中国电力 ›› 2014, Vol. 47 ›› Issue (7): 21-25.DOI: 10.11930/j.issn.1004-9649.2014.7.21.4

• 发电 • 上一篇    下一篇

基于在线支持向量回归算法的电站热耗率模型

李辉   

  1. 华能国际电力股份有限公司,北京 100031
  • 收稿日期:2014-03-13 出版日期:2014-07-18 发布日期:2015-12-10
  • 作者简介:李辉(1983—),男,河南商丘人,工学硕士,工程师,从事电站生产管理、性能优化和污染物控制等技术研究。E-mail: lihui@hpi.com.cn

The Turbine Heat Rate Model Based on Accurate Online Support Vector Regression Algorithm

LI Hui   

  1. Huaneng Power International, Inc., Beijing 100031, China
  • Received:2014-03-13 Online:2014-07-18 Published:2015-12-10

摘要: 利用在线支持向量回归算法(AOSVR),建立了机组热耗率的在线回归模型。介绍了模型的更新过程,包括增加新样本的递增和冗余样本的删除。对某1 000 MW机组的热耗率计算进行了建模,并与常用的离线式模型SVR和LS-SVR进行了对比,结果表明AOSVR模型能够根据新样本对模型不断进行更新,具有较强的自适应能力和泛化性能,适合在线应用。进一步通过输入参数扰动分析得出AOSVR具有较强的鲁棒性,能够克服输入参数的非正常波动,保证热耗率计算的可靠性。

关键词: 汽轮发电机组, 热耗率, 在线支持向量机, 回归预测, 鲁棒性

Abstract: An online regression model of the steam turbine heat rate is established by using the accurate online support vector regression (AOSVR) algorithm. The model updating process, which includes how to add new samples and delete redundant samples, is also presented. Moreover, the AOSVR-based regression model of the heat rate for a 1 000-MW steam turbine is built and compared with the normally used offline models of SVR and LS-SVR. The comparison indicates that the AOSVR-based regression model can be continuously updated according to the new samples, which shows strong self-adaptive and generalizing capabilities and is suitable for online application. Then, further analysis by adding a disturbance to the input parameter shows that the AOSVR model has strong robustness and can overcome the abnormal fluctuation of the input parameters and guarantee the reliability of the heat rate computation.

Key words: steam-turbine generator unit, heat rate, online support vector, regression prediction, robustness

中图分类号: