考虑气象特征提取的短期风速预测方法

Short-term Wind Speed Prediction Considering Feature Selection

为提高短期风速预测精度,研究了考虑气象特征提取的短期风速预测方法。针对输入气象特征较多且难以提取,提出一种简易气象特征提取方法,通过极限学习机和改进珊瑚礁算法,从较多气象特征中提取最优气象特征。以最优气象特征为预测模型输入,能够有效增强模型泛化能力。对墨西哥某风电场风速预测结果表明,改进珊瑚礁算法结合极限学习机的方法能够有效提取气象特征,提高预测精度,具有一定的实用价值。

It is significant to predict short-term wind speed precisely for optimizing the wind farm. This paper presents a method for short-term wind speed prediction considering the feature selection. Feature selection is an important task because irrelevant features can increase the cost of the prediction model, and make the system performance poorer. In this paper, a simple method of feature selection is proposed based on extreme learning machine （ELM） and coral reefs optimization algorithm （CRO）. Features are chosen to be the input of an ELM, which achieves the lowest root mean square error （RMSE）. In order to obtain the global optimal solution of all situations, a modified CRO algorithm is proposed. Together, these algorithms are able to select the optimum features successfully in short-term wind speed prediction. The sample wind farm in Mexico demonstrates the MCRO-ELM model has better performance in short-term wind speed prediction.