Page - 176 - in Short-Term Load Forecasting by Artificial Intelligent Technologies

Energies2018,11, 2038 Table10.Comparisonofcosts in fourcases: averagefinalprice (AFP),pessimist,DMC,andretailer. Month ConsumptionkWh AFP(in€) Pessimist (in€) DMC(in€) Retailer (in€) Saving% January 125,702 6643 6677 6567 7434 12 February 136,620 5834 5821 5772 6760 15 March 119,103 4778 4628 4551 5338 15 April 108,475 3965 3874 3804 4346 12 May 130,149 5164 5001 4948 5571 11 June 157,785 7953 7802 7752 8815 12 July 160,212 8423 8361 8313 9315 11 August 100,343 5133 4957 4936 5477 10 September 167,116 9272 9040 8989 10,036 10 October 141,077 9410 9213 9176 9953 8 November 127,613 8818 8691 8665 9534 9 December 130,583 9717 9634 9575 10,524 9 Total2016 1,604,778 85,111 83,698 83,048 93,103 11 5.Conclusions Loadforecastinghasbeenanimportantconcerntoprovideaccurateestimates for theoperation andplanningofPowerSystems,but it canalsoariseasan important tool toengageandempower customers inmarkets, forexample fordecisionmaking inelectricitymarkets. In thispaper,wepropose theusingofdifferentensemblemethods thatarebasedonregression treesasalternative tools toobtainshort-termloadpredictions. Themainadvantagesof thisapproach are theflexibilityof themodel (suitable for linearandnon-linear relationships), they take intoaccount interactionsamongthepredictorsatdifferent levels,noassumptionor transformationsonthedataare needed,andtheyprovideveryaccuratepredictions. Fourensemblemethods (bagging, randomforest, conditional forest, andboosting)wereapplied to theelectricityconsumptionof thecampusAlfonsoXIIIof theTechnicalUniversityofCartagena (Spain). In addition to historical load data, some calendar variables and historical temperatures were considered, aswell as dummyvariables representing different types of special days in the academiccontext (suchasexamsperiods, tutorialperiods,oracademic festivities). Theresults show theeffectivenessof theensemblemethods,mainly randomforest, anda recentvariantofgradient boosting called the XGBoostmethod. It is alsoworth tomention the fast computational time of the latter. To illustrate the utility of this load-forecasting tool for amedium-size customer (a campus university),predictionswithahorizonof48hwereobtainedtoevaluate thebenefits thatare involved inthechangefromtariffstopriceofwholesalemarketsinSpain. Thispossibilityprovidesaninteresting optionfor thecustomer (areductionofaround11%inelectricitycosts). AuthorContributions:M.d.C.R.-A.andA.G.(AntonioGabaldón)conceived,designedtheexperimentsandwrote thepartconcerning loadforecasting.A.G. (AntonioGuillamón)andM.d.C.R.-A.collectedthedata,developed andwrote thepartconcerningthedirectmarketconsumer.Allauthorshaveapprovedthefinalmanuscript. Funding:This researchwas fundedbytheMinisteriodeEconomía, IndustriayCompetitividad(AgenciaEstatal de Investigación,SpanishGovernment)underresearchprojectENE-2016-78509-C3-2-P,andEUFEDERfunds. The thirdauthor isalsopartially fundedbytheSpanishGovernment throughResearchProjectMTM2017-84079-P (AgenciaEstatalde InvestigaciónandFondoEuropeodeDesarrolloRegional).Authorshavealsoreceivedfunds fromthesegrants forcoveringthecosts topublish inopenaccess. Acknowledgments: Thisworkwas supported by theMinisterio de Economía, Industria y Competitividad (AgenciaEstatalde Investigación,SpanishGovernment)underresearchprojectENE-2016-78509-C3-2-P,andEU FEDERfunds. The thirdauthor isalsopartially fundedbytheSpanishGovernment throughResearchProject MTM2017-84079-P(AgenciaEstatalde InvestigaciónandFondoEuropeodeDesarrolloRegional).Authorshave alsoreceivedfundsfromthesegrants forcoveringthecosts topublish inopenaccess. Conflictsof Interest:Theauthorsdeclarenoconflictof interest. 176