부산대학교 도서관

In this paper, a ensemble weather forecasting method based on multi-initial value, multi-mode and multi-physical process, with multiple algorithm perturbations, and BMA+EMOS statistical modeling is proposed. Based on the ensemble weather forecasting method, a new energy probabilistic power prediction method combining taboo algorithm and BP neural network algorithm is proposed. Through the design example analysis, the proposed method can effectively reduce the prediction bias, reduce the upper and lower limit bandwidth of probability prediction by 25%, and overcome the power distribution fat tail and multimodal anomalies, making the power prediction results more stable and accurate.