基于小波神经网络的铝电解槽状态预测  

Aluminum Reduction Cell Situation′s Forecast Based on Wavelet Neural Networks

作　　者：林景栋[1] 王丰[1] 廖孝勇[1] 

机构地区：[1]重庆大学自动化学院,重庆400044

出　　处：《控制工程》2012年第2期290-293,共4页Control Engineering of China

基　　金：重庆市自然科学基金资助项目(CSTC 2009BB3209)

摘　　要：针对目前国内对铝电解槽运行状况诊断存在的难度大、效率低等问题,着眼于与实时工况相区别而反应电解槽电解能力和稳定运行的电解槽状态的研究,设计了一种以小波包算法提取槽状态信息和建立了用非线性Morlet小波基取代传统神经元非线性激励函数的紧致型小波神经网络的槽状态预测模型。利用小波变换的时域局部化性质和神经网络的自学习能力,对铝电解槽的槽状态进行分析预测,克服了传统神经网络收敛速度慢,容易陷入局部最优等缺点。通过Matlab对状态预测算法进行编程。结果显示,相比传统的神经网络预测模型,铝电解槽的槽状态预测更加准确。Aiming at the problems of aluminum cell operation situation diagnosis, which is of such as great difficulty and low efficien- cy, the aluminum cell reduction situation which is different from real-time work condition. A wavelet packet state information extraction algorithm and a wavelet neural networks Whose excitation function is Morlet wavelet base are established. Aluminum reduction cell situa- tion is analyzed and forecast using time localization and self-learning ability of wavelet neural networks. Wavelet neural networks over- comes the slow convergence speed and easily get the most excellent local faults of traditional neural network. Forecast algorithm is pro- grammed by Matlab. The results showed that the forecast of aluminum reduction ceil situation is more accurate than traditional neural networks.

关 键 词：小波包 小波神经网络 槽状态预测 

分 类 号：TP27[自动化与计算机技术—检测技术与自动化装置]