Dual Unscented Kalman Filter for Natural Gas Pipeline Leak Detection: Non-Isothermal Modeling and Effect of Thermal Properties

Model-based pipeline leak detection is one of the most widely used software methods for leak identification and isolation. Unlike methods based on statistical analysis of measurements, model-based algorithms apply a dynamic model to estimate the flow features in a pipeline. Many studies have been done towards the leak detection in natural gas pipelines based on dynamic models such as Extended Kalman Filter ⁽¹⁾ and Adaptive Particle Filter ⁽²⁾ to estimate the leak location. Most of the models used in the above-mentioned research were under either an isothermal or an adiabatic assumption. However, these idealistic and over-simplified models do not accurately represent the real system, especially the temperature change along the pipeline associated with a leak. A non-isothermal natural gas flow model has been developed for investigating the effect of temperature distribution ⁽³⁾, but the flow process with a leak has not yet been studied. In this research non-isothermal state equations were derived for natural gas pipeline flow process considering the effect of gas compressibility, heat transfer coefficient, ground temperature, and leak in the pipeline. The effect of ground temperature, heat transfer coefficient, and leak in the pipeline on the pressure, flow rate and temperature profile were studied using Matlab^™simulation. The result showed that the ground temperature and heat transfer coefficient would affect the steady state flow rate. The Unscented Kalman Filter was applied to estimate the leak location and the magnitude of the leak. Dual Unscented Kalman Filter (DUKF) combines parameter estimation and leak detection. For the practice of leak detection using a model-based method, the model parameter needs adjustment due to the applications in different environments. Hence we used DUKF to update the parameter and estimate the leak under different thermal properties including ground temperature and heat transfer coefficient. Using a simplified isothermal model for the DUKF, the parameter was successfully tested using various applicable situations.

Reference:

(1) H E Emara-Shabaik, et al.,  A non-linear multiple-model state estimation scheme for pipeline leak detection and isolation, Proceedings of the Institution of Mechanical Engineers, Part I: Journal of Systems and Control Engineering 2002 vol.216(6), 497

(2) Ming Liu, et al., Fast leak detection and location of gas pipeline based on an adaptive particle filter, International Journal of Applied Mathematics & Computer Science, 2005,15(4),541

(3) M. Abbaspour, K.S.Chapman, Nonisothermal Transient Flow in Natural Gas Pipeline. Journal of Applied Mechanics, 2008, 75, 031018