PULSE-72: Single-Cell Pulse and Drive-Cycle Dataset for ARGUS (Mixed-Training Deep Learning for Equivalent Circuit Parameter Identification in Lithium-Ion Batteries)

The dataset was acquired from tests of a commercial LG INR18650 MJ1 lithium-ion cell (nominal capacity 3500 mAh).For equivalent-circuit parameterization, we used voltage responses to rectangular galvanostatic current pulses with durations from 9 to 144 s and amplitudes of approximately 0.5C, 1C, 2C, and 3C, in both charge and discharge directions, within the mid state-of-charge region (45–55% SoC).In total, the pulse library contains 72 measured windows (24 excitation templates × 3 SoC levels), with reference ECM-2 parameters obtained by least-squares voltage-error minimization. To validate parameter transfer under dynamic operation, the dataset also includes 18 drive-cycle profiles from the same cell: UDDS, NEDC, and WLTC at 25°C and 35°C with three replicates per stratum.These data support a full workflow: pulse-based ECM identification followed by forward voltage validation on realistic load trajectories. This Zenodo record is the data companion of the ARGUS (Mixed-Training Deep Learning for Equivalent Circuit Parameter Identification in Lithium-Ion Batteries) article and is intended for reproducible benchmarking of physics-based and hybrid data-driven ECM identification methods.