Application of BP neural network to power prediction of wind power generation unit in microgrid

With the intensifying world energy crisis and environmental concerns, new energy power generation is favored by all countries. Wind power generation, in particular, has become a focus of research for the academic community. However, the intermittence and stochasticity inherent in wind energy itself also challenge the security, stable operation and energy quality of the microgrid system. Accurate short-term wind power prediction can effectively resolve this problem. Conventional wind power prediction methods mostly involve the use of a single neural network tool, which frequently leads to problems like complex model structure and excess sensitivity of the prediction results to the sample data used. Therefore, this paper proposes an idea of using fuzzy clustering technique to select similar days before establishing the BP neural network. The simulation experiment in this paper is based on the August wind speed data of an area, selecting August 27 as the forecasting day. The relative error of the forecasting results is within 5%, with a couple of singular points though. The experimental results show that the method of combining fuzzy clustering with BP neural network provides better prediction accuracy, and can be applied to wind power prediction.