摘要
在电网大数据结合环保监管的背景下,研究了以企业用电和纳税数据为特征的企业生产状态识别方法。实际输入特征通常存在异常和部分缺失的情况,采用回归分析解决数据异常和部分缺失的问题,提高了分析结果的鲁棒性。建立支持向量回归模型来识别污染企业生产状况,通过网格搜索选择多个支持向量回归组合模型来识别污染企业生产状况,增加了模型的泛化能力,提高了分类精度。最后,实际测试结果验证了所提出的基于组合支持向量回归的排污企业生产识别方法的精确性和适用性。
Under the background of big power data and environmental supervision,the identification of enterprise production status is studied,which is characterized by enterprise electricity consumption and tax payment.Regression analysis is adopted to solve the problem of data anomaly and partial missing,which improves the robustness of analysis results.Support vector regression model is established to identify the production status of pollution emission enterprises,and multiple support vector regression models are selected through grid search to identify the production status of pollution emission enterprises,which increases the generalization ability of the model and improved the classification accuracy.Finally,the simulations confirm the accuracy and applicability of the proposed method based on ensemble support vector regression(e-SVR).
作者
靳旦
唐伟
Jin Dan;Tang Wei(State Grid Sichuan Electric Power Research Institute, Chengdu 610041, Sichuan, China)
出处
《四川电力技术》
2020年第3期29-32,77,共5页
Sichuan Electric Power Technology
关键词
组合支持向量回归
网格搜索
生产识别
ensemble support vector regression
gird search
identification of production