基于机器学习算法的能源总线系统数据驱动建模研究  

Study on Data-driven Modeling of Energy Bus System Based on Machine Learning Algorithms

作　　者：孙可欣 范蕊[1] 郑彬 步婷 周亿冰 史洁[1] SUN Kexin;FAN Rui;ZHENG Bin;BU Ting;ZHOU Yibing;SHI Jie(Tongji University,Shanghai 200092;SPIC Northeast Electric Power Co.,Ltd,Shenyang 110181;Shanghai Research Institute of Building Sciences CO.,LTD,Shanghai 200032)

机构地区：[1]同济大学,上海200092 [2]国家电投集团东北电力有限公司,沈阳110181 [3]上海市建筑科学研究院有限公司,上海200032

出　　处：《建筑科学》2024年第4期248-259,共12页Building Science

基　　金：国家自然科学基金项目“基于不确定性的集成低品位能源总线系统运行机理与多目标优化研究”(52078356);福建省住房和城乡建设行业科学技术计划项目“装配式建筑碳排放评价指标体系与减排策略研究及示范应用(2022-K-94)”。

摘　　要：区域能源总线系统(EBS)集成多种可再生能源、余热废热资源等,成为当前低碳能源转型背景下实现建筑供热/冷的一项核心基础设施。为探讨适用于能源总线系统的数据驱动建模方法,本文以夏热冬冷地区典型酒店建筑群为研究对象,利用多元线性回归、支持向量回归、随机森林回归、人工神经网络和高斯过程回归等机器学习方法,基于典型系统工况下的Dymola仿真数据建立模型,对比分析各种算法模型的适用性,为能源总线系统的建模和优化提供基础。经过数据模型比较分析可知,人工神经网络、随机森林回归、基于Matern内核的高斯过程回归、基于径向基函数内核的高斯过程回归和基于径向基函数的支持向量回归较适用于本文能源总线系统的建模方法。上述模型的R2在多数情况下能保持在0.90以上,从而预测能源总线系统性能变化趋势,及时作出策略响应。The district energy bus system(EBS)integrates a variety of renewable energy and waste heat resources,and has become a core infrastructure for realizing building heating/cooling in the context of the current low-carbon energy transition.To explore the data-driven modeling methods applicable to EBS,this paper took typical hotel buildings in hot summer and cold winter areas as the research objects and used machine learning methods such as multiple linear regression,support vector regression,random forest regression,artificial neural network,and Gaussian process regression to build models based on simulation data in Dymola under typical system working conditions.Then,the applicability of different algorithmic models was compared and analyzed to provide a basis for modeling and optimization of EBS.The results of comparative analysis of data models show that artificial neural network,random forest regression,Gaussian process regression based on Matern kernel,Gaussian process regression based on radial basis function kernel,and support vector regression based on radial basis function are more suitable for EBS.The R2 of the above model can be kept above 0.90 in most cases,so as to predict the performance variation trend of the EBS and make timely strategic responses.

关 键 词：能源总线系统 夏热冬冷地区 数据驱动 酒店建筑 

分 类 号：TU831[建筑科学—供热、供燃气、通风及空调工程]