Hybrid framework for surrogate modelling of massive solar collectors in road pavements

Abstract

This paper investigates the application of surrogate modelling in the design and thermal response assessment of Pavement Solar Collectors (PSCs). The PSC system is a sustainable infrastructure solution that utilises both solar and shallow geothermal energy. PSCs incorporate a network of pipes embedded in the asphalt layer to create a heat exchange layer. During warm months, water circulating through this layer captures solar heat, which can then be used for snow melting in winter, enhancing road safety, or for domestic and industrial heating applications. Finite Element (FE) analysis is a widely used method for evaluating the thermal response of PSCs to optimize their design. However, the substantial computational requirements of numerical modelling, especially for long-term time-dependent analyses, pose significant challenges in assessing the long-term thermal behaviour of PSCs. Surrogate models, approximating complex physics-based simulations, drastically reduce computational demands, enabling rapid and accurate evaluations of various design parameters and scenarios. In this study, a validated FE simulation framework was employed to generate data, which was then used to develop a data-driven surrogate model for PSCs. In order to refine the surrogate model's performance to its optimal level, hyperparameter optimisation was carried out. The comparison of outlet water temperature results between finite element and surrogate models showed a high correlation, with a coefficient of determination of 0.97 observed for both training and test data sets. Subsequently, the surrogate model was integrated as an objective function in a Particle Swarm Optimization (PSO) algorithm to automate the Heat Harvesting Capacity (HHC) optimisation of PSCs. The PSO algorithm demonstrates robust performance in identifying optimal solutions while also offering a substantial reduction in computational costs compared to FE simulations.