基于粗差判别的参数优化自适应加权最小二乘支持向量机在PX氧化过程参数估计中的应用

Adaptive weighted least square support vector machine regression with gross error detection and its application to estimate kinetic parameters for industrial oxidation of p-xylene

针对软测量建模过程中数据可能存在粗大误差以及粗差数据对模型的性能产生的影响,提出了一种基于粗差判别的自适应加权最小二乘支持向量机回归方法 (WLS-SVM)。该方法首先根据3δ法则检测出样本中的显著误差并加以剔除,然后根据样本误差的大小自适应地调整权值,使得非显著误差对模型性能的影响大大降低。另外,由于最小二乘支持向量机的正则化参数和核宽度参数对模型的拟合精度和泛化能力有较大的影响,一般依靠经验和试算的方法进行估计,耗时且不准确,本文将模型的参数作为进化算法的优化问题,应用自适应免疫算法(AIGA)对参数进行优化选择。仿真实验表明,该方法对非线性系统的建模具有很好的效果。同时,将该方法应用于工业PX氧化建模过程中动力学参数的估计中,结果表明,基于粗差判别的参数优化自适应最小二乘支持向量机预测精度高,取得了较好的效果。

The presence of gross errors can corrupt a model＇s performance,giving undesirable results.A novel weighted least square support vector machine regression（WLS-SVM）is proposed,which combines gross error detection and adaptive weight value for the training sample.First,the 3δ principle is applied to detect the gross error.Second,the initial weight is obtained according to the fitting error of each sample.Then,an adaptive immune algorithm（AIGA）is applied to obtain the optimal parameters of the WLS-SVM.To illustrate the performance of the WLS-SVM,simulation experiment is designed to produce the training sample.The results showed that the predicting performance of AIGA-WLS-SVM is the best. Furthermore,the AIGA-WLS-SVM method was applied to estimate the rate constants of an industrial p-xylene oxidation model,and the satisfactory result was obtained.