Dust, data and when to clean.
How to measure what dust costs a solar plant, and how to decide when cleaning pays.
GridMarshal, November 2026
A slow loss that nobody sees.
Dust builds up a little each day, so output falls without a single alarm.
Much of Pakistan has long dry spells, dust storms and farmland, roads or construction close to plants. Dust settles on the glass and blocks part of the light before it reaches the cells.
Rain washes some of it off, but not evenly. Low-tilt panels often keep a line of dirt along the bottom edge, and bird droppings stay until someone removes them.
Because the loss grows slowly, the plant still looks healthy on any single day. The cost only shows when you compare what the plant made with what it should have made.
Soiling has a shape.
Compare strings and inverters against expected output and against each other.
Soiling
A slow, steady fall across most strings and inverters, then a step back up after rain or cleaning.
A fault
A sudden drop on one string or one inverter while its neighbours carry on as before.
Haze or cloud
Output falls, but measured irradiance falls with it, so performance against the light received stays the same.
Heat
Output dips on the hottest afternoons and recovers when it cools. Temperature-corrected performance ratio removes most of this effect.
One trap: if the irradiance sensor is dirty too, it reads less light and hides the loss. Clean the sensor on its own schedule and before any test.
Measure the gain from one cleaning.
The step in output after a cleaning is the size of the soiling loss.
- Pick clear days just before and just after the cleaning, with similar irradiance.
- Compare normalised output, in kWh per kWp or as performance ratio, not raw kWh.
- Keep a reference: clean one inverter's strings first and leave a similar one dirty for a short while.
- Read the gap: the difference between the clean and dirty inverter, on the same day, is the soiling loss.
- Repeat through the year, because the rate of soiling changes with the season.
Putting a number on the dust.
An example with assumed figures, to show the arithmetic.
Example: a 1 MW plant. Assume it makes 4,500 kWh on a clear day when it is clean. After two weeks without rain, a comparison with a clean reference inverter shows the plant would now make 4,230 kWh on the same kind of day.
Daily loss now: 4,500 − 4,230 = 270 kWh, which is 270 ÷ 4,500 = 6%.
If the loss grew evenly from 0% to 6% over those 14 days, the average loss was about 3%. Energy lost so far: 14 × 4,500 × 3% = 1,890 kWh.
If the plant is not cleaned for another 10 days, it loses at least 270 × 10 = 2,700 kWh more, and more than that because the dust keeps building.
Clean when the energy is worth more than the wash.
Compare the value of the energy you would recover with the cost of one cleaning round.
Take the energy you expect to recover before the next cleaning or the next good rain. Multiply it by your own price per kWh. If that is more than the cost of one cleaning round, clean now.
In the example above, the question is whether at least 2,700 kWh, at your price, is worth more than one round of cleaning. Each plant has its own answer, because water, labour and energy prices differ.
- Watch the trend, not a single day.
- Check the forecast: if rain is likely soon, waiting may cost less.
- Clean the worst areas first, such as rows near a road or the downwind edge.
- Record every cleaning so the next before and after comparison is easy.
See the dust in the data.
For operations and maintenance (O&M) companies, EPC firms (engineering, procurement and construction) and plant owners.
- String evaluation: each string compared with the mean, within 5% green, 10% yellow, above that red
- Inverter and string heat maps in kW per kWp, plus inverter normalised and yield views
- Expected output per inverter compared with actual, with a deviation alarm
- Temperature-corrected performance ratio and irradiance next to output
- Recurring cleaning tasks from templates, and tickets that record the energy lost
Know when your plants need cleaning.
A walkthrough of GridMarshal with your own plants in mind.