Leak Detection And Location Of Gas Pipelines Based On Unscented Kalman Filter

Pipeline accidents and subsequent leaks cause huge economic loss, and environmental pollutions that can even endanger life safety. Economics, safety, and environmental issues associated with pipeline leaks are a growing concern around the world.The real time model method based on EKF is widely used to detect and locate small leaks. The extended Kalman filter must linearize the nonlinear model before estimating the sates of this model. And the nonlinear degree of gas pipeline model is serious, which generates great error during linearization; in each prediction step of the estimation, it needs to compute the Jacob matrix, which slows the convergence.The method presented in this paper is based on unscented Kalman filter (UKF), which does not need to linearize nonlinear gas pipeline model. So it can improve the precision. And it only includes some simple linear computation in the estimation, so the detection speed is greatly enhanced.In order to further improve the detection precision based on UKF, this paper modifies the UKF with introducing IKF into it-called IUKF. IUKF can obtain more accurate probability characteristics of states in the process of updating, which improves the estimation precision without increasing the dimensions. And the simulation results prove that the IUKF makes a better performance comparing with EKF and UKF, which makes great sense in leak detection and location of long gas pipelines.