System-level data-driven identification of complex closed Brayton cycles via dynamic mode decomposition

System identification is crucial for understanding complex systems, and as a powerful data-driven identification tool, dynamic mode decomposition deserves an expanded focus, from applications in individual components to full-system analysis. This paper develops a Modelica-based multi-loop helium-xenon closed Brayton cycle system model, symmetrically arranged and visualized, capable of simulating and generating full-system datasets for identification. The dynamic performance of the cycle, using a nuclear power system as the heat source, is investigated under varying loads. The steady-state simulation results show a maximum error of about 3 %, and the transient response aligns well with the references, demonstrating the model's reliability. Data-driven identification techniques, including dynamic mode decomposition and its extension with control, were then employed to predict system evolution under dynamic and transient conditions, with prediction errors within 1 % and 5 %, respectively. These results highlight the effectiveness of these dynamic identification approaches, which maintain numerical accuracy while enhancing computational efficiency. This study provides a reference for modelling and transient operation of multi-loop closed Brayton cycle systems and proposes a general, data-driven system identification and rapid prediction framework for transient and real-world scenarios with noise in complex energy systems, highlighting the potential to extend such approaches to comprehensive system-wide analysis, and offering profound implications for optimization and operations management at broader system levels.