This study proposes a real-time methodology for estimating the energy efficiency of battery-powered motor drive systems under dynamic conditions. Instantaneous power estimation can be distorted by energy storage components-such as capacitors, inductors, and the moment of inertia-combined with position-dependent effects. To address this, physics-based models of each storage component are developed to compute energy flows at the component level, including the battery, inverter, and motor. The method further incorporates variations in driver input commands, enabling stable and accurate efficiency measurements even during transient operations. The proposed approach can also be extended to support component-level loss analysis and separation, fault diagnosis, and prognostics and health management (PHM) techniques.