An electrical power system must maintain the balance between generation and consumption at every instant. When a large generation unit disconnects, or when renewable production changes unexpectedly, the system frequency begins to deviate from its nominal value. This deviation must be contained rapidly, since an uncontrolled event can activate protection schemes, disconnect additional equipment and, in extreme cases, contribute to a wider system disturbance.
Traditionally, a significant share of this stabilising behaviour was provided by the rotating masses of conventional power plants, including hydro, gas, coal and nuclear units. The increasing integration of wind and solar generation changes this picture. These resources are mainly connected through power converters and, unless specifically controlled to do so, do not naturally provide the same inertial response. Accordingly, future power systems require new and controllable sources of flexibility.
Hydrogen electrolysers are interesting in this context because, from the grid point of view, they are large electrical loads whose consumption can be adjusted. During an underfrequency event, an electrolyser can temporarily reduce its absorbed power. A reduction of 10 MW in consumption has, in the first moments of the disturbance, the same balancing effect as an additional 10 MW of generation. During an overfrequency event, the opposite response may be provided by increasing consumption, provided that sufficient operating margin is available.
Polymer electrolyte membrane, or PEM, electrolysers are particularly suitable for this type of operation owing to their fast ramp-up and ramp-down capabilities. This opens the door towards the provision of services such as Frequency Containment Reserve, in which power changes proportionally with the frequency deviation, and Fast Frequency Reserve, in which a pre-allocated response is activated shortly after a severe disturbance.
However, flexibility does not mean unlimited operation. The electrolyser stack must remain within acceptable current, voltage, temperature and pressure limits. Hydrogen production also changes when the electrical power is modified, although a hydrogen buffer can decouple short electrical events from downstream users. Moreover, the plant must preserve operating headroom. An electrolyser already operating at full power cannot further increase its consumption, while a unit operating close to its minimum load has little room to reduce it.
The electrolyser stack operates with direct current (DC), while the public grid normally operates with alternating current (AC). Therefore, a power conversion chain is required between the two systems. This interface is responsible for determining how rapidly, cleanly and safely the electrolyser can respond to a grid request.
One of the main concerns is the current ripple supplied to the stack, meaning the small and fast variations superimposed on the intended current. Excessive ripple can increase power consumption, reduce efficiency and accelerate degradation. Consequently, the objective is not simply to move from one power setpoint to another. The transition must be carried out while limiting overshoot, avoiding instability and respecting the electrochemical constraints of the stack.
To address this challenge, the research developed a three-leg interleaved buck converter. The total current is shared between three parallel paths, while the switching signals are phase-shifted so that a large part of the individual current ripples cancel at the electrolyser terminals. In addition, an adaptive lead-lag controller recalculates its parameters according to the operating point, instead of relying on a fixed tuning that is only adequate around one current level.
The experimental results showed current ripple below 1%, a consistent transient response over the operating range, and improved behaviour during a 50% reduction of the input voltage when compared with a conventional PI controller. The converter also maintained the required output quality under current sharing and phase failure tests. These results are relevant because the provision of a grid service is only useful if the electrical response does not create additional stress or shorten the lifetime of the hydrogen equipment.
The frequency-support capability was validated using a real 5 kW PEM electrolyser, a custom-built converter and a Power-Hardware-in-the-Loop (PHIL) platform. In this setup, the converter and electrolyser were physical equipment, while a future low-inertia representation of the Iberian Peninsula and Continental European power systems was simulated in real time. The reference disturbance represented a large 3 GW loss of renewable sources generation.
The electrolyser successfully provided both Frequency Containment Reserve (FCR) and Fast Frequency Reserve (FFR) at different provision levels. Its response improved the minimum frequency (nadir) reached after the event and reduced the initial rate of change of frequency. The response also reduced the temporary need for imported balancing power through the simulated interconnection with the Continental Europe power system area.
It is important to notice that a faster command does not always produce a better system response. When the requested reserve exceeded the ramp rate capability of the stack, increasing the reference resulted in limited additional benefit. This confirms that ancillary services must be designed around measured plant capabilities and protection limits, rather than around idealised assumptions regarding the speed of the electrolyser.
A flexible load should also remain controllable during voltage disturbances. A separate converter architecture study investigated low voltage ride through (LVRT), which is the ability of the system to remain connected while the grid voltage temporarily decreases. In the simulated test, the input voltage was reduced to 20% of its nominal value for half a second. The proposed intermediate conversion stage largely isolated the electrolyser stack from this disturbance and maintained the required conditions at its terminals.
Nevertheless, maintaining the same power during a deep voltage reduction comes with an important consequence: the grid current can increase significantly. In the AC test, the current rose to approximately seven times its pre-fault value. Therefore, fault ride-through cannot be considered only from the perspective of protecting the electrolyser stack. Current limiting, semiconductor ratings, cable limits, protection coordination and, when required, temporary power reduction must also be included in the design.
The technical capability is becoming increasingly credible, but commercial deployment requires more than a fast response in a laboratory test. Megawatt scale plants must demonstrate the same behaviour over long operating periods, under realistic thermal conditions and with the full balance of plant (BoP) equipment active. In addition, system operators require clear procedures for verifying response speed, available headroom, recovery behaviour, power quality and the long term effect on stack degradation.
Grid codes and electricity markets must also recognise controllable demand as a provider of ancillary services. The service design should reward useful flexibility without creating inefficient operating incentives, such as maintaining excessive unused capacity only to participate in a reserve market. At a larger scale, the control of fleets of electrolysers must also be coordinated to avoid thousands of units reacting in exactly the same manner and recovering their consumption at the same instant.
Finally, the project economics should account for both outputs of the plant: hydrogen and electrical flexibility. Ancillary-service revenue will not transform an uneconomic hydrogen project into a viable one. However, it can improve asset utilisation, reduce balancing costs and facilitate the integration of renewable generation. The actual value will depend on the grid location, market rules, converter capabilities and the operating freedom allowed by the hydrogen-production process.
Electrolysers should not be considered only as passive consumers waiting for periods of low cost renewable electricity. When properly designed, they can absorb surplus generation, reduce consumption during shortages and remain controlled during grid disturbances, while continuing to perform their main function of producing hydrogen.
This does not mean that electrolysers can replace every battery, conventional generator or network reinforcement. It means that future hydrogen plants can be designed as an active part of the electrical system rather than being connected to it as an afterthought. In this manner, the same infrastructure can contribute to two objectives: producing low-carbon hydrogen and supporting the stability of the grid that supplies it.
[1] A. M. Elhawash, R. E. Araújo, and J. A. Peças Lopes, “Frequency support from PEM hydrogen electrolysers using Power-Hardware-in-the-Loop validation,” Int. J. Hydrogen Energy, vol. 175, Art. no. 151203, 2025, doi: 10.1016/j.ijhydene.2025.151203.
[2] A. M. Elhawash, A. S. Hussein, R. E. Araújo, and J. A. Peças Lopes, “Low ripple adaptive lead lag current controlled interleaved buck converter for PEM hydrogen electrolyzers,” Control Eng. Pract., vol. 174, Art. no. 107029, 2026, doi: 10.1016/j.conengprac.2026.107029.
[3] A. M. Elhawash, R. E. Araújo, and J. A. Peças Lopes, “A new adaptive lead-lag control scheme for high current PEM hydrogen electrolyzers,” in Proc. 2023 IEEE Vehicle Power and Propulsion Conf. (VPPC), Milan, Italy, 2023, pp. 1–6, doi: 10.1109/VPPC60535.2023.10403130.
[4] A. M. Elhawash, R. E. Araújo, and J. A. Peças Lopes, “A new design for an electrolyzer power converter architecture capable of fault ride through,” in Proc. 2025 IEEE Kiel PowerTech, Kiel, Germany, 2025, pp. 1–6, doi: 10.1109/PowerTech59965.2025.11180384.
