Hydrogen is a sought after energy carrier within modern energy systems because it addresses a critical challenge faced today: decarbonizing the energy sector. When produced using renewable energy methods, hydrogen can be stored and utilized when other weather dependent renewable energy sources, such as wind or solar, are not available. Due to its long-term storage capability, hydrogen is particularly useful for addressing seasonal mismatches in energy supply and for providing grid backup during extended outages.

One approach to utilize hydrogen for energy generation is through a Proton Exchange Membrane Fuel Cell (PEMFC). This device uses hydrogen and oxygen (air) to generate electricity, with heat and water produced as byproducts. The zero-emission nature of this process makes it very compelling. However, PEMFCs operate optimally only within specific environmental conditions and struggle to respond to dynamic load changes, which reduces their feasibility in real-world applications. Environmental control, such as temperature-regulated enclosures, and electrical energy storage devices, including batteries and supercapacitors, help overcome these practical challenges and enable real-world implementation.

Despite this, PEMFCs remain highly complex systems that require precise control of hydrogen and oxygen (air) mass flow rates, their partial pressures, stack temperature, and overall energy management between the fuel cell and associated storage devices. This is where state-of-the-art machine learning control can be utilized to help transform PEMFCs into practical and robust energy solutions.

In this project, the control of PEMFC stacks will be investigated to optimize the efficient production of electricity. This will be achieved through a machine learning (ML)-based control strategy paired with Model Predictive Control (ML-MPC), which will serve as the primary control approach for the PEMFC. To ensure real-world feasibility, the control strategies will be experimentally implemented and evaluated. The focus of this LBP is the expirimental implementation of the ML-MPC developed in FES project T06-Q11.

The goals of this work are to maximize fuel cell efficiency, optimize system performance, and prolong stack life while operating in conjunction with a battery or supercapacitor. This will result in an operational renewable energy system capable of handling transient loads and meeting real-world energy demands. The outcomes of this work will establish a strong foundation for further development, with only environmental control - such as enclosure temperature regulation - remaining to be studied before application to a broader market, where the system must operate reliably under both cold winter conditions and hot summer environments.