Short term performance investigation of solar PV module: A machine learning based approach

Abstract

This study presents a short-term performance analysis of the photovoltaic (PV) module considering weather impact in the context of Bangladesh by using machine learning. A Multilayer perceptron model is used to analyze the data and to predict the output. To collect the weather data and the output data, a weather station has been developed and deployed on the rooftop of a 7-story building in Gabtoli, Dhaka, Bangladesh. All the sensor data can be accessed remotely. In this study data from 1st November 2019 to 28th February 2020 are used in four separate data set for training purpose. It is observed that the output energy prediction improves with the increase in training data. The result shows that the temperature has the highest linear correlation with the module short circuit current among all the weather parameters i.e. humidity, wind speed, and air pressure.