<p>Accurately identifying ice loads acting on a ship is crucial for ensuring the structural safety of polar vessels. To mitigate the severe ill-posedness caused by limited measurement data and high sensitivity to noise, regularization techniques are widely used in ice load identification. However, existing regularization strategies often struggle to balance the preservation of transient impact peaks with the reconstruction of smooth background fluctuations, leading to ice load identification results that lack physical plausibility. To address this, this study proposes an ice load identification method based on a hybrid regularization strategy. First, Green’s kernel functions are used to establish a dynamic mapping between the structural strain response and impact ice loads, thereby formulating the inverse problem model for ice load identification. On this basis, a hybrid regularization model combining the advantages of the L1 and L2 norms is introduced, designed to simultaneously capture the peaks of impact loads while maintain the continuity of low-amplitude background loads. The corresponding objective function is efficiently solved using a coordinate descent algorithm, and the optimal set of regularization parameters is selected based on the Bayesian Information Criterion (BIC). Finally, the performance of the proposed method is evaluated through high-fidelity numerical simulations of the local structure of an ice-class research vessel, and its engineering applicability is validated through scaled model experiments. The results indicate that the hybrid regularization strategy not only enhances the accuracy of peak load identification but also effectively mitigates signal truncation in low-amplitude regions, demonstrating a balanced overall performance. In multi-source impact scenarios, it exhibits high robustness and low sensitivity to interference. The proposed method provides an effective solution for ice load identification in polar vessels.</p>