A power price forecasting method based on nonparametric GARCH model

Based on nonparametric conditional heteroscedasticity estimation theory, an improved power price forecasting method is proposed in this paper. In this method, according to the practical power price curve, conditional variance function is built and the model is determined by nonparametric estimation method. Besides, in the nonparametric estimation process, the iterative estimation algorithm is introduced to deal with the problem where conditional standard deviation is unpredictable. And the estimation credence of conditional variance function is improved by continually modifying the estimation value of conditional standard deviation as input variance. On the research of day-ahead time series fluctuation characteristics of power price in California in 2000, the time series of Humb node is modeled and forecasted. And the test results show that the model proposed in this paper can better reflect time series volatility clustering characteristics of power price, and it can improve forecasting effect of peak power price by using nonparametric estimation theory.