基于XGBoost算法的新型短期负荷预测模型研究

A new short-term load forecasting model based on XGBoost algorithm

针对目前电网在负荷预测中所采集到的数据普遍存在着特征维度较少、特征关系不明、有效数据量较少的特点,为了提高电网短期负荷预测精度,提出一种基于XGBoost算法的新型负荷预测模型。基于XGBoost算法的负荷预测模型采用CART树作为基学习器,输入预处理后的历史负荷和特征数据,通过构建多个弱学习器逐层训练并得到模型,最后向模型输入测试集特征得到最终的预测结果。所搭建的负荷预测模型具有避免对数据特征的标准化、处理字段缺失的数据、不用关心特征间是否相互依赖、学习效果好的优点。根据真实电网数据实验结果,基于XGBoost算法的负荷预测平均绝对误差百分比下降到3.46%,对比基于BP、GRNN、DBN神经网络的负荷模型预测值精度更高,表明所提模型的优越性。

In order to improve the short-term load forecasting accuracy of the power grid for the present environment where there are generally few feature dimensions,the feature relationship is unknown and the few effective data exists,a new load forecasting model based on XGBoost algorithm is proposed in this paper.The load prediction model based on XGBoost algorithm adopts CART tree as the basic learner,inputs the preprocessed historical load and characteristic data,then,several weak learner are built and trained layer by layer to get the model,and finally,the features of test set into model are input to get the final predicted results.The load forecasting model proposed in this paper has the advantages,such as avoiding the standardization of data features,processing data that missing fields automatically,not caring whether the features are interdependent or not,and a better learning effect.According to the experimental results of real power grid data,the MAPE of load prediction based on XGBoost algorithm drops to 3.46%,which is more accurate than the predicted value of load model based on BP,GRNN and DBN neural network,which shows the superiority of proposed model.