Product
Cell voltage monitoring for electrolysers
Stack voltage is the sum of every cell in the stack. By the time that total moves, the cell that caused it has usually been drifting for weeks. Yunify CVM measures each cell, so the drift is visible while there is still time to act on it.
What stack-level measurement hides
A stack contains hundreds of cells in series. The control system reports one voltage across the whole assembly, so a single cell drifting by 100 mV is diluted into a stack reading that barely moves. The cell continues to degrade, the local current density shifts, and the first clear signal often arrives as a purity excursion or a trip.
Cell-level measurement changes what is observable. A voltage rising against its neighbours under identical conditions is a strong, early, local signal, and it can be attributed to a mechanism rather than to the stack as a whole.
What cell-level data makes visible
Uneven degradation
Cells departing from the fleet baseline while stack voltage stays inside its alarm band.
Gas crossover risk
Voltage and behaviour patterns associated with membrane condition, ahead of a hydrogen-in-oxygen alarm.
Electrolyte and contamination effects
Local deviations consistent with contamination rather than with load.
Shunt and leakage paths
Distribution patterns across the stack that a single aggregate reading cannot show.
Evidence for a guarantee claim
Per-cell history that separates stack degradation from balance-of-plant losses and from operating-regime effects.
Three ways to get the measurement
Cell-level visibility is the requirement. Dedicated hardware is one route to it. Where a stack already reports enough for the physics to work from, or already measures cell voltage through somebody else's system, there are two more, so the route is scoped per site rather than assumed.
Yunify CVM hardware
A dedicated acquisition layer wired to the stack. The direct measurement, and the right answer where nothing usable is installed and the cell count is high.
Soft sensing
Where a stack is not instrumented cell by cell and cannot be, the Engine estimates per-cell behaviour from the signals that do exist, being stack voltage, current, temperatures and flows, against a physics model of the stack. Coarser than a direct measurement, and frequently enough to rank cells and see a drift start.
On third-party hardware
Where an OEM or an existing monitoring system already measures cell voltage, Yunify reads it rather than duplicating it. The analytics do not require our acquisition layer to be the source.
Designed to retrofit
CVM is intended for stacks that are already running. Where hardware is installed it sits as an acquisition layer feeding Yunify Edge, which passes the data to the physics models without touching the control system. Where the measurement comes from a soft sensor or from a cell-voltage system the plant already owns, the path is the same from Edge onward.
That matters commercially as well as technically. Most electrolyser capacity that needs this measurement has already been commissioned, and the alternative is usually to wait for the next stack replacement.
Frequently asked questions
What is cell voltage monitoring?
Measuring the voltage of each individual cell in an electrolyser stack, rather than the single aggregate voltage across the whole stack that a control system normally reports.
Can it be fitted to an electrolyser that is already running?
Yes. CVM is designed as a retrofit and does not require changes to the installed PLC, DCS or SCADA.
Do we have to install CVM hardware to get cell-level diagnostics?
Not always. There are three routes: our acquisition hardware, soft sensing from the signals a stack already reports, or reading a cell-voltage system the plant already has. Which one applies is a site question, answered from what is installed and how many cells are in the stack.
How many cells can be monitored?
Hundreds of cells per stack. Scope is set per site during deployment.
Does cell voltage monitoring need Yunify Engine?
No, the measurement stands alone. The physics models add attribution, so a deviation can be tied to a mechanism and an action rather than only flagged.