关于铝电解过程节能模型优化仿真研究  

Simulation Study on the Optimization of Energy Saving Model about Aluminum Electrolysis Process

作　　者：姚立忠[1] 李太福[1,2] 苏盈盈[1] 张恒健[1] 

机构地区：[1]重庆科技学院电气与信息工程学院,重庆401331 [2]重庆大学自动化学院,重庆400030

出　　处：《计算机仿真》2016年第12期360-365,共6页Computer Simulation

基　　金：国家自然科学基金(51374268;51375520);重庆市教委科学技术项目(KJ1401301);重庆科技学院校内科研基金项目(CK2015Z26);重庆市重大应用技术开发项目(cstc2013yykf C0034);重庆高校优秀成果转化项目(KJZH14218)

摘　　要：铝电解过程能耗大,建立精确的节能模型是降低能耗的重要前提。由于上述过程参数多,耦合性强,很难定性/定量地分析出它们与节能目标之间的关系及重要性,致使模型建立困难。提出一种结合无迹卡尔曼神经网络(Unscented Kakman Filter Neural Network,UKFNN)和改进型随机化测验的节能模型优化分析方法。首先利用UKFNN建立铝电解过程的动态节能模型。然后利用UKFNN释义图和连接权法,计算各参数的贡献率。最后,利用改进的动态随机化测验对节能模型连接权进行显著性检验,剔除冗余参数,并与优化前的节能模型比较。仿真结果表明,优化后模型性能优良。上述方法为分析节能模型参数关系和优化节能模型提供了一种有效途径。An analytic method about the optimization of energy saving model was proposed based on Unscented Kalman Filter Neural Network（UKFNN） and improved randomization test. Firstly, the dynamic model between energy parameters and the direct current consumption of aluminum per ton was established using UKFNN. Then, the neural network interpretation chart and connection weight method were used to calculate the on the relative contribution of the direct current consumption. Lastly, the connection weight and energy parameters were tested by the improved dynamic randomization test, to eliminate redundant connection weights and energy saving parameters. The experimental results show that the optimized model has an excellent performance. So, the method for the analysis of model parameter, optimization of energy saving model provides an effective way.

关 键 词：铝电解 节能模型 无迹卡尔曼滤波 神经网络 随机化 耦合性 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程] TP391.9[自动化与计算机技术—控制科学与工程]