Electrolysers: load follows wind, stacks age faster than planned

The electrolyser follows wind and sun

What happens. The electrolyser is fed from wind or solar plants, so its load swings with the weather. Stack temperature, gas flows and pressure are always catching up with the power.

What it costs. Lower efficiency at poor operating points and faster wear of the stacks. Energy is also lost when the plant is turned down to stay inside its limits.

Why the usual fix fails. A battery as buffer is expensive. Running at a fixed safe power gives up renewable energy, and rules of thumb for ramp rates stay cautious because the stack limits are not well known.

How a fixed-price remote diagnosis checks it. From your power and stack logs (load, voltage, temperatures, gas and water flows) I look at how load and temperature behave together. The fixed-price remote diagnosis tells you whether load control within the stack limits could raise efficiency and what data it needs.

Background: Economic predictive control

Stack temperature swings with load

What happens. When the load changes, the stack heats up or cools down faster than the cooling loop can follow. Temperature overshoots and sags, and operators watch it closely.

What it costs. Degradation of membranes and catalysts is faster at the temperature extremes, and efficiency drops when the stack runs cooler than it should.

Why the usual fix fails. A standard controller tuned for steady load reacts only after the temperature has moved. Wider alarm limits only hide the swing.

How a fixed-price remote diagnosis checks it. From stack temperature, coolant flow, load and ambient logs I check how the thermal behaviour responds to load steps. The fixed-price remote diagnosis tells you whether predictive thermal control, which acts ahead of the load, would keep the stack steadier.

Background: Predictive control of process plants

Gas crossover at low load triggers safety shutdowns

What happens. At low load, hydrogen can cross into the oxygen side or the other way round. The gas purity falls towards the alarm limit, and the safety system shuts the plant down.

What it costs. Lost production, restarts that wear the stacks, and a minimum load that is set high to be safe and so limits how far the plant can follow the power.

Why the usual fix fails. A purity analyser reacts to the result, not to the cause, and a higher fixed minimum load simply removes the operating range where you would like to run.

How a fixed-price remote diagnosis checks it. From purity readings, load, pressure and temperature logs I look at how crossover builds up. The fixed-price remote diagnosis tells you whether purity can be estimated ahead of the analyser and the minimum load managed instead of fixed.

Background: Computing unmeasured states for control

The ageing of the stack is unknown

What happens. The stack gets worse with operating hours, but nobody knows by how much. Voltage at a given current slowly rises, and the plant is run as if the stack were new.

What it costs. Replacement is planned by calendar or by a failure, not by condition. Stacks are either replaced too early or run beyond their useful efficiency, which raises the energy used per kilogram.

Why the usual fix fails. The manufacturer’s lifetime figure is an average under test conditions. Measuring one voltage trend is noisy, because load, temperature and pressure change it as much as ageing does.

How a fixed-price remote diagnosis checks it. From long voltage, current, temperature and load histories I separate ageing from operating effects. The fixed-price remote diagnosis tells you whether degradation can be estimated well enough to plan replacements and adapt operation.

Background: Control that adapts to drift and wear

Pumps, dryers and compressors are oversized

What happens. The plant around the stack, with its pumps, gas dryers and compressors, is designed for full load. Most of the time it runs at part load, and its own power use stays high.

What it costs. The plant’s own consumption eats into the efficiency of the whole chain, so the cost per kilogram of hydrogen is higher than the stack alone suggests.

Why the usual fix fails. Replacing equipment with smaller units costs money and limits full-load operation. Variable-speed drives help only if someone sets them well.

How a fixed-price remote diagnosis checks it. From the power measurements of each unit and the load history I show where the own consumption sits and when. The fixed-price remote diagnosis tells you whether plant-wide optimisation of the auxiliaries would save energy without touching the hardware.

Background: Economic predictive control

Frequent starts damage the stacks

What happens. With fluctuating power, the plant is started and stopped often, and held in standby in between. Each cycle takes the stack through temperature and potential changes it was not made for.

What it costs. The stack lifetime shortens, and with it the business case, since the stack is a large part of the cost of the plant.

Why the usual fix fails. A fixed minimum run time or standby rule is either too cautious, so that cheap power is lost, or too loose, so that the stack ages quickly.

How a fixed-price remote diagnosis checks it. From start and stop logs, standby times and voltage trends I check which cycles hurt most. The fixed-price remote diagnosis tells you whether an optimised start-stop strategy can trade power price against stack wear.

Background: Economic predictive control

Pressure swings at load changes

What happens. When the load steps up or down, gas production changes quickly and the pressure in separators and lines swings. Valves and compressors work against it with delay.

What it costs. Pressure trips and shutdowns, and stress on the stack from pressure differences between its two sides.

Why the usual fix fails. Tight standard controllers on each pressure loop fight each other, and loosening them lets the pressure run closer to the limit.

How a fixed-price remote diagnosis checks it. From pressure, valve and load logs I look at how the loops interact during load changes. The fixed-price remote diagnosis tells you whether predictive pressure control coordinated across the plant would keep pressure inside the limits.

Background: Predictive control of process plants

Control must scale from kilowatt to megawatt units

What happens. What worked on a small pilot stack behaves differently on a large unit. Gas volumes, heat flows and delays are bigger, and the dynamics are new.

What it costs. Commissioning takes longer than planned, the first months run with cautious settings, and surprises show up at the customer.

Why the usual fix fails. Copying the pilot’s settings fails because the large unit has other time constants. A new tuning round in operation costs production time on an expensive plant.

How a fixed-price remote diagnosis checks it. From pilot data, design documents and the specification of the large unit I check which dynamics change with size. The fixed-price remote diagnosis tells you what a control concept for the large unit has to cover and what to test beforehand.

Background: Predictive control of process plants

Storage and offtake are not coordinated

What happens. Hydrogen is produced when power is cheap or plentiful, but the customer takes it on another schedule. The storage is filled and emptied by simple level rules.

What it costs. Production is curtailed when the storage is full, and the customer is short when it is empty. Both cost money.

Why the usual fix fails. A larger tank costs investment. Fixed level thresholds do not know tomorrow’s power or demand.

How a fixed-price remote diagnosis checks it. From production, storage level and offtake logs, together with power price history, I check how often the rules fail. The fixed-price remote diagnosis tells you whether a storage dispatch that looks ahead could reduce curtailment and shortfalls.

Background: Economic predictive control

Production ignores the power price

What happens. The electrolyser runs when there is a production order or when it can, not according to the price of power. Contract terms and spot prices are known, but the operating plan does not use them.

What it costs. A higher cost per kilogram of hydrogen than necessary. Cheap hours go by unused, and expensive hours are paid for.

Why the usual fix fails. Simple time-of-day rules do not follow the market, and a manual plan cannot be made fast enough when prices change.

How a fixed-price remote diagnosis checks it. From your power contract, price history and operating logs I check how much of the production fell into expensive hours. The fixed-price remote diagnosis tells you whether price-led operation within your delivery commitments is worth building.

Background: Economic predictive control

Feed water quality varies

What happens. The water treatment delivers feed water whose quality changes with source, season and the state of the treatment. Conductivity trends up and down.

What it costs. Impurities damage membranes and catalysts, and the damage is slow and shows up only as a lower stack life.

Why the usual fix fails. A conductivity alarm at the limit reacts when the damage is already possible. Replacing resin or membranes on a fixed calendar is wasteful in good times and late in bad ones.

How a fixed-price remote diagnosis checks it. From conductivity, flow and regeneration logs of the treatment I look at when quality drifts. The fixed-price remote diagnosis tells you whether monitoring and control of the treatment can keep the feed water inside the limit.

There are few sensors inside the stack

What happens. The stack is closed. Operators see total voltage, current, inlet and outlet temperatures and little else. What happens in individual cells stays hidden.

What it costs. A weak cell can hold back the whole stack or fail first, and the plant is run blind to it.

Why the usual fix fails. Adding sensors inside the stack is hard and changes the product. Reading only total voltage gives the average and hides the outlier.

How a fixed-price remote diagnosis checks it. From the existing voltage, current and temperature signals, and what is known of the cell design, I check how much of the internal state can be computed. The fixed-price remote diagnosis tells you whether cell states can be estimated well enough to act on.

Background: Computing unmeasured states for control

First plants are commissioned with little experience

What happens. Plants of a new kind are started up by teams that have not run one before. Procedures are written, but behaviour in transients is unfamiliar.

What it costs. Delays, restarts and trips in the first months, with the plant earning little while costs run.

Why the usual fix fails. Learning on the real plant is slow and expensive. Documentation from the supplier describes the normal case, not the awkward ones.

How a fixed-price remote diagnosis checks it. From the piping and instrument drawings, the planned control logic and the process data of similar units I check which start-up and shutdown situations are risky. The fixed-price remote diagnosis tells you whether a simulation before start-up would catch problems early, and what it would need.

Waste heat is not used

What happens. The stack gives off heat that is cooled away. Nearby, a building, process or network needs heat at the same time or later.

What it costs. Lost energy, and a worse overall efficiency, since the heat is a large part of the energy not ending up in the hydrogen.

Why the usual fix fails. A heat exchanger on its own gives heat at times when nobody wants it. Heat demand and electrolyser load do not follow each other.

How a fixed-price remote diagnosis checks it. From the heat flows, the load of the electrolyser and the heat demand of the user I check when heat could be used. The fixed-price remote diagnosis tells you whether heat integration control with storage would raise the share of heat that is used.

Background: Economic predictive control

Hydrogen compressors cycle

What happens. The compressors start, stop and unload as the gas production follows the power. They cycle more than they were designed for.

What it costs. Wear on the machines, maintenance work and unplanned stops, which reach back to production.

Why the usual fix fails. A larger buffer vessel smooths the flow but is costly, and simple on-off rules cycle at the worst moments.

How a fixed-price remote diagnosis checks it. From compressor logs, pressures and gas flows I count the cycles and look at how the flow arrives. The fixed-price remote diagnosis tells you whether a load plan, which smooths the compressor duty ahead of time, would reduce the cycling.

Background: Optimal paths, schedules and setpoint plans

Downstream synthesis needs steady hydrogen

What happens. A methanation or ammonia unit downstream wants a steady feed. The electrolyser upstream delivers what the power allows, which varies.

What it costs. A mismatch between supply and use: the synthesis runs at part load, is shut down or burns off hydrogen, all of which waste money.

Why the usual fix fails. A large buffer would solve it, but costs a lot and the synthesis units cannot follow quickly anyway.

How a fixed-price remote diagnosis checks it. From production, buffer and synthesis logs and the allowed load ramps of the synthesis I check how the two sides interact. The fixed-price remote diagnosis tells you whether coordinated buffer and synthesis control could keep both within their limits.

Background: Predictive control of process plants

Leaks and faults are found late

What happens. Leaks of hydrogen and faults in equipment are found when a gas detector triggers or a pressure drop is noticed. Slow leaks and creeping faults go unseen for a while.

What it costs. Safety risk, lost product and unplanned shutdowns.

Why the usual fix fails. More gas detectors cover only the places where they sit, and fixed alarm thresholds either trigger too often or too late.

How a fixed-price remote diagnosis checks it. From pressure, flow and balance data of the plant I check whether a model of the normal behaviour can flag deviations early. The fixed-price remote diagnosis tells you whether model-based detection of leaks and faults is realistic with your instruments.

Background: Computing unmeasured states for control

Start-ups begin without a control architecture

What happens. A young company has a stack and a promising design, but the control grew with the prototype. Nobody wrote down what the control must do, which signals exist and how the pieces fit.

What it costs. Rework when the plant grows, delays in qualification and customers who ask questions that cannot be answered.

Why the usual fix fails. Hiring a controls engineer takes time in a market with few of them. Letting the software team grow the control as they go gives a structure that is hard to change later.

How a fixed-price remote diagnosis checks it. From your design documents, signal lists and prototype logs I put together what the control must do and how it could be structured. The fixed-price remote diagnosis gives you a requirements and architecture outline to work from and shows where deeper work is needed.

Stack testing is slow

What happens. Characterising a stack on the test bench takes long: many operating points, each held until it settles, and tests repeated for each design change.

What it costs. Development time, bench occupancy and the staff who run the tests.

Why the usual fix fails. More benches cost money. Testing fewer points saves time but leaves gaps in the picture.

How a fixed-price remote diagnosis checks it. From past test series and bench records I check which points carried the information and which were redundant. The fixed-price remote diagnosis tells you whether planned experiments with estimation of the stack parameters would give the same picture in fewer runs.