Research On Intelligent Forecasting Of Short Term Power Load Considering Meteorological Factors

With the continuous and rapid development of the smart grid,supplying users with safe,reliable and cost-effective electric power has become the necessity of economic growth and social progress,as well as the basic goal of the power industry.A short-term load forecasting for electric power system lays a solid foundation for achieving this goal.A scientific load forecasting plays a significant role in improving the energy efficiency and reliability of electric power.Meanwhile,with the improvement of people's living conditions,the extensive use of heating and cooling equipment makes the load considering meteorological factors take a larger proportion of the total electric load.Meteorological factors have more remarkable effect on the electric load.Therefore,researches on the relationship between meteorological factors and the electric load as well as the precision improvement of short-term load forecasting have become hot and difficult,drawing continuous attention of scholars and insiders at present.In this thesis,the electric load in a certain area of Hubei Province is taken as the research object.The study on the correlation between load and meteorological factors is made based on the characteristic analysis of the electric load in the thesis.Meanwhile,two optimization models of the neural network are built and applied to short-term load forecasting in this area.Firstly,this thesis makes a classification and analysis of electric load characteristics in this certain area based on historical load data,exploring the law of the periodic change of load.This thesis also discusses the factors which affect the electrical load preliminary,and modifies the historical abnormal data.Then,further research and quantitative analysis are made on the correlation between load and meteorological factors.The modified model on temperature of meteorological forecast is established.Besides,similar historical day is chosen based on the meteorological factors.All these efforts lay a good foundation for the building of subsequent forecasting models.Next,for the problems of non-stationarity and uncertainty of electric load data,empirical mode decomposition algorithm is employed for the stabilized and hierarchical decomposition of the data and different high-frequency,intermediate-frequency and low-frequency components are obtained.Meanwhile,the BP neural network models considering daily characteristics of meteorological factors are built respectively based on different components.Relevant case studies show that the models are helpful to improve the forecasting precision.Lastly,in order to improve the forecasting efficiency,the improved forecasting model of BP neural network is built based on the bacterial foraging optimization algorithm in this thesis.Compared with the predicted results which don't consider the meteorological factors,the predicted results are more precise by using this model when the meteorological factors are taken into consideration.Aimed at the problem that predicted results in some days are not of high precision in August,the establishment of BFO-BP model considering the real-time meteorological factors makes a further contribution to the forecasting precision.