Research On The Long Straight Gas Pipeline Leakage Detection And Location

The demand for energy, especially for natural gas resources, is becoming more urgent now. With the exploitation of natural gas in western China and the import of foreign natural gas, pipelines connecting natural gas fields to demand centers are becoming longer, of increased pressure and with larger diameter pipes. In this case, pipeline leakage accidents occur frequently, and lead to serious life and property loss and environmental pollution. Therefore, it is vital to detect and locate the leaks in time and to carry out countermeasures to prevent leaks. Therefore, this thesis presents research toward the detection and locating of leakage in long straight pipelines using both experimental and theoretical techniques.For normal conditions and leakage conditions, differential equations of basic gas flow equations are deduced respectively using the method of characteristic line. Pressure and temperature distribution along the pipeline are obtained based on the equations of motion and the energy equation. The gas pipeline leakage models (a storage tank model, a small hole model, a large hole model and a pipe model) are establised, these four models are verified using simulation data, thus, leakage rate can be calculated.Toward to this research object, consider the physical model and the experimental condition, a gas pipeline leakage detection test bench is built. The principle of Fiber Bragg grating (FBG) strain sensors detecting pipeline internal pressure changes is studied, and also experimental tested. Combined with hydrodynamic theory, the leakage test results were analyzed to verify the feasibility of experimental studies.In order to use the extend kalman filter (EKF) based method to detect and locate leaks in natural gas pipeline, the basic idea of real-time model for filter and the established discrete transient pipe flow model including one point leakage are combined. In this condition, assuming the leak points are distributed on the segment, so a discrete transient flow model includes multi-point leakage are obtained. The leakage amount of segment points is included in this discrectizaion model and taken as a component of state variables. Consider the impact of system noise and measurement noise, the EKF is employed to estimate the flow of hydraulic elements, such as leakage amout et al. The leakage amount and leakage position formulas are eatablished. Simulation example and experimental data are used to test and verify this method.The negative pressure wave (NPW) method is currently the most widely used method for pipeline leak detection and location. However, high rates of false positives and low accuracy of positioning are the main drawbacks of this method. In this thesis, the method of using FBG strain sensors to detect negative pressure waves is studied and experimental tested. From this, an improved NPW method for leak detection and location is proposed and the principle of this method is studied. Using this method, the signal from FBGs is de-noised by a wavelet threshold method. Based on the comparison of experimental results and the study of the negative pressure wave attenuation law, the reasons why an improved NPW method is superior to a conventional NPW method is technically and theoretically analyzed.In order to locate the leakage position using the improved NPW method, firstly, wavelet analysis is employed to find the signal point mutation. Therefore, the time difference between the arrivals of the negative pressure wave propagates at the upstream and downstream FBG strain sensors can be calculated. Secondly, an improved leak location formula is proposed based on the variation in the negative pressure wave propagation velocity and the gas velocity variation, and Compound Simpson and Dichotomy Searching are employed to solve this formula. Finally, the positioning effect is verified by using experimental data.To reduce the false alarm rate and labor, the pattern recognition theory is introduced into the leak detection field. Using this theory, the features of the normal signal and leakage signal, as captured by FBG strain sensors are extracted as input feature vectors, and thus a least squares support vector machine (LS-SVM) based gas pipeline leak detection model is built, a real-time intelligent gas pipeline leak detection system is achieved.