MBR中膜污染的人工神经网络预测研究进展  

Research progress of artificial neural network for membrane fouling prediction in MBR

作　　者：张浩良 刘聪 洪乾坤 王侃鸣 王红宇 ZHANG Haoliang;LIU Cong;HONG Qiankun;WANG Kanming;WANG Hongyu(College of Environment,Zhejiang University of Technology,Hangzhou 310014,China)

机构地区：[1]浙江工业大学环境学院,浙江杭州310014

出　　处：《工业水处理》2022年第7期15-23,共9页Industrial Water Treatment

基　　金：国家自然科学基金面上项目(21776262);浙江省教育厅一般科研项目(自然科学类)(Y201941086)。

摘　　要：膜生物反应器(MBR)作为一种新型废水处理技术在污水处理方面具有广阔的应用前景。但是,膜污染是制约MBR进一步发展的瓶颈性问题。近年来,随着数学算法及计算机技术的发展,将人工神经网络(ANN)等机器学习算法应用于MBR的膜污染预测成为研究的热点。总结了膜污染的影响因素,探讨了基于经典数学模型膜污染预测的优缺点,综述了近年来国内外学者运用简单ANN、优化算法ANN和深度学习ANN对MBR膜污染预测的研究,提出优化算法ANN与深度学习ANN在面对复杂环境下更具优势。此外,还探讨了当前运用ANN机器学习算法进行膜污染预测存在的缺陷,指出ANN模型在中试和工业化规模的MBR膜污染预测中应用较少,并对其未来的发展进行了展望。As a new type of wastewater treatment technology,membrane bioreactors(MBR)have a broad application prospect for wastewater treatment.However,membrane fouling has becoming the bottleneck problem for wide MBR application.With the development of mathematical algorithms and computer technology in recent years,application of artificial neural network(ANN)as a machine learning algorithm for membrane fouling prediction has become the research hotspot.The factors influencing membrane fouling were summarized,the advantages and disadvantages of membrane fouling prediction based on classical mathematical models were discussed.And the recent studies on MBR membrane fouling prediction using simple ANN,optimization algorithm ANN and deep learning ANN were re⁃viewed.It was proposed that optimization algorithm ANN and deep learning ANN had more advantages under the complex environment.In addition,the drawbacks of the current ANN machine learning algorithm for membrane foul⁃ing prediction were discussed.It was pointed out that this model had been limited used in pilot and industrial scale MBR,and the future directions of ANN models for membrane fouling prediction were put forword.

关 键 词：膜生物反应器 膜污染 传统数学模型 人工神经网络 机器学习 

分 类 号：X703[环境科学与工程—环境工程]