Prediction Intervals Forecasts of Wind Power based on PSO-KELM

Abstract

Wind power forecasting is essential to the management of power system. However, traditional point forecasting of wind power can hardly accurate due to the stochastic nature of wind. Providing prediction intervals(PIs) by giving a range of potential outcomes for a given probability was important to assess the potential uncertainties for such forecasting result. In this paper, to achieve better short-term wind power PIs, a probabilistic PI model was built based on particle swarm optimization kernel extreme learning machine(PSO-KELM). It was established through intelligent algorithm KELM so as to finding the best upper and lower limit of optimal PIs, which optimized the output weights of KELM by PSO algorithm. Finally realized directly forecasting prediction interval of wind power, and it made full use of the advantages of KELM, including fast learning speed and generalization ability. Case studies using real wind farm data from Gansu had been conducted. By comparison with based on particle swarm optimization extreme learning machine(PSO-ELM) probability prediction model, the results demonstrate that PSO-KELM has higher prediction accuracy and faster optimization with different confidence level. And it will help power systems to make right decision.