自适应差分进化算法优化BP网络的炼焦能耗模型  

BP Network Incorporating Adaptive Differential Evolution Algorithm for Coking Energy Consumption Model

作　　者：都吉东 陶文华[1] 李邵鹏 DU Ji-dong;TAO Wen-hua;LI Shao-peng(School of Information and Control Engineering,Liaoning Shihua University,Fushun 113001,China;Fushun Qilong Chemical Go.,Ltd.,Fushun 113001,China)

机构地区：[1]辽宁石油化工大学信息与控制工程学院,辽宁抚顺113001 [2]抚顺齐隆化工有限公司,辽宁抚顺113001

出　　处：《测控技术》2018年第4期20-23,36,共5页Measurement & Control Technology

基　　金：国家自然科学基金资助项目(61203021)

摘　　要：为了解决炼焦生产过程中炼焦能耗难以实时监测的问题,基于面向实际建立的输入输出指标体系,依据焦化厂实际炼焦生产过程中的历史数据,建立了BP网络的炼焦能耗预测模型。为了提高模型的预测精度,利用自适应差分进化算法(ADE)对复杂函数的优化能力,对BP网络权值阈值随机产生易陷入局部极小值的缺点进行优化,找出了BP网络的最佳连接权值和阈值,建立了自适应差分进化算法(ADE)优化BP网络的炼焦能耗模型。仿真结果表明,自适应差分进化算法(ADE)优化BP网络的炼焦能耗模型的预测值与实际值的接近程度达90%以上。该研究为炼焦生产过程中炼焦能耗难以实时监测的难题提供了一种新思路,可为炼焦行业高效低耗生产提供理论依据。In order to solve the problem that coking energy consumption is difficult to be monitored in real time during coking production,a BP network model for coking energy consumption prediction was established based on the actual input and output index system and the historical data of coking production process in coking plant.In order to improve the prediction accuracy of the model,the optimization ability of adaptive differential evolution algorithm(ADE)for complex function was used to optimize the BP network whose weights and thresholds were easy falling into local minimum,then the optimal connection weights and thresholds of BP net-work were found,and the coking energy consumption model of ADE-BP was established.The simulation results show that the prediction value of ADE-BP coking energy consumption model was more than 90%close to the actual value.This study provides a new idea for the difficult problem of real-time monitoring coking energy con-sumption in coking process,which can provide a theoretical basis for the high efficiency and low consumption of coking industry.

关 键 词：炼焦能耗 自适应差分进化算法 BP网络 预测精度 

分 类 号：TP18[自动化与计算机技术—控制理论与控制工程]