Iterative Fuzzy Modeling of Hydrogen Fuel Cells by the Extended Kalman Filter

Abstract

Hydrogen economy is one of the recently opened alternatives in the field of non-polluting energy. Hydrogen fuel cells show high performance, high reliability in stationary applications and minimal environmental impact. To increase the efficiency of the hydrogen fuel cell it is very important to have a good model to predict its dynamic behavior. In addition, this model must be able to adapt iteratively to the changes that occur in its performance due to operating conditions and even to the degradation through its lifespan. This paper presents the application of an iterative fuzzy modeling methodology based on the extended Kalman filter applied to a real hydrogen fuel cell. Two algorithms based on the Kalman filter will be compared with the well-known backpropagation algorithm from three different initializations: by uniform partitioning, subtractive clustering and CMeans clustering. The used data have been collected during the actual operation of a real 3.4 kW proton exchange membrane fuel cell. As the article experimentally shows, the Takagi-Sugeno type fuzzy model allows to create a very accurate nonlinear dynamic model of the fuel cell, which can be very useful to design an efficient fuel cell control system