Short-Term Power Load Forecasting Based On Chaotic And Neural Network

Electric short-term load forecasting is a quite important work in the Power System. As Power System is a huge system having a dynamic behavior, and with the development of power market, the research of load forecasting has attracted more and more people's attention and has become an important field of modern power system.In the beginning of this article, we will see the introduction and presentation for the development of chaos and phase-space reconstruction theories, and the compare of several methods of computing the phase-space reconstruction parameters. It is found that the phase-space reconstruction leads chaotic theory into power load time series analysis, renews the formerly dynamics system in the phase-space, and helps to research chaotic attracts.Secondly, it will come with the development of neural network and its application in power system, the elements of BP arithmetic, and then, the way how to construct logical forecasting model based on the character of short-term load forecasting, especially, the choose of neural network.Then, due to the reference field has great effect on the precision of load forecasting, along with the summarized from previous studies, this article presents a "two step search" method to replace Euclid distance. Using this method on selecting neural network's input data, combining chaotic multi-step forecasting model based on neural network, then, we can find this model has better effect on forecasting through analyzing the load data of ShenZhen's power market by the Matlab simulation.Finally, based on three cubed multinomial models of electric load curve, a data smoothness formula is deduced and carried on the forecasting result in order to eliminate the error of the curve and make the result have better robustness. It has proved that the expressions could further improve the precision of load forecasting.