Alkaline, PEM or AEM: choosing for variable load rather than for a datasheet

A typical comparison table ranks alkaline, PEM and AEM on specific energy at rated load, current density, operating pressure, footprint and capital cost per kilowatt. Every figure in it is quoted at a single steady operating point.

A plant coupled to solar spends a large share of its operating hours below half load, passes through the low-load region twice a day, and stops and starts on a daily cycle. A wind-coupled plant does something less predictable but no more steady. The load duration curve, not the rated point, is what the equipment actually experiences.

That does not make the datasheet wrong. It makes it incomplete for this application, and the missing part is where the operating cost and the degradation actually come from.

Minimum stable load. Often a gas quality limit rather than an electrical one, though not always. As current density falls, hydrogen production falls with it, but hydrogen permeation through the separator or membrane does not fall proportionally. The hydrogen fraction in the oxygen stream therefore rises as load drops, and the minimum load is often set where that fraction approaches its safety limit, conventionally a margin below the lower explosive limit. It is whichever validated limit binds first, though, and thermal management, electrolyte circulation, gas separation, the rectifier's own turndown and the OEM's stated operating window can each get there before crossover does. A technology with a lower achievable minimum load captures more of a solar morning and evening.

Cold start and ramp. Electrical response is fast in all three technologies, often faster than the power supply feeding them. The real constraints are thermal, because stacks have an efficient operating temperature and reaching it takes time and energy, and gas quality, because purity has to be established before product gas can be accepted. A start from cold and a restart from warm standby are very different events and are often quoted interchangeably.

Degradation per cycle. Every technology pays something for start-stop cycling, through different mechanisms, and the size of that cost is what determines whether idling overnight is cheaper than shutting down. It is seldom stated in a datasheet, which is precisely why it is worth asking about explicitly.

Crossover behaviour across the load range. Related to minimum load but not identical. What matters operationally is how quickly the hydrogen-in-oxygen reading moves as load drops, because that determines how much usable turndown exists before the safety system intervenes, and how much margin remains as the separator ages.

The mature technology, with the longest industrial track record and the lowest cost per kilowatt at scale. Uses a liquid alkaline electrolyte and nickel-based electrodes, and conventional designs avoid platinum group metals, which matters for cost and for supply chain exposure. Catalyst and coating materials vary by product, so confirm them for the design being compared.

Its characteristic weakness under variable load is the low end. Alkaline systems have historically had a higher minimum stable load than PEM, driven by crossover in the diaphragm and by the need to maintain electrolyte circulation and temperature. Modern pressurised alkaline designs have improved this considerably, and the range quoted by suppliers now varies widely, which is itself a reason to ask for the number rather than assume it.

Cold start is slower, because the electrolyte has thermal mass and has to be brought to temperature. For a plant that stops every night, that time and the energy behind it are a recurring cost rather than a commissioning detail.

A solid polymer electrolyte, higher current density, a smaller footprint for the same output, and the ability to operate at meaningful differential pressure, which can simplify downstream compression.

Its advantage under renewable coupling is the low end and the transient response. PEM stacks generally hold acceptable gas quality further down the load range and respond quickly, which suits a plant chasing a variable resource.

The costs are capital and materials. PEM depends on iridium and platinum, which carries both a price and a supply concentration. It is also less tolerant of feedwater quality departures, and the consequences of getting that wrong are expensive rather than merely inefficient.

The newest of the three, and the one most often misread. AEM aims to combine the membrane architecture that gives PEM its turndown and response with alkaline chemistry that avoids platinum group metals. If it delivers on both, it addresses the main weakness of each of the others.

The open question is durability at industrial scale and duration. Membrane and ionomer stability under alkaline conditions is the active research problem, unit sizes are smaller, and the operating record is shorter than a project financier is usually comfortable with.

That does not make it the wrong choice. It makes it a choice that should be made with the degradation assumption treated as uncertain, with a measurement plan behind it rather than a supplier curve, and with the commercial terms reflecting that uncertainty.

Take your expected generation profile and turn it into a load duration curve. Then ask each supplier the same questions against that curve rather than against their rated point.

What is the minimum stable load, and what sets it. What is the hydrogen-in-oxygen fraction at that load, when new and at end of guaranteed life. What is the start time from cold and from warm standby, and what does each consume. What is the effect of a start-stop cycle on guaranteed stack life, and how is a cycle defined. What is the specific energy at 30, 50 and 100 per cent load rather than at rated. What is excluded from the performance guarantee if the plant follows this curve.

The last question is usually the most revealing, because it is where the supplier states which parts of your intended operation they are unwilling to stand behind. If the answers are hard to obtain before purchase, they will not become easier to obtain during a dispute, which is a subject covered in more depth on the performance guarantee page.