基于K均值模式划分改进模糊聚类与BP神经网络的风力发电预测研究

Wind Power Prediction Based on K-means Pattern Classification Improved Fuzzy Clustering and BP Neural Network

为了克服模糊聚类不能反应数据内部差别的不足,提出了一种基于模式划分改进的模糊聚类与BP神经网络的风电功率预测算法。该算法首先确定最佳的模式划分数,接着将不同的风速模式下的历史风速数据进行模糊聚类以确定关联系数,并对关联系数高的历史风速数据和发电数据进行训练,然后利用BP神经网络得出风电预测功率。以新疆某地区实际风力发电数据作为仿真算例,对比分析了所提算法与未改进模糊聚类与BP神经网络在风力发电预测中的误差,结果表明所提算法克服了模糊聚类的缺点,具有更高的精度,对地区发电计划安排具有较高的价值。

In order to overcome the disadvantage of fuzzy clustering in reflecting the inner difference of data,a wind power prediction algorithm is put forward based on K-meas pattern classification improved fuzzy clustering and BP neural network . The pattern of wind speed is defined by K-means method firstly, the correlation coefficient of the history wind speed data is defined by fuzzy clustering method under different wind speed pattern ,the high correlation coefficient historical wind speed and wind power electricity data would be trained,and obtaining the predicting wind electricity power in BP neural network. Taking actual wind power generation data in a certain region of Xinjiang as simulation example, the errors of the proposed algorithm and the unimproved fuzzy clustering and BP neural network in wind power generation prediction are compared and analyzed. The results show that the proposed algorithm can overcome the disadvantage of fuzzy clustering and have higher accuracy,which can provide reference for the regional power generating plan.