基于遗传神经网络热轧带钢卷曲温度预报系统  

Genetic Neural Network Based Temperature Prediction System of Hot Rolling Strip Steel Roll

作　　者：孙铁军[1] 高海[1] 王洪希[2] 徐昭云[2] 

机构地区：[1]北京科技大学信息工程学院,北京100083 [2]北华大学电气信息工程学院,吉林吉林132021

出　　处：《自动化仪表》2008年第4期23-25,共3页Process Automation Instrumentation

基　　金：吉林省教育厅科研计划项目(编号:200688)

摘　　要：带钢冷却过程的热交换是非常复杂的多变量、非线性、强耦合、大滞后的过程,难以用数学模型精确表达。提出了一种将遗传算法与神经网络结合起来,能够提高卷取温度预报精度的系统。利用热连轧现场生产过程中的实际历史数据,对预报带钢卷取温度的遗传神经网络进行离线学习和测试。结果表明,达到了卷取温度预报的精度要求,同时具有较快的收敛速度,能够满足在线实时控制的要求。Heat exchange in cooling process of strip steel is a very complicated multiple variable, non-linear, closed coupling and large time lag process that is impossible to be accurately expressed with mathematical model. A coiling temperature prediction system combining generic algorithm and neural network is proposed to enhance the accuracy of prediction is proposed. By using practical historical data in production of hot rolling field, the generic neural network predicting coiling temperature of strip steel is tested and trained off-line. The results shows that the expected accuracy of predicting coiling temperature is obtained and faster convergence speed is presented, the system meets the requirements of on-line real time control.

关 键 词：遗传神经网络 迭代 训练集样本 非线性 温度控制 

分 类 号：TP273[自动化与计算机技术—检测技术与自动化装置]