Availability and energy-loss attribution
Availability answers “what share of the fleet’s capacity-time was able to generate?” and is a percentage. Energy-loss attribution answers “how many kWh does each cause explain?” and is an estimate in kWh. The two do not move together: the energy cost of downtime depends on when it happens, so one inverter down during the sunniest hour costs more than three inverters down at dawn.
How SolarSENS classifies an interval
Section titled “How SolarSENS classifies an interval”SolarSENS evaluates each inverter in five-minute intervals inside the plant’s configured operating window (window start included, end excluded). Intervals outside the window — night, for a typical configuration — are dropped before classification and never enter the calculation.
Each in-window interval is then classified as available, unavailable, or unknown by combining the inverter’s active power and status, irradiance against the plant’s configured minimum, and whether a dispatch instruction is active. The thresholds are plant-specific configuration, not published constants. An active dispatch instruction does not by itself exclude an interval: a curtailed inverter still producing above threshold counts as available.
Unknown intervals — chiefly missing power and status telemetry, and most below-threshold-irradiance combinations — are excluded from the denominator entirely; this page calls the remaining classified intervals applicable. Exclusion is not downtime: a plant with poor data coverage can show high availability computed from few intervals.
This is why availability differs from connectivity, uptime, and performance ratio. A communicating inverter can be unavailable for generation, and an available inverter can still underperform.
How availability is aggregated
Section titled “How availability is aggregated”Availability is capacity-weighted when inverter capacity metadata exists: an unavailable 500 kW inverter moves the plant result five times as much as an unavailable 100 kW inverter. Without capacity metadata, the calculation averages applicable intervals unweighted.
Longer periods recombine the underlying weighted counts; they are not averages of the displayed daily percentages. Days with more applicable intervals carry more weight.
For example, a day at 90% availability from 90 of 100 weighted intervals and a day at 80% from 40 of 50 weighted intervals combine as:
From unavailable intervals to estimated kWh
Section titled “From unavailable intervals to estimated kWh”Unavailability loss estimates what the unavailable intervals would have produced, using the same five-minute intervals plus weather, meter and grid context, curtailment context, and the soiling model where one exists for the plant. Interval estimates are summed per plant-local day; negative interval estimates are clamped to zero.
The result is a model output, not a meter reading. It carries the uncertainty of its inputs, and it assigns no root cause — it says the energy was likely lost, not why the equipment was down.
How the loss bridge is composed
Section titled “How the loss bridge is composed”The Insights loss bridge shows four loss categories, each calculated independently:
| Category | What it estimates |
|---|---|
| Irradiation | Energy effect of actual solar resource differing from the budget assumption |
| Overspill | Energy not delivered during curtailment |
| Unavailability | Energy lost during unavailable intervals, as described above |
| Soiling | Energy effect of the plant’s soiling model |
Because the categories are independent estimates, they do not mechanically sum to the total gap. The bridge closes that gap with two presentation stages: Variance (whatever difference remains after the four categories) and Meter to Inverter (aggregated inverter yield versus metered yield). A large Variance means the four models explain little of the gap for that range — a signal to investigate, not a fifth loss cause.
For the related calculations, see Performance ratio calculation and Cleaning and soiling analytics.
Next step
Section titled “Next step”Continue with Availability and loss analysis for the operator workflow.