基于代理模型模式匹配的锅炉实时操作优化  

Optimization of boiler real-time operation based on pattern-matching of agent model

作　　者：钟崴[1,2] 林雪茹 林小杰 周懿[1] ZHONG Wei;LIN Xue-ru;LIN Xiao-jie;ZHOU Yi(Key Laboratory of Clean Energy and Carbon Neutrality of Zhejiang Province,Zhejiang University,Hangzhou 310027,China;College of Energy Engineering,Zhejiang University,Hangzhou 310027,China;Jiaxing Research Institute,Zhejiang University,Jiaxing 314024,China)

机构地区：[1]浙江大学浙江省清洁能源与碳中和重点实验室,浙江杭州310027 [2]浙江大学能源工程学院,浙江杭州310027 [3]浙江大学嘉兴研究院,浙江嘉兴314024

出　　处：《浙江大学学报（工学版）》2023年第7期1428-1438,共11页Journal of Zhejiang University：Engineering Science

基　　金：国家重点研发计划资助项目(2019YFE0126000);国家自然科学基金资助项目(51806190)。

摘　　要：为了提高电厂燃煤锅炉操作优化的有效性与实时性,提出新的基于代理模型模式匹配(PMAM)的建模框架.提出主蒸汽流量的滞后性计算方法.采用改进的模式匹配优化模型,计算历史优化操作库.引入工况注意力机制参数、状态参数区间频率法及调控最小的3层方案优化机制,确保模式匹配方案的有效性.采用神经网络算法预建模构建锅炉操作优化的代理模型,基于代理模型表征模式匹配步骤,使得本文方法可以适用于在线应用.案例结果表明,利用提出的模式匹配优化模型,能够有效地寻找优化的锅炉操作方案,工况相似度大于95%,可以使得锅炉效率提升1.92%;训练的代理模型均方误差小于0.35%.与传统方法相比,本文方法避免了优化求解带来的泛化误差影响,在提升锅炉效率的同时,具有高可靠性及实时性.A novel framework for modeling coal-fired power plant boiler operations was proposed based on pattern matching with an agent model(PMAM)in order to enhance the effectiveness and real-time performance.A new method for calculating the lag of the main steam flow rate was proposed.An improved pattern-matching optimization model was introduced to calculate the optimal operational database for historical optimization.A threelevel scheme optimization mechanism was incorporated in order to ensure the effectiveness of the pattern-matching approach.The mechanism includes attention parameters,state parameter interval frequency and regulation minimum.An agent model for boiler operation optimization was constructed offline by using a neural network algorithm,and pattern-matching steps were represented based on the agent model to enable online applications.The case results show that the proposed pattern-matching optimization model can effectively find the optimized boiler operation scheme,and the similarity of working conditions is more than 95%,which can improve the boiler efficiency by1.92% in practice.The mean square error of the trained agent model is less than 0.35%.The method avoids the influence of generalization error caused by optimization solutions compared with traditional methods,and has high reliability and real-time performance while improving boiler efficiency.

关 键 词：模糊C均值聚类 数据驱动 模式匹配 操作优化 在线优化 电厂锅炉 

分 类 号：TK227[动力工程及工程热物理—动力机械及工程]