Pemfc Modeling And Identification Based On Fractional Calculus

The proton exchange membrane fuel cell (proton Exchange membranefuel cell, referred to as PEMFC) is the representative of the modern cleanenergy. The fuel cell technology is also undergoing rapid changes. A suitablemathematical model of proton exchange membrane fuel cell is essential in theanalysis of fuel cell-based power system. So far, various forms of fuel cellmodels have been proposed to describe the static or dynamic characteristics ofthe fuel cell under a certain working conditions. However, most of the modelsare of the shortcoming of complex parameters and structure, which makes themodel applicability poor. Although some these models can be used for the fuelcell control system design, they are still inadequate in the description of thefuel cell electrical characteristics. Furthermore, an important factor that has acritical impact on the fuel cell electrical characteristics—double charge layer,can be described by a fractional model. As a starting point, a fractional-ordermodel can be established in view of the inherent advantages of the fractionalorder to the integer model. The main content of this paper is as follows: 1. Three methods to establish the fuel cell fractional-order transfer modelhave been proposed: based on the internal mechanism, based on systemstructure and based on the fuel cell equivalent circuit. The models based onthe internal mechanism and equivalent circuit model has a fixed form whilethe one based on the model of the system structure is similar to the analysis ofelectrochemical impedance spectroscopy method, the resulting modelstructure is not unique. In order to determine the model parameter values, thefuel cell actual input and output values can be used and then the problem isturned into a the fractional order system identification problem.2. An improved NLJ algorithm has been proposed to completeidentification of the fractional system parameters. The improved algorithm notonly eliminates the impact of the initialization values of system parameters onthe identification results, which means in the case of random initialization, theidentification efficiency has been significantly improved and the stability ofthe recognition results after continuous identification calculation is also undera very good guarantee. The correctness and validity of the improved NLJalgorithm have been verified with the simulation identification with a set ofknown parameters and the optimal curve fitting validation with a set ofunknown parameters.