基于BP神经网络的污水格栅间气体监测*

为实现对城市污水泵站格栅间内H2S和CO有害气体浓度实时准确监测分析，设计气体传感器阵列监测系统，建立单隐层BP神经网络模型对气体浓度样本进行训练，确定网络隐含层数为9，训练函数为traindm，模型优化后输出的H2S和CO气体浓度与实测值的平均相对误差分别为1.3%与0.4%。研究结果表明:采用BP神经网络模型对传感器阵列采集到的样本数据进行训练，可提高其测量精度，提升精度范围在57%~75%之间，能够满足对污水泵站有害气体监测的基本要求，所建立模型表现出良好效果。研究结果可为污水格栅间有害气体监测提供新方法。

Monitoring of gas in sewage grille room based on BP neural network

To realize the real-time and accurate monitoring and analysis on the concentrations of H2S and CO harmful gases in the grille room of urban sewage pump station,a gas sensor array monitoring system was designed.A single hidden layer BP neural network model was established to train the gas concentration sample.It was determined that the number of hidden layers of the network was 9 and the training function was traindm.The average relative errors of the concentrations of H2S and CO between the optimized model output and the measured values were 1.3% and 0.4%,respectively.The results showed that adopting the BP neural network model to train the sample data collected by the sensor array could improve its measurement accuracy,and the improve accuracy range was 57%~75%,which could meet the basic requirements of harmful gases monitoring of sewage pump station,the established model presented good effect.The results can provide a new method for monitoring the harmful gases in the sewage grille room.