Materials & Energy
Dirty Panels or a Weather-Model Error?
2026-09-10
A gradual solar-output decline may suggest soiling, but weather inputs, sensor drift and equipment availability can produce similar residuals. A computational review can compare these explanations using existing records. The useful outcome is a better attribution boundary, not an automatic cleaning recommendation or a promised recovery in generation.
The computational starting point
Start with archived power, irradiance, temperature, availability and maintenance records for an authorized site, or a suitable documented public example. pvlib provides soiling-model examples. A model example is not site evidence. Check timestamps and clipping before treating a daily performance ratio as a clean measure of panel condition.
Sources: pvlib soiling models.
Where intuition enters
An intuition that the system is accumulating a removable loss becomes a prediction of gradual decline and a defined reset pattern. The rival is a changing irradiance bias or temperature-model error. Specify what observation would distinguish those stories before selecting convenient rain or maintenance dates from the record.
A test that can disagree
Reproduce a weather-normalized baseline and compare a cumulative-loss model with drift and seasonal alternatives. Use weather context with known provenance; reanalysis can be useful but may miss local irradiance conditions. Hold out contiguous periods and avoid fitting the reset magnitude separately to every favorable event.
Inspect whether the proposed loss pattern appears in independent sensors or only in one denominator of a ratio. Treat maintenance annotations as potentially incomplete. If modest weather-input changes reverse the attribution, report that uncertainty. A residual model can describe lost fit without establishing a physical cause.
Sources: Copernicus climate reanalysis.
An illustrative decision
Imagine that a supposed cleaning reset coincides with an irradiance-sensor recalibration. The records cannot attribute the improvement to soiling removal without further existing evidence. A useful result may be a small list of ambiguous events that should be excluded from a confident historical savings claim.
What the research would deliver
The deliverable would be a residual-attribution audit, sensitivity analysis and a recommendation about the next computational model. A solar analytics buyer can use it to avoid encoding a false cause into a product. The scope involves existing data only and does not direct physical maintenance or electrical work.
Questions this raises
Does a power increase after rain prove cleaning?
No. Rain may coincide with weather, temperature or measurement changes that also affect the comparison.
Can the preview quantify guaranteed savings?
No. It can identify an attribution route and its evidence needs. Realized savings require separately supported operational and financial assumptions.
Sources and their limits
- pvlib soiling models. Soiling-model examples, not evidence of losses at an actual installation.
- Copernicus climate reanalysis. Reanalysis combines observations with models; it is not independent ground truth.
Prepared with AI assistance. The linked sources support the specified technical points; they do not validate applied psionics as a whole or guarantee a result for a client.
Read the editorial and evidence standard.
Continue reading
Explore Scientific Oracle consultingfor a scoped review of an existing-data research decision. Start with a non-confidential outline of the question, available evidence and the decision it needs to inform.