Research And Application On Daily Run-off & Load Power Forecasting By Chaos Theory &combination Model

Recently, Chaos forecasting, which becomes a new research area, starts to make far-reaching influence to forecasting theory science. This thesis tries to apply Chaos forecasting technology to predict runoff and load power on the hydraulic and electric system, through reconstructing phase space to get the forecasting value. And adding the combination method to avoid the drawback that only forecasting method is used to daily runoff, it makes deeply a research on this.Firstly, this thesis reviews the theory and methods of forecasting to the runoff and load power, and introduces the development of the chaos theory and its application on the hydropower and electrical system. Chaos exists not only in nature, but also in the hydropower and electric system.Secondly, the cause of the variety of the parameters, the initial value of the system, the function overlap, and etc, is expounded, which can probably leads to Chaos. Although the chaos is similar with stochastic process, it is a certainly process after all, which has fractional dimension, maximal Lyapunov exponent, etc. This paper presents the phase space reconstruction theory and algorithm, by reconstructing the chaos attractor, can get the maximal Lyapunov exponent, then predicts the chaotic series. In addition, this thesis briefly introduces the identification method of chaos.Then, the dissertation applies chaos theory to analyze the process of runoff and the time series of load power. The chaos prediction method based on the largest Lyapunov exponent from small data sets is presented mainly, and many simulative experiment is made on it. Then, the paper analyzes the chaotic time series of load power and runoff of the Qingjiang valley, and builds the forecasting model to get the predict value.From the example, it shows that the small data sets algorithm calculate very quickly and is very useful.In addition, whereas the complexity of the hydraulic system, a optimal model to predict daily runoff is established in this paper, based on the theory of optimum weighted composition modeling, which integrates two runoff forecasting models, and its weighting coefficients are derived according to least square principle. Simulation results show that the model can reduce the prediction error and yield higher forecast precision compared to using only one method, and it therefore provides a valuable means for forecasting runoff.