Analysis Of Short-term Forecast For The Electricity In Power System Considering Influence Factorsof Historical Data

With the development of electricity market,electricity forecasting has been used tohelp the power system operator make the right decision.Accurate prediction of electricpower could arrange the start-stop of internal power generation units economically andreasonably.It can also reduce generating cost and improve the economic benefit and socialbenefit.Therefor, the research on how to predict the quantity of electricity precisely hasgreat realistic significant.This paper is based on traditional support vector machines(SVM),by means of D-S evidence theory the influencing factor of historical data and errorprecision of the electricity forecasting are the key point in the research. The main work isas follows:First of all is to find three factors by whom to built four frameworks of fusionrecognition for the evidence of influencing factor of the historical data,they are theconsumption coefficient,the chain relative ratio of curve graphs in the past years and thetemperature difference. then fusion of historical data by means of evidence theory isproposed.It avoids just considering some single influence factors like the similarity ofpower load curves or temperature in some traditional methods.And then the obtainedelectricity value which is similar to the electricity feature in the forecasting month byproof fusion,and the completion of data by the recognition framework guarantee theaccuracy and the integrity of the training data of support vector machines.Secondly, the arithmetic of support vector machines based on the D-S evidencetheory is focused.This method pretreats initial electricity data on the basis of proof fusionand solves the problem that original data fluctuates obviously. It rejects some initial datawhich is far from the electricity feature in the forecasting month and simplifies changingpatterns of the historical data .And then the prediction model of support vector machines isbuilt . Meanwhile ,simulation experiments is made in parameters selecting and Parametersbest which is suitable for this sampleis obtained.At last,the arithmetic of support vector machines by means of D-S evidence theoryand normal arithmetic of support vector machines is used in electricity forecasting respectively.By comparing and analyzing two sets of data obtained and contrasting themto normal support vector machines,the method proposed in this paper is much moreprecise in forecasting.