摘 要：随着环保意识的提高，火电厂氮氧化物的排放问题引起了社会的关注。选择性催化还原方法（selective catalytic reduction,SCR）普遍采用于火电厂的脱硝系统中，喷氨量的控制是脱硝系统的关键问题，若不能得到有效控制，不仅影响火电厂脱硝效率，而且氨逃逸有可能会造成二次污染。SCR系统的准确动态建模是对喷氨量进行有效控制的前提。本文通过分析SCR系统化学反应机理建立了SCR动态模型，并利用实际运行数据对其进行验证，平均百分比误差为10.73%，仿真结果表明该模型精度高。因该系统具有较强的非线性，传统比例、积分、微分（PID）控制器难以得到满意的控制效果，本文设计了神经网络预测控制器，对SCR系统的喷氨量进行控制，并与传统PID控制器进行对比，利用火电厂实际运行数据对该方法进行了验证，提出的控制策略在两个典型工况下的脱硝效率比传统PID控制策略高约10%，结果表明该方法具有较好的控制效果。

Abstract:Environmental protection problems are attracting more and more attention and now becoming an essential social issue.The selective catalytic approach for NOx reduction is commonly apllied in thermal power plant.Adjusting injection of ammonia,which is a key in SCR system,would affect the de NOx efficiency.Futhermore,excessive ammonia injection would lead to ammonia escape in outlet of SCR system,resulting in environmental pollutant.A mechanism model of SCR system based on chemical analysis is proposed,and it is verified by using historical data of power plant which owning 10.73% MAPE(mean absolute percentage error)compared with historical data,showing a high accuracy.Owing to the nonlinear characteristic of SCR system,traditional PID controller could not achieve a satisfactive control performance.Therefore,a model-based predictive control strategy is needed to optimize the control effect,simulation results reveal that the effeciency of denitration of novel strategy is almost 10% higher than the traditional one,which indicates that the novel strategy achieves a superior control performance.

TK411+.5

国家自然科学基金（51606063）