Prediction interval estimations for electricity demands and prices: a multi-objective approach

Electricity price and demand forecasting are becoming essential practices for the deregulated market participants such as system operators, generation companies, industries and end use consumers. With a prominent growth in the uncertainty aspect of the energy sources, climatic changes and demand patterns, it is essential to supplement the traditional point forecasts with prediction intervals (PIs) which are an important tool for quantifying the uncertainty of forecasted entities. This study proposes a novel approach for generation of PIs using a differential evolution-based multi-objective approach. The traditional PI generation is framed as a multi-objective problem and a set of Pareto-optimal solutions are generated. The proposed technique is validated using electricity price and demand data from the Ontario electricity market. Experimental results indicate that the proposed technique can successfully generate high-quality PIs.