Basic Problems In Short-term Electric Load And Wind Forecasting

The research of short term load forecasting is more and more important, especially in electric power. Separating white noise from short time load series, and reducing forecasting errors are more significant for the safe running of the power system and the economic development.The white noise series make an upper limit of load forecasting accuracy, and is an essential data for finding better forecasting models. In the real complex systems, temperature, and the finite digital computer can produce additive errors, reduce forecasting accuracy. How to solve those problems is the study of this article. The main contents and results are:(1) According to the Cramér's decomposition theorem, the load time series can be decomposed signal and noise. When the load time series have the frequency signal and high frequency noise, the variance of white noise can be estimated by using the method of difference.(2) Based on the method of wavelet, the time load series can be easily decomposed a series of signal and noise, but it is hard to identify which is the best white noise. Use the theory of Statistics Method, we can find the best white noise. This article study more numberical tests(3) According to the Shannon sampling theorem, the time load series which were sampled from complex systems in the real world almost have white noise series within them. For the weather sensitive loads, the Shannon sampling theorem predicates that the traditional"day"sampling methods which were used to eliminate the fluctuant/non-stationary of load series can usually enlarge the white noises. Propose the method of low order nonlinear transformations to reduce the forecasting errors.(4) Employ the same prediction model to test the performance of power function transformation and logarithmic transformation (a low order nonlinear transformation). This article study more numerical tests.