Artificial Neural Network Based Predictive Performance Analysis of Photovoltaic Output Under Various Solar Radiation Ranges

Abstract

Accurate assessment of photovoltaic (PV) power output under varying environmental conditions is essential for evaluating solar energy performance and optimizing its application. This study employs an Artificial Neural Network (ANN) approach to predict the output power of a solar panel using a combination of electrical and weather-related parameters. The input variables include short-circuit current (), open-circuit voltage (), maximum voltage (), maximum current (), efficiency (EFF), fill factor (FF), solar radiation (G), wind speed (v), ambient temperature (Tₐₘ), and panel temperature (Tₚₐₙₑₗ) under solar radiation intensity in the ranges of 200–400, 400–600, 600–800, and 800–1000 W/m². Three ANN models, ANN-10 (10 inputs), ANN-8 (8 inputs except and ), and ANN-6 (6 inputs , , , and ), are developed and compared based on their prediction accuracy using the coefficient of correlation (R) by varying a number of neurons in the hidden layer from 1 to 15. Among them, the ANN-8 model (8-10-1) demonstrates the highest prediction accuracy (R value of training, testing, validation, and overall model exceeding 0.999), particularly at high solar radiation levels (800–1000 W/m²). The results confirm that ANN-based modelling is a reliable and effective tool for forecasting PV output power under diverse operating conditions. This approach supports solar energy potential assessment and performance optimization across various geographical and climatic contexts.