Artificial Neural Network-based digital twin for a flat plate solar collector field

Abstract

In this study, a digital twin for a flat plate solar collector field is proposed. This kind of system is used to reduce carbon dioxide emissions in bioclimatic buildings to convert them into Zero Energy Buildings. The core of the digital twin is an Artificial Neural Network prediction model, which is a good alternative to models based on physical equations for modeling systems with strong non-linearities, such as the ones found in flat plate solar collectors. The Artificial Neural Network prediction model is calibrated and validated with data saved during one year of operation comprising sunny days, cloudy days, partially cloudy days and non-operation days. Validation shows good results using several statistical metrics, suggesting that the Artificial Neural Network model is suitable for operation and control purposes. With a highly accurate virtual representation, the Artificial Neural Network model allows data analysis of the plant operator, prediction of behavior, and offers recommendations for optimizing system performance. In addition, the digital twin presented as part of this work is not just limited to the model, but is also enriched by the integration of data acquisition technologies and a user interface into a web page. This innovative integration establishes a robust framework for proactive, real-time decision-making and efficient management of the plant, ensuring enhanced system operation and sustainability.