The global transition towards sustainability has accelerated the demand for hydrogen as a clean energy resource. Its primary use is in industrial processes to replace grey hydrogen, but it also plays a growing role in hard-to-electrify sectors such as heavy transport, long-duration energy storage, and synthetic fuel production for aviation and shipping. A key technology enabling this transition is the electrolyzer, which is the dominant method for producing green hydrogen by splitting water using electricity.

Since electricity is a core input to the electrolyzer, the interface to the electrical grid plays a crucial role. This interface must not only supply power with high reliability and low losses, but also handle rapid changes in power demand arising from fluctuating hydrogen production requirements and variable grid conditions. Furthermore, it must offer precise control to manage voltage levels, limit harmonic distortion, and adapt to grid-side constraints in real time. Simultaneously, the electrolyzer itself can operate as a controllable and responsive load, capable of providing ancillary services such as frequency regulation and absorbing surplus renewable energy during periods of overproduction.

A promising solution to meet these interface requirements is the Modular Multilevel Resonant (MMR) converter. This architecture combines the scalability and modularity of conventional modular multilevel converters (MMCs) with resonant soft-switching techniques that significantly reduce switching losses. As a result, MMR converters are well suited for compact, high-power, and high-efficiency applications such as grid-connected electrolyzers.

A recent contribution by Li et al. presents a complete real-time simulation of an MMR-based Solid-State Transformer (SST) interfaced with an electrolyzer load. Their study uses an OPAL-RT platform to model both the power converter and the electrolyzer plant in real time, validating a continuous modulation method based on continuous modulation index control. While their work demonstrates promising performance, the controller is executed entirely within the real-time simulator, where plant and control share an idealized computational environment. In that setting the communication link behaves as if it were instantaneous, the timing is perfectly deterministic, and none of the practical constraints of embedded execution appear. Real hardware introduces round-trip communication delay, packet-level timing variability, and limits on the achievable update rate. These effects shape the behavior of the controller and determine whether a control method that performs well in simulation can also function on real hardware.

To address this gap, this thesis implements the continuous modulation index control strategy on a Xilinx-based MultiProcessor System-on-Chip (MPSoC) and evaluates its feasibility in a hardware-in-the-loop setup in which the control runs on external embedded hardware and the electrolyzer system is simulated on an OPAL-RT platform. This configuration exposes the controller to realistic timing and communication conditions, enabling a more representative assessment of the method. The control algorithm and system model are discussed in detail in the Literature Review and Theory sections.