基于改进粒子群优化LSSVM的污水COD软测量建模

Soft measurement modeling of wastewater COD based on improved particleswarm optimization LSSVM

针对在污水处理过程中水质参数(如出水化学需氧量(COD),pH值)变化过程的高度时变性、非线性和复杂性等特点,提出一种基于改进粒子群优化最小二乘支持向量机(IPSO-LSSVM)的软测量模型。该模型将小样本机器学习——最小二乘支持向量机(LSSVM)引入工业污水处理过程水质参数预测,网络训练过程中采用粒子群优化算法,使得该算法能够自适应获取最优超参数,形成IPSO-LSSVM算法,对工业污水处理出水COD参数进行回归预测。实验结果表明:与LSSVM和PSO-LSSVM模型相比,IPSO-LSSVM模型预测结果的均方根误差分别降低了40.9%和30.5%;相关系数分别提高了13.0%和6.6%。这表明IPSO-LSSVM模型在预测精度、收敛速度和抗干扰能力等方面明显优于LSSVM和PSO-LSSVM模型。

Aiming at the problem of highly time-varying,non-linear and complex changes of water quality parameters(such as effluent chemical oxygen demond(COD),pH value)in new wastewater treatment plants,a soft sensor modeling method based on improved particle swarm optimization least squares support vector machine(IPSO-LSSVM)is proposed.The small sample learning machines-Least Squares Support Vector Machine(LSSVM)is introduced to predict the effluent chemical oxygen demand(COD)during wastewater treatment.In order to improve the prediction accuracy of the model,the improved particle swarm optimization algorithm is used in the network training process,which makes the algorithm adaptively obtain the optimal model parameters.The resucts show that compared with the traditional LSSVM model and PSO-LSSVM model,the prediction results of root mean square error by the IPSO-LSSVM model was reduced by 40.9%and 30.5%,respeetively;the correlation coefficient was increased by 13.0%and 6.6%respectively.This shows that IPSO-LSSVM model is significantly better than the LSSVM and PSO-LSSVM models in terms of prediction accuracy,convergence speed and auti-interference ability.