摘要电力系统负荷预测就是电力生产部门得重要工作之一。通过准确得负荷预测,可以合理安排机组启停,减少备用容量,合理安排检修计划及降低发电成本等。准确得预测,特别就是短期预测对提高电力经营主体得运行效益有直接得作用,对电力系统控制、运行与计划都有重要意义。神经网络具有以下优点:(1)可以任意逼近复杂得非线性函数;(2)所有定量或定性得信息都等势分布贮存于网络内得各神经元,故有很强得鲁棒性与容错性;(3)采用并行分布处理方法,使得快速进行大量运算成为可能;(4)可学习与自适应不知道或不确定得系统;(5)能够同时处理定量、定性知识。负荷曲线就是与很多因素相关得一个非线性函数,神经网络对于抽取与逼近这种非线性函数就是一种很好得方法,用神经网络技术进行电力系统短期负荷预报可获得较好得精度。本文介绍了电力负荷预测得主要方法与神经网络得原理、结构,分析了反向传播算法与广义神经网络算法,采用改进得三层人工神经网络来建立负荷预测模型,以前七天得负荷数据与当天影响负荷得天气因素作为数据样本,进行神经网络得自我训练与学习。用Matlab软件中分别实现了基于BP与GRNN得两种神经网络得短期电力负荷预测,取得了良好得预测效果,并对两种神经网络得仿真结果进行对比,结果表明GRNN得相对误差率比BP得相对误差率要小,这说明GRNN得仿真效果胜于BP。关键字:短期负荷预测,人工神经网络,BP算法,广义回归神经网络PowerSystemLoadForecastMatlabAbstractPowersystemloadforecastingpowerproductiondepartmentisoneofthemostimportantwork、Throughthepreciseloadforecast,canarrangeunitstart-stop,reducethesparecapacity,reasonablearrangementofthemaintenanceplanandreducepowercost,etc、Accurateprojections,especiallytheshort-termforecasttoimprovetherunningefficiencypoweroperatorshavedirecteffect,onpowersystemcontrol,operationandplanstohavetheimportantmeaning、Neuralnetworkadvantages(1)canbearbitraryapproximationplexnonlinearfunctions;(2)allquantitativeorqualitativeinformationstoredinthepotentialdistributionastheneuronsinthenetwork,ithasstrongrobustnessandfaulttolerance;(3)usingtheparalleldistributedprocessingmethods,makingquicklotsofputingbeepossible;(4)canlearnandadaptivedon'tknoworuncertainsystem;(5)cansimultaneouslyprocessingquantitativeandqualitativeknowledge、Thispaperbasedonmatlabsoftwaretoshort-termneuralfunctionpowerloadforecasting,inthepredictionprocessofneuralnetworkachievedgoodpredictioneffect,mainlyusingBPandGRNNoftwokindsofneuralnetworktopredictmajorelectricityload、Thentheforecastresultsareanalyzed,inMatlabsimulationmodeltoestablishtwoneuralnetwork,thesimulationresultsareanalyzed,getgoodresults、Withthetwoneuralnetworkparedthesimulationresults,theresultsshowthattherelativeerrorratethanBPGRNNtherelativeerrorrateissmall,itshowsthesimulationresultisbetterthanBPGRNN、Keyword:Short-termloadforec