Cleaning and soiling analytics

Cleaning Analytics helps operations teams identify plants that may benefit from cleaning, then inspect the evidence behind each candidate. It combines weekly equipment performance, estimated soiling, weather context, recorded cleaning events, and estimated energy loss.

Use it as a decision-support tool. A suggested cleaning date is an analytical signal, not a scheduled appointment or an automatic instruction to clean.

Core indicators

Indicator	What it tells you
Soiling rate	The estimated effect of dirt on equipment performance. A higher value indicates dirtier equipment.
Adjusted performance	Performance normalized for temperature and recent operating history. A higher value indicates cleaner or stronger relative performance.
Cleaning status	A summary label—Clean, Dirty, or V. Dirty—used to rank plants and inverters for review.
Estimated soiling loss	Energy estimated to have been lost because of soiling, shown for recent portfolio periods and individual plants.
Suggested cleaning	A modeled date derived from the direction and pace of normalized performance. It indicates when a plant may cross its cleaning target.
Last cleaned	The most recent cleaning event available to the model.

Status labels make the portfolio easier to scan, but they should not be used alone. Confirm the recent soiling rate, adjusted performance, estimated loss, and plant context before deciding what to do.

The indicators use three time grains. Soiling observations and trends are weekly. Estimated soiling loss is calculated per day. Portfolio summaries combine daily results over a selected period. Only compare values on the same grain — a weekly soiling rate is not one day’s energy loss.

flowchart LR
    A[Weekly plant observations] --> B[Soiling trend and status]
    A --> C[Suggested cleaning date]
    A --> D[Daily soiling-loss estimate]
    D --> E[Selected-period summary]

Weather context and modeled results

Cleaning Analytics brings observed or otherwise available plant context together with derived analytical results.

Type	Examples	How to interpret it
Observed or available context	Irradiance, precipitation, humidity, module temperature, inverter operation, and recorded cleaning events	Evidence about the conditions around the plant and the model period. Availability and source can vary by plant.
Modeled results	Adjusted performance, soiling rate and status, suggested cleaning, and estimated soiling loss	Analytical interpretations built from the available inputs. They are estimates, not direct measurements of dirt or lost energy.

Weather is supporting context, not proof of soiling by itself. Rainfall, humidity, irradiance, module temperature, outages, and data gaps can all affect the pattern you see.

See Availability and energy-loss attribution for how soiling fits into the wider loss interpretation.

Recorded cleaning events

Recorded cleaning events appear as context in the plant timeline and influence how SolarSENS interprets performance after cleaning. Cleaning Analytics reads those events; it does not create or schedule cleaning work.

After a cleaning event is recorded and new operating data becomes available, compare the post-cleaning trend with the earlier baseline. Improvement is consistent with soiling having contributed to the earlier decline. Little or no improvement suggests that other causes may need investigation.

Data and decision boundaries

Signal	Use it for	Do not infer
Cleaning status	Ranking plants or inverters for review.	A field work order has been scheduled.
Suggested cleaning	Seeing when a plant may cross the cleaning target.	Crew availability, weather safety, cost, or contract approval.
Estimated soiling loss	Understanding the modeled energy impact.	A directly measured loss value.

• Suggestions depend on sufficient, valid plant and inverter data. A suggested date can change as new weekly observations arrive.
• The analytics view does not confirm crew availability, access restrictions, weather safety, cleaning cost, or contractual requirements — check those before scheduling work.
• Clean, Dirty, and V. Dirty use SolarSENS-defined thresholds applied consistently across plants. Compare these labels within SolarSENS, not against external soiling benchmarks or site-specific cleaning targets.

Related workflow

• Use Cleaning Analytics to prioritize plants and inspect plant-level evidence.
• Availability and energy-loss attribution for the wider loss model.