华电技术

• •    

基于机器学习的空冷火电机组负荷预测研究

彭维珂1,聂椿明1,徐钢2,陈衡2   

  1. 1. 华北电力大学能源动力与机械工程学院
    2. 华北电力大学
  • 收稿日期:2021-01-08 修回日期:2021-01-30 发布日期:2021-02-04
  • 通讯作者: 徐钢
  • 基金资助:
    国家重点研发计划项目;国家自然科学基金

Study on load forecasting of air cooling thermal power unit based on machine learning

Wei-Ke PENG1, 1,Xu gang2, 1   

  1. 1.
    2. NCEPU
  • Received:2021-01-08 Revised:2021-01-30 Published:2021-02-04
  • Contact: Xu gang

摘要: 为及时准确地预测空冷机组整机性能,本文引入了基于机器学习的大数据分析方法。针对某600 MW空冷火电机组全年的历史运行数据进行预处理与稳态工况筛选,分别建立了基于BP神经网络和随机森林算法的机组负荷预测模型。通过预测结果比较分析和模型敏感性分析,表明随机森林预测模型具有精度高、泛化能力强、训练时间短等优点。最后,利用特征参数筛选和分负荷工况建模来对随机森林模型进行性能优化,研究结果为电厂实时大数据分析提供一定的参考。

关键词: 燃煤电站, 负荷预测, 机器学习, 大数据, 直接空冷

Abstract: In order to predict the performance of the air-cooling power plant unit timely and accurately, this paper introduces the big data analysis method based on machine learning. In this paper, a load forecasting model based on BP Neural Network and Random Forest algorithm are established for a 600 MW Air-cooled thermal power unit. Through the comparative analysis of the prediction results and the sensitivity analysis of the models, it shows that the Random Forest prediction model has the advantages of high precision, strong generalization ability and high calculating speed. Finally, characteristic parameter selection and load condition classification are used to optimize the Random Forest model. The work of this paper may offer some reference and gist for research on real-time big data analysis of the power plant.

Key words: Coal fired power plants, Load forecasting, Machine learning, Big data, Direct air cooling