BP网络优化类芬顿系统中H_2O_2与TCP的质量浓度比

Optimization of mass concentration rate between H_2O_2 and TCP by BP network

在改性活性炭与H2O2联用去除三氯酚(TCP)动力学研究的基础上,采用BP网络模型对反应过程条件进行优化。以TCP质量浓度、H2O2与TCP质量浓度比值作为BP网络的输入因子,一级反应动力学模型参数作为输出因子,采用Levenberg-Marquardt算法,以已知的11组数据作为样本,对模型进行优化训练,获取的模型对20 mg·L-1的TCP、H2O2/TCP质量浓度比值为0.1时的拟一级动力学参数进行校核,发现实测值与网络预测值相对误差为3.9%。采用该法预测效果较好,对于给定的TCP质量浓度可确定最佳的H2O2与TCP质量浓度比值。

Base on the reaction kinetics of 2,4,6 -Trichlorophenol （TCP）, the scheme was optimized using Back Propagation（BP） network on TCP removal by modifying GAC combined with H2O2. The input factors were TCP mass concentration and the mass concentration ratio of H2O2 and TCP, the output factor was pseudo-first-order kinetic coefficient. Levenberg-Marquardt calculating method was adopted and the known 11 groups data was trained to gain optimal model. The pseudo-first-order kinetic coefficient was checked when TCP mass concentration was 20 mg. L -1 andρ （ H2O2 ）/ρ （TCP） ratio was 0.1, the results showed that the relative discrepancy was 3.9 % between the experiment value and network predicting value. The predict effect was excellent and the optimal ρ （ H2O2 ）/ρ （TCP） ratio could be established with the given initial TCP mass concentration.