A surrogate-based computational framework for optimizing thermal strategies in large multi-subsystem borehole heat exchanger sites

At large multi-subsystem borehole heat exchanger (BHE) sites, the overall thermal efficiency and heat production of the system can be significantly enhanced by optimizing the distribution of heating and cooling loads among subsystems, while honoring the technical operating parameters. However, practical factors such as heat regeneration, site-specific conditions, and complex BHE layouts complicate the optimization task. To address these challenges, this study proposes a surrogate-based computational framework designed to efficiently and globally optimize thermal strategies, including maximizing geothermal heat production and optimally distributing heat loads among subsystems. The proposed framework consists of two main components: surrogate model development and optimal thermal strategy search. As a demonstration, it is applied to an operational large BHE site. To efficiently generate training samples, a support vector classifier is used to refine the potential parameter space of thermal strategies. After obtaining sufficient samples, the support vector regressor is selected as the surrogate model, following a performance comparison with other regressors. These efficient mathematic techniques minimize computational costs while improving model prediction accuracy. Finally, by integrating a particle swarm optimization algorithm, the framework identifies the maximum allowable heat extraction for the study site and determines the thermal load configuration for each subsystem. By leveraging measured operational data, the computational framework considers critical site-specific factors such as groundwater flow, heat regeneration, and geological and hydrogeological stratigraphy, making it highly practical and adaptable. This approach ensures that the system operates efficiently and sustainably, offering a robust solution for managing complex BHE arrays within diverse real-world scenarios.