基于多智能集成学习的中短期电煤价格预测

Medium and short-term electricity coal price prediction based on multi-intelligence ensemble learning

中文关键词  集成学习  电煤价格  滚动预测  灵敏性差异  XGBoost  长短神经网络

英文关键词  ensemble learning  electricity coal price  rolling prediction  sensitivity difference  XGBoost  long-short term memory

中文摘要

      为扩大电力市场交易量与下调市场电价, 需要提升电煤价格预测的可靠性与准确性. 为此本文提出了多智 能集成学习的中短期电煤价格预测方法. 首先, 阐述了Stacking集成学习的结构和原理; 然后, 介绍了数种智能电煤 价格的预测模型, 并通过算例证明了不同单智能模型对数据的感知能力存在差异性; 进而, 通过比较单智能模型预 测结果的差异值均差, 筛选出预测性能优异并且数据感知角度差异性明显的智能模型组. 为了充分发挥个模型感 知能力差异性的优势, 利用Stacking融合各模型, 得到一种适用于电煤价格滚动预测的集成模型. 最后, 通过滚动预 测2019至2020年的电煤价格, 对集成模型的有效性进行验证.

英文摘要

      In order to expand the trading volume of the electricity market and lower the electricity price, it is necessary to improve the reliability and accuracy of electricity coal price prediction. Therefore, a multi-intelligence ensemble learning method for medium and short-term electricity coal price prediction is proposed in this paper. First, the structure and principle of Stacking ensemble learning model are explained. Then, several intelligent coal price prediction models are introduced, and the difference of data perception ability of different single intelligent models is proved through calculation examples. Furthermore, by comparing the average difference of the prediction results of a single intelligent model, the intelligent model group with excellent prediction performance and obvious difference in data perception is selected. To give full play to the advantages of the differences in the perception ability of each model, the Stacking is used to fuse the models to obtain an ensemble model suitable for rolling prediction of electricity coal prices. Finally, the effectiveness of the ensemble model is verified by rolling prediction of the electricity coal price from 2019 to 2020.