چكيده به لاتين
Pipelines are known as one of the most economical methods of transporting fluids over long distances and distributing them in large areas and cities. The fluids transported and distributed by these pipelines are of high economic value and sometimes hazardous and pollute the environment. In this situation, it is necessary to develop a system to monitor these pipelines, responsible for detecting, locating, and estimating leakage size. In this research, leak detection techniques have been divided into hardware and computational methods, and their strengths, weaknesses, and limitations have been investigated. Since pipeline leakage is a very harmful complication and leads to much environmental and economic damage, the chosen method should identify and locate the leakage in the least possible time and cost. Computational leak detection methods are generally methods that detect leaks either online or with minimal delay. These methods require much less measurement and monitoring equipment to detect, locate, and estimate leaks sizes than hardware methods. Computational methods do not have the same performance in the leak detection process in different operational and environmental conditions. It is vital that for each pipeline to prevent severe environmental and economic damage, appropriate method(s) with specific operational and environmental conditions for the same pipeline to develop the Monitoring system. This study evaluates the performance and capability of Real-Time transient modeling methods, the classification method, and the negative pressure wave method in detecting, locating, and estimating the leakage size of three gas pipelines with lengths of 21 m, 62/1 km, and 50 km were studied. Since it is impossible to repeat leak detection tests for each method in pipelines, OLGA software has been used to simulate various leakage and operating conditions. Among the studied methods, the negative pressure wave method recorded the best result among the studied methods with 0/84, and 1/32% relative mean error in leak detection, respectively, for the second and third pipeline cases. Twenty-four important and practical methods in this field were examined for the classification method. The Fine KNN method gave the best result with 8/1%, and 1/7% mean relative error for the second and third cases leakage localization, respectively. In the presence of environmental noise, the Fine KNN method with a relatively minimal loss of accuracy recorded a relatively stable performance in detecting, locating, and leak size estimating. The online modeling method for leak detection recorded 15/30%, 18/36%, and 31/10% of the mean relative error for the first, second and third case studies, respectively. Since due to the limitations of computing systems, it is not yet possible to develop a single comprehensive method for detecting, locating, and estimating leakage sizes that can perform reliably in a variety of operating and leakage conditions, it is recommended that a set of computational methods be used to perform the leak detection process in different conditions with high accuracy.
