Time series analysis of the strategic petroleum reserve's brine pipeline test data

Abstract

Ability to detect leaks in the Strategic Petroleum Reserve's brine pipeline depends on the ability to observe small drops in pressure, e.g. changes of the order of 0.3 psi (pounds per square inch). Typical pressure variation includes a random component (referred to as measurement noise) due primarily to measurement error and a systematic component (referred to as process noise) due to various internal and external disturbances such as offshore tides, temperature changes, and pump action. Much of the systematic component can be removed through time series modeling, with residuals from the model representing the random component. This paper addresses the estimation of the noise components through time series models applied to test data. Effectiveness of leak-detection algorithms based on test statistics (e.g. two-minute averages of pressure readings) can be determined from known or estimated standard deviations of the process and measurement noise components. The U.S. Department of Energy (operator of the Strategic Petroleum Reserve) plans to use the results of the time series analysis, together with hydraulic models, in order to establish leak-detection procedures that will meet Environmental Protection Agency requirements.