Abstract
This paper investigates an innovative state estimation technique implemented within an advanced distributed framework, aimed at reducing computational complexity while detecting multiple sensor faults in hydrogen-blended natural gas pipelines. The novel distributed estimation technique is based on the ensemble Kalman filter (EnKF) and is referred to as partial-distributed multi-sensor fault detection, isolation, and accommodation. The architecture includes a set of local EnKFs and an information fusion center. These local filters operate simultaneously to generate unique local state estimates based on a distinct set of sensor measurements, which are subsequently transmitted to the information fusion center for the computation of fault-free state estimates. To reduce computational complexity, the PD approach segregates nonlinear computations from the local filters and delegates them to the main filter. Additionally, a fault diagnosis strategy is developed based on local state residuals. Since each local filter generates a distinct local state estimate based on its unique set of sensor measurements, comparing the local state residual against a threshold facilitates the identification and isolation of faulty sensors. Furthermore, an adaptive thresholding approach is incorporated to facilitate effective fault identification and isolation. The proposed technique has proven to be effective in highly nonlinear, and high-dimensional systems with simultaneous multiple sensor faults. The effectiveness of the proposed approach is demonstrated through extensive simulations and comparative analyses.
