The increasing computational demands of high-fidelity simulations in underground mine ventilation, particularly under transient, turbulent, and multi-parameter operating conditions pose significant challenges for real-time prediction and digital twin deployment. Proper Orthogonal Decomposition (POD) offers an intrusive projection-based model reduction strategy capable of extracting low-dimensional structures governing the dominant flow dynamics. In this work, we develop a POD-based reduced-order model (ROM) specifically tailored to a large-scale mine ventilation system. The methodology includes systematic snapshot generation across varying conditions, eigen-decomposition of the correlation matrix to extract the most energetic modes and Galerkin projection onto the POD subspace to obtain a reduced dynamical system that preserves the key transient behaviors of the full-order equations. The proposed POD-ROM is evaluated against high-resolution simulations of an industrial axial ventilation system. This demonstrates that the POD representation effectively captures the dominant flow dynamics, including the characteristic unsteady structures generated by the ventilation system, as well as the overall performance trends of the airflow network while reducing computational costs by several orders of magnitude. A detailed error analysis highlights the influence of mode truncation and nonlinear term reconstruction on model fidelity. Limitations of the classical POD-Galerkin formulation in handling strongly nonlinear or off-design conditions are discussed, motivating future integration of others reduction techniques and parametric mode interpolation to extend the robustness of POD-ROMs. Overall, this study confirms the relevance of POD as a foundational step toward real-time digital twins for sustainable and energy-efficient mine ventilation systems.