摘要
针对BP网络建模易陷入局部极小、收敛速度慢等缺点,建立GA-BP网络预测模型,为混凝投药系统生产指导提供决策依据。利用遗传学习算法具有全局寻优的特点,同时优化BP网络的初始权值和网络结构,建立GA-BPNN混凝投药的预测控制模型。通过算法比较和模型仿真结果分析,GA-BP混合模型较BP模型收敛速度快,其平均预测相对误差仅为9.94%,预测精度远高于BP模型。表明GA-BP模型可以有效、可靠地用于混凝剂投加量预测控制系统的生产指导中。
Aimed at the shortages of the back propagation （BP）network modeling such as easily to fall into local minimum, slow convergence, ete, a GA-BP network prediction model was established to provide decision basis for production guidance of coagulation dosage system. By using the characteristics of global optimization of genetic algorithm （ GA）, at the same time optimizing BP network＇ s initial weight and network structure, a predictive control model GABPNN was established for coagulation dosage. Through algorithm comparison and analysis of model simulation results, GA-BP hybrid model convergence speed is quicker than the BP model ;the average relative prediction error is only 9. 94% ,prediction accuracy is far higher than the BP model. The results show that the model has high prediction accuracy in the coagulation dosage and can be applied to the operating process.
出处
《化工自动化及仪表》
CAS
北大核心
2009年第2期75-78,共4页
Control and Instruments in Chemical Industry
