基于煤仓分层和软测量的入炉煤实时监测  

Real-time Monitoring of Furnace Feed Coal Based on Layering Model and Soft Measurement of Coal Bunker

作　　者：陈亚平[1] 王焕明 孙胡彬 魏勇 王灵敏 赵敏 周晓亮 徐明祥 赵虹[1] CHEN Yaping;WANG Huanming;SUN Hubin;WEI Yong;WANG Lingmin;ZHAO Min;ZHOU Xiaoliang;XU Mingxiang;ZHAO Hong(State Key Laboratory of Clean Energy Utilization,Zhejiang University,Hangzhou 310027,Zhejiang Province,China;Zhejiang Zheneng Taizhou Second Power Generation Co.,Ltd.,Taizhou 317109,Zhejiang Province,China;Hangzhou Jiyi Technology Co.,Ltd.,Hangzhou 310012,Zhejiang Province,China)

机构地区：[1]浙江大学能源高效清洁利用全国重点实验室,浙江杭州310027 [2]浙江浙能台州第二发电有限责任公司,浙江台州317109 [3]杭州集益科技有限公司,浙江杭州310012

出　　处：《动力工程学报》2025年第2期190-197,213,共9页Journal of Chinese Society of Power Engineering

摘　　要：为了实现入炉煤的实时监测,在燃料唯一特征码全过程跟踪模型的基础上构建了煤仓分层模型,初步粗略地监测到煤种分层情况,并根据水分质量分数软测量结果和煤仓分层模型进行水分质量分数的匹配,将成功匹配的历史数据作为入炉煤煤种的真实值,将磨煤机电流、出力等运行参数作为特征输入,建立了基于随机森林算法的入炉煤预测模型,结合模型预测结果和水分质量分数软测量结果对煤仓分层模型错误部分进行了修正。结果表明:分类模型评估指标的测试结果为0.880,满足工程需求;所提出的将机理分析与机器学习相结合的燃煤信息监测模型实现了入炉煤的全时段识别,为锅炉燃烧优化及智能锅炉的建设打下基础,同时实现了燃料特征码全过程跟踪的闭环。In order to achieve real-time monitoring of coal fed into the furnace,a coal bunker layering model was developed based on the tracking model of unique fuel feature code throughout the entire process.The coal type layering situation was roughly monitored,and the matching of moisture contents was performed based on the soft measurement results of moisture content and the coal bunker layering model.The successfully matched historical data were used as the true coal type entering the furnace,and the operating parameters such as the current and output of the coal mills were taken as feature inputs.The prediction model of the furnace feed coal based on the random forest algorithm was developed,and the erroneous parts of the coal bunker layering model were corrected by combining the model prediction results and moisture soft measurement results.Results show that test result of evaluation indicators for classification models is 0.880,which meets the engineering requirements.The proposed coal information monitoring model based on mechanism analysis and machine learning can achieve full-time identification of furnace feed coal.It lays a foundation for combustion optimization and the construction of intelligent boilers,and also realizes the closed loop of the whole-process tracking of fuel feature code.

关 键 词：入炉煤监测 煤仓分层 磨煤机 软测量 随机森林 

分 类 号：TK16[动力工程及工程热物理—热能工程]