Quantifying soiling and environmental stress impacts on rooftop photovoltaic performance in a coastal industrial environment.

This paper presents a field-based performance evaluation of the rooftop photovoltaic (PV) systems operating under hot-arid coastal industrial environmental conditions. Two identical 5-kilowatt (kW) grid-connected PV systems were installed and commissioned for continuous monitoring of system parameters using a high-resolution data-acquisition system to quantify the effects of soiling, humidity, temperature, and industrial atmospheric pollutants on energy yield and system efficiency. A controlled comparison was conducted by applying a cleaning regime to one array, enabling direct assessment of soiling-induced performance losses under real operating conditions. Cleaning was done every 60 days for module 1, while module 2 was left uncleaned to simulate soiling accumulation. Continuous logging of irradiance, temperature, wind speed, voltage, current, and energy yield allowed direct correlation of environmental conditions on the performance of the PV systems. Finally, four machine learning models optimized using Improved Particle Swarm Optimization (IPSO) are used to classify rooftop photovoltaics (PV) as clean or dirty. The proposed models are implemented to estimate the power yield of PV systems in two categories (cleaned and non-cleaned). Results highlight the significant influence of soiling and environmental stress on PV system performance, demonstrate the role of maintenance frequency in mitigating efficiency losses, and provide empirical evidence to support optimized cleaning strategies. The findings support maintenance planning and provide site-specific evidence for cleaning decisions in coastal industrial PV installations.