基于粒子群优化的溶解氧质量浓度支持向量回归机  

Support Vector Regression Using Particle Swarm Optimization for Dissolved Oxygen Concentration

作　　者：安爱民[1] 祁丽春 丑永新[1] 张浩琛[1] 宋厚彬[1] 

机构地区：[1]兰州理工大学电气工程与信息工程学院,兰州730050

出　　处：《北京工业大学学报》2016年第9期1318-1323,共6页Journal of Beijing University of Technology

基　　金：国家自然科学基金资助项目(61563032)

摘　　要：针对污水处理中溶解氧质量浓度无法在线精确测量的问题,提出基于粒子群算法优化支持向量回归机(PSO-SVR)的溶解氧质量浓度软测量模型.为了提高溶解氧的预测精度和效率,采用粒子群算法对支持向量回归机的模型参数进行优化,并以自动获取的最佳参数组合构建溶解氧与其影响因子间的非线性软测量模型,利用该软测量模型对国际基准仿真模型BSM1的溶解氧质量浓度进行预测.仿真结果表明:该模型能得到较好的预测效果,与SVR、RBF神经网络相比,PSO-SVR模型不仅计算复杂度低,而且收敛速度快,预测精度高,泛化能力强.To improve the precision and efficiency of dissolved oxygen,a particle swarm optimization( PSO) of support vector regression( SVR) method was proposed to predict the concentration of dissolved oxygen in the soft measurement model because of the problem of the dissolved oxygen concentration cannot be measured online in the wastewater treatment. The particle swarm algorithm was used to optimize the parameters of support vector regression,and the construction of dissolved oxygen and its impact factors in nonlinear soft sensor prediction model were automatically obtained by the best combination of parameters by using the model of dissolved oxygen concentration on the international BSM1 benchmark simulation model. The simulation shows that the prediction model has good prediction effect compared with SVR,RBF neural network model,the PSO-SVR model not only has low computational complexity,but also has the advantages of fast convergence,high forecast precision,strong generalization ability etc.

关 键 词：溶解氧质量浓度 粒子群算法 支持向量回归机 污水处理 软测量 

分 类 号：TQ015.9[化学工程] TP391.9[自动化与计算机技术—计算机应用技术]