中文摘要：

为了合理利用风电，提高电网的稳定性、经济性，需要对风电的输出功率进行有效预测；然而，单一模型的预测结果精度不高。提出一种基于Kalman滤波相空间重构的Elman神经网络短期风速组合预测模型。该模型采用Kalman滤波算法对风速进行滤波处理，通过相空间重构来确定风速序列的延时时间和嵌入维数；并利用Elman神经网络建立了预测模型。仿真实验表明，该模型预测精度有了明显提高。

英文摘要：

To use wind power rationally and improve efficiency and stability, it is necessary to predict the output power of wind. As the single model prediction is inaccurate, a combination of different models is proposed. The wind speed data are processed with a Kalman filter. The phase space is reconstructed to set delay and dimensions of wind speed time series before constructing an Elman neural network. A prediction model and evaluation standard are established with training samples chosen. Tests show improvement in the accuracy of the model.