Photovoltaic panel dust monitoring based on hyperspectral remote sensing

The Benban Solar Power Plant is located in the Benban region of Aswan Governorate, Egypt, a desert area approximately 650 kilometers south of Cairo and 40 kilometers northwest of Aswan. With a total capacity of 1650 megawatts, it is Egypt's first large-scale photovoltaic power plant and currently the world's fourth largest solar power plant. Construction began in February 2018 and was completed in November 2019. The plant covers an area of ​​37.2 square kilometers and cost US$4 billion.

Project Introduction

Located in a desert region, this power generation project is constantly plagued by sandstorms. Dust easily accumulates on the surface of the photovoltaic modules. This dust not only blocks sunlight, reducing the transmittance of the glass and decreasing the amount of solar radiation reaching the cells, but also alters the heat transfer process. Dust absorbs solar radiation and converts it into its own heat, while simultaneously hindering heat dissipation from the photovoltaic module's cover glass. In short, dust is one of the main issues affecting the performance ratio of photovoltaic power plants. This is mainly reflected in the following aspects:

1. The impact of dust on solar radiation

△Figure 1. Increased dust levels cause diffuse reflection from photovoltaic panels.

Dust on glass surfaces not only significantly affects solar radiation transmittance but also causes substantial diffuse reflection of solar radiation reaching photovoltaic modules, increasing the reflectivity of the modules. Experimental measurements show that accumulated dust on glass surfaces can lead to a loss of 5%-30% of solar radiation.

2. The effect of dust calcification and deposition on light transmittance

Dust contains a large amount of calcium and magnesium oxides. When dust on the surface of photovoltaic modules comes into contact with rainwater, a small amount of calcium and magnesium ions dissolve into the rainwater and re-adhere to the glass surface of the photovoltaic modules, forming a thick and hard layer of calcium and magnesium scale. Once scale forms, it is not easy to remove and will seriously affect the power generation performance of photovoltaic modules. In severe cases, it can even cause problems such as hot spots on some photovoltaic cells.

• The impact of dust on the temperature of photovoltaic modules

Dust adhering to the surface of tempered glass can absorb some solar radiation and convert it into heat energy, thereby increasing the operating temperature of photovoltaic modules, affecting heat dissipation, and further amplifying the heat-temperature effect of photovoltaic modules, thus affecting photovoltaic power generation performance.

Therefore, during the operation and maintenance of photovoltaic systems, special attention should be paid to the impact of dust on photovoltaic surface on photovoltaic power generation. Timely cleaning is necessary to keep the surface of photovoltaic modules clean and ensure the efficient operation of the power station.

To address the issue of dust monitoring in photovoltaic modules, we used the XIGUANG-003 hyperspectral satellite to image the target area on January 1, 2024. The resulting image had a ground resolution of 10 meters and a wavelength of 150 spectral bands. We identified and extracted the photovoltaic panel areas and assessed the dust coverage based on reflectivity.

△Figure 2 Schematic diagram of a photovoltaic power station in the desert

△Figure 3 Regular cleaning of photovoltaic panels

△Figure 4 Location of Benban Photovoltaic Power Station

△Figure 5. Hyperspectral true-color image of the study area after fusion

We fuse high-resolution panchromatic data with hyperspectral data to obtain high-resolution hyperspectral data, and then perform radiometric calibration and atmospheric correction to obtain surface reflectance data, thereby analyzing the reflectance of each band in the photovoltaic panel area.

△Figure 6 Landmarks near Benban Photovoltaic Power Station

△Figure 7 Reflectance of ground features near the Benban photovoltaic power station (reflectance values ​​magnified 1000 times)

As shown in Figures 6 and 7, samples of photovoltaic panels, centrally irrigated farmland, and other land surfaces were labeled. It can be seen that the reflectance curves of these three types of land features show significant differences. The reflectance of desert surfaces is high across all wavelengths, while the reflectance of farmland is higher than that of photovoltaic panels in the 450-560nm range, and then decreases. Therefore, based on these differences in reflectance, machine learning methods can be used to classify land features.

△Figure 8 Wind rose diagram of the study area in 2023

We also obtained daily wind direction data for the study area in 2023, and calculated the wind direction frequency to create a wind rose diagram to verify the spatial distribution of dust accumulation, as shown in Figure 8. The prevailing wind directions in the study area are northwest and north-northwest. We marked the prevailing wind directions on the dust coverage map in Figure 10.

*Data source: Prediction of Worldwide Energy Resource

△Figure 9 Differences in reflectivity of photovoltaic panels with different levels of dust accumulation

Figure 10. Dust cover map of the study area on January 1, 2024

Based on the differences in reflectivity of photovoltaic panels, we created a dust coverage map. We then selected two regions, northwest and southeast, to verify these reflectivity differences, as shown in Figure 9. Photovoltaic panels with higher dust coverage exhibited significantly increased reflectivity. Furthermore, the spatial distribution of dust coverage followed a decreasing trend from northwest to southeast, largely coinciding with the prevailing wind direction. While actual conditions may be affected by variations in the angle of the photovoltaic panels or different zoning patterns, this map provides a relatively intuitive overview of dust accumulation.

After acquiring multi-scene data, it is possible to further analyze dust accumulation patterns through time-series information, predict dust coverage by combining meteorological data, accurately estimate the next cleaning time, and calculate appropriate cleaning cycles by combining power generation statistics, thereby rationally scheduling cleaning tasks in advance, improving production efficiency, and reducing cleaning costs. Simultaneously, it is possible to combine ground experiments to analyze the relationship between reflectivity and the actual quantification of dust coverage, further improving the accuracy of dust coverage monitoring.