基于遗传算法变量选择的子空间辨识方法研究  

Study on Subspace Identification Method Based on Genetic Algorithm Variable Selection

作　　者：李明珠[1] 程思宁[1] 张守兴[1] LI Ming-zhu;CHENG Si-ning;ZHANG Shou-xing(Haikou University of Economics,Hainan Haikou 571127,China)

机构地区：[1]海口经济学院,海南海口571127

出　　处：《广州化工》2021年第15期8-9,18,共3页GuangZhou Chemical Industry

基　　金：海南省自然科学基金资助(618QN254);海口经济学院校级科研资助项目(HJKY(ZD)20-10);海口经济学院校级科研资助项目(HJKY(ZD)21-01)。

摘　　要：针对污水处理过程中重要水质参数BOD_(5)(5天生化需氧量)和COD(化学需氧量)的难以实时在线测量问题,提出了基于遗传算法变量选择的子空间辨识软测量建模方法。先利用遗传算法选出与目标变量密切相关的辅助变量,再结合子空间算法建立目标函数的预测模型。基于加州大学的UCI数据库进行仿真实验,结果表明,结合变量选择后的模型比仅使用子空间建模的模型,其预测精度得到了提高,进一步验证了文中方法的有效性。Aiming at the problem that it was difficult to measure the important water quality parameters BOD_(5)(5-Day Biochemical Oxygen Demand)and COD(Chemical Oxygen Demand)in the process of wastewater treatment in real time,a soft sensor modeling method of subspace identification based on genetic algorithm variable selection was proposed.Firstly,the genetic algorithm was used to select the auxiliary variables which were closely related to the objective variables,and then the prediction model of the objective function was established by combining the subspace algorithm.Simulation experiments were conducted based on UCI database of University of California.The results showed that the prediction accuracy of the model combined with the variable selection was improved compared with the model using only subspace modeling,which further verifiedthe effectiveness of the proposed method.

关 键 词：遗传算法 变量选择 子空间辨识 污水处理 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]