What is electrolyzer system integration?

Electrolyzer system integration is the engineering work of turning a stack into an operable plant: adding water treatment, power electronics, cooling, gas separation and purification, safety systems and control, then connecting that unit to compression, storage, dispensing, the grid and the site. The stack typically accounts for a minority of the engineering effort; the surrounding system and its interfaces account for most of it.

What is actually inside a containerized PEM electrolyzer

A containerized PEM electrolyzer is a complete process plant. The stacks are one component among ten.

Tap water enters a water purification stage and is buffered in a deionised water storage tank. A transformer and rectifier convert grid AC into the low-voltage, high-current DC the stacks need. Inside the stacks, water splits into hydrogen and oxygen; both gases leave carrying water, so each side needs a separator. The hydrogen then passes through purification and gas analysis to reach the specified purity — 99.999% in FEST’s case. Around all of this sit a closed-loop cooling system, a nitrogen and instrument air supply for purging and valve actuation, and the measurement and control equipment that runs the whole sequence automatically.

Physically this is split across two 40-foot containers: a power container holding the rectifier room, medium-voltage transformer and utility room, and a power-to-gas container holding the stacks, pumps, heat exchangers, ultrapure water storage and gas purification. From 5 MW upwards the transformer moves outside for space reasons.

Two stack suppliers is a deliberate choice: it gives a second source and covers two product classes. It also reflects a wider position — as a systems integrator rather than a stack developer, the company can take the best available stack from the market. A 2.5 MW stack generation is on the roadmap and is expected to become the industry standard, but roadmap and buildable product are not the same thing.

Why an efficiency figure is meaningless without its operating point

Three numbers for the same electrolyzer can differ by a wide margin depending on where and how they were measured.

FEST states below 5.05 kWh/Nm³, equivalent to 56.7 kWh/kg, and attaches three conditions to it: at nominal load, at beginning of life, and measured across the whole battery limits of the system. Change any one of those and the figure improves. DC stack efficiency alone is considerably better because it excludes the rectifier, pumps, cooling and purification. Running at 30–40% of nominal load is also better, because cell voltage falls with current density. Beginning-of-life figures are better than end-of-life figures by definition.

The practical consequence for anyone evaluating offers: an efficiency number without a stated operating point, lifetime assumption and measurement boundary is not a specification. It is marketing.

Frankfurt: what changes in the step to megawatt scale

Moving from a few hundred kilowatts to megawatt scale is a change in safety and automation category, not a change in size.

FEST’s entry into the MW class was a 5 MW plant for Infraserv Höchst in Frankfurt, built as two 2.5 MW units producing 1,000 Nm³/h, or 2,160 kg per day, and commissioned in Q1 2024. The hydrogen refuels fuel-cell trains in the Rhein-Main region. The plant replaced an older alkaline electrolysis unit inside an existing chemical park — which meant the surrounding infrastructure, permitting basis and industrial safety culture were already in place.

That turned out to matter more than the technology step. The challenges were MW-scale PEM integration, residential proximity, availability and a much higher degree of automation than smaller systems require. The lesson the team drew: new hydrogen technology becomes manageable faster when it is embedded in proven industrial processes rather than built on a greenfield with everything unfamiliar at once.

Hellesylt: the location determines the system design

Site conditions do not just constrain a standard product — they reach back into the electrical design.

The 3 MW plant for Norwegian Hydrogen AS in Hellesylt, on the Geirangerfjord, produces 600 Nm³/h (1,281 kg/day) and was commissioned in Q2 2024. FEST acted as general contractor for the full scope: electrolyzer, storage, compression, trailer filling, a 350/700 bar refuelling station and heat recovery. It supplies hydrogen for vessels in a fjord where fossil-powered shipping is being phased out.

The grid forced a change in power electronics

Norwegian fjord grids are weak — low connected power and low short-circuit power compared with a central European grid. The original thyristor-based rectifier could not meet the grid harmonics requirement, so the design switched to an active front-end rectifier. The consequence is worth noting: the electrolyzer now improves grid stability for everyone connected at that point, rather than degrading it.

Climate, logistics and non-EU certification

The plant operates in a −30 °C environment, under half a metre of snow on the container roofs during installation, in a remote location, and under Norwegian national codes outside the EU regulatory framework. The waste heat was put to use: it feeds an underfloor heating system that keeps the site surface clear, because nobody wants to shovel snow to keep trailer access open.

Bamberg: a stack is not a plug-and-play component

Introducing a new stack platform is a system engineering project, not a procurement decision.

The 2.5 MW outdoor system at Robert Bosch in Bamberg, commissioned in Q4 2025, was the first plant built on Bosch’s Hybrion PEM stacks and launched the gEL 500 platform. It produces 500 Nm³/h (1,068 kg/day) and supplies the factory — including the endurance test benches for mobile fuel-cell stacks, so the hydrogen is partly consumed by the fuel cells whose electricity returns to the process.

The unusual part of the setup was that Bosch was simultaneously the stack supplier and the customer. The engineering work was in the new stack interfaces, the dynamic behaviour of the system, and the interaction between stack and balance of plant — all of it converging on a safety and explosion-protection concept that had to be certified as an integrated unit.

Stuttgart: projects fail between the interfaces

The single most expensive planning mistake is laying out a plant with placeholder equipment instead of equipment that exists.

The Green Hydrogen Hub for Stadtwerke Stuttgart is 3 × 3 MW, built with sister company Maximator Hydrogen, with commissioning scheduled for Q2 2027. It combines a battery storage system, electrolysis, three tonnes of hydrogen storage, trailer filling and a pipeline connection feeding industrial offtakers in Stuttgart — arguably the most complete distribution setup of any European hydrogen hub currently in engineering.

It is also built on a very tight plot, split across four contract packages, with multiple stakeholders and numerous handovers. The pre-planning had been done vendor-neutrally, as an ideal-world tender process would suggest: blocks placed on a layout without reference to equipment that actually exists. Real equipment has real dimensions, real service clearances and real safety distances.

The countermeasure FEST has standardised since the Norwegian project is a single 3D model covering every engineering lot — electrical, piping and civil — so collisions, logistics routes and service access are resolved on screen rather than on site. In Stuttgart that model is now also being used to reorganise engineering across the lots after the fact.

The permitting risk nobody prices in

Planning with hypothetical equipment carries a second cost. If a permit is applied for on the basis of a system that is then replaced by a different one, the change may be treated as a minor amendment — or it may require the whole approval process to be repeated with every authority involved. In Germany, that can add many months to a schedule that was already the critical path.

The overarching control layer that most tenders forget

An electrolyzer, a compressor, a storage system and a refuelling station each arrive with their own controller. Somebody has to sit above them.

FEST’s architecture puts an aggregation PLC (Level 1) above the sub-controls for battery storage, each electrolyzer, gas analytics and the refuelling or trailer-filling station, with a business management platform (Level 2) above that handling electricity prices, direct marketing, grid operator signals and ERP integration. The operator gets one human-machine interface for the whole plant: start, stop, set electrolyzer load, run a trailer filling, see the state of every package unit.

This layer is routinely absent from tender documents. Early project phases concentrate on technology, business case and capital cost; how the plant will actually be operated for the next 15 to 20 years gets deferred. The result is a plant that works and an operating team that has to work across four unconnected screens to run it.

The same point applies to software generally. Asked what the industry underestimated most over the past decade, the answer was unambiguous: the complexity of the automation, by everyone involved — manufacturers included.

How electrolyzer plants scale to 20 and 50 MW

Scaling is achieved by repeating a proven module, not by designing a bigger one.

A 20 MW plant is built as four 5 MW gEL 1000 units with redundant auxiliaries, so a fault in one module does not take the plant down. At 50 MW the arrangement changes: five 10 MW blocks sharing centralised auxiliaries, which reduces the integrator’s scope per megawatt but requires the balance-of-plant to be bundled rather than replicated. Below that threshold, replication wins because each module is already type-tested.

The industrialisation behind that sits in the factory: four parallel assembly lines in Goslar, every unit tested and pre-qualified before it ships, and a scope split that leaves civil works and incoming medium-voltage supply with the customer while everything hydrogen-carrying stays under one contract. FEST calls this a package-unit EPCm approach, positioned between a single all-in EPC contract — which prices in every shared risk — and a self-managed multi-lot tender, which prices in every interface.

What it costs and how long it takes

Roughly €7 million total installed cost for a 5 MW system in Germany, and about 18 months from purchase order to operation.

The €7 million figure covers the complete installed system including site acceptance testing and trial run, not the ex-works equipment price. A 50 MW plant does not cost ten times as much — the on-site work scales sub-linearly — but the exact figure depends heavily on whether balance of plant is bundled or replicated.

On schedule, the manufacturing side is the predictable part: 12 months to factory acceptance test as standard, 8 months on an express track, and three months from delivery to a fully operational plant. The variance sits outside the factory, in construction, grid connection and permitting. One 2 MW project in Austria ran seven months from order to acceptance — but only because an approval was already in place and could be amended rather than reapplied for.

Watch a clip from the Webinar

Read next

Article based on a Mission Hydrogen webinar. Technical figures are the speaker’s field estimates, presented for general understanding rather than as design specifications.