Dust analysis in photo-voltaic solar plants with satellite data

Abstract

Dust is a complex problem in evaluating photo-voltaic (PV) solar plants as it requires analog or digital sensors, pyranometers, heliostats, particle matter (PM) sensors, or similar devices for the dust or to consider the indirect reduction in the output or direct current. Novel techniques based on image analysis have been introduced in different fields, such as agriculture and the control of the vegetation, and in the last two years, PV panel evaluation techniques using satellite data have been considered. This paper presents the development of a new evaluation method for PV panels using satellites Landsat 8 and Sentinel-2 calibrated with dust sensors. The PV plant was spread across 400 ha and comprised 600,800 panels with a capacity of 192.26 MW. A meteorological station and a random forest (RF) model use information from 2019 to 2022 in the analysis. Our findings quantify the loss production per panel with detailed satellite image evaluation of the soiling, primarily owing to the dust covering each panel in the PV solar plant, using three years of data from 2019 to 2022. Accordingly, we obtain a mean absolute error (MAE) of 0.22, mean squared error (MSE) of 0.07, explain variance score of 0.88, and R2 score of 0.88.

Keywords

• Panels
• Photo-voltaic
• Satellite data
• Dust analysis
• Soiling evaluation