(560g) Process Monitoring and Leakage Diagnosis for Distributed Pipeline System Based on Discrete Observer and Moving Horizon Estimation

Distributed pipeline systems are increasingly constructed onshore and offshore for oil, gas and water transportation, which makes cheaper, safer and better product delivery available to local suppliers and consumers [1]. However, such benefits heavily rely on security working condition of the distributed pipeline systems, and any singularity (such as leakage, break and bend) may lead to catastrophe including casualties, property damage, resource loss, and trigger stringent environmental issues. Considering that, this work focuses on distributed pipeline system working condition monitoring and operating state estimation by using Luenberger observer design, moving horizon estimation (MHE) [2] and extreme learning machine (ELM) [3].

In terms of so-called water hammer equation based modelling, a linearized first-order hyperbolic partial differential equations (PDEs) based model is derived around equilibrium points of velocity, pressure and density by applying the linearization method. Given that Euler discretization method cannot guarantee the spatial discretization accuracy, Cayley-Tustin discretization scheme (Crank-Nicolson integration scheme) is applied in this work so that the energy and quadratic invariants are preserved without any model reduction or spatial approximation of distributed parameter system [4-5]. Given that a discretized-time distributed pipeline model is a infinite-dimensional system, a discrete Luenberger observer [6] is designed for pipeline process monitoring and ultimately the-state-of-art moving horizon estimation is utilized for multi-input and multi-output (MIMO) infinite-dimensional pipeline system state estimation under coloured noise and physical constraints. In particular, various leakage scenarios are generated by adjusting the boundary condition of upstream velocity and downstream pressure and ultimately extreme learning machine algorithm is utilized for leak detection, localization and size estimation intelligently. Finally, a series of simulations are presented based on a long-range infinite dimensional pipeline system and the proposed method is tested and demonstrated.

[1] Billmann L., Isermann R., 1987. Leak detection methods for pipelines. Automatica. 23 (3): 381-385.

[2] Rawlings J. B. 2009. Model Predictive Control: Theory and Design. Mathematics in Science and Engineering. Madison, WI: Nob Hill Publishing, LLC. p. 576. ISBN 978-0-9759377-0-9.

[3] Huang G., Zhu Q., Siew, C. K. 2006. Extreme learning machine: theory and applications. Neurocomputing. 70 (1): 489–501.

[4] Xu Q., Dubljevic S., 2017. Linear model predictive control for transport-reaction processes. AIChE Journal 63 (7): 2644-2659.

[5] Havu V., Malinen J., 2007. The cayley transform as a time discretization scheme. Numerical Functional Analysis and Optimization 28 (7-8): 825-851.

[6] Luenberger D., 1971. An Introduction to Observers, IEEE Transactions on Automatic Control, 16 (6): 596-602.