Research On Short Term Load Forecasting Of Power System

Power system short-term load forecasting is a very important job,the prediction accuracy has a direct impact on the economy and stability of power system.Therefore, the research on short term load forecasting of power system has great significance.In this paper, in order to improve the accuracy of load forecasting, the pretreatment methods based on improved fuzzy C-means and the combination forecasting method are put forward.Historical data has a lot of bad data, and the traditional methods can not meet the requires. So the bad data identification correction model based on fuzzy C-means clustering is proposed, using the similarity of daily load curve.For a single load forecasting method can not meet the actual load forecasting accuracy, the combination forecasting method based on a variety of single prediction method is put forward:①Wavelet decomposition is used to separate different components of the load sequences.The low frequency sequences have long period, it is forecasted by using nonlinear extrapolation method of least squares fitting.The high frequency sequences have short period, it is forecasted by using a combination of intelligent prediction methods;②The high frequency sequences are forecasted by using three forecasting methods respectively, that is the method based on gaussian chaos particle swarm optimazition and feed-forward neural network, the method based on the improved chaos theory and Elman neural network, the method based on kernel principal component analysis and least squares support vector regression based on adaptive particle swarm optimization;③According to the error of "virtual predict", three single forecasting methods are combinated based on the idea of "sub-period variable weight".The programming of this paper is based on MATLAB7.1 platform.The actual load sequence(a Sichuan area electric power load data, 288 sampling time) are analyzed, the analysis results show that the proposed short term forecasting model is practical.