为了充分利用BP神经网络的函数逼近能力并克服其易陷入局部极值点的主要缺陷 ,提出了利用遗传算法 (GA)和BP算法联合进行网络权值优化调整的思想 ,并把改进的遗传算法 (MGA)和BP网络有机结合 ,建立了MGA—BP网络系统。该网络系统具有遗传算法性能参数优选、网络结构参数优选、网络性能参数优选以及GA—BP算法联合进行网络权值修改几种功能。利用MGA—BP网络进行了带钢冷连轧机轧制压力预报 ,结果良好。 In order to make full use of the function approximation capability of BP neural network and to overcome the main defects that it is easy to fall into the local extremum point, the idea of ​​using genetic algorithm (GA) and BP algorithm to jointly optimize the network weights is proposed. The improved genetic algorithm (MGA) and BP network, an MGA-BP network system was established. The network system has several functions of optimizing genetic algorithm parameters, optimizing network structure parameters, optimizing network performance parameters, and jointly modifying network weights with GA-BP algorithm. The MGA-BP network was used to predict the rolling pressure of cold tandem mill. The result is good.