Representing the complex operational and navigational behavior of an ocean-going vessel with a single linear model is challenging due to inherent nonlinearities and diverse operational conditions. However, vessel operations can be segmented into distinct regions, where localized linear models can effectively capture its dynamic behavior. This study presents a data-driven localized model for predicting ship performance, supporting Digital Twin (DT) development. The presented method leverages Dynamic Mode Decomposition with control (DMDc) to capture localized linear dynamics. To achieve this, vessel operation data is first partitioned into distinct operating regions using Gaussian Mixture Models (GMM) coupled with the Expectation Maximization (EM) algorithm. Applying Singular Value Decomposition (SVD) on these operating regions reveals that the data within each region lies on a low-dimensional hyperplane, enabling linear approximations. Subsequently, DMDc is applied within each operating region to derive localized state-space models. The resulting framework delivers an interpretable, computationally efficient DT capable of predicting ship performance and fuel consumption across the full operational range, supporting green shipping strategies and emission reduction initiatives.