This paper presents a self-adaptive digital twin (DT) framework based on the broad learning system. Existing DT methods in integrated energy systems (IES) are constrained by high data demand, limited time resolution, and poor adaptability to unseen conditions. To address these limitations, we develop an online parameter adaptation algorithm that maintains long-term consistency between the DT and the physical system. The framework is validated through a Hardware-in-the-Loop experiment, establishing a dynamic DT with second-level time resolution for a commercial air-source heat pump. Experimental results demonstrate strong self-adaptation and generalization capabilities, with high accuracy achieved under both experienced and unseen conditions. • Adaptive DT ensures consistency with physical system across varying conditions. • Online incremental updates complete in ∼ 40 s, enabling real-time DT adaptation. • DT snapshot with second-level resolution keeps 90.1% of power errors within 5%. • Hardware-in-the-Loop experiments bridge the gap between DT and real-world deployment.
Fu et al. (Sun,) studied this question.