基于人工神经网络的燃煤烟气脱汞效率预测  

Expectation of the mercury removal efficiency from the coal-fired flue gas based on the artificial neural network

作　　者：宋军[1] 郭庆[1] 张安超[1] 盛伟[1] 邢微波 刘志超[1] 

机构地区：[1]河南理工大学机械与动力工程学院,河南焦作454000

出　　处：《安全与环境学报》2016年第4期277-282,共6页Journal of Safety and Environment

基　　金：国家自然科学基金项目(51306046);煤燃烧国家重点实验室开放基金项目(FSKLCC1108);河南省自然科学基金项目(12A470003);动力工程及工程热物理河南省一级重点学科项目

摘　　要：针对燃煤烟气单质汞(Hg0)的脱除效率,采用神经网络遗传算法模拟寻优的方法获得最佳运行参数,并探讨多影响因子同时改变时脱汞效率的变化规律。利用田口试验方法设计试验,在固定床试验台架上考察负载量、煅烧温度、反应温度对脱汞效率的影响;采用64组试验数据对设计的多层神经网络进行训练、验证及测试,关联系数(R)在0.96以上,均方误差(Mean Squared Error,MSE)为0.007;对训练后的神经网络采用遗传算法进行多影响因子极值寻优。结果表明:采用硝酸锰浸渍催化剂的最优脱汞效率为95.1%;负载量为15%时,最佳脱汞效率、反应温度、煅烧温度分别为94.6%、141.8℃、445.2℃,进一步用试验验证了寻优结果的准确性。To evaluate the efficiency of elemental mercury removal efficiency from coal-fired flue gas,the paper is to introduce the artificial neural network（ ANN） method and the genetic algorithms（ GA） to the finding of the optimal operation parametersand the effects of the numerous factors. In promoting the research and treatment,we have explored the Taguchi method by laying out an experiment through comparing the aforementioned traditional methods with the artificial neural network（ ANN） and the genetic algorithms（ GA） using the single factor test,which lets us include a number of variables in the same trial and improve the test efficiency but reduce the corresponding costs. For example,we have worked out an experiment by using Taguchi L16,in which we have carried out a fixed-bed experimental bench to test the effects of the MnO xloading,the calcinations＇ temperature,and reaction temperature on Hg0,finding 64 sets of samples in the experiment of L16. The results of our study show that it is possible to reach a higher accuracy of the 3-10-1 structure by using the Levenberg-Marquardt（ LM） algorithm. As a result,the performance r can be promoted by above 0. 96 and the MSE about0. 007. At the same time,the successful application of the genetic algorithms for forecasting the results of the nonlinear neural network system helps to optimize the utilization of 2D and 3D.And,in turn,the optimization results help to make the relation between the calcination temperature,the reaction temperature and the mercury removal efficiency clarified. The results of the genetic algorithm optimization of the multi-factors prove to promote the maximum removal efficiency of mercury up to 95. 1% due to the use of the manganese nitrate impregnated catalyst. The optimal mercury removal efficiency,the calcination temperature and the reaction temperature have thus been made raised by 94. 6%,141. 8 ℃,and 445. 2 ℃,respectively,when the loading stays at15%. Further experiments have also verified the accuracy of the simulation r

关 键 词：环境工程学 脱汞效率 田口试验方法 多层神经网络 极值寻优 

分 类 号：X701[环境科学与工程—环境工程]