If you have closed out a utility-scale or large C&I solar project recently, you have probably seen this pattern in the EPC contract: Substantial Completion is gated on a performance exhibit, and that exhibit requires not one test but two. A capacity test — almost always referencing ASTM E2848 — and an availability test, usually defined in a few paragraphs of contract language plus an Excel "availability calculator" the owner's engineer hands over.
The capacity test gets most of the attention, and most of the tooling. The availability test usually gets a spreadsheet, a data dump, and a long email thread about which hours count.
This article walks through what each test actually proves, why availability testing has no equivalent of ASTM E2848, and where availability tests go wrong in practice.
Two Tests, Two Questions
The two tests answer different questions, and neither substitutes for the other.
A capacity test asks: does the plant deliver the power that was promised? You take measured power, irradiance, and weather data, filter it to stable operating conditions, fit a regression, and evaluate it at agreed reporting conditions. ASTM E2848 defines the method, and the result is a capacity ratio — measured versus expected power. If you want the full picture, our practical guide to ASTM capacity testing covers the method end to end.
An availability test asks: was the equipment running whenever it should have been? A plant can pass a capacity test on a good week and still have an inverter that trips every afternoon. Availability testing catches that: it counts, inverter by inverter, how much of the eligible test time each unit was actually operational.
| Capacity test | Availability test | |
|---|---|---|
| Question answered | Does the plant deliver the promised power? | Was the equipment running when it should have been? |
| Method | ASTM E2848 regression at reporting conditions | Defined by your contract |
| Typical pass criterion | Capacity ratio ≥ agreed percentage | Availability ≥ 98–99% over consecutive test days |
| Input data | Power, irradiance, weather, plus a model of expected output | Inverter power and status, irradiance, meter export |
Both are typically conditions of Substantial Completion, and they often run over the same days on the same data feeds. Contractually, though, they are separate hurdles: passing one does not waive the other.
Why There Is No "ASTM E2848 for Availability"
Here is the part that surprises engineers coming from the capacity side: there is no single standard that defines how to calculate PV availability.
The standards that exist are frameworks, not formulas. IEC TS 63019 provides an information model — a consistent vocabulary for categorizing time — but deliberately does not prescribe one calculation. IEC TS 61724-3 defines an energy-based evaluation method but leaves the expected-energy model to the user. And when Sandia National Laboratories reviewed real PV O&M contracts, they found multiple competing availability definitions in the wild — some time-based, some energy-based, some hybrid — which is exactly why their best-practice report on availability guarantee language exists.
The practical consequence: your contract is the standard. The formula, the irradiance threshold, the exclusion list, the guaranteed percentage, the measurement boundary (inverter vs. plant), the rounding convention — all of it is defined per deal. Two plants a mile apart can have materially different availability definitions, and both are "correct."
This is also why the generic "uptime" number on your monitoring dashboard cannot settle a contractual question. That number was computed with someone else's definition — not your contract's.
Anatomy of a Typical Commissioning Availability Test
Contract language varies, but commissioning-stage availability tests tend to share a common skeleton:
1. Eligible time. The clock only runs when the plant should be producing — typically when plane-of-array irradiance is above an agreed threshold and the plant is exporting to the grid. Nighttime, deep overcast, and grid-down periods do not count against anyone.
2. Per-inverter accounting. For every eligible interval, each inverter is classified as operational or not — from its status codes where available, or from its power output. Unavailable time is "charged" against a downtime budget.
3. A consecutive-day window. Commissioning tests commonly demand a set number of consecutive days — often three to five — so a single good afternoon cannot carry the result.
4. A guaranteed percentage. Commissioning tests are often set at 99% or higher, since a freshly commissioned plant with no degraded components should run essentially clean. Annual availability guarantees in O&M contracts typically sit at 98–99%.
5. Exclusions. Grid outages, owner-caused downtime, and force majeure typically do not count against the contractor. Some contracts also extend the test window to compensate for excluded time.
Put those five pieces together and a completed test looks like this:
Every minute of the window is classified: grey is ineligible time (night and below the irradiance threshold), the hatched band is a documented grid outage excluded from the test, and the red slivers are charged downtime.
Run the numbers and the margins are thin. At a 99% guarantee over a five-day test on an 18-inverter site like the one above, the total permitted downtime across all inverters is a few hundred inverter-minutes. One inverter that takes a morning to restart — the red block on INV 07, day 1 — can consume most of the budget on its own, and that plant lands at 98.76%: a miss.
Where the Arguments Start
Nearly every disputed availability test comes down to one of four things:
The exclusion list. Who decides whether a two-hour outage was a grid event (excluded) or a plant fault (charged)? If downtime can be reclassified after the fact without documentation, the number becomes negotiable — and negotiable numbers breed disputes. The fix is procedural: every exclusion should be a logged event with a time span, a category, and a justification that both parties can inspect.
Missing data. Data loggers fail, cell modems drop out, and suddenly there is a three-hour gap in the middle of the test window. Does that gap count as downtime, get thrown out, or get papered over? Whatever the answer, it should be a stated policy applied consistently — not a case-by-case decision made after seeing which side it helps. The conservative reading is that missing data counts as unavailable unless documented evidence shows otherwise; anything looser lets data gaps quietly improve the result.
Status-code interpretation. Is an inverter in "standby" down or waiting for irradiance? Does "throttled" count as operational? Vendor status codes do not map cleanly onto contract categories, and an unstated mapping is a dispute waiting to happen. The mapping from each status code to operational/down/excused should be written down and agreed before the test, not reverse-engineered afterward.
The irradiance threshold. The eligibility threshold decides when the clock runs. Set it too low and inverters get charged for minutes when there was not enough light to operate; measure it with a partially shaded sensor and the clock runs at the wrong times entirely. The threshold value and its source belong in the test documentation.
None of these are calculation problems — a spreadsheet can compute a ratio. They are documentation and auditability problems, which is precisely what ad hoc spreadsheet workflows are worst at. If you have seen what a failed capacity test does to a project timeline, an ambiguous availability result is the same story with more line items to argue about.
Two Tests, One Workflow
Since both tests draw on the same site configuration and largely the same data, running them in two disconnected tools — one platform for capacity, one spreadsheet for availability — duplicates effort and splits the audit trail.
That is why HelioTest runs availability tests alongside its ASTM E2848 capacity tests, on the same sites and the same data uploads. The contract's definitions become explicit settings — irradiance threshold, export condition, status-code classification, charging policy, guaranteed percentage — and exclusions become documented events with quantified effects. Every minute of the test window gets a machine-readable classification, and the deliverables are built for the counterparty, not just for you: a PDF report ready for owner sign-off, the minute-by-minute table as CSV, and an Excel workbook with live formulas that recomputes the result from the raw data — so the owner's engineer can audit the arithmetic without trusting a black box.
Key Takeaways
Capacity and availability tests are twin conditions of Substantial Completion in many EPC contracts, and they answer different questions: promised power versus equipment uptime. Capacity testing has a genuine standard in ASTM E2848; availability testing does not — the contract itself defines the formula, thresholds, and exclusions, which is why a monitoring dashboard's generic uptime number cannot settle a contractual availability question.
The tests that close cleanly are the ones where the definitions were pinned down before the test started — the status-code mapping, the irradiance threshold, the missing-data policy, and an event log that documents every exclusion. The tests that end in email threads are the ones where those questions were left for after the result came in.
If you have an availability test coming up at handover, talk to us — or start with the capacity side and our sample ASTM E2848 test report to see what an audit-ready deliverable looks like.
Peter is an engineer with a PhD in renewable energy management and over a decade of experience in software development for renewable energy applications. He built HelioTest to replace the fragile spreadsheet workflows he encountered across dozens of ASTM E2848 capacity tests.
