With A Confidence Interval Of The Short-term Price Forecasting Methods

As the kernel index in electricity market, electricity price forecasting and information attaining are very important for all the market participants and researchers. While price series of the electricity market present some particular characteristics, such as mean reversion, heteroscedasticity, multicycle, jump and peak, and forecasting precision is often imprecise and is difficult to improve, all of which derive from the formation of price. In the meanwhile, the single point prediction can't effectively instruct the market participants' trading strategies or provide the electricity regulatory institution with the basis for supervising and controlling, whose application is doubted. Therefore, how to forecast electricity price effectively and getting useful information as much as possible is one of the most important thing in power market.Having fully known the research background, the formation and the characteristic of the price series, I choose ANFIS and WARIMA with a confidence interval as tools. After introducing both models in details, I experiment the historical data of both the crisis and stable periods from California in the case study. Through result analyzing ,it is shown that the latter is much better than the former in the forecasting precision and usefulness. The result with a confidence interval is obtained by using the muti-resolution analysis of wavelet and ARIMA models. It is proved that by analysing the results, more precise forecasting price is gotten and the historical data is included in the confidence intervals in the stable time. During the crisis periods, WARIMA model can provide a confidence interval by a certain confidence level, which makes electricity forecasting meanful, and research valuable. We emphasize the value of the confidence intervals and the successful application of the wavelet as a tool before modeling.In the last part of the paper, we discuss the electricity price forecasting and its prospect in order to provide some useful information for the future research.