基于DDPG的锅炉NO_(x)排放和屏式过热器超温的多目标优化  

Multiobjective Optimization of Boiler NO_(x) Emissions and Platen Superheater Overheating Based on DDPG

作　　者：王赫阳 刘骁 樊昱晨 刘欣[1,2] 张超群 Wang Heyang;Liu Xiao;Fan Yuchen;Liu Xin;Zhang Chaoqun(School of Mechanical Engineering,Tianjin University,Tianjin 300350,China;Yantai Longyuan Power Technology Co.,Ltd.,Yantai 264006,China)

机构地区：[1]天津大学机械工程学院,天津300350 [2]烟台龙源电力技术股份有限公司,烟台264006

出　　处：《天津大学学报（自然科学与工程技术版）》2024年第8期787-797,共11页Journal of Tianjin University：Science and Technology

基　　金：国家能源集团科技资助项目(GJNY-21-168).

摘　　要：锅炉空气分级燃烧技术的使用虽降低了NO_(x)排放,但同时造成了炉内高温火焰上移,导致位于炉膛上部的屏式过热器吸热量增加和超温加剧,影响机组的安全运行.因此,锅炉亟需一个可对NO_(x)排放和屏式过热器超温进行协调优化的多目标控制策略.针对目前基于机器学习的锅炉优化模型普遍局限于针对单一锅炉运行目标的优化,提出了基于深度强化学习的锅炉多目标优化模型,包括预测模型和优化模型:预测模型采用深度神经网络构建锅炉运行参数与NO_(x)浓度和屏式过热器温度的非线性映射;优化模型采用深度确定性策略梯度(DDPG)算法训练策略网络,通过优化运行参数实现锅炉的多目标协同控制.对某600MW锅炉的研究结果表明,通过锅炉配风和过热器减温水量等参数的调整,可实现NO_(x)排放和屏式过热器超温率的协同优化,NO_(x)排放平均降低22.6 mg/m^(3),屏式过热器超温率平均降低0.161.Air staging combustion technology is used to reduce boiler NO_(x) emissions.However,it causes the high-temperature flame to move upward in the furnace,thereby increasing the heat absorption and aggravated overheating of the platen superheater in the upper furnace,which substantially affects the safe operation of the boiler.Therefore,a multiobjective optimization strategy is required to realize coordinated control of boiler NO_(x) emissions and overheating of the platen superheater.Considering the current limitations of machine learning-based boiler optimization models with only a single boiler objective,a multiobjective optimization model framework based on deep reinforcement learning is proposed,including a prediction model and an optimization model.The prediction model adopts a deep neural network to construct the nonlinear relationships between the boiler operating parameters,NO_(x) emissions,and temperatures of the platen superheater.The optimization model adopts the deep deterministic policy gradient(DDPG)algorithm to train the policy network and realize multiobjective coordinated control by optimizing the boiler operating parameters.The application of this model to a 600MW boiler shows that the coordinated optimization of NO_(x) emissions and the overheating rate of the platen superheater can be realized by adjusting the boiler air distribution parameters and attemperation spray of the superheater.The average NO_(x) emissions are reduced by 22.6 mg/m^(3),and the average overheating rate of the platen superheater is reduced by 0.161.

关 键 词：燃煤锅炉 NO_(x)排放 屏式过热器 多目标优化 深度确定性策略梯度 

分 类 号：TM621.2[电气工程—电力系统及自动化]