Study On The Model With Ability Of Explaning For Market Clearing Price Forecasting

To solve the problems of artificial neural network model which structure is noteasy to be understood or be explained and the results can't provide more extrainformation about the indeterminacy of the forecasting price. So a novel hybrid modelis presented in this paper, using the decision tree and the artificial neural networkmethods. First, the electricity price forecasting model is transformed into theelectricity price rising ratio forecasting model. Second, the electricity price rising ratioforecasting model is built for each period based on the decision tree approach, whichnot only provides the rising ratio of the electricity price but also gives the probabilityand the mainly influence factors of the ratio. Then, the artificial neural networkmodels are built for the clusters which have bad accuracy of price forecasting thatclassified by the decision tree forecasting models. Finally, real-word data ofQueensland spot market in Australia is employed to demonstrate the validity of theproposed approach.