摘要
针对湿法脱硫装置运行参数多且相互高度耦合,脱硫效率定量描述困难的问题,以及传统BP网络存在的问题,提出一种基于自适应优化多层GA-BP的脱硫效率预测模型。将基于主成分分析后的降维数据作为输入变量,采用双层基因优化BP网络结构,并引入自适应变异和交叉概率,对BP网络初始权值、阈值进行改进,利用优化后的网络对脱硫效率进行预测。该模型已成功应用于大唐三门峡1000MW机组脱硫装置,结果表明:实际脱硫效率平均绝对误差小于0.5%,较传统BP算法与GA-BP算法分别降低25.82%和16.10%,具有更高的预测精度。
On the basis of the problems existing in the traditional BP neural network,including the difficulty of quantitatively describing the desulfurization efficiency,as well as many operation parameters and high coupling between them,a prediction model of desulfurization efficiency based on adaptive multi-layer GA-BP is proposed.Firstly,the reduced dimension data based on principal component analysis(PCA)is used as the input variable,and the structure of BP network is optimized by using double-layer gene.Adaptive mutation and crossover probability are introduced to improve the initial weight and threshold value of BP network,and the optimized network is used to predict the desulfurization efficiency.The model has been successfully applied to the desulfurization unit of Datang Sanmenxia 1000MW unit.The results show that the average absolute error of actual desulfurization efficiency is less than 0.5%,which is 25.82%and 16.10%lower than that of traditional BP algorithm and GA-BP algorithm respectively,and has higher prediction accuracy.
作者
章文涛
张东平
郑淑馨
ZHANG Wen-tao;ZHANG Dong-ping;ZHENG Shu-xin(School of Electrical Engineering,Nanjing Institute of Technology,Nanjing 211167,China;School of Environmental Engineering,Nanjing Institute of Engineering,Nanjing 211167,China)
出处
《信息技术》
2022年第2期53-58,共6页
Information Technology
基金
江苏省自然科学基金(BK20181023)
企业重大科研攻关项目(科18-168)。
关键词
神经网络
自适应优化
遗传算法
脱硫效率
主成分分析
neural network
adaptive optimization
genetic algorithm
desulfurization efficiency
principal component analysis
