基于变分模态分解和集成学习的光伏发电预测

Photovoltaic Power Generation Forecasting Based on Variational Modal Decomposition and Ensemble Learning

针对光伏发电量数据的非平稳性造成的发电量预测性能问题,提出一种基于改进变分模态分解和集成学习的光伏发电量预测方法。采用改进变分模态分解方法分解光伏发电量数据获得发电量分量,通过集成学习方法构建发电量分量预测模型;将发电量分量预测值进行组合,获得最终发电量预测结果。实验结果表明,所提方法在公开数据集上对光伏发电量进行预测的均方误差、平均绝对误差、决定系数值分别为0.2232,0.3387,0.9797,与其他方法相比具有更高的预测准确率和更小的误差。

Targeting the problem of power generation prediction performance caused by the non-stationarity of photovoltaic power generation data,this paper proposes a photovoltaic power generation prediction method based on improved variational mode decomposition and ensemble learning.The photovoltaic power generation data is decomposed to obtain the power generation components by the improved variational mode decomposition,and the power generation component prediction model is established with the ensemble learning.Furthermore,the predicted values of the power generation components are combined to obtain the final power generation prediction results.The experimental results show that the mean square error,average absolute error,and determinant coefficient values of the proposed method for the photovoltaic power generation prediction on public datasets are 0.2232,0.3387,and 0.9797,respectively,indicating that the method has higher prediction accuracy and smaller errors than other methods.