Water Electrolysis: Stack Curves; Degradation and System Efficiency

Water electrolysis projects are bought on stack efficiency and delivered on balance of plant. The stack vendor quotes a number per normal cubic metre; the project discovers the cost of water treatment, rectifiers and compression months later.

This note is for project buyers and engineering teams specifying electrolysers. It covers what the efficiency figure hides, how to read a stack datasheet against real operating hours, and the acceptance tests that reveal whether a system will hold its curve.

The market context sits in this analysis of the water electrolysis market, which tracks the shift from demonstration capacity toward industrial scale.

Stack efficiency is quoted at a single point, and a real plant spends its life running along a curve

A single efficiency number describes one load point, usually near nominal. Real operation spends time at part load following renewable input, where the specific energy consumption is different.

Ask for the curve across the operating window, not the point. A system that is excellent at nominal and poor at low load is a poor match for an intermittent supply.

Degradation rate decides the lifetime cost

Performance falls as membranes and catalysts age, and the rate matters more than the starting efficiency. Two stacks with the same first-day number diverge sharply over five years.

Require a degradation rate per thousand hours and a definition of end of life. Without both, the warranty has no measurable trigger.

Water quality is a process requirement, not a utility detail

Electrolysers need water at a purity that ordinary plant supply does not reach. The polishing train, its consumables and its downtime belong in the project budget from the start.

Ask what the stack tolerates and what the vendor recommends, then design to the stricter figure. Water problems show up as stack problems and the warranty conversation is unkind.

Rectifier and power electronics set the real efficiency

The stack sees DC, and the conversion from AC carries losses that the stack datasheet excludes. Transformer, rectifier and cabling losses belong in the same calculation.

Ask for system efficiency at the connection point, not stack efficiency at the terminals. The gap between the two is where project business cases quietly fail.

What to test on arrival and at commissioning

Start with the documentation package: pressure test certificates, leak test records, and the membrane conditioning record. Then verify the stack against a performance test at defined temperature, pressure and current density.

Run the test at more than one load point and record the specific energy consumption at each. A single-point test at nominal is what the vendor expects; a two-point test tells you about the shape of the curve.

Record everything with the instrument calibration. A performance number without a calibration trace cannot be defended in a warranty discussion two years later.

The electrolysis cell sits inside a wider chemical process

Chlor-alkali plants have run industrial electrolysis for decades, and much of the operating discipline transfers: brine purity, membrane care, cell voltage monitoring and the habit of trending cell performance rather than reacting to failures.

The industrial base behind that chemistry is described in this overview of process chemicals and refining supply, which covers the sectors where electrolysis has been in continuous use.

Balance of plant is where projects overrun

Compression, drying, storage and the gas analysis train are usually bought separately and scheduled separately. Each has its own lead time and its own commissioning requirements.

Map the interfaces before the purchase order, including pressure levels and purity at each boundary. Most integration problems are interface problems, not equipment problems.

Hydrogen purity affects the offtake contract

The purity specification follows the offtake, and the purification step that achieves it costs both capital and energy. A project designed around stack output and not around the gas specification will need equipment it did not budget.

Confirm the measurement method as well as the limit. Purity measured at the stack outlet and purity measured at the delivery point are different numbers.

Safety documentation is part of the deliverable

Hydrogen at pressure brings area classification, ventilation and detection requirements that shape the civil works as much as the process design.

Ask for the hazard review as a deliverable with the equipment, not as a separate study afterwards. The layout decisions it drives are expensive to reverse.

Spares and service strategy from day one

Stacks and membranes have long lead times, and a plant that waits until failure to order will be down for months. The spares list should be agreed with the performance guarantees.

Ask which components the vendor expects to replace within five years and at what price. That answer is a more honest total cost figure than the headline efficiency.

Load following is a controls problem before it is a stack problem

A stack that follows renewable input spends much of its time ramping. Ramp rate limits, minimum load and the number of start-stop cycles per year all belong in the specification.

Ask how many cycles the stack is designed for and what the minimum stable load is. A plant designed for baseload operation and run intermittently will age faster than its guarantee assumes.

Instrumentation determines whether the guarantee is enforceable

Cell voltage monitoring, temperature distribution and pressure differentials are what make a performance claim checkable. Without them, a degradation dispute becomes an opinion.

Specify the instrumentation with the stack, not as an option. Retrofitting monitoring into a commissioned plant is expensive and always happens after the first poor reading.

Factory acceptance testing versus site acceptance

The factory test proves the stack left the works in specification. The site test proves the system performs with the actual water, power and ambient conditions.

Budget for both, and define the acceptance criteria separately. A factory result quoted as a site guarantee is a common and expensive confusion.

Site conditions change the number the datasheet promises

Altitude, ambient temperature and cooling water availability all affect a system that was specified on standard conditions. A plant at elevation derates before it ever reaches nameplate.

Ask for the correction method and apply it to your own site data. A guaranteed figure that only holds at sea level is a figure you cannot claim.

Operator competence is a commissioning deliverable

Electrolysis plants fail on water chemistry and on response to alarms far more often than on the stack itself. Training delivered as a slide deck does not change that.

Require hands-on training with the actual control system, and record who attended. The first week of operation is where the difference shows.

FAQ: questions buyers ask

Is higher current density always better?

No. It reduces capital cost per unit of capacity but raises specific energy consumption and often the degradation rate. The right point depends on electricity price and load profile.

How long should a stack last?

It depends on the technology, the load profile and the water quality. The useful question is the guaranteed degradation rate and the defined end-of-life threshold.

Can I specify on stack efficiency alone?

Not if you are paying the electricity bill. System efficiency at the connection point is the number that determines operating cost.

What is the most common commissioning surprise?

Water quality and interface pressure levels between the stack and the downstream train. Both are cheap to fix at design stage and expensive afterwards.

Which test proves the guarantee?

A multi-point performance test at defined conditions with calibrated instruments, recorded against the contractual curve. Anything less is not evidence.

The pattern is consistent across these decisions: the stack is the visible purchase, while the plant is a set of interfaces, and projects that price the interfaces at design stage are the ones that hold their budget, while the ones that do not spend the first operating year renegotiating scope, and the second explaining why the efficiency figure in the business case has never appeared at the connection point.