What lenders need after commercial operation, and how plants evidence it

Before commercial operation the reporting regime is dense. A lender's technical advisor reviews the technology, the EPC contract, the schedule and the budget, produces a report that supports financial close, then monitors construction against milestones. There is a large advisory industry around this and a great deal of published guidance.

After commercial operation, most of that apparatus stands down. What remains is periodic reporting against covenants, an annual or semi-annual technical review, and whatever the lender asks for when something looks wrong. The frequency drops, the scrutiny becomes exception-driven, and the reporting pack is usually compiled by the operations and maintenance contractor.

That last point is the structural feature worth noticing. The party generating the performance report is frequently the party whose performance it describes. Nobody is being dishonest; the incentive simply is not neutral, and both the lender and the owner are aware of it.

Debt service cover ratio. Cash available for debt service against the obligation for the period. The covenant test is historic, but the forecast that sits beside it is what determines whether a lock-up or a cure is coming.

Availability. Against a contractual definition that is rarely as clear as it looks. Planned versus unplanned, grid versus plant, full outage versus derating, and the treatment of force majeure all sit inside the definition rather than in the measurement.

Production against forecast. For a hydrogen plant, output and specific energy; for solar, the pair of metrics covered in performance ratio against capacity utilisation factor. Deviation matters less than whether the deviation is explained, and whether the explanation is consistent with the resource data.

Reserve account balances. Debt service and major maintenance reserves. The major maintenance reserve is sized on an assumption about when large expenditure falls due, which for an electrolyser plant usually means stack replacement.

Insurance and incidents. Cover in force, claims, and reportable events. An insurer asking what happened after a thermal event wants a traceable sequence, not a narrative, which is the subject of what monitoring evidence an insurer can price. For Indian sites the same records serve the CEA safety framework taking effect in 2027.

Degradation appears twice in a project model and both instances are assumptions at close.

It appears as a production decline, reducing output or raising specific energy over time and therefore reducing revenue or raising operating cost. And it appears as timing for the largest recurring expenditure the plant will face, which for an electrolyser is stack replacement, sized into the major maintenance reserve. Turning a measured trend into a date that survives review is covered in remaining useful life reporting for lenders.

Both were set from OEM curves and comparable projects before the plant existed. Once it is running, measured degradation is what confirms or corrects them. A stack degrading more slowly than assumed is an argument for revisiting a reserve that is over-funded and constraining distributions. One degrading faster is a problem that is considerably cheaper to raise early than to discover at the point of replacement.

This is why a degradation trend answers a different question for a lender than a single period's ratio. The ratio reports what happened. The trend changes the forecast, and it is the same record that an electrolyser performance guarantee is argued from.

Availability disputes are rarely about arithmetic. They are about definitions applied to events after the fact, and evidencing availability guarantee compliance is a problem in its own right.

Recurring questions: does a plant that is technically capable but not producing because the grid curtailed it count as available. Is an outage caused by feedwater quality a plant fault or a supply fault. How is a partial derating treated, and does running at seventy per cent count proportionally or not at all. When a fault has more than one contributing cause, whose is it. Was the maintenance that caused the outage planned, and was it planned within the notice period.

Each of those is decided by the contract and then evidenced from the data. The side with a continuous, timestamped record of plant state, ambient conditions, grid signals and alarm history is the side that can support its reading of the definition. The other side is arguing from recollection.

The reporting that survives scrutiny has three properties. It is continuous rather than sampled, so a question about a specific hour eighteen months ago can be answered. It is traceable, so a figure in a report can be followed back to the measurements behind it. And its provenance is separable from the party whose performance it describes.

In practice that means a commissioning baseline captured properly, retention aligned to the tenor of the debt rather than to the default retention of the historian, and enough resolution that averaging has not already destroyed the answer.

Yunify supports this reporting rather than replacing the parties who produce it. It runs on-premise, holds the operating record, and attributes movement in performance to a mechanism, so a figure in a quarterly pack can be traced to something physical rather than asserted. For certification and audit purposes the same principle applies: it supports the reporting, and the certifying body certifies. What that body actually samples is set out in what a certification auditor samples.

The timing point is the one that gets missed. Every question a lender asks after commercial operation is retrospective. No instrument installed today can answer what happened last year, and the covenant conversation that makes people wish they had the data is not usually the first one.