基于改进WOA和BiLSTM的MBR膜污染预测研究

打开文本图片集

中图分类号：X703；TP183 文献标识码：A DOI： 10.7535/hbgykj. 2025yx06008

Research on MBR membrane fouling prediction based on improved WOA and BiLSTM

XUE Tonglai'， ZHU Zhicheng²，LIU Xiangcen 1 ， ZHANG Zheng1， ZHOU Meng -1 （1.Collegeof Electricaland ControlEngineeing，Nrth China Universityof Technology，BeijinglOo144，China；2.Iner Mongolia Ecological Environment Supervision Technical Support Center，Hohhot， Inner Mongolia Olool1，China）

Abstract：Torealize real-time prediction and intellgent monitoring of membrane fouling in Membrane Bioreactor（MBR） systems，a membranefouling prediction model based onanimproved Whale Optimization Algorithm （the Whale Optimization Algorithmintegrated withaglobalsearchstrategy，referred toastheGravitational Search WhaleOptimizationAlgorithm，

GS-WOA）anda Bidirectional Long Short-Term Memory（BiLSTM） neural network wasdeveloped.First，monitoring data samples werestandardized，and the BiLSTMneural network was adoptedas the basic prediction framework tofullutilizeits bidirectional temporalfeatureextractioncapabilityforcapturing thedynamicvariationofthefouling process.Then，the Whale Optimization Algorithm（WOA）was improved by introducing agravitational search mechanismand adaptive inertia weight to globallyoptimize BiLSTMhyperparameters such as learning rate，numberof hidden neurons，and timestep，thereby balancing global explorationandlocalexploitation.Finally，theoptimized model was trainedandvalidatedusingactualoperationaldata. The results show that the GS-WOA-BiLSTM model achieves a prediction accuracy of R2=0.9837 ，improving by approximately 6.6% compared with the LSTM model，while the mean absolute error and root mean square error are reduced by 28.1% and 19.9% ，respectively. The predicted values exhibited excellent agreement with measured data. This method enables high-precision predictionand trend forecastingof membrane fluxandtransmembrane pressure，providing reliable technical support for intelligent monitoring and optimized operation of MBR systems.

Keywords：water pollution control engineering；MBR membrane；membrane fouling prediction;improved Whale Optimization Algorithm；BiLSTM neural network

膜生物反应器技术（membranebio-reactor，MBR）通过将高效膜分离技术与传统活性污泥法有机结合实现了污水的高效净化与回用。（剩余10274字）