A fuzzy inference system approach for diagnostics and prognostics of polymer electrolyte fuel cells

Climate change is one of the most important global issues currently and is one of this generations’ greatest challenges. With the world’s growing population, energy demand continues to increase which leads to increased carbon emissions due to primary energy sources coming from fossil fuels. Alternative sources of energy and low carbon technologies are needed to meet this demand in a sustainable and commercially feasible manner [1]. Fuel cells can provide an efficient means of energy conversion for electricity production without CO2 emissions and are considered as a promising technology for moving towards a sustainable, low carbon future. Polymer Electrolyte fuel cells (PEFCs) are one of the main types of fuel cells which have received much research and investment due several advantages including high power density, versatility, and scalability for a range of applications. To achieve global consumer adoption, fuel cells must be able to provide equivalent or improved services compared to conventional technologies. For example, in automotive applications, the lifetime of an internal combustion engine (ICE) vehicle has been estimated to range between 11 to 13 years [2]. Therefore, fuel cells in electric vehicles (FCEV’s) are expected to have similar lifetimes. One of the main challenges, however, to widespread commercialisation of PEFC technology is their reliability and durability. Fuel cells are required to perform under different operating and environmental conditions over a certain number of years with acceptable performance losses. Lifetime system efficiency losses below 10% and a degradation rate of 2-10μV/h are considered to be acceptable for the majority of applications [3]. Current fuel cell lifetimes are in the range of 2,500 to 3,000 hours which is approximately 6 to 8 years assuming 1 hour of usage per day. The U.S Department of Energy (DoE) have outlined lifetime targets of 5,000hrs for transportation applications and lifetime targets for stationary applications at 40,000hrs [1]. In order to achieve these targets, improving reliability and durability is crucial. Diagnostics and prognostics are so important for improving these attributes as the capability of diagnosing degradation and predicting lifetime can provide highly valuable information. For example, the early detection of operating conditions deviating outside of nominal ranges and diagnosing how this deviation impacts performance can inform the user to adjust operation to avoid or minimise component degradation therefore reducing potential downtime and improving  reliability and durability. Prognostics can provide the user with information on predicted degradation rates based on current operating conditions and predict future failures and remaining useful life. This can enable appropriate operation and maintenance schedules to be developed to avoid unscheduled failures and maintenance thereby improving reliability and durability. By utilising diagnostics and prognostics to improve the reliability and durability of fuel cells, this increases the technology readiness level and accelerates the commercial feasibility of the technology. This thesis presents a fuzzy inference system approach for diagnostics and prognostics of polymer electrolyte fuel cells. The fuzzy inference system (FIS) was validated experimentally under steady state and dynamic fuel cell operating conditions to simulate the environment which a fuel cell would experience in either automotive or stationary applications. The novel diagnostic and prognostic FIS involves establishing the relationships between fuel cell operating conditions with the consequential impact on degradation phenomena, state of health, and ultimately fuel cell lifetime. This is performed using a knowledge-based fuzzy logic approach to enable simplified analyses with linguistic classifications for input and output parameters. The developed fuzzy inference system contributes to the existing literature because it can provide low complexity, in-situ diagnostic and prognostic capabilities and has the potential for online application. These attributes can enable improved fuel cell reliability and durability in addition to improved understanding and analysis of fuel cell state of health for non-specialist users. This would be become particularly beneficial as fuel cell power applications become more available to a wider non-technical audience. The combination of diagnostic and prognostic capabilities can facilitate online health monitoring and enable improved decision-making for developing maintenance strategies which can increase fuel cell reliability and durability. Improving these can accelerate the utilisation of fuel cells as a viable zero carbon alternative to fossil fuel-based energy sources thereby increasing the progress towards a low carbon industry and society to reduce our impact on climate change and the environment for future generations.