Availability: PV systems
Availability answers: how much of the plant was actually producing, weighted by how much sunlight it missed?
An inverter down at noon costs more than the same inverter down at 6pm. PV availability is weighted by irradiance for exactly that reason β it measures lost opportunity, not lost clock time.
The formula
Ξ£k [ Ξt Γ (Gk / 1000) Γ C_down,k ]
Availability = ( 1 β ββββββββββββββββββββββββββββββββββ ) Γ 100
P_total Γ Ξ£k [ Ξt Γ (Gk / 1000) ]
Term | Meaning | Unit |
|---|---|---|
C_down,k | Capacity not producing at interval k | kWp |
P_total | Sum of installed capacity across all devices | kWp |
Gk | Irradiance at interval k | W/mΒ² |
Ξt | Interval length in hours (0.25 for 15-minute data) | hours |
Read it as: of all the sunlight-weighted capacity the plant could have offered, what fraction was offline. Subtract from 1 and you have availability.
Intervals below 50 W/mΒ² are dropped, so a plant is never marked unavailable for not producing at night.
What counts as "not producing"
A device is counted in C_down,k at a given interval when:
- It reported zero output, or
- It was not communicating at all, or
- The power plant controller was limiting the site at that interval β a curtailed device is producing, but not producing freely, so it lowers availability.
Capacity weighting means the numbers behave the way you would expect. On a 500 kWp site with one 50 kWp inverter down, availability is 90% β not 50% β regardless of how many inverters the site has.
Two availability figures
Figure | Curtailment and agreed downtime |
|---|---|
Availability | Counted against you. This is the raw measure of what the plant delivered. |
Internal Availability | Excluded, where an availability adjustment event covers the interval. |
Availability adjustment events are logged periods β planned maintenance, grid constraints, owner-instructed curtailment β that are agreed not to count against the plant. Internal Availability is the figure to use when a contract excludes those causes.
Both are calculated over the same intervals from the same data. They differ only in which intervals are excused.
Availability and PR
They are deliberately separate numbers:
- Availability β how much of the plant was running.
- PR β how well the part that was running performed.
PR already discounts the sunlight that fell on offline equipment, so a day with heavy downtime shows low availability and normal PR. Read them together: low availability with normal PR is a downtime problem; full availability with low PR is a performance problem.