资源描述:
《模式识别作业—人工神经网络》由会员上传分享,免费在线阅读,更多相关内容在教育资源-天天文库。
1、模式识别大作业——外文翻译ArtificialNeuralNetworksinShortTermloadForecasting人工神经网络在短期负荷预测中的应用姓名:刘德龙学号:03081413班级:030814日期:2011.05外文文献原文:ArtificialNeuralNetworksinShortTermloadForecastingK.F.Reinschmidt,PresidentB.LingStonehWebsterAdvancedSystemsDevelopmentServices,Inc.245SummerStreetBoston,U02210Phone:617-
2、589-1841Abstract:Wediscusstheuseofartificialneuralnetworkstotheshorttermforecastingofloads.Inthissystem,therearetwotypesofneuralnetworks:non-linearandlinearneuralnetworks.Thenonlinearneuralnetworkisusedtocapturethehighlynon-linearrelationbetweentheloadandvariousinputparameters.Aneuralnetworkba
3、sedARMAmodelismainlyusedtocapturetheloadvariationoveraveryshorttimeperiod.Oursystemcanachieveagoodaccuracyinshorttermloadforecasting.Keywords:short-termloadforecasting,artificialneuralnetwork1、IntroductionShortterm(hourly)loadforecastingisanessentialhctioninelectricpoweroperations.Accurateshoi
4、rttermloadforecastsareessentialforefficientgenerationdispatch,unitcommitment,demandsidemanagement,shorttermmaintenanceschedulingandotherpurposes.Improvementsintheaccuracyofshorttermloadforecastscanresultinsignificantfinancialsavingsforutilitiesandcogenerators.Variousteclmiquesforpowersystemloa
5、dforecastinghavebeenreportedinliterature.Thoseinclude:multiplelinearregression,timeseries,generalexponentialsmoothing,Kalmanfiltering,expertsystem,andartificialneuralnetworks.Duetothehighlynonlinearrelationsbetweenpowerloadandvariousparameters(whethertemperature,humidity,windspeed,etc.),non-li
6、neartechniques,bothformodelingandforecasting,tendtoplaymajorrolesinthepowerloadforecasting.Theartificialneuralnetwork(A")representsoneofthosepotentialnon-lineartechniques.However,theneuralnetworksusedinloadforecastingtendtobelargeinsizeduetothecomplexityofthesystem.Therefore,trainingofsuchalar
7、genetbecomesamajorissuesincetheenduserisexpectedtorunthissystematdailyorevenhourlybasis.Inthispaper,weconsiderahybridneuralnetworkbasedloadforecastingsystem.Inthisnetwork,therearetwotypesofneuralnetworks:non-linearandlinearneuralnetwork