A Multi-model Combination Approach for Probabilistic Wind Power Forecasting

摘要

Short-term probabilistic wind power forecasting can provide criticalquantified uncertainty information of wind generation for power systemoperation and control. As the complicated characteristics of wind powerprediction error, it would be difficult to develop a universal forecastingmodel dominating over other alternative models. Therefore, a novel multi-modelcombination (MMC) approach for short-term probabilistic wind generationforecasting is proposed in this paper to exploit the advantages of differentforecasting models. The proposed approach can combine different forecastingmodels those provide different kinds of probability density functions toimprove the probabilistic forecast accuracy. Three probabilistic forecastingmodels based on the sparse Bayesian learning, kernel density estimation andbeta distribution fitting are used to form the combined model. The parametersof the MMC model are solved based on Bayesian framework. Numerical testsillustrate the effectiveness of the proposed MMC approach.