神经网络预测催化超临界水氧化废水效果的研究

Study on the Neural Network in Predicting the Degradation Efficiency of Wastewater by Catalytic Supercritcial Water Oxidation

在间歇式反应器中进行催化超临界水氧化DDNP废水试验,考察催化剂浓度、反应温度、压力、停留时间对氧化效果的影响。在实验基础上采用BP神经网络算法,建立以催化剂浓度、反应温度、压力、停留时间作为网络模型的输入层,COD去除率作为输出层的双隐层BP神经网络预测模型,预测催化超临界水氧化废水的效果,仿真结果表明模型预测效果较好。

The catalytic supercritical water oxidation （SCWO） of DDNP wastewater is performed in an intennittent reactor, to investigate the oxidation decomposition efficiency of organic compounds. The decomposition efficiency is influenced by the concentration of catalyst, reaction temperature, pressure, residence time. Based on the experimental results, a BP Elman network prediction model with two hiddenlayer is established using the concentration of catalyst, reaction pressure, temperature, residence time as input variables, and the COD removal rate as output. The simulation results show that the Elman model is of excellent forecasting ability.