Neural Network Techniques for forecasting in Real-Time of Photo Voltaic system

Abstract

Effective forecasting of solar photovoltaic (PV) power in very short time intervals is essential for addressing the challenges of variability and uncertainty in renewable energy generation. Improved prediction accuracy directly contributes to greater grid stability, better operational planning, and reduced costs. PV power output is highly influenced by fluctuating environmental variables such as temperature, relative humidity, solar radiation, and wind speed. These conditions introduce significant unpredictability, making accurate modeling a priority. This paper investigates the application of deep learning approaches to forecast PV power in the very short term. Specifically, it explores the use of the Long Short-Term Memory (LSTM) neural network and assesses its performance in forecasting solar power output. The model’s effectiveness is benchmarked against other architectures, including Gated Recurrent Unit (GRU), Multi-layer Perceptron (MLP), and Convolutional Neural Network (CNN), using performance indicators such as Mean Absolute Error (MAE), Root Mean Square Error (RMSE), and the coefficient of determination (R²).