Model Predictive Control For Oxygen Starvation In Air Supply System Of PEM Fuel Cell

With the development of society and standard of living, people pay more attention to the problems of environmental protection and energy shortage. Fuel cells become the focus of most research activities in the world because of its high efficiency and low pollution. But oxygen starvation problem in air supply system of PEM fuel cell affects system performance directly. This paper uses two control schemes, linear quadratic regulator and quasi-infinite horizon model predictive control, to solve above problem based on mathematic model of fuel cell system and obtains approving effect under MATLAB/SIMULINK simulation environment.Firstly, we introduce the working principle and kinds of fuel cell, pointing that proton exchange membrane(PEM) fuel cell is considered as promising vehicle power source because of its outstanding performance and compact structure. But the character of multi-input and multi-output, nonlinear, strong coupling, hysteresis, constraint, uncertainty and random disturbance makes conventional control technique not suitable for it. Next, we present the research actuality of fuel cell control and show that the oxygen starvation in air supply system is an important factor influencing system performance.Mathematic model is an effective tool to analyze system performance. Fuel cell's model is used for researching the relationships of operating variables, working condition, interior status and cell performance. With its help, we can comprehend the process of reaction and the influence of parameters deeply. But PEM fuel cell is a complex object. It relates to many subjects such as electrochemistry, hydrodynamics, thermodynamics and electrotechnics. So the course of modeling is very difficult. Pukrushpan proposed a airflow dynamic model suitable for control of PEM fuel cell transient behaviour. Because this paper utilizesthat model to solve oxygen starvation problem, we then explain each part of it in detail and elucidate control requirement and control task. After that, we establish simulation model under MATLAB/SIMULINK environment and accomplishsystem simulation imitating the running of fuel cell vehicle. The result shows that step disturbance of load current affect system performance greatly. It causes oxygen excess ratio in cathode deviating far from its desired output. So our task is to regulate the pressure of supply manifold and increase the flow rate of cathode inlet in order to reducing the loss of system performance. Because the nonlinear model of system is not suitable for analysis and computation, we then linearize it and prepare for the design of control system.It is convenient for analysis and design of multi-input and multi-output system with state space expression, which is the most obvious character in modern control theory. Linear quadratic regulator(LQR) is the earliest and maturest technique in state space design. So the theory of LQR is introduced at first. With the help of it, we design the system controller on the basis of PEM fuel cell linear model. Then we analyze the influence of weight matrix Q and R on system performance. Compared with open-loop system, the effect of closed-loop system controlled by LQR is much better. But the existence of system constraints makes LQR not applicable. We must try a new method to deal with it.Model predictive control(MPC) obtains wide application with its advantages in flexible problem description, capability of dealing with constraints on-line, moving horizon strategy and multivariable control. Firstly, we introduce the basic principle, problem description and stability of MPC, pointing that quasi-infinite horizon MPC is considered as one of the most developmental predictive control arithmetic. Next, the stability theorems of disperse system quasi-infinite horizon MPC are deduced. With its help, we design the system controller on the basis ofPEM fuel cell linear model. After that, we accomplish system simulation with step disturbance of load current under MATLAB/SIMULINK environment. The result shows that the effect of closed-loop system controlled by MPC is much better than open-loop system. Especially, the outstanding capability of dealing with system constraints makes MPC more suitable for solving oxygen starvation problem. The validity and superiority are proved.