基于IPSO算法的回转窑煅烧带温度D-FNN预测控制  

D-FNN predictive control for burning zone temperature in rotary kiln with IPSO algorithm

作　　者：田中大[1] 高宪文[2] 李树江[1] 王艳红[1] TIAN Zhongda GAO Xianwen LI Shujiang WANG Yanhong(College of Information Science and Engineering, Shenyang University of Technology, Shenyang 110870, China College of Information Science and Engineering, Northeastern University, Shenyang 110819, China)

机构地区：[1]沈阳工业大学信息科学与工程学院,辽宁沈阳110870 [2]东北大学信息科学与工程学院,辽宁沈阳110819

出　　处：《中南大学学报（自然科学版）》2016年第10期3409-3416,共8页Journal of Central South University:Science and Technology

基　　金：国家自然科学基金资助项目(61034005);辽宁省博士启动基金资助项目(20141070)~~

摘　　要：为了提高石灰回转窑煅烧带温度的控制性能,提出一种基于改进的粒子群优化算法(IPSO)与动态模糊神经网络(D-FNN)的预测控制方法。该方法利用动态模糊神经网络建立石灰回转窑煅烧带温度的非线性预测模型,通过输出温度的预测值,引入输出反馈与偏差来校正预测误差,建立偏差与控制量的控制性能指标,通过改进的粒子群优化算法滚动优化得到系统最优控制量。对控制方法的稳定性进行分析。仿真实验结果表明动态模糊神经网络的温度预测误差在±10℃之内,具有较高的预测精度。提出的预测控制方法能使输出煅烧带温度快速稳定地跟踪设定值的变化,同时在系统输出有扰动的情况下也能较好地跟踪设定值。控制量的平均单步滚动优化需0.31 s,可满足实际应用。In order to improve the control performance of burning zone temperature in lime rotary kiln,a predictive control method based on an improved particle swarm optimization（IPSO） and dynamic fuzzy neural network（D-FNN） was proposed. This predictive control method utilizes dynamic fuzzy neural network to build a nonlinear predictive model for burning temperature in lime rotary kiln.Through predictive output temperature, performance indicators were established by deviation and control value to reduce the error in feedback output error and error correction. The optimal control value was obtained by rolling optimization of improved particle swarm optimization algorithm. The stability of the control method was analyzed. The simulation results show that the temperature prediction error of the dynamic fuzzy neural network is within ±10 ℃, and has high prediction accuracy. The proposed predictive control method can make the burning zone output temperature fast and stable track the change of the setting value. The system can also track the setting value well with the disturbance of the system output. The average single step rolling optimization of control value needs 0.31 s, which can meet the practical application.

关 键 词：回转窑 煅烧带温度 粒子群优化算法 动态模糊神经网络 预测控制 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]