基于RCNN-SVR组合模型短期风电功率预测

Short-term Wind Power Prediction Based on RCNN-SVR Combined Model

提出一种回归卷积神经网络与支持向量回归组合模型(RCNN-SVR),采用该模型预测短期风力发电功率。首先搭建了一种回归卷积神经网络(RCNN)模型;由于RCNN存在计算量大的问题,因此利用RCNN从数据集中提取特征因素,并用特征因素训练支持向量回归(SVR)对风电输出功率进行预测;最后采用某风电场数据集进行验证,结果表明RCNN-SVR模型比单独使用的传统RCNN模型或支持向量机具有更高的准确率。其中,RCNN-SVR模型的CV-RMSE、MAE和MAPE分别为0.0998、0.3928和0.5468,说明RCNN-SVR模型有效地提高了预测精度和输出结果的稳定性。

A combined regression convolutional neural network and support vector regression model(RCNN-SVR)is proposed,and the model is used to predict short-term wind power generation.Firstly,a regression convolution neural network(RCNN)model is built.Then,RCNN is used to extract the characteristic factors from the data set,and the characteristic factors are used to train support vector regression(SVR)to predict the wind power output.Finally,a wind farm data set is used for verification.The results show that RCNN-SVR model has higher accuracy than the traditional RCNN model or support vector machine used alone.Among them,the CV-RMSE,MAE and MAPE of RCNN-SVR model are 0.0998,0.3928 and 0.5468 respectively,indicating that RCNN-SVR model effectively improves the prediction accuracy and the stability of output results.