The Application Of Chaos Algorithm In Short-term Load Forecasting

When the power system is controlled on the line, we should use the short-termload forecasting to realize reasonable dispatch of sending and supplying power. Theshort-term load forecasting is the important basis of security dispatching andeconomical operation in the power system, so the precision of the load forecastingdirectly affects not only reliability and the economical efficiency of the power systemoperation, but the quality of supplying power.The artificial neural network has developed a lot for the sake of researchinghumanity's cognition process. Its central question is the cognition and the simulation.Because the humanity only understood a limited part of the real nervous system, theconsummation and development of artificial neural network depends on the researchof neurophysiology and neuroanatomy which presents some more detailedinformation and evidence. The chaos neural network is a new science whichdeveloped recently. The people had discovered the chaos phenomenon in the brain,and the chaos theory can be used to understand certain irregular activities in the brain.Then, the chaotic dynamics supplies a new turning point for people's studying ofneural network, so the research of the chaos neural network has become a new projectwhich people front. The paper do further research to chaos short-term load forecastingIn the base of analyses on common time series loading forecasting, it entirelyanalyses the elementary theory of chaos and phase space reconstruction of time series.And it calculates and analyses largest Lyapunov exponent, delay time of 24 hours incertain real network, so chaos characteristic is discovered. Consequently, it make afurther foundation to the below research.According to the basic principle of neural network, the paper establishes the BPnetwork model, and educes the BP algorithm which adopts S shape function with thesteepest descent method. Using BP network function in Matlab nerve network toolbox,it realizes load forecasting, enhances study speed and increases the reliability andaccuracy of algorithm.Applying the BP neural network, it proposes short-term load forecasting of chaostime series based on the BP network. According to the denoise pretreatment ofhistorical load series and estimation of Lyapunov exponent, the paper confirm trainingsample of network and network parameter. Through doing short-term load forecastingin real network, the result shows that the method could be used in our forecasting.The method still haves a lot of deficiency. For instance, how to achieve theconvergence of network quickly; how to consider the relative factor that affects loadforecasting thoroughly; how to modify the load which is affected by other greatincident(e.g, accident). Further research should be done in the future.