Improving Solar Energy Monitoring: Advanced Deep Learning Predictive Model for Photovoltaic Power Generation

Abstract

This paper investigates an approach to improve the reliability of photovoltaic (PV) systems by developing advanced predictive models. With a particular focus on early detection of possible deficiency or beneficiary picks in solar PV production, this research explores techniques based on Deep Learning (DL) for refining predictive models, as well as an integrated mechanism for measuring the deviation of photovoltaic production from its normal output. Thereby providing enhanced monitoring of PV installations. The results suggest significant improvements in the ability to predict and prevent malfunctions, paving the way for improved performance and sustainability of PV systems. In addition, this research establishes a comparison between the Long Short-Term Memory (LSTM) and Bidirectional LSTM (Bi-LSTM) models, highlighting the effectiveness of the latter in the advanced prediction of solar PV production with an MAE of 1.24 kWh, and a MAPE of $13.80 \%$. This contribution offers significant perspectives for the scientific community, providing improved monitoring and prediction methods to ensure the efficiency of PV systems.