To further improve the accuracy and reliability of the short-term electricity price forecasting results, this paper proposes a Stacking learning method to integrate different basic learner short-term electricity price forecasting models.First, the data is carried out by J-Fold and cross-validation for segmentation and training.Then, feature transformation is conducted for the original features to reconstruct the secondary features.The new features are used build to train the Meta learner for the final prediction of the sample data.The experimental results show that the Stacking integrated model has smaller error and good stability compared with a single regression model, which provides a new method for short-term electricity price forecasting.%为进一步提高短期电价预测结果的准确性和可靠性,本文提出了一种运用Stacking学习方式去集成不同基础学习器的短期电价预测模型.首先采用J-Fold和交叉验证的方式对数据进行分割和训练,将原始特征进行特征变换,重新构建二级特征;然后再使用构建的新特征去训练Meta学习器,用于样本数据的最终预测.实验结果表明,相比较于单一的回归模型,Stacking集成模型具有更小的误差和良好的稳定性,为短期电价预测提供了新方法.
展开▼