Point and interval forecasting of real-time and day-ahead electricity prices by a novel hybrid approach

Accurate forecasting of electricity market prices presents important information to market participants. This provides forward planning of their bidding strategies in order to maximise revenue, profit, and utility perspectives. Nevertheless, due to the non-stationarities involved in market clearing price, an accurate forecasting of these prices is very complex. In this case, transformation from traditional point forecasts to probabilistic interval ones is of great importance to quantify the uncertainties of potential forecasts. In this study, interval forecasting of market clearing prices is conducted based on a novel approach within two consecutive steps. In the first step, a new hybrid method is proposed to estimate point forecasts: combination of wavelet transformation (Wt), feature selection based on Mutual Information (MI), extreme learning machine (ELM), and bootstrap approaches in an ensemble structure is employed. The second step consists of the following stepwise parts: calculating the variance of the model uncertainties based on the extracted data from the ensemble structure, estimating the noise variance by using the maximum-likelihood estimation (MLE), and improving the accuracy of interval forecasting by using particle swarm optimisation (PSO) algorithm. The effectiveness of the proposed approach termed as Wt–mutual information–ELM–MLE–PSO is validated through electricity market real data of Australian electricity network from real-time and day-ahead market viewpoints.