神经网络预测在饮用水粉末活性炭应急处理中的应用

针对粉末活性炭应急处理工艺影响因素众多,经典数学模型很难精确描述多因素共同作用时的污染物去除情况,以出水的剩余污染物浓度为预测指标,结合实际试验情况,选取投加量、吸附时间、污染物种类、污染物浓度、pH、温度、有机物浓度等主要影响因素,使用BP神经网络建立粉末活性炭应急处理技术出水水质仿真及预测模型.结果显示,试验值和模拟值之间有紧密的相关性且离散程度不明显,该BP神经网络模型对粉末活性炭应急处理工艺有较好的预测能力.

Application of neural network prediction in powdered activated carbon emergency treatment of drinking water

The classical mathematical model was difficult to accurately describe the pollutant removal when many factors affected powdered activated carbon emergency treatment process. With the residual contaminant concentration of outflow as prediction indexes and the actual test conditions, selecting the main influencing factors, such as dosage, adsorption time, types of pollutants, pollutant concentration, pH, temperature and concentration of organic matters, the simulation and prediction model of outflow quality of powdered activated carbon emergency treatment process was established by BP neural network. The results showed that there was a close correlation between the experiment values and simulation values, and the degree of dispersion was not obvious, which indicated that the BP neural network model had a good predictive ability on the powdered activated carbon emergency treatment process.