综合智慧能源 ›› 2023, Vol. 45 ›› Issue (12): 36-42.doi: 10.3969/j.issn.2097-0706.2023.12.005

• 智慧与清洁供热 • 上一篇    下一篇

基于XGBoost的供热系统热力站调控预测方法

雍和忠1(), 章宁2, 黄伟3, 魏峥1, 吴燕玲3,4,*(), 钟崴3   

  1. 1.盐城热电有限责任公司,江苏 盐城 224002
    2.浙江大学 工程师学院,杭州 310015
    3.浙江大学 能源工程学院,杭州 310027
    4.浙江大学 常州工业技术研究院,江苏 常州 213022
  • 收稿日期:2023-06-19 修回日期:2023-08-21 出版日期:2023-12-25 发布日期:2023-08-25
  • 通讯作者: *吴燕玲(1982),女,副研究员,博士,从事锅炉机组建模与性能分析、复杂热力系统建模与优化分析、能源动力系统稳态等方面的研究,shelleywu@zju.edu.cn
  • 作者简介:雍和忠(1968),男,工程师,从事供热服务相关工作,hzyong2001@sina.cn
  • 基金资助:
    国家重点研发计划项目(2019YFE0126000)

XGBoost-based regulation and prediction method for the heating station's heating system

YONG Hezhong1(), ZHANG Ning2, HUANG Wei3, WEI Zheng1, WU Yanling3,4,*(), ZHONG Wei3   

  1. 1. Yancheng Thermal Power Company Limited,Yancheng,224002,China
    2. Polytechnic Institute, Zhejiang University,Hangzhou 310015,China
    3. College of Energy Engineering, Zhejiang University,Hangzhou 310027,China
    4. Changzhou Industrial Technology Research Institute, Zhejiang University,Changzhou 213022,China
  • Received:2023-06-19 Revised:2023-08-21 Online:2023-12-25 Published:2023-08-25
  • Supported by:
    National Key R&D Program of China(2019YFE0126000)

摘要:

热力站精准预测调控为降低供热系统供热能耗、提升用户热舒适性和降低污染物排放提供关键数据支撑。为了充分利用热力站历史运行状态、阀门未来设定开度、天气特征等关键数据对热力站二次供、回水温度进行精准调控,提出了一种基于极限梯度提升(XGBoost)的供热系统热力站调控预测方法。对上述数据进行预处理、特征构造、模型训练和预测生成热力站温度响应预测模型,结合二次网温度延迟时间进而确定阀门调控策略。提出的方法在24 h的预测温度和实际温度对比中,最大预测误差为0.47 ℃,验证了热力站温度响应模型的准确度和可靠性,可用于指导热力站的温度调控。

关键词: 热力站, XGBoost, 模型训练, 阀门调控, 预测精度, 智慧供热

Abstract:

Accurate prediction and control of heating stations provide solid data support for the heating systems' energy consumption reduction, users' thermal comfort improvement and pollutant emission alleviation. Making full use of a heating station's key data, such as historical operating states,the set openings of valves and the weather characteristics, a heating station control and prediction method based on Extreme Gradient Boost(XGBoost) is proposed to make accurate regulation on secondary heating supply and return water temperature. Preprocessing, feature construction and model training are conducted on the data above,then a heating station temperature response prediction model is generated. Combining the model and temperature delay time of the secondary network, the valve control strategy is thereafter made. The error between the 24 h predicted temperature and the measured temperature is kept within 0.47 ℃, which verifies the accuracy and reliability of the heating station temperature response prediction model. The proposed strategy can guide the temperature control of heating stations.

Key words: heating station, XGBoost, model training, valve control, prediction accuracy, intelligent heating

中图分类号: