Estimation of Partial Shading and Soiling Percentage for a Solar Panel using an Artificial Neural Network Model

The power generated by solar panels is affected by irradiation levels, ambient temperature, shading (nearby trees and objects which can fall on panels), and faults (open circuit, and short circuit faults). The data of annual shading helps to decide whether the panels need to be shifted to a different location or to identify the cause of shading and eliminate it. This paper presents a neural network model to calculate the percentage shading over a particular module using output power, irradiation, and temperature as inputs. A comparison using the various training-to-validation ratio is carried to decide the best ratio for the model. This paper also proposes a monitoring system of a PV panel using Arduino UNO microcontroller, connected to the voltage sensor, current sensor, temperature sensor, and humidity sensor. The data acquired from the sensors are stored in an SD card using an SD card module connected to the Arduino. Further analysis and development of mathematical and prediction models consider the data acquired for a window of 10 seconds over a particular period.