Research On MBR Simulation Based On DE-GSA-LSTM

MBR(membrane bioreactor)has many advantages,such as good effluent quality,small floor area and so on.However,membrane fouling has become a bottleneck in the development of MBR technology.The substances in wastewater will adhere and deposit on the membrane surface and inside the membrane pore,resulting in membrane fouling,which leads to the decline of MBR filtration performance and the improvement of operation power consumption.For this problem,how to prevent the occurrence of membrane fouling,how to slow down the aggravation of membrane fouling,and how to clean or replace the membrane after fouling have always been the research hotspots in MBR field.In the process of MBR system operation,it is difficult to directly measure the pollution degree of membrane module.How to accurately measure the operation state of membrane module,so as to take measures to prevent and control membrane pollution,has always been a research difficulty.In this paper,LSTM(long-term and long-term memory network)is used as the core to build a deep learning model to simulate the operation state of membrane module in MBR.In MBR system,hydrodynamics and biochemical reactions are complex,so it is difficult to accurately predict operating state parameters by traditional mathematical model.ANN(artificial neural network)links input and output without considering the specific changes in the process.In recent years,many research results have been achieved in the field of MBR.On the basis of standard cyclic neural network,LSTM adds gating memory unit,which makes the network have long-term memory function and is suitable for solving time-dependent problems.However,LSTM is easy to fall into local minimum value point,easy to occur gradient disappearance and gradient explosion.In this paper,DE(differential evolution algorithm)and GSA(gravity search algorithm)are used to optimize the initial weights and thresholds of LSTM.DE is an evolutionary algorithm with strong global search ability and simple program implementation.GSA is an intelligent optimization algorithm based on population,which has the advantages of fast convergence speed and high precision.The combination of DE and GSA can keep the strong global search ability of DE and avoid the slow convergence rate of DE in the later stage,and use GSA to accelerate the convergence speed.Using DE-GSA to optimize LSTM can avoid the network model falling into local minimum and accelerate the convergence speed.The DE-GSA-LSTM model was established to predict the membrane flux of membrane module in MBR system.At the same time,the training DE-LSTM model and LSTM model were compared.The experimental results show that the DE-GSA-LSTM model has higher accuracy.Based on the previous single parameter prediction model of MBR,a multi parameter prediction model based on LSTM is established with the core of DEGSA-LSTM,which can predict multiple parameters such as membrane flux,sludge apparent yield coefficient and membrane resistance,so as to describe the operation status of membrane module more comprehensively.The experimental results show that the multi parameter prediction model based on LSTM has high accuracy,which can accurately describe the membrane fouling state in MBR,and has certain reference value for the research in this field.