Electrical power distribution systems are composed of heterogeneous components, which include continuous power sources, discrete relays, passive and active loads, and fastswitching power conversion subsystems. This heterogeneity introduces significant challenges for model-based diagnosis, such as building accurate models, and generating fast and accurate diagnoses while ensuring robustness to measurement noise and modeling errors. In this paper, we present a comprehensive methodology for the diagnosis of parametric and discrete faults in electrical power distribution systems that include dc and ac components. We use a hybrid bond graph modeling language to systematically develop computational models and algorithms for hybrid state estimation, robust fault detection, and efficient fault isolation. Simulation and experimental results on a real-world electrical power distribution system demonstrate the effectiveness of our methodology.