引用本文:张金金,张倩,马愿,李智.基于改进的随机森林和密度聚类的短期负荷频域预测方法[J].控制理论与应用,2020,37(10):2257~2265.[点击复制]
ZHANG Jin-jin,zhangqian,mayuan,lizhi.Short-term load frequency domain method prediction method based on IRF and DBSCAN[J].Control Theory and Technology,2020,37(10):2257~2265.[点击复制]
基于改进的随机森林和密度聚类的短期负荷频域预测方法
Short-term load frequency domain method prediction method based on IRF and DBSCAN
摘要点击 1693  全文点击 515  投稿时间：2019-12-06  修订日期：2020-08-10
查看全文  查看/发表评论  下载PDF阅读器
DOI编号  10.7641/CTA.2020.90991
  2020,37(10):2257-2265
中文关键词  负荷预测  IRF  DBSCAN  EWT
英文关键词  load forecasting  IRF  DBSCAN  EWT
基金项目  国家重点研发计划项目(2016YFB0900400),国家自然科学基金(51507001), 分布式可再生能源发电集群并网消纳与用采、配自、调控系统数据融合(5212N0180061), 安徽省电力公司科技项目(5212001700), 安徽省博士后研究基金(Z01011804).
作者单位E-mail
张金金 安徽大学电气工程与自动化学院 1095610420@qq.com 
张倩* 安徽大学电气工程与自动化学院  
马愿 安徽大学电气工程与自动化学院  
李智 国网安徽省电力有限公司  
中文摘要
      精确的负荷预测对于电力系统的有效调度和安全运行至关重要. 本文提出基于改进的随机森林(IRF)和密度的聚类 (DBSCAN)的频域组合预测方法. 首先, 采用经验小波变换 (EWT)分解负荷, 得到不同的固有模态分量 (IMFs); 其次，根据各分量特征采用合理的方法进行预测。其中, 低频、中频分量采用IRF预测; 高频分量使用DBSCAN根据气象因素温度和湿度聚类, 再根据每类的样本特性选择处理方法. 最后, 叠加各分量的预测值, 获取负荷预测值. 根据某地市现场负荷数据进行实验，预测结果分别与EWT-IRF、EWT-随机森林 (RF)、经验模态分解(EMD)- IRF模型的预测结果进行对比. 结果表明, 提出的模型具有更高的预测精度, 反映了实际负荷的随机性.
英文摘要
      Accurate load forecasting is of great significance for effective scheduling and safe operation of power systems. A frequency domain combined prediction method based on Improved Random Forest (IRF) and Density-Based Spatial Clustering of Applications with Noise (DBSCAN) is proposed in this paper. Firstly, Empirical Wavelet Transform (EWT) is used to decompose load to obtain different Intrinsic Mode Functions (IMFs). Secondly, a reasonable method is used for prediction according to the characteristics of each component. Among them, the low frequency and intermediate frequency components are predicted by IRF. The high frequency component is clustered using DBSCAN according to the meteorological factors temperature and humidity，then we select the processing method according to the sample characteristics of each class. Finally, the prediction results of the respective components are superimposed to obtain the total prediction result. Experiments are carried out based on the measured load data of a certain city. The prediction results are compared with the prediction results of EWT-IRF, EWT- Random Forest (RF), and Empirical Mode Decomposition (EMD)-IRF model. The results indicate that the proposed model has higher prediction accuracy and reflects the randomness of the actual load.