基于特征选择和改进深度森林的短期风电功率预测

Short-term wind power prediction based on feature selection and improved deep forest

针对风电数据特征维数高、数据冗余性大和有效特征挖掘不充分,最终导致预测精度偏低的问题,提出了一种基于特征选择和改进深度森林的短期风电功率预测方法.首先使用Kendall Rank相关系数、灰色关联度和随机森林特征重要性三种方法进行特征选择,选择有效特征,综合确定最佳输入特征集.然后,在深度森林的基础上,引入优良预测性能的极端随机树构建级联层,提出改进深度森林模型以提高模型泛化能力和预测性能.最后,建立改进深度森林模型进行风电功率短期预测.以新疆某风电场实测数据进行了算例仿真,验证了该方法的有效性.结果表明其与对比方法相比,具有更好的准确性和拟合效果.

Aiming at the problem of low prediction accuracy caused by high feature dimension,large data redundancy and insufficient effective feature mining of wind power data,a short-term wind power prediction method based on feature selection and improved deep forest is proposed.Firstly,Kendall rank correlation coefficient,grey relation analysis and random forest feature importance are used to select features,and effective features are selected to determine the best input feature set.Then,on the basis of deep forest,the extreme random tree with excellent prediction performance is introduced to construct cascade layer,and the improved deep forest model is proposed to improve the generalization ability and prediction performance of the model.Finally,an improved deep forest model is established for short-term prediction of wind power.The simulation results of a wind farm in Xinjiang verify the effectiveness of the method.The results show that the method has better accuracy and fitting effect than the comparison method.