An Improved Method For Short-term Load Forecasting Based On FFT And Chaos

Short-term load forecasting is the basis of power systems,the accuracy of forecasting will have a direct impact on the economic safe operation of the entire power grid. The research of short-term load forecasting has been gradually concerned.The research showed that load has the characteristics of periods and chaos. So the method of synthesizing double periods and chaotic components,which is the method for load forecasting based on FFT and choas has become hotspots.The method for load forecasting based on FFT and choas mainly study on that FFT is used to separate the periodic part form the load ,and then the chaotic method in which the multi-step prediction method with adding-weight one-rank local- region model is widely used, is used to forcast the rest of the load. Because the part of periodic frequency is totally extracted, and the predicted load would be used, the algorithms will have the problem of the extraction over and accumulated error. Mainly production in this paper has been as follows:1) The load is the mix of periodic,chaotic and random, which are also mixed up in frequency spectrum,and the load can be transformed from time domain by FFT. The part of periodic frequency is automatically identified, and the part of periodic frequency is totally extracted, so this would cause excessive extraction. In this paper, the contribution of periodic , chaotic and random in frequency spectrum is considered, and the quantity of seperating periodic part is determined which are based on the spectrum of chaotic and random.2) In multi-step prediction method with adding-weight one-rank local- region model, the neighboring points in more steps used for the forecast can be predicted value, which also can cause accumulated error . In this paper, a factor adaptive modified the weight of the neighboring points is used to reduce the accumulated error.3)In this paper, it makes an analysis on the rest of the load which is random noise.Meanwhile ,the wavelet transform is used to separate the choas and noise,which would be studied individually. Otherwise, Testing results of test functions show that the improved methods have better precision. Using these improved methods in electric short-term load forecasting, the Testing results of Hainan electric load forecasting are much better.