广义动态模糊神经网络在铝电解预测中的研究与仿真  

Research and simulation of general dynamic fuzzy neural network for forecast in aluminum electrolysis process

作　　者：曾水平[1] 倪亚超[1] 

机构地区：[1]北方工业大学自动化系,北京100041

出　　处：《冶金自动化》2014年第1期21-27,共7页Metallurgical Industry Automation

基　　金：国家自然科学基金资助项目(51174007)

摘　　要：铝电解过程中,每日出铝量、氟化铝添加量的决策对铝电解过程温度和过热度的准确控制非常重要,是影响铝电解技术经济指标的重要因素之一。本文基于广义动态模糊神经网络算法(GD-FNN),构造了铝电解预测系统,在给定铝电解槽所需温度之后,对铝电解中的出铝量和氟化铝添加量进行预测。此算法通过对高电流效率、低能耗的电解槽的运行规律进行自适应分析,训练出对应的决策规则,运用到效率低的电解槽,可以实现铝电解槽温度和过热度控制,提高铝电解电流效率。通过对某铝电解厂实际数据进行仿真实验,证明了该算法在铝电解控制中的有效性。In the aluminum electrolysis process, the determination of aluminum tapping volume and A1F3 addition is very important to control the temperature of aluminum production cells and the over- heated temperature precisely. Based on general dynamic fuzzy neural network, an aluminum electrolytic prediction system is constructed to forecast the aluminum tapping volume and A1F3 addition at a given temperature of aluminum production cells that we need. Through the adaptive analysis of operation law of high current efficiency and low-energy electrolytic cell, the corresponding rules are trained and ap- plied to the inefficient electrolytic cell algorithmically, so as to control the temperature of aluminum production cells and the overheated temperature and improve the efficiency of aluminum electrolysis. The simulation shows the algorithm validity in the industrial control systems.

关 键 词：铝电解 出铝量 氟化铝添加量 广义动态模糊神经网络 

分 类 号：TF821[冶金工程—有色金属冶金] TP183[自动化与计算机技术—控制理论与控制工程]