Online pipeline models must address the problem of what to do with the extra pressure, flow, and possibly other measurements beyond what is needed for boundary conditions for the underlying set of partial differential equations (PDEs). Existing state estimation techniques attempt to estimate the new state as if the model errors in each time step were completely uncorrelated, creating a necessity for external automatic tuning mechanisms if there are slowly-time-varying or unvarying sources of error such as unknown pipe roughness or unknown ground thermal properties. In other fields, the Kalman Filter has been used to produce good estimates of the states of small dynamic systems, but it is computationally intractable for systems as large as typical pipeline models. A new approximate form of Kalman Filter which is presently used in weather simulation called the Ensemble Kalman Filter (EnKF) provides much greater computational efficiency. This article demonstrates the implementation of EnKF for state estimation on artificial gas and liquid pipelines with known errors, and compares it to existing methods of state estimation and automatic tuning (including the problem of leak location in a leak detection system) for speed and accuracy.

This content is only available via PDF.
You can access this article if you purchase or spend a download.