Design of an improved aerosol-aware intelligent framework for dust severity prediction and performance recovery in solar photovoltaic (PV) plants

To ensure the reliable operation of solar panels, it is essential to understand the impacts of aerosol and dust pollution. These pollutants can adversely affect solar irradiance transfer, power generation, performance ratio, cleaning expenses, and the project's economic viability. In the framework of the current investigations on photovoltaic (PV) soiling, PV panels are categorized as either clean or dusty. The predominant research on PV soiling has depended on established threshold values to assess the severity of soiling. Nevertheless, only a limited number of studies have demonstrated a correlation between the chemical composition of particulate matter (dust/aerosol) that contaminates PV modules and its impact on power derating, cleaning, and recovery at the PV plant level. This research introduced AeroDust-PVNet, an innovative integrated artificial intelligence system employing five distinct analysis methodologies. Spectral, photothermal, environmental, chemical, thermal, and radiometric analysis-based DustFormer employs optical, thermal, aerosol, meteorological, and electrical data to identify four distinct categories of aerosol and dust pollution. The contaminants comprise mineral dust, soot, pollen, cement particles, and various aerosol pollutants. The dust–aerosol severity index network system generates the type-conditioned severity index and the power derating factor for dust-aerosol pollution. Compared to traditional cleaning methods, the CLEAN reinforcement learning optimizer enhances cleaning duration, frequency, and intensity while utilizing reduced water and energy. The PV-recovery evaluation and calibration Digital Twin assesses yield and residual loss post-cleaning. The results of AeroDust-PVNet are assessed using the VALI-X Benchmark Engine, in contrast to previous methodologies. The AeroDust-PVNet framework enhances precision, cleaning efficacy, PV system outputs, and dependability.