Operating Space Design Of Data Driven Microbial Fuel Cell And Control

As a major bioelectrochemical system,microbial fuel cell(MFC)converts biomass into electric energy through microbial metabolic activities and can be combined with wastewater treatment to processing polluted water.Microbial fuel cells(MFC)have become a promising but challenging technology in recent years.It can meet urgent energy needs,especially using wastewaters as substrates,which can get new energy and solving environmental problems.At present,most researchers are committed to work on how to improve the performance of microbial fuel cells and application in order to put into practical production.In this paper,data-driven method is used to study the operation design and on-line control of MFC.The main work is as follows:Proper design of operation variable space is of great significance for developing new MFC device and improving the performance of MFC process.This paper presents a space design method based on data-driven model which makes full use of the historical experimental data.Compared with the traditional mechanism model design method,it is easy to realize quickly and economically.The support vector regression(SVR)forward and inverse model are deduced with the quadratic kernel function,in which the quadratic kernel function is suitable for the mathematical formula in the inversion stage.And the space design of operation variables is proposed to calculate directly from the inverse model with the effect of confidence interval when the model prediction uncertainty is considered.The proposed design method is verified in the real MFC-A2/O equipment.It is shown that the designated operation space is a narrow and effective region of the knowledge space which brackets the entire fraction of the MFC experiment space.And in general terms,the possible product quality from the designated operation space is more densely concentrated on the desired value compared to the tradition forward model design method.Aiming at the on-line control problem of MFC,this paper proposes a class of explicit model predictive control method based on machine learning data model,which is divided into two stages:off-line design and on-line control.In the off-line design phase:(1)sampling the admissible state space by the mathematical model and solving for optimal model predictive control actions at each sampled data point after determining feasible region of the nonlinear programming problem;(2)constructing the feasibility sample discriminator to identify the feasibility sample using support vector machines,and(3)constructing the control surface of explicit model predictive controller using different method such as artificial neural network and partial least squares.In the process of on-line control,the process data are collected in real time,and the feasible control output is calculated by using the existing predictive control surface.The method is validated in a simulation model of a class of microbial desalination fuel cell(MDC).The simulation results show that the proposed explicit model predictive control method avoids on-line optimization,is easy to realize,and has a good control effect.