Proton Exchange Membrane Fuel Cell Modeling And Simulation

Proton exchange membrane fuel cell (PEMFC), one type of power gener-ators using renewable energy source with the advantages of clean generation,high efficiency, stable performance, and wide application, is regarded as a keycomponent of green energy system in the21stcentury. The model of PEMFCis the basis of the entire fuel cell system design, control and optimization, be-ing of great significance for theoretical research and engineering applications.Against the background of Integrated Automation of the Complex SystemsInformation and Control,one of the subprojects in the3rdperiod of Project211in Beijing University of Chemical Technology, the research of this thesis onthe focus of the Hydrogen Reactor Power System Integration and IntegratedControl Platform, covers the PEMFC models based on empirical formulae,neural network and adaptive neuro-fuzzy inference system (ANFIS) to man-age an in-depth study on fuel cell modeling approaches, proceeding the basicwork for the implementation of control algorithms and optimization strategieson the platform. The main content is as follows.First, a PEMFC modeling approach based on empirical formulae is dis-cussed, which includes analyzing the electrochemical mechanism of PEMFC to define the influence of each factor on fuel cell performance, interconnectingthe performance indices with the major factors through the empirical formulae,using least squares fitting to obtain the undetermined coefficients, and simu-lating the fuel cell system on the basis of the developed empirical model.Second, two PEMFC neural network modeling approaches are investi-gated. After accomplishing of the PEMFC static neural network model, a kindof dynamic neural network model is constructed by introducing the sensitivityanalysis based on Fourier transform to evaluate the neural network structureand optimizing the number and weight values of neurons in the hidden layeraccording to the evaluation to realize network dynamic adjustment.At last, a PEMFC modeling approach based on ANFIS is design. By em-ploying data clustering technology and hybrid learning algorithm, the model isbuilt on a small data set and then used to verify the feasibility and advantagesof ANFIS modeling approach in the field of PEMFC model research.This thesis makes great efforts to PEMFC model research by implement-ing and comparing4types of modeling approaches. The simulation resultsshow all of them are feasible, which cover the range from simple analysis tothe entire input-output relationships mapping of the system, could contributeto fuel cell model development, and also lay the foundation for further studyon the PEMFC design, control and optimization.