Impedance characteristics of grid-connected inverters are an important basis for analyzing inverter-grid interaction stability, and accurately obtaining impedance characteristics under different operating points is essential for stable operation. To address the problem that practical measurements can cover only a limited number of operating points, a multi-operating-point impedance identification method for grid-connected inverters based on a deep operator network structure is proposed to accurately characterize impedance under multiple operating points. Firstly, for multi-operating-point impedance identification, an identification model is constructed in which the trunk network performs frequency feature mapping and the branch network represents operating-point information. During the training stage, a frequency-band-weighted loss function is constructed, and constraints on identification errors in key frequency bands are introduced to achieve impedance identification under multiple operating points. Secondly, a small-sample impedance transfer model for parameter variation scenarios is proposed. A source-domain pretrained model and a target-domain fine-tuning strategy are introduced to improve the impedance identification performance under small-sample conditions. Finally, case studies verify that the proposed method achieves satisfactory multi-operating-point impedance identification performance and maintains good applicability in parameter variation and multi-inverter parallel scenarios.