A Digital Twin Design Applied to a Photovoltaic Conversion System

Abstract

This article presents the creation of a digital twin for a photovoltaic conversion chain and proposes a predictive model for the electrical parameters of this chain using the LSTM (Long Short-Term Memory) model. The digital twin enables accurate,

real-time simulation of the photovoltaic system, integrating short- term forecasting to optimize performance. The LSTM model is

employed to predict critical electrical parameters such as DC current, DC voltage, AC current, and AC voltage, based on

environmental and electrical data collected. The study demon- strates that the LSTM model can effectively capture temporal

dependencies and provide accurate predictions, contributing to better management and optimization of the photovoltaic system. The results show excellent model performance with minimal errors, and the article discusses the challenges related to the observed discrepancies as well as opportunities to improve forecast accuracy.