Annual Forecast of Photovoltaic Power Generation Based on MLP Artificial Neural Networks

Abstract

The intermittency of solar energy resources presents a significant challenge in balancing power generation and load demand. To enhance system consistency, forecasting photovoltaic solar energy is crucial. Among numerous techniques, Artificial Neural Network (ANN) is an efficient tool that can help simplify this problem and predict photovoltaic power generation based on various inputs such as weather data and panel characteristics. In this paper, we present the results of an annual forecast of photovoltaic power generation based on Multilayer Perceptrons (MLP), which provides valuable insights into the potential of MLP ANN for accurate and reliable prediction of photovoltaic power generation, thereby improving the efficiency and reliability of photovoltaic systems. The results were obtained based on data collected over a year and validated with data from the following year. Mean Squared Error (MSE) was utilized to quantify the error between the predicted and measured photovoltaic solar energy generation. The analysis demonstrated that this annual forecast of photovoltaic power generation is highly accurate.