Design optimization and development of an explainable multi-objective optimization framework: A-CAES system integrated with advanced thermal energy storage

Abstract

Adiabatic compressed air energy storage (A-CAES) system integrated with advanced packed bed latent thermal energy storage (PBLTES) subsystem is of great interest for next generation energy storage technology. However, the integration of A-CAES system with PBLTES subsystems currently faces heat accumulation problems and lacks a comprehensive optimization framework. To overcome the above challenges, the present study presents an indirect design approach for the thermal oil-based cascade PBLTES system integrated with A-CAES system. Based on this approach, we conduct a parametric study on the interaction between the PBLTES system and A-CAES system in terms of operating parameters and structural design. Furthermore, a multi-objective optimization framework is constructed using an explainable hybrid machine learning (ML) model. At last, a systematical comparison with other A-CAES with different configurations PBLTES system is made to evaluate its final economic benefits. Results indicate that no linear relationship exists between most parameters and the round-trip/exergy round-trip efficiency (RTE/ERTE). The SHAP analysis further reveals that the capsule size is the most influential factor while considering the thermodynamic indicator for RTE. Based on the TOPSIS, the optimal results of RTE, the annual CO₂ emission reduction (ACER) and total annual costs (TAC) of the new system are 71.62%, 8336 t and 1.13 m$/year. The final profit of the optimal system is 22.24% and 12.69% higher than that of the thermal oil configurations and direct design of PBLTES in A-CAES, respectively. This paper provides reference for different designs and optimizations of A-CAES systems.