Short-term Load Forecasting Based On Fuzzy Combination Weighting Reconstruction Decomposition Theory

Electric power industry is the national important basic industry in the field of energy and the lifeline of the national economy.After the founding of new China,especially in recent years,electricity consumption has been growing rapidly.According to statistics,by the end of 2012,China's annual electricity consumption had reached 4.9571 trillion kilowatt hours;The new power generation equipment capacity had surpassed 87 million kilowatts;At the same time,new energy power generation technology has been used;new electrical equipment,for example,electric vehicle,has been joined.In the process of expanding of power grid construction,electric power system load is more complex,and the accurate load forecasting is also becoming more difficult.In spite of the difficulty of accurate load forecasting,it impacts significantly on making power plan,reducing the power storage capacity,avoiding major accidents,protecting the safety of production and daily life,promoting economic and social benefits,etc..Therefore,it is mentioned that more important to achieve accurate short-term load forecasting.In conclusion,based on the short-term load forecasting for the subject,this paper mainly has done the following research work and innovation:1.Due to the load sequence is a kind of random and fluctuation time series,empirical model decomposition(EMD)is used to decompose it stably into different scales.There are also more obvious characteristics on different scales of sequence,which determine different input vector and internal parameters in every predictive model,and improve the final prediction accuracy.2.Bat algorithm(BA)is used to optimize the parameters of Support vector machine(SVM)and construct BA-SVM model.In order to improve the prediction accuracy of SVM,it is necessary to solve the problem of the choice of internal parameters of SVM.Therefore,this paper puts forward using BA to optimize the main parameters of the SVM,and build the BA-SVM model.At the same time,it introduces a kalman filter(KF)to get further prediction accuracy of SVM.Thus,through KF state space model and operation principle of minimum mean square error estimation,the SVM prediction is used as observed value of KF to correct the SVM prediction and KF-BA-SVM combination forecast model is established.3.In view of the existing disadvantages in similar day selection of short-term load forecasting,fuzzy combination weight similar day selection scheme is put forward and use EMD and KF-BA-SVM,the fuzzy combination weighting reconstruction decomposition theory is established for short-term load forecasting model.Firstly,it analyzes the load characteristics and the main factors influencing the load changes in test region.According to the result of the analysis,main factors are fuzzily quantified;secondly,according to the different factors,objective weight value of factors influencing the load change and subjective weight value of load variation characteristics are determined by the entropy weight method and the Euclid distance with weights of k-means algorithm;finally,according to the set of intersection,the similar days are obtained and short-term load foresting model of fuzzy combination weighting reconstruction decomposition theory is built.4.The short-term load forecasting software is preliminary designed by VB and MATLAB.