Short-term Load Forcasting Of Power System Based On Chaotic And Wavelet Transform

With the development of power market, the power load forecasting plays a more important role than ever in power system. In practice, the planning,designing and the dispatching automation of power system require load forecasting greatly. The research of load forecasting attracts more people's attention and has been an important field of morden power system.Power system is a very strong nonlinear system, and appears as chaos behavior. This paper reviews and comments the theories and methods of the presently electrical loads prediction, and introduce the application and development of chaos theory. Combined with the processing method of chaotic times series, this dissertation presents higher forecasting precision method through wavelet singularity detection and wavelet eliminating noise. Robustness of the local linear regression method is poor because the illness-state matrix is sensitive to noise when the model is linearized. An adaptive-self prediction filter was proposed to solve the influence of illness-state matrix. The dissertation main focus on follows item:Historical load date is the basis of load forecasting and load feature analysis. False data and noises among load data will disturb load forecasting and load feature analysis. This paper illustrated the influence of different level noise on predict accuracy through the chaotic system of Logistic map. So the load date must be corrected and smoothed before using especially for chaotic method. Processed by the method through adjusting amplitude of their wavelet modulus maxima and processing the wavelet decomposed detail signal by soft Rigorous SURE threshold based on wavelet analysis and singularity theory, fault date be eliminated. The real historical information and regulation data can be gained for load forecasting. At the same time noises are removed. The validity of the method is proved by the application in the one-rank local-region method.Local linear regression method was widely used in chaotic power system short term load forecasting. But there are two shortages in the method. First, the prediction accuracy is sensitive to the embedding dimensions. The prediction error become larger and the algorithm become unstable if the embedding dimension is not correct. Second, Robustness of the method is poor because the illness-state matrix is sensitive to noise when the model is linearized. An adaptive-self prediction filter was proposed...