Start with the right reference
Every loss is measured against something. The three common references answer different questions:
| Reference | Built from | Use it to |
|---|---|---|
| Budget or target (P50) | Design simulation for typical weather | Explain variance to the business plan |
| Weather-adjusted expected | Design model run with actual irradiation and temperature | Isolate losses the O&M team can influence |
| Availability-adjusted | Expected output for the capacity that was actually available | Separate downtime from performance losses |
For day-to-day O&M, the weather-adjusted expected output is usually the most useful, because it removes the weather the team cannot control.
Build a loss waterfall
A loss waterfall walks from the target to the actual, one category at a time, so every kilowatt-hour of difference is attributed once.
Worked example · illustrative numbersA month at a 40 MWp plant
| Step | MWh |
|---|---|
| Target generation | 6,000 |
| Irradiation above budget | +310 |
| Thermal (temperature) loss | −180 |
| Soiling | −95 |
| Plant downtime (inverters, strings, SCBs) | −120 |
| Grid unavailability and curtailment | −40 |
| Unexplained residual | −25 |
| Actual generation | 5,850 |
The plant missed target by 150 MWh, but the weather helped by 310 MWh. Controllable losses (soiling and downtime) cost 215 MWh: that is where O&M attention should go.
Loss categories that work
| Category | Typical causes | Data needed | Usual owner |
|---|---|---|---|
| Irradiation variance | Weather better or worse than budget | POA irradiation vs budget | None (not controllable) |
| Thermal | High module temperature | Module temperature, temperature coefficient | Design; partly O&M (ventilation) |
| Soiling | Dust, bird droppings, pollen | Soiling sensor or PR trend between cleanings | Cleaning team |
| Tripping / downtime | Inverter trips, string and SCB faults, transformer outages | Inverter status, string currents, alarm log | Field and OEM teams |
| Efficiency | Inverter or component running below rated efficiency | DC and AC power by inverter | Asset manager, OEM |
| Clipping | DC power above inverter AC rating | Inverter output at rated limit | Design |
| Grid / curtailment | Grid outage, export limit, instructed curtailment | Grid availability, set-point records | Grid operator, commercial team |
| Communication / data | Data logger or network outage | Data availability per device | SCADA team |
Avoid double counting
- Attribute in a fixed order. For example: grid first, then downtime, then soiling, then thermal. Each step uses the output left after the previous one.
- Don't count data gaps as generation loss. A logger outage loses data, not necessarily energy. Check the meter before booking a loss.
- Keep a residual. Whatever can't be explained stays in an “unexplained” bucket. If it grows, your model or your data needs attention.
- Make categories mutually exclusive wherever possible, and write the rules down.
What to record for every loss event
- Start and end time
- Affected equipment and capacity
- Expected and measured output during the event
- Estimated energy lost (kWh)
- Data source used
- Category and probable cause
- Owner and linked work order
- Confidence level (measured, estimated, assumed)
Turn losses into action
- Rank events by energy lost and by how often they recur. A small loss that happens every day can beat one big outage.
- Validate the data before dispatching anyone: rule out sensor and communication problems first.
- Assign the issue with an owner, priority and due date.
- Close it only when the evidence is in and output has recovered on comparable days.
How nxtGrid helps
How nxtGrid helps
- Loss bucketing by parameter, including tripping, efficiency and thermal losses
- A loss waterfall from target to actual generation for any period
- Loss reports and loss charts at portfolio and plant level
- Alerts that become tickets and work orders, so losses have owners