Power Load Forecasting, Mathematical Methods And Applied Research

Load forecast is one of the important tasks of the electric system, theaccurate ones are the assurances which can elevate the benefits of theelectric system economy and the society. This paper focuses on theresearch of the electric load forecast model, and advances a newcombination forecast model which is based on the Maximal informationcompression index after studying the existent forecast method deeply.This paper is divided into four chapters, the first part generalizes thesignificances of the research on Change forecast in the electric system,the characteristics and the research actualities of the charge forecastelectric system.The second chapter researches on econometric model,stepwiseregression model and fuzzy clustering algorithm model deeply.The third chapter presents a new explanation of prediction from theview point of information theory, that is, prediction is to extractinformation which can foretell the future from the existent information ofthe forecast object, and Maximal information compression index isintroduced to measure the loss of information produced by singleprediction model for the first time. After fuzzy evaluation of the loss ofinformation, each prediction model is assigned a corresponding weightcoefficient on the principle that the less the loss of information, the bigger the weight coefficient. In this way, we establish a new combinationforecast model, which is based on the Maximal information compressionindex, and apply it to predict the electric consume in the east China areafor a long-term, we also analyses the forecast errors comparatively. Theresults of the experiment indicate that the combination forecast model,which is based on the Maximal information compression index, performsbetter than any other forecast model, and manifests preferable forecastperformance.The fourth chapter is the conclusion which summarizes the work andthe production of the paper and puts forward the further researchdirection.