Machine learning applications for predicting liquid fraction in a PV system with NEPCM and fins

Abstract

The heat-absorbing and heat-releasing properties of phase-change materials (PCMs) at specific temperatures make them ideal for improving the thermal management of solar systems. Thermal conductivity of PCMs is increased when combined with nanoparticles, leading to significant improvements of their ability to manage and transfer heat efficiently. The objective of this article is to anticipate liquid fraction (LF) with a suitable level of accuracy by utilizing machine learning predictive models in a building-integrated photovoltaic (PV) system. Four different system configurations were proposed, namely PCM without fins, PCM with longitudinal fins, PCM with the additional of nanomaterials and longitudinal fins, and PCM with the additional of nanomaterials and Y-shaped fins. After accomplishing the numerical simulation data, in order to save computational cost and maintain accurate predictions, the following models were chosen to accomplish this objective: linear regression, lasso regression, polynomial regression, and an Auto-Regressive Integrated Moving Average (ARIMA) model. Then, the comparison of them in all cases, using the root mean squared error (RMSE), as well as mean absolute error (MAE). The ARIMA model had superior performance in predicting PCM LF, achieving an RMSE of 0.0149 and an MAE of 0.0093. This technique has been used to predict the LF of the other situations. The results indicated that the case incorporating PCM with the inclusion of nanoparticles and longitudinal fins exhibited the shortest time required to reach the LF value of 1. The approach used in this study reduces the reliability on computationally expensive simulations, which provides a more efficient method for optimizing PV systems. The study also highlights the potential of machine learning to enhance thermal management of PV cells, contributing to more efficient and sustainable energy solutions.

Keywords

• Machine learning
• Prediction models
• Time series analysis
• Phase change materials