Page - 202 - in Emerging Technologies for Electric and Hybrid Vehicles

Energies 2016,9, 594 Thereareseveral typesofDSMapproachesdescribed in the literature. In thisarticle,weconsider two important classesof suchapproaches. Thefirst class, called control-basedDSM,makes control decisions forappliances for thenext timeperiodbasedonavailabledataat thegiven timesuchas consumption, flexibility, priorities, etc. The secondclass, calledplanning-basedDSM, plans control decisions fora longerperiod in the future (e.g.,onedayahead). Becausecontrol-basedDSMdoesnot planahead, it risksusingtheavailableflexibility tooearly, suchthatnoflexibility is leftwhenlateron, e.g., a largepeakoccurs. Planning-basedDSMdoes take this intoaccount,but requirespredictionsof futureproduction/consumption. Forexample, theplanning-basedDSMapproaches in [2,3] require suchpredictionsatahouse level. InSection2,wediscusscontrolandplanning-basedDSMinmore detail, andarguethatbothapproacheshaveseriousdisadvantages. InSection3,wedescribehowaHEMScanbeused to control the chargingof a single electric vehicle (EV)at thehouse level to followsomedesired loadprofile. In contrast to the relatedwork (e.g., [4]),nopredictionofa loadprofile is required.Weshowthatapredictionofasingleparameter characterizing theoptimalsolutionandapredictionof the loadfor theupcoming interval is sufficient tomakeanear-optimalplanningofanelectricvehicle.Weprovideananalysis thatstudieshowthe resultsofouronlineapproachapproximates theoptimalsolution,andgiveaboundonthemaximal relativedeviation. Thisboundshowsthat,underreasonablecircumstances, thecosts for theachieved solutionareonlya fewpercentmore thanthatof theoptimalsolution. Section4generalizes theseresults to thecaseofaneighborhoodwithmultipleEVs. Thepresented solution isapeak-shavingapproachthatkeeps theoverall loadof theneighborhoodbelowacertain level, andsimultaneouslykeeps the loadprofilesofeach individualhouseasflataspossible. Bythis, also the risk that thevoltageexceeds the legalbounds (e.g., 207V–253V inTheNetherlands [5]) is minimized, thecable loadiskept lowandtherebyunnecessary lossesareavoided.Note, thataDSM approachthatdoesnot take theseaspects intoaccountmaycausemoreproblemsthan it solves,aswas shownbyHoogsteenetal. [6]. SinceSections3and4dependonpredictionsofa fewcharacterizingvalues,westudyhowthese predictionsareobtained. Section5usesmeasureddataanddiscusseshowtherequiredpredictionsof thecharacteristicscanbeeasilyobtained. ThepotentialofourapproachisevaluatedinSection6andcomparedtothestate-of-the-artresearch. We show that the fewparametersweneed canbepredicted accurately andare sufficient for near optimaloperationofagroupofhouses. Section7concludes thisarticle. Summarizing, thecontributionof thisarticle isademandsidemanagementmethodologywith the followingproperties: • Itonlyusesapredictionofasingleparameterthatcharacterizestheoptimalsolution, togetherwith apowerpredictionof thehouse for theupcoming interval toplan the chargingof an electric vehicle inahouse. • Apeakatthetransformercanbecounteractedwithlowcommunicationoverheadusingadecision makingprocess thatalsorequirespredictionsofcharacteristics thataidatmakingtrade-offsat the neighborhoodlevel. • Apredictionschemefor therequiredparameters, combinedwithasensitivitystudythatshows that theresultsdonotsuffermuchfrompredictionerrors. 2.RelatedWork There are several classes of DSM approaches in the literature. In Section 2.1, we discuss control-basedDSMapproaches, and in Section 2.2 planning-basedDSMapproaches. Asweuse electricvehiclesasaproofofconcept,wetreatDSMapplied foragroupofelectricvehicles specifically inSection2.3. 202