Electricity Price Forecasting And Study Based On ARMA-GARCH Model

The electricity price is very important in the electricity market, it can not only reflect the supply-demand relationship, but also can adjust and control the transaction of the electric power market. So as a core part of the efficiency of the electricity competition, the electricity price is the most important part for all the market participants. With the reform of the electricity market sweeping across the world,the market participants pay great attention to the electricity price forecasting. Because the accurate electricity price forecasting has a very important significant to the market participants, they can use the prediction electricity price as a reference when making decision in the electricity market competition, so as to be in a favorable position in the transaction of the electricity market. Therefore, how to predict the future electricity price according to the historical price data and the characteristics of the electric power market, has become a hot research direction at home and abroad, so an exact price forecast is becoming more and more important.The electricity price has many different characteristics with other commodities, as it is impacted by many factors, the electricity price has the characteristics such as mean reversion, strong volatility, high jump, price spikes and leverage effect, these characteristics increase the difficulty of electricity price forecasting. At present, several methods of price forecasting have been proposed, mainly including time sequence method, artificial neural network method, wavelet analysis method and combination forecasting model etc. This paper mainly analyzed and compared the prediction model based on time sequence method.According to the changing characteristics of the electric price, this paper used GARCH model, TGARCH model and EGARCH model to establish the electric price’s prediction model of PJM power market, MISO electric power market, and the new England electric market respectively. The model assumed that the residuals obeying normal distribution, t distribution and generalized error distribution, thus compared the prediction accuracy of different models of different electric power market. By comparison we can see that, the ARMA-GARCH model had different prediction effects due to different characteristics of the electric price data. It was hard to say that which GARCH model had better prediction effect, GARCH model, TGARCH model, EGARCH model and PARCH model were suitable for prediction of the electricity in different electric power market according to the different characteristics of electric power market and the different distribution of the residual error.