Abstract
A significant quantity of sensors distributed throughout the natural gas pipeline are susceptible to errors. Timely diagnosis of sensor faults in such scenarios holds great significance in averting catastrophic failures. This paper proposes a novel approach termed as model-based multi-sensor fault detection, isolation, and accommodation (MM-SFDIA) technique to mitigate multiple sensor faults occurring simultaneously in large-scale distributed systems. The proposed approach leverages a distributed filtering framework, employing multiple local ensemble Kalman filters (EnKFs). Each individual local filter generates a distinct local state estimation using a distinct set of sensor measurements. By analyzing the differences among these local state estimates, a strategy based on state consistency, the faulty sensors are identified. Furthermore, an adaptive thresholding technique is devised to ensure resilient fault detection and identification. Compared to the existing state-of-the-art techniques, the proposed approach offers a lower computational burden and is applicable to high dimensional nonlinear systems with numerous sensor faults. Moreover, the results affirm the effectiveness of the proposed architecture, demonstrating a high accuracy and low execution time in detecting and isolating multiple sensor faults.