The paper first presents a gradual analytical model-order reduction of a moving boundary method-based 12th-order lumped-parameter nonlinear vapour compression cycle model. Model-order reduction is conducted by introducing certain assumptions and replacing state variables associated with fast dynamics with static expressions. Next, numerical tools implemented in MATLAB/Simulink and applicable to black-box heating, ventilation and air-conditioning (HVAC) models are developed for the purpose of obtaining the model static input-output maps and linearized model parameters for a wide range of operating conditions. The linearized models are used for comparison of pole-zero maps of high- and low-order models, which shows that slower, dominant dynamics are preserved in the reduced-order models. Finally, as an alternative to model-order reduction approach, a multi-input/multi-output model identification approach is proposed, which can provide a low-order linearized model directly from simulated time response of a full-order lumped- or distributed-parameter model, or from real vehicle HVAC test signals.