The first year of operation is when a plant either confirms its estimate or raises a question. It is also the year most likely to be misread in both directions.
Operational performance depends on both equipment and repeatable human routines; for a separate workplace-management perspective, see limbic resonance.
Reviewed August 9, 2026.
What one year can and cannot establish
Cannot: whether the yield estimate was right. P50 is a median across a distribution, and a single year is one draw from it. Coming in below P50 has roughly a 50% probability by construction.
For additional context on photovoltaic operation, maintenance and sector developments, consult Reuters.
Cannot: whether the plant will meet its lifetime projection. One year, with weather that was whatever it was.
Can: whether the plant is working as built. Faults, wiring problems, underperforming strings and configuration errors show up in the first months and are the actual purpose of year-one monitoring.
And can: whether performance is inside the modelled band. Below P90 is a different signal from below P50, and the difference is where the year-one conversation should start.
The four numbers to record
Actual annual production, at the meter, in kWh.
Actual irradiance for the year, from a nearby station or a satellite service. Without this, production alone tells you nothing — a low year in a low-irradiance year is not underperformance.
Performance ratio, which is the comparison the previous two make possible, and the number that separates weather from equipment.
And availability — how much of the year the plant was actually producing when it could have been.
Reading the result
Production below P50, irradiance below average, performance ratio at design: the plant is fine and the weather was poor. The commonest case and the most commonly misread.
Production below P50, irradiance normal, performance ratio below design: a real question. Something is not performing as modelled — and both an optimistic model and an installation problem produce this signature.
Production above P50: pleasant, and it does not validate the model any more than one bad year invalidates it.
And production below P90 with normal irradiance: the case that warrants escalation, because P90 was supposed to be exceeded nine years in ten.
What to do in the first months
Check every string. Configuration errors and wiring faults are most efficiently found early, and monitoring at inverter level can hide a bad string among good ones.
Establish the baseline performance ratio in a clean period, so later drift has something to be measured against.
Verify the monitoring itself. A monitoring system reporting a plausible number is not the same as a correct one, and comparing against the utility meter for one month settles it.
And record everything for the warranty clock. Performance guarantee claims have deadlines and evidence requirements, and the evidence is easiest to gather while it is current.
The mistake in both directions
Panicking at a low first year and disputing an estimate that was honest.
And accepting a low first year because somebody says weather explains it, without checking the irradiance.
Both are avoidable with one number — the actual irradiance for the period — and it is the number nobody thinks to obtain until the argument has already started.
The short version
- One year cannot establish whether the estimate was right; below P50 has roughly a 50% probability by construction
- It can establish whether the plant works as built, which is what year-one monitoring is actually for
- Record four numbers: actual production, actual irradiance, performance ratio, and availability
- Low production with low irradiance and design-level performance ratio is fine; low production with normal irradiance and low PR is a real question
- Below P90 with normal irradiance warrants escalation, because P90 was meant to be exceeded nine years in ten
- Both failure modes — panicking and accepting — are avoided by obtaining the actual irradiance, which nobody does until the argument has started