融合多类人工智能模型的电力系统负荷短期预测技术研究  

Research on Power System Load Short-Term Forecasting Technology Integrating Multiple Artificial Intelligence Models

作　　者：梁凌宇 赵翔宇 黄文琦 袁红霞 曹尚 周锐烨 郭尧 LIANG Lingyu;ZHAO Xiangyu;HUANG Wenqi;YUAN Hongxia;CAO Shang;ZHOU Ruiye;GUO Yao(China Southern Power Grid Digital Grid Research Institute Co.,Ltd.,Guangzhou 510663,Guangdong,China)

机构地区：[1]南方电网数字电网研究院有限公司,广东广州510663

出　　处：《电力大数据》2022年第6期16-23,共8页Power Systems and Big Data

摘　　要：系统负荷短期预测作为电网调度及规划中必不可少的一环,在安排发电机组启停及检修计划、保持电网运行和降低发电成本等领域起到重要作用。针对系统负荷预测,近年来国内外已提出一系列智能预测方法,但其大多采用单一模型进行实现。而单一模型在同一预测对象前提下,只擅长挖掘数据的某一类特征,进而使得预测结果存在不同偏好。因此本文提出一种基于Stacking集成学习的多类型人工智能模型融合方法,即利用多个不同类型的基础模型构成弱分类器,使其在相同样本基础上单独训练、单独预测后,再利用另一个人工智能模型作为强分类器对所有弱分类器的预测结果进行拟合,最终输出一个唯一的系统负荷预测结果。最后,以某网5个省的真实负荷作为实验对象,并抽取一段时间的平均准确率进行对比。结果表明,本文提出的预测方法准确率要高于单一人工智能模型。As an indispensable part of power grid dispatching and planning,system load short-term forecasting plays an important role in arranging the start,shutdown and maintenance plans of generating units,maintaining the operation of the power grid,and reducing power generation costs.For system load forecasting,a series of intelligent forecasting methods have been proposed at home and abroad in recent years,but most of them are realized by a single model.On the premise of the same prediction object,a single model is only good at mining a certain type of characteristics of the data,which makes the prediction results have different preferences.Therefore,this paper proposes a multi-type artificial intelligence model fusion method based on stacking ensemble learning,which uses multiple different types of basic models to form a weak classifier,which is trained and predicted separately on the same sample basis,and then uses another one.The artificial intelligence model is used as a strong classifier to fit the prediction results of all weak classifiers,and finally output a unique system load prediction result.Finally,the real load of 5 provinces of a certain grid is used as the experimental object,and the average accuracy rate over a period of time is selected for comparison.The results show that the accuracy of the prediction method proposed in this paper is higher than that of a single artificial intelligence model.

关 键 词：系统负荷 短期预测 人工智能 集成学习 模型融合 

分 类 号：TM715[电气工程—电力系统及自动化]