| Oil and gas resources are the most important sources of energy. They are transported by road, rail, water, air and pipeline. These five ways, in which pipeline transportation is the main method. However, with the pipe material itself, the aging, corrosion, condensate pipe, natural disaster, misusing etc., there is a large number of potential causes of accidents in pipeline. Situations such as burst pipes, broken pipes, leakage, punch stolen, become more serious. Therefore, the development of stable and reliable leak detection and localization system is an urgent need for pipeline safety.Development of pipeline transportation industry is analized after readed associated literatures. Leak detection methods widely used are also discussed. The project of leak detection and localization on the pipeline between Nanyi depot and Nansan depot with its length 30km in Daqing Oilfield, the largest in China. It is the center of this thesis, which mainly covers the following aspects:(1) Acoustics principle based negative pressure wave method is used to detect leaks. During the experiment, in order to overcome high false alarm rate caused by work condition adjustment and complicated interference, dual pressure transducers technology is proposed to determine the direction of propagation of negative pressure wave and reduce false alarm rate. Two pressure transducers are placed on each end of the pipe, the distance between them about 500 meters. This is not sufficient to signal attenuation. So, it is reliable of the correlation result between signals acquired by the two pressure transducers. The cause of negative pressure wave, work condition adjustment or leaks, will be determined after synthesizing the propogation direction of the pressure wave on each end.(2) Multivariate WaveShrink desoising method is applied to prefilter the four-dimensional signals generated by dual pressure transducers after rsearching on wavelet transform based denoising technique. This method improves the signal noise ratio more than that by traditional wavelet threshold denoising due to consider the noise correlation between multichannel signals in the space. Multivariate signals at the same time point are deemed as an integrity, which is different from the univariate signal wavelet thresholding.(3) Sparse decomposition based leak detection is proposed after analyzed signal representation theory. Pressure signal is divided into four types:stationary signal, slow descent signal, mutation descent signal and ascent signal. Overcomplete dictionary is designed to classify the pressure signal by improved orthogonal matching pursuit sparse decomposition algorithm. Experiments show that the recognition rate of pressure signal can achieve over 90%.(4) Correlation analysis between leak signal and that from another station is used to determine the leak point in the pipeline. The source and different forms of correlation are also discussed. Performance of four time delay estimation methods under different SNR simulation signals and practical signals from application are compared, which are basic cross correlation, phase transformation generalized cross correlation, maximum likelihood function generalized cross correlation and least mean square. Basic correlation is used to estimate the time difference between the negative pressure wave arriving at two ends of the pipeline. Correlated analysis result of signals can localize the leak point and its error is less than 5%. This performs well in the implementation of project.(5) Software system for Daqing oilfield leak detection is realized based on the methods mentioned above, finally. |
PDF Full Text Request