Marine Navig Safety Sea Transport.6(4):539–546.

Sciberras E, Norman R. 2012. Multi-objective design of ahybrid propulsion system for marine vessels. IET ElectricalSystems in Transportation. 2(3):148–157.

Silvas E. 2015. Integrated optimal design for hybrid electricvehicles [dissertation]. Technische Universiteit Eindhoven.p. 19–36.

Silvas E, Bergshoeff E, Hofman T, Steinbuch M. 2014. Com-parison of bi-level optimization frameworks for sizing andcontrol of a hybrid electric vehicle. In: Vehicle Power andPropulsion Conference (VPPC), 2014 IEEE. IEEE. p. 1–6.

Skinner BA, Parks GT, Palmer PR. 2009. Comparison of sub-marine drive topologies using multiobjective genetic algo-rithms. IEEE Trans Veh Technol. 58(1):57–68.

Solem S, Fagerholt K, Erikstad SO. 2015. Optimization ofdiesel electric machinery system configuration in concep-tual ship design. J Mar Sci Technol (Japan). 20(3): 406–416.http://dx.doi.org/10.1007/s00773-015-0307-4.

Soleymani M, Yoosofi A, Kandi DM. 2015. Sizing and energymanagement of a medium hybrid electric boat. J Mar SciTechnol (Japan). 20(4):739–751.

Unknown. 2016. Homer pro-microgrid software for designoptimized hybrid microgrids. [accessed 2017 Dec 18].https://www.homerenergy.com/homer-pro.html.

Völker T. 2013. Hybrid propulsion concepts on ships. Sci JGdynia Maritime Univ. 79:66–76.

Vu TL, Ayu AA, Dhupia JS, Kennedy L, Adnanes AK. 2015.Power management for electric tugboats through oper-ating load estimation. IEEE Trans Control Syst Technol.23(6):2375–2382.

Vu TL, Dhupia JS, Ayu AA, Kennedy L, Adnanes AK. 2014a.Control optimization for electric tugboats powertrain with agiven load profile. In: International Symposium on FlexibleAutomation ISFA. ASME.

Vu TL, Dhupia JS, Ayu AA, Kennedy L, Adnanes AK. 2014b.Optimal power management for electric tugboats withunknown load demand. In: American Control Conference(ACC); IEEE. p. 1578–1583.

Waschl H, Kolmanovsky I, Steinbuch M, Del Re L. 2014. Opti-mization and optimal control in automotive systems; vol.455; Springer. p. 205.

Wen S, Lan H, Hong YY, David CY, Zhang L, Cheng P. 2016.Allocation of ess by interval optimization method consider-ing impact of ship swinging on hybrid pv/diesel ship powersystem. Appl Energy. 175:158–167.

Yum KK, Tasker B, Skjong S, Pedersen E, Steen S. 2016. Sim-ulation of a hybrid marine propulsion system in waves. In:Proceedings of 28th CIMACWorld Congress.

Zahedi B, Norum LE. 2013. Modeling and simulation of all-electric ships with low-voltage dc hybrid power systems.IEEE T Power Electr. 28(10):4525–4537.

Zahedi B, Norum LE, Ludvigsen KB. 2014. Optimized effi-ciency of all-electric ships by dc hybrid power systems. JPower Sources. 255:341–354.

Zhan K, Gao H, Chen H, Lin Z. 2015. Optimal retrofitting of ahybrid propulsion systemusingNSGA-II algorithm for trail-ing suction hopper dredger. In: Electric Ship TechnologiesSymposium (ESTS); IEEE. p. 201–206.