Direct methanol fuel cell modeling based on the norm optimal iterative learning control

Abstract

Direct methanol fuel cells are one of the most promisingly critical fuel cell technologies for portable applications. Due to the strong dependency between actual operating conditions and electrical power, acquiring an explicit model becomes difficult. In this article, the behavioral model of direct methanol fuel cell is proposed with satisfactory accuracy, using only input/output measurement data. First, using the generated data which are tested on the direct methanol fuel cell, the frequency response of the direct methanol fuel cell is estimated as a primary model in lower accuracy. Then, the norm optimal iterative learning control is used to improve the estimated model of the direct methanol fuel cell with a predictive trial information algorithm. Iterative learning control can be used for controlling systems with imprecise models as it is capable of correcting the input control signal in each trial. The proposed algorithm uses not only the past trial information but also the future trials which are predicted. It is found that better performance, as well as much more convergence speed, can be achieved with the predicted future trials. In addition, applying the norm optimal iterative learning control on the proposed procedure, resulted from the solution of a quadratic optimization problem, leads to the optimal selection of the control inputs. Simulation results demonstrate the effectiveness of the proposed approach by practical data.