Robust unknown input observer based estimation and mitigation against sensor spoofing attacks in cyber-physical microgrids

Microgrids are increasingly dependent on real-time sensing and network communication for the monitoring and control of distributed energy resources (DERs), exposing them to numerous cyber attacks such as sensor spoofing. In such attacks, additive biases are injected into sensor measurements without directly spoofing the communication network and control logic. Conventional observer-based bias estimation methods face several challenges. They typically rely on accurate system models, and their estimation performance deteriorates in the presence of system uncertainties, including parameter uncertainties, process and measurement noise, and unknown inputs. These factors make sensor spoofing-induced biases difficult to distinguish from normal state variations, causing interaction between sensor biases and system uncertainties and leading to degraded control performance in practical microgrid operation. To overcome these limitations, this paper proposes a Robust Unknown Input Observer (RUIO) for the accurate estimation of sensor spoofing-induced biases in the presence of various system uncertainties. The sensor bias is modeled as an augmented state, and the observer gain is synthesized using Linear Matrix Inequality (LMI) optimization under an H∞ performance criterion to ensure robustness against system uncertainties. Given the structural observability limitation of the Unknown Input Observer (UIO) formulation, which prevents simultaneous estimation of independent current and voltage sensor biases, an event-triggered bias estimation and mitigation framework is designed to counter any single sensor spoofing attack. The dual-RUIO scheme is switched adaptively based on the current-sharing error after mitigation, with an appropriate holding interval in place to ensure stable operation. The effectiveness of the proposed RUIO-based framework is validated through extensive experiments on a cyber-physical microgrid testbed. Comparative evaluations against conventional state estimation methods demonstrate improved estimation accuracy and mitigation performance.