基于粒子群算法的BP神经网络在催化裂化反应再生过程中的应用

The application of BP neural network based on a particle swarm optimization to the catalytic cracking reaction-regeneration process

催化裂化反应再生装置是一个高度非线性、强耦合的系统,工艺的复杂性使对其难以建立准确模型。基于BP神经网络强大的自适应、自学习、泛化和非线性映射能力,以加工负荷、操作条件为输入变量,柴油产出为输出变量,建立5-11-1的BP神经网络结构的催化裂化反应再生过程柴油产出关于加工负荷,操作条件的模型。然后利用粒子群算法寻优BP神经网络初始最优权值和阈值,提高神经网络的预测精度。结果证明:基于PSO-BP神经网络的催化裂化反应再生过程的预测模型在预测精确度比未经优化的BP神经网络大大提高。

The catalytic cracking reaction-regeneration process is a highly nonlinear and strongly coupled operating productive process. It is difficult to accurately describe the model due to its complex process. Owing to artificial neural network＇s powerful self-adaptive, self-organizing, self-learning and nonlinear prediction ability, a 5-11-1 BP neural network structure was built for modelling the catalytic cracking reaction regeneration process in which the machining load, 5 operating conditions were set as input variables, and diesel production was set as the output variables. Then the optimal weight and threshold of BP neural network are optimized by a particle swarm algorithm （PSO） for improving the prediction accuracy of neural network. The results show that the prediction model of the catalytic cracking reaction regeneration process based on PSO-BP neural network is significantly higher than that of BP neural network without optimization.