Machine Learning-Based Prediction of PV Power Considering Dust and Environmental Factors

Solar energy is one of the most important forms of renewable energy for generating electricity. The efficiency of Photovoltaic (PV) panels that produce electricity depends on the continuously changing weather parameters like dust, wind speed, wind direction, ambient temperature, module temperature, relative humidity and most importantly the solar irradiation. The dust particles in the atmosphere, like the other weather parameters, is also changing continuously and plays a major role in reducing the sunlight on the PV panels by scattering, absorbing or reflecting the incoming sunlight. This leads to reduced efficiency in the production of electricity and consequently the DC power generated from the PV panels will be less. This paper aims to build a model using machine learning algorithms to predict the output DC power due to the dust (PM 2.5 and PM 10 concentrations are considered in the dataset), wind speed, wind direction, the ambient temperature of the region where the PV panels are placed, the module temperature, the relative humidity. Using this dataset, a random forest model (by implementing a machine learning algorithm known as the Random Forest Algorithm) was developed and implemented. This model achieved a mean absolute error (MAE) of 482.02 Watts and an R-squared score of 0.873 and can be used to achieve optimized performance and efficient energy conversion of PV systems.