基于结构改进RBF神经网络的NO_(x)预测模型比较  

Comparison of NO_(x) Prediction Models Based on Structurally Improved RBF Neural Network

作　　者：于静 金秀章[1] 刘岳 YU Jing;JIN Xiuzhang;LIU Yue(Department of Automation,North China Electric Power University,Baoding 071003,China)

机构地区：[1]华北电力大学自动化系,河北保定071003

出　　处：《控制工程》2023年第9期1616-1623,共8页Control Engineering of China

基　　金：国家重点专项资助项目(2016YFB0600701)。

摘　　要：针对燃煤电厂选择性催化还原(selective catalytic reduction,SCR)脱硝系统入口氮氧化物浓度测量准确性的问题,提出基于结构改进的径向基函数神经网络(improvedradial basis functionneural network,IRBFNN)预测模型。采用互信息选取辅助变量作为模型的输入变量,避免变量过多或过少导致模型精度降低;利用k近邻互信息估计辅助变量的延迟时间,解决时序问题;采用调整时序的辅助变量和主导变量建立结构改进的RBF神经网络(RBFNN)预测模型;采用人工鱼群算法(artificial fish swarm algorithm,AFSA)和粒子群优化(particle swarm optimization,PSO)算法加速验证结构改进模型的优越性,并对2种优化算法的优化能力进行分析。仿真结果表明,结构改进的RBF神经网络模型的均方根误差和平均绝对百分比误差明显高于原模型;AFSA优化后的模型精度高于PSO算法,然而其需要调节的参数较多。For the measurement accuracy of NO_(x) concentration at SCR inlet of coal-fired power plant,the RBF neural network prediction model based on structure improvement is proposed.It uses mutual information to select auxiliary variables as the input variables of the model to avoid reducing the accuracy of the model due to too many or too few variables;k-nearest neighbor mutual information is used to estimate the delay time of auxiliary variables to solve the timing problem;the auxiliary variables and leading variables are adjusted to establish the structure improved RBF neural network(RBFNN)prediction model;artificial fish swarm algorithm(AFSA)and particle swarm optimization(PSO)algorithm are used to accelerate the verification of the superiority of the structural improvement model,and the optimization ability of the two optimization algorithms is analyzed.The simulation results show that the root mean square error and average absolute percentage error of the improved RBF neural network model are significantly higher than those of the original model;the accuracy of the optimized model of AFSA algorithm is higher than that of PSO algorithm,but there are many parameters to be adjusted.

关 键 词：改进RBF神经网络 互信息 k近邻互信息 人工鱼群算法 粒子群优化算法 

分 类 号：TP29[自动化与计算机技术—检测技术与自动化装置] TK39[自动化与计算机技术—控制科学与工程]