Digital-Twin-in-the-Loop Dual-Layer Time–Frequency Decomposition and Capacity Configuration of Hybrid Energy Storage for Wind Power

INTRODUCTION: Conventional hybrid energy storage sizing often relies on fixed frequency thresholds and insufficiently considers device constraints and economic objectives. This study proposes a digital-twin-in-the-loop capacity configuration method based on dual-layer time–frequency decoupling. METHODS: ICEEMDAN was used to decompose wind power, while power, energy, and duration characteristics were extracted for mode classification. SGMD was applied only to transition modes identified from battery, supercapacitor, and PCS capabilities. SAC determined the SGMD parameters and critical-mode allocation ratios. Initial capacities were obtained from the 95th-percentile power and cumulative energy trajectories and corrected through multi-scenario digital-twin evaluation. RESULTS: For a 50 MW wind farm, the final battery and supercapacitor capacities were 1.53 MW/2.65 MWh and 0.50 MW/0.10 MWh, respectively. The total cost was RMB 31.2561 million, corresponding to reductions of 15.75% and 8.76% relative to the single-layer ICEEMDAN configuration and the fixed-parameter ICEEMDAN+SGMD configuration considered in the component-wise comparison. Rapid grid-power fluctuations were also reduced. CONCLUSION: The proposed method coordinates time-scale separation, device constraints, and configuration economy, providing an effective approach for wind-power hybrid energy storage planning.