DetectionMethod
Soiling model residual analysis
- ID
method:soiling-model-residuals- IRI
https://solarlattice.org/kb/method/soiling-model-residuals- Status
- active
Soiling model residual analysis compares measured plant output against a soiling model during dry periods, when dust builds up, to confirm granular dust soiling losses. Start with normalized production trends and find dry periods where output falls even though availability, clipping, curtailment and irradiance sensors do not explain the drop. Check that trend against soiling station data, rainfall, cleaning logs and nearby reference blocks if they are available. Some models, such as the Kimber approach, build the soiling profile from power and rainfall data. They need a minimum cleaning threshold for rain and a grace period after each rain event. Results can differ between analyses because there is no agreement on grace periods and cleaning thresholds. Global models have also correlated poorly at sites with low average soiling losses. Do not rely on one dirty irradiance sensor. Sensor drift, poor cleaning practice or a local obstruction can produce what looks like soiling but does not represent the array.
<!-- review notes
- sentence 1: support check: partial - Detection method mentions model residuals for dry accumulation periods, but doesn't say it compares measured output to a model to confirm dust soiling in these words.
- sentence 5: support check: partial - Lack of agreement on grace periods and cleaning thresholds causing differing results is stated, but global models correlating poorly at low-loss sites is overstated: the passage says two models correlated poorly with each other at sites under 5% average loss.
-->
Claims
produces
c1 verified · fact
Long-term normalized output decline
Look for output falling during dry accumulation periods with no availability, clipping, curtailment or irradiance sensor problem to explain it.
verified by: joshuahamsaverified on: 2026-10-20review note: support check: partial - Passages support looking for dry-period output decline without availability/clipping/curtailment/irradiance sensor issues, but not the specific signals (rain absence, POAI drift check) or that it is framed as a residual-analysis output.
c2 drafted · lead
Production recovery after rain
The recovery seen after rain is compared with the recovery the assumed soiling model predicts. If the two do not match, the local wash-off assumptions are wrong.
review note: support check: partial - Passage says to test whether the model explains recovery after rain and update wash-off assumptions if mismatched, but doesn't mention a step-up in normalized power or specific signals.
c3 verified · fact
Production step after cleaning
The step seen after cleaning is compared with the step the model predicts, to check the model's accumulation rate.
verified by: joshuahamsaverified on: 2026-10-20review note: support check: partial - Passage covers testing model against recovery after cleaning and updating accumulation assumptions, but not the step on the cleaning date or specific signal.
typical_power_loss
c4 drafted · lead
nominal 28.4 PERCENT
Conditions: Soiling loss with natural cleaning only (2 precipitation events, 0 cleaning events), in a technical-risk quantification example
review note: support check: partial - Index row gives natural cleaning 28.4% with 2 precipitation and 0 cleaning events, but nothing links it to soiling model residual analysis or 'typical power loss'.; possible conflict with c5 (same source; check the conditions); possible conflict with c6 (same source; check the conditions)
c5 verified · fact
nominal 4.1 PERCENT
Conditions: Soiling loss with periodic cleaning (2 precipitation events, 12 cleaning events), same example
verified by: joshuahamsaverified on: 2026-10-20review note: support check: partial - Row gives periodic cleaning 4.1% with 2 precipitation and 12 cleaning events, but no link to residual analysis or 'typical'.; possible conflict with c4 (same source; check the conditions); possible conflict with c6 (same source; check the conditions)
c6 verified · fact
nominal 2.3 PERCENT
Conditions: Soiling loss with triggered cleaning (2 precipitation events, 20 cleaning events), same example
verified by: joshuahamsaverified on: 2026-10-20review note: support check: partial - Row gives triggered cleaning 2.3% with 2 precipitation and 20 cleaning events, but no link to residual analysis or 'typical'.; possible conflict with c4 (same source; check the conditions); possible conflict with c5 (same source; check the conditions)
Referenced by
Data