Sensor Fault Detection in a Distributed Power System via Decentralized Kalman Filters

The problem of fault detection (FD) of large-scale systems has been considered in this paper. The proposed FD approach employs decentralized Kalman observers for simultaneous estimation of states and detecting sensor faults in the presence of noises in the model and sensory measurements. Then, decentralized model predictive controllers (MPC) has been designed based on the estimation of states. Also, the residual signals will be constructed via Kalman filters. The proposed FD scheme does not rely on the availability of full state measurements. The residual evaluation function is derived from a change detection algorithm called cumulative sum (CUSUM). The performance of the proposed FD approach is evaluated through simulations performed for a case study, namely FD of a distributed power system.