The Research On Methods Of Load Regional Forecasting

Power load forecasting is a method using the prediction theory, according to theparts of electric power load and various factors influencing the load changes, which takethe change law of historical data and the influence rule of intrinsic factor load intoaccount. It summarizes the effects of future load and change trend, and uses the law topredict the future load. Load forecasting provides the basic data and plays a veryimportant role in the work and the safe operation of the power system especially thebasic position. How to further improve the prediction precision is necessary to study.This article analyses the advantages and disadvantages and applicability of variousmethods currently used in power load forecasting, prediction in grey model, artificialneural network. It also to study a comparative analysis of several different neuralnetwork prediction model in the application of combination forecasting model on thisbasic theories.The variable weight combination, based on the grey model as well as the artificialneural network model, is established using the least square method to solve the predictionmodel in the article. With the characteristic analysis based on actual data, the articleselects the appropriate variables as input. Combined forecast model using BP neuralnetwork, RBF neural network and Elman neural network those three kinds of neuralnetwork and grey GM (1,1) is set up on an designated area witch provides the data. Themodel used to calculate social total electricity demand forecasting of the arearespectively in two cases, short-term and long-term total electricity demand forecasting.Comparisons of the value and the actual load forecasting results obtained by the analysisof the relative error, the predictive value and the actual graphics to value and threemodels of the prediction results are used to investigate the application in quality of thethree kinds of neural network in power load forecasting method, also the applicability forthe area offering load data. In the MATLAB simulation, the actual load data are used asinput to obtain the forecast value curve. The results showed that short-term loadforecasting, using a combination of the results of Elman neural network model and grayprediction combined with the accuracy of the results of the two models is superior to theother two, more suitable for the text in the selected region; in the medium and long termload forecasting, using the accuracy of the results combined model RBF neural networkand gray prediction combined results of the two models is superior to the other twomethods, more suitable for the actual situation of the text in the selected region.