【摘要】 考虑气象数据与风电场的历史风功率数据特征，提出一种基于邻近传播聚类算法的长短期记忆神经网络风功率预测模型。利用邻近传播聚类算法(Affinity Propagation, AP)对NWP数据与风功率数据聚类，分别得到多个子集；同时，为了获得能更好反映风功率的特征，采用主成分分析法(Principal Components Analysis, PCA)对NWP子集数据降维处理；最后利用特征匹配方法分别建立基于长短期记忆神经网络预测模型。根据预测日的NWP数据与前一日功率数据选取最优匹配模型进行风功率预测。使用宁夏某风电厂数据进行仿真验证，实验表明所提出方法可以提高短期风功率的预测精度。更多还原

【Abstract】 A Long Short Term Memory(LSTM) neural network wind power prediction model based on the Affinity Propagation(AP) Algorithm is proposed considering meteorological data and historical wind power data of wind farms. The NWP data and wind power data are clustered using the AP Algorithm to obtain several subsets of wind power data, at the same time, the Principal Components Analysis(PCA) is used to reduce the dimensionality of the NWP subsets in order to obtain features that better reflect the wind power. Finally, the feature matching method is used to build a LSTM based neural network prediction model. The best matching model is selected for wind power prediction based on the NWP data of the prediction day and the power data of the previous day. Simulations were carried out using data from a wind power plant in Ningxia, and the experiments shows that the proposed method could improve the prediction accuracy of short-term wind power.更多还原