人工神经网络用于电除盐产水水质预测研究

为了实现化工行业高纯水全自动生产,本文利用电除盐（electrodeionization,EDI）技术替代传统的混床技术,并研究了人工神经网络对EDI产水过程模拟仿真的可行性,采用误差反向传播网络（BP网络）建立了进水流量、电导率、pH值以及工作电压与EDI透过水电阻率之间关系的动态模型,并对不同的训练样本归一化方法和训练方法进行比较。结果表明,在网络隐含层层数为1、节点数为13时,采用归一化方法三能够较好的预测EDI透过水电阻率,且该模型可用于EDI除盐过程的动态描述,为实现化工行业高纯水全自动生产奠定了基础。

Prediction and Application of Artificial Neural Network for the Electrodeionization Product Water Quality

Abstract： In order to achieve the automatic controlled process of producing high purity water in chemical engineering industries, ion exchange mixed bed was instead by electrodeionization （EDI） for boiler feedwater. The feasibility of simu- lation based on artificial neural network （ANN） for EDI producing pure water was also investigated. With error back prop- agation （BP） network, a dynamic simulation model showing the relationship between feed water flow rate, conductivity,pH value, operating voltage and outlet resistivity was established. Furthermore, the different training sample normalization methods were compared at different radbas node number. The results showed that the third normalization method to 1 rad- bas and 13 node number ANN were able to predict resistivity of EDI outlet accurately, and the simulation model can effec- tively describe the process of EDI producing high pure water, which laid the ground work for achieving the automatic con- trolled process of producing high purity water in chemical engineering industries.