基于信息融合和堆叠模型的超短期风电功率预测

针对超短期风电功率预测,准确捕捉功率变化因素和建立混合预测模型是提高预测精度的有效手段之一。为了能够继承和整合单个模型的优点以及增强历史信息的表示和利用能力,文章提出了一种基于信息融合和堆叠模型的超短期风电功率预测模型。首先,利用相关性方法选择历史功率序列和历史测风塔数据的特征,作为预测模型的输入;然后,建立两层堆叠的集成模型作为预测模型,并使用交叉验证和超参数优化以增强预测模型的泛化性能;最后,以每个基学习器的输出作为元学习器获得最终预测值的新输入。通过东北某风电场真实数据的验证,以及与单一模型、深度神经网络模型和集成学习模型的对比,验证了所提模型的可行性和有效性。

Ultra-short-term wind power forecasting based on information fusion and stacking model

For ultra-short-term wind power forecasting,accurately capturing the factors that affect power changes,and establishing a hybrid forecasting model is one of efficient means to improve forecasting accuracy.In order to inherit and integrate the advantages of single model and enhance the ability to express and utilize historical information,an ultra-short-term wind power prediction method based on information fusion and stacking model is proposed.Firstly,the features of historical power series and historical wind tower data are selected by using correlation method,which are used as the input of the forecasting model.On this basis,a two-layer stacked ensemble model is established as the prediction model.In addition,cross-validation and hyper-parameter optimization are used to enhance the generalization performance of the prediction model.Finally,the output of each base learner is used as a new input for the meta-learner to obtain the final predicted value.This method is verified by real data of a wind farm in the Northeast China,and compared with single model,depth neural network model and popular ensemble learning model,the validity of the proposed model is verified.