主管:中国科学院
主办:中国优选法统筹法与经济数学研究会
   中国科学院科技战略咨询研究院

中国管理科学 ›› 2026, Vol. 34 ›› Issue (10): 386-396.doi: 10.16381/j.cnki.issn1003-207x.2023.1457

• • 上一篇    

基于特征选择的火电机组燃煤量预测模型比较研究

贺昊1, 周健2, 朱磊1()   

  1. 1.北京航空航天大学经济管理学院,北京 100191
    2.青岛理工大学管理工程学院,山东 青岛 266520
  • 收稿日期:2023-09-05 修回日期:2024-06-04 出版日期:2026-10-25 发布日期:2026-10-09
  • 通讯作者: 朱磊 E-mail:leizhu@buaa.edu.cn
  • 基金资助:
    国家自然科学基金项目(72342002);国家自然科学基金项目(72122002);国家自然科学基金项目(72403137)

Comparative Study on Coal Consumption Prediction Models of Thermal Power Units Based on Feature Selection

Hao He1, Jian Zhou2, Lei Zhu1()   

  1. 1.School of Economics and Management,Beihang University,Beijing 100191,China
    2.School of Management Engineering,Qingdao University of Technology,Qingdao 266520,China
  • Received:2023-09-05 Revised:2024-06-04 Online:2026-10-25 Published:2026-10-09
  • Contact: Lei Zhu E-mail:leizhu@buaa.edu.cn

摘要:

清洁能源的间歇性和不确定性增加了发电系统的运行复杂性,因此,需要利用具有调峰能力的火电机组为风电、光伏等可再生能源进行调峰,提高电网的整体稳定性,建立准确有效的火电机组调峰煤耗预测模型,对火电企业实现高效运营有重要意义。本文基于火电机组运营大数据,针对火电运行数据的高维和强耦合的特点,首先使用聚类算法划分工况,再使用特征选择方法处理数据集,降低机组数据维度和过拟合的风险,最后建立集成树模型实现对火电机组燃煤量的智能预测。通过对不同模型的预测效果进行分析,研究发现,不同的机器学习算法在不同的火电运营工况下的煤耗预测效果存在差异,因此,在进行煤耗管理时,需要基于工况差异,选择预测精度高的机器学习算法,以最大限度地发挥算法带来的火电机组煤耗节约效果。

关键词: 火电机组, 燃煤量, 特征选择, 集成学习, 预测模型

Abstract:

The intermittency and uncertainty of clean energy have increased the operational complexity of the power generation system. It is necessary to utilize thermal power units with peak load regulation capability to balance the fluctuation of renewable energy sources such as wind and solar power, thereby enhancing the overall stability of the grid. Establishing an accurate and effective coal consumption prediction model for thermal power units is of great significance for the efficient operation of thermal power enterprises. The method used in this paper is based on the operational big data of thermal power units, considering the high-dimensional and strong coupling characteristics of the data. Firstly, clustering algorithms are employed to divide operational conditions. Secondly, feature selection methods are applied to process the dataset, reducing data dimensionality and the risk of overfitting. Finally, ensemble tree models are established for the intelligent prediction of coal consumption in thermal power units. By analyzing the prediction effects of different models, it is found that the coal consumption prediction performance of various machine learning algorithms varies under different operational conditions of thermal power. Therefore, when managing coal consumption, it is necessary to select machine learning algorithms with high prediction accuracy based on the differences in operational conditions to maximize the coal-saving effects brought by the algorithms.

Key words: thermal power unit, coal consumption, feature selection, ensemble learning, prediction model

中图分类号: