An Artificial Intelligence and Machine Learning Model to Estimate the Cleaning Periodicity for Dusty Solar Photovoltaic (PV) Modules in A Composite Environment

Abstract

Solar energy is harnessed on a considerable scale nowadays. By 2030, the solar power output is expected to increase to 2500 GW marginally. High cell temperatures and soiling significantly affect the performance of solar photovoltaic systems. This study clarifies the effect of dust deposition on the transmission and output power of photovoltaic modules. The analytical and machine-learning models were developed to analyze the effects of soil deposition on the photovoltaic panels. The field data were used to train and test the algorithm for developing the machine-learning model. An optimum cleaning and maintenance schedule is then proposed based on the site's environmental conditions. The novelty of the research was to gather environmental parameters in real-time conditions that affect the soiling rate of photovoltaic panels, further affecting the conversion efficiency of photovoltaic panels. Based on the theoretical model developed, the cleaning frequency of the module was observed to be 18 days, considering 5% power loss and dust density accumulation of 2g/m2. A random forest model was developed considering ambient temperature, solar irradiance, relative humidity, wind speed, dust concentration, and energy generated. The predicted cleaning frequency is observed to be 25 days using the random forest model.