Model Predictive Control for Electrochemical Impedance Spectroscopy Measurement of Fuel Cells Based on Neural Network Optimization

Abstract

Electrochemical impedance spectroscopy (EIS) is a key specification of fuel cells, which can reflect the healthy state. EIS could be measured by using the ripple modulation of dc/dc converter connected to fuel cells. This method has the advantage of no external excitation sources and low volume. In ripple modulation, the mixed signals composed of direct current and alternate current are difficult to track accurately, so a model predictive control (MPC) method with favorable stability and dynamic response is proposed. Considering that the traditional optimization algorithm takes a long time to calculate, a recurrent neural network (RNN) optimization is used to find a solution of the quadratic programming (QP) problem in order to reduce online computation time. Moreover, field programmable gate array (FPGA) is employed to implement the proposed MPC method. Experimental results demonstrate that the proposed method could measure the EIS of fuel cell effectively.