Multi-leak Detection And Location In Oil Pipelines Based On Unscented Kalman Filter

Pipeline transportation occupies a large proportion in petroleum transportation industry because of its convenience and stability.Due to the unbalanced energy distribution in China and the economic development in recent decades,the construction of pipeline network has also entered a period of rapid development.Maintenance problems,material selection,natural disasters,man-made damage and some other reasons may lead to our country's pipeline leakage accidents occur frequently,causing great damage to the national economy and environment.There are not only single leak point but also multiple leak points in the leak,but because of the complexity of the multi-leak diagnosis,researches on the problem are still in its infancy at home and abroad.Therefore,researching on oil and gas pipeline multi-leak diagnosis technology has a very important practical significance.This thesis takes oil pipeline as a research object,presenting a method of multi-leak diagnosis and positioning based on unscented kalman filter and then improves it.Through only setting pressure and flow sensors on both ends of pipeline,thisthesis do researches on the multileak oil pipeline diagnosis methods based on nonlinear state observers,the main work is divided into the following several parts:1.Established the state space model based on transient flow of oil pipeline,and analyze the influence of each parameter on the model.In this thesis,a simulation is carried out on the steady-state operation of the pipeline and the conditions of multiple leakages respectively,with gaussian white noise is added to simulate real measured values,analyzing characteristics of numerical calculation.2.According to the recursive principle of classical kalman filter,applying taylor expansion to nonlinear system model,using the extended kalman filter algorithm to estimate the state of pipeline model,locating multi-leak condition diagnosis of pipeline.By linearizing the nonlinear system,the kalman recursive method can be applied to nonlinear systems.The system state quantity is expanded to realize the calculation of pipeline leakage and estimation of leakage location.It is proved that this method is feasible for the multi-leak diagnosis of oil pipeline by setting up the simulation experiment in different locations.3.This thesis designed a multi-leak diagnosis method based on UKF.In the oil pipeline leakage diagnosis method based on EKF,due to the need to calculate the jacobian matrix and truncation errors are produced during the process of pipe model linearization,making model become inaccurate and increasing the amount of calculation.Unscented kalman filter algorithm directly using the nonlinear model of pipeline system without linearization approximation and calculating jacobian matrix,the calculation accuracy is higher and easy to realize,improved the timeliness of leak diagnosis.4.This thesis designed a multi-leak diagnosis method based on SR-UKF.In the process of unscented kalman recursive algorithm,the state error covariance matrix P sometimes is not positive,which leads to the interruption of the algorithm.Combine square root kalman filtering with unscented kalman filter algorithm,in the process of unscented kalman recursion method,transmit the square root of the covariance matrix of state error directly,effectively guarantee that the covariance matrix is positive,reducing the amount of calculation at the same time.Simulation results show that this method has good performance in the multi-leak diagnosis of oil pipeline,and it is helpful to apply the unscented kalman filter to practical problems.