Multi-Domain Cyber-Physical State Estimation Using Unscented Kalman Filter: Application to Battery-Supported DC Microgrids

Cyber-physical systems require the simultaneous estimation of heterogeneous states associated with physical dynamics, component health, environmental conditions, and cyber disturbances. However, their heterogeneous dynamics and interacting effects make joint estimation challenging. This paper presents a unified multi-domain state estimation framework based on an augmented unscented Kalman filter (UKF) for the joint estimation of physical, health, environmental, and cyber states. The formulation accommodates physics-based, data-driven, and hybrid state transition and measurement models, allowing adaptation to different levels of model knowledge. A battery-supported DC microgrid is considered as a representative application, with the augmented state capturing the electrical dynamics, battery state of charge and state of health, available photovoltaic power, load demand, and cyber-induced voltage measurement corruption. Domain-specific process uncertainty accounts for heterogeneous state evolution, while a pragmatic post-update projection prevents physically inadmissible state values from propagating through the nonlinear model. Simulation studies under individual and combined multi-domain variations demonstrate the ability of the framework to jointly estimate and distinguish among interacting states across the four domains, supporting its applicability to nonlinear cyber-physical systems.