Research For Load Forecasting Method Of Electric Power System

In the modernization of the electric power system management and even in thedevelopment of the national economy, load forecasting of power has a significant importance,particularly accurate short-term load forecasting of power. It is an important part of the energymanagement system, it is the basis of the electricity prices for electric power dispatching, andit is a prerequisite for security, economic and reliable operation of power system. It isimportant for load forecasting of power system to research the load characteristic, analyze theload theory, optimize the load forecasting algorithm, improve the forecasting accuracy, andestablish the efficient and practical load forecasting model.This paper mainly studies electric power system short-term load forecasting method andits forecasting model. Adaptive AR model based on electric power load measurement series,chaos forecasting model based on electric power load measurement series and the optimalcombination forecasting model are respectively established based on the analysis of load timeseries.For the adaptive AR model based on electric power load measurement series, threedifferent recursive algorithms of model estimation are respectively established based on theleast squares method, the Wiener filter theory and the Kalman filter theory. The optimalforecasting model is obtained by improving the basic model and analysis and comparison ofthe effect of each model.For building forecasting model based on the phase space reconstruction of chaotic timeseries, firstly the methods of calculating parameters such as delay time, embedding dimension,maximum Lyapunov exponent etc are analyzed and advanced. Secondly improved euclideandistance formula is applied to build forecasting model based on maximum Lyapunovexponent, which makes predictable accuracy much higher. Chaos Kalman filter algorithm forforecasting is proposed, which applies Kalman filter algorithm to chaos time series. State andspace model of chaotic system is directly built by evolution relationship of phase and point.To load series of power system which has the characteristic of chaos, recursive prediction ofphase and point evolution is used to predict timely in advantage of real-time correctiontechnology of Kalman filter algorithm.For the combination forecasting model, in order to overcome the limitations of eachsingle forecasting model and comprehensively utilizted prediction results of each model, thecombination forecasting method is proposed by weighting for the least squares method, Kalman filter algorithm, and chaos Kalman filter algorithm. The weight coefficient isdetermined by variance-covariance method (MV) and the optimal weighting method. A moreappropriate model is derived by analyzing the advantages and disadvantages of the twomethods. In order to avoid mutations of forecasting result based on a single method, animproved combination forecasting method is proposed to correct the model and advance theprediction accuracy.