基于主成分分析优化BP神经网络模型的厌氧膜生物反应器膜污染预测*

膜污染是厌氧膜生物反应器运行中不可避免的问题,制约了工艺技术的推广应用,分析膜污染的形成过程是控制膜污染的重要内容。基于主成分分析（PCA）和反向传播神经网络（BPNN）的理论,提出了一种采用主成分分析优化BP神经网络的膜污染预测模型。以反应器连续运行试验数据为样本,利用相关性分析确定模型的输入变量,并基于输入变量间存在信息重叠问题,采用主成分分析法对输入因素进行降维处理,提取贡献率为70.4%的第一主成分和贡献率为17.7%的第二主成分作为输入特征。结合模型的贡献度分析和主成分分析发现,反应器内的污泥浓度是膜污染影响因素中最主要的特征变量,贡献度为34.9%。对比分析优化模型和单一模型的预测结果,发现PCA-BPNN模型的拟合效果更好,平均相对误差仅为3.8%,可用于膜污染分析研究,为后续研究提供参考。

Membrane Fouling Prediction of Anaerobic Membrane Bioreactor Based on BP Neural Network Model Optimized by Principal Component Analysis

Membrane fouling is an inevitable problem in the operation of anaerobic membrane bioreactors, which seriously hinders the application of membrane technology. An in-depth analysis of its formation mechanism is an important measure to manage membrane fouling. Based on the theory of principal component analysis (PCA) and back propagation neural network (BPNN), a prediction model for membrane fouling based on BP neural network optimized by principal component analysis was proposed. By inputting the continuous reactor experiment data, the correlation analysis was used to determine the input variables of the model. Considering the information overlap between the input variables, principal component analysis was used to reduce the dimension of the input factors, and the first principal component with a contribution rate of 70.4% and the second principal component with a contribution rate of 17.7% were extracted as the input characteristics. Combined with the contribution analysis and principal component analysis of the model, it was found that sludge concentration was the most important characteristic variable among the influencing factors of membrane pollution, with a contribution of 34.9%. By comparing the prediction results of the optimized model and the single model, it was found that the PCA-BPNN model had a better fitting effect, and the average relative error was only 3.8%, which can be effectively used in membrane pollution analysis and provide an reference for subsequent research.