INTERNATIONAL JOURNAL of RENEWABLE ENERGY RESEARCH W. Ben SALEM et al., Vol.11, No.1, March, 2021 Robust Diagnosis of a Fuel Cell Pumping System by Bond Graph Modeling Approach Wafa Ben SALEM *‡, Houssem CHAOUALI*, Dhafer MEZGHANI*, Abdelkader MAMI* *Department of Physics, Faculty of Sciences of Tunis, UR-LAPER (UR17ES11), University of Tunis El Manar, 2092 Tunis, Tunisia (wafa.bensalem@fst.utm.tn, houssem.chaouali@fst.utm.tn, dhafer.mezghanni@gmail.com, abdelkader.mami@fst.utm.tn) ‡ Corresponding Author; Wafa Ben SALEM, University of Tunis El Manar, 2092 Tunis, Tunisia, wafa.bensalem@gmail.com Received: 03.01.2021 Accepted:24.02.2021 Abstract- This article presents works relized to verify the efficiency of a developed diagnosis system dedicated for a PEM Fuel Cell pumping system. The studied system is composed of PEM fuel cell powering an electrical pumping system via an electronic adaptation stage which consists of a cascade-connected DC-DC boost converter and a voltage inverter. The diagnosis system is developed based on Bond Graph approach which allows to model the studied system graphically in order to facilitate to investigate its working performances by simulating the developed model under the 20-Sim software. Obtained results show an efficient fault detection for different simulation scenarios carried out by imposing both mechanical and hydraulic faults to the system to test the reponse of the presented diagnosis algortihm. Keywords PEM Fuel Cell generator; Motor Pump; Static Converters; Diagnosis System; Bond Graph; 20-Sim Software. 1. Introduction The field of energies, mainly based on fossil resources, is expected to undergo a strong evolution due to the problems posed by the massive exploitation of these fuels. Environmental degradation and declining reserves, as well as new stand-alone or portable applications, are spurring the development of new energy technologies. Hydrogen, one of the ideal energy carriers to replace fossil fuels in the long term. The fuel cell, an electrochemical element allowing to produce electricity from hydrogen, is one of the best ways to use this energy vector in order to supply electric charges. Thus, fuel cells are currently experiencing a renewed interest, both industrially and in research. Industrialists from different sectors (electronics, automotive, heating, pumping, etc.) are investing in the development of this technology with low emissions of harmful gases and low noise pollution. The fuel cell pumping system is the ideal solution for supplying water in isolated areas that are not connected to the electrical grid. A fuel cell-based pumping installation is generally made of these different parts: - a fuel cell: many types of fuel cells have been developed. Each one presents its advantages and withdrawals compared to other types. Researches are mostly interested with Proton Exchange Membrane (PEM) type for its portability and temperature working range advantages. - A pumping unit: Electrical pumps have become the most used ones nowadays. An electrical pumping system is generally made of two parts (electrical motor and a pump). - Static converters: a PEMFC is a DC generator and they are generally used for low voltage applications. Thus, a DC- DC Boost converter is generally used to adapt the PEMFC voltage to the voltage level required by the pumping motor. Also, a voltage inverter is used because the most used motors for pumping systems (Induction motors) are generally AC type motors. From another hand, the importance of diagnosis systems in such applications resides in alarming supervisors in case of degradation in working performance of a single part of the system which may be helpful to avoid a full system shutdown or to take necessary precautions before stopping the system for repair. For example, in a water pumping system, this can help humans to store necessary water quantity before the system completely stops or even to avoid water waste in some cases by repairing the faulty part. different types of detection algorithms dedicated to physical systems have been designed by researchers in the Automation community [10-13]. These methods are basically arranged in two categories: model-based and model non model-based methods and each one of these categories is split into two sub-categories (qualitative and quantitative methods). Methods based on Bond Graph (BG) techniques are considered as model-based qualitative ones. BG modeling