Dynamic models represent the temporal responses needed for control, grid interaction, and fault diagnosis in commercial-building heating, ventilation, and air-conditioning (HVAC) systems. Existing reviews have largely examined these applications separately, leaving cross-functional model transferability and implemented operational interfaces unclear. This systematic review compares the roles, requirements, and supporting evidence for dynamic models in control optimization (CO), operational energy flexibility (EF), and fault detection and diagnosis (FDD). The literature is synthesized according to functional requirements, transferable structures and interfaces, implemented health-to-action coupling, model grounding, evaluation settings, and deployment stages. CO and operational EF share a need for action-conditioned prediction, while EF additionally requires grid-responsive estimates of response magnitude, duration, and recovery. Reuse in FDD is conditional on whether the model preserves fault-sensitive states, outputs, parameters, or residuals, so shared model services do not eliminate the need for application-specific outputs and decision logic. Implemented coupling is confined to a small set of subsystem-specific fault-aware control interfaces, and no eligible end-to-end implementation coupling FDD with operational EF was identified within the peer-reviewed commercial-building corpus. Evaluation remains dominated by simulation, while laboratory, hardware-in-the-loop, and field studies are less common and provide different evidence about real-time operation and deployment. The findings support shared data and model-service infrastructure with application-specific models, explicit health-to-action interfaces, and evaluation settings matched to the intended claims.
