Efficient Design Optimization of Thermal Battery Using Multi-Fidelity Surrogate Modeling

Abstract

Thermal battery (TB), a promising energy reserve system, has been widely studied due to its long-term storage capacity. However, due to exceptionally high computing costs of TB simulation models, research on the optimization of TB systems has received limited attention. To address these issues, an efficient TB design optimization methodology is developed to maximize the volumetric energy density of TB while satisfying target performances. In order to accurately and efficiently predict engineering performances of the TB system, a multi-fidelity (MF) surrogate method, which integrates high-fidelity (HF) and low-fidelity (LF) data, is applied. First, the detailed and effective heat transfer models of TB are used as respective sources of HF and LF data for a single working condition. Second, the MF surrogate model is generated for different working conditions utilizing a small number of HF data in the corresponding working condition and pre-obtained LF data from a different working condition. Third, a modified MF dataset selection indicator is proposed to maximize the use of MF surrogate models. The numerical results demonstrate that the proposed approaches greatly enhance the optimization efficiency by up to 60% while sacrificing little accuracy compared to the conventional one.

Keywords: Thermal battery, heat transfer, Multi-fidelity surrogate-based optimization, Modified indicator for multi-fidelity dataset selection, Prior knowledge