Leak Detection and Localization in Oil Pipelines Using a Hybrid CNN–LSTM Deep Learning Model

Leak detection and localization in oil pipelines are essential for maintaining operational safety and preventing environmental and economic losses in the oil and gas industry. In this study, a hybrid deep learning model based on convolutional neural networks (CNN) and long short-term memory (LSTM) networks is proposed for accurate leak detection and localization. The Khisht to Nargesi oil pipeline, located in southern Iran, with a length of ۱۲.۸ km and a diameter of ۱۲ inches, is simulated using OLGA software to generate realistic operational data. Pressure and flow rate signals are collected from pipeline sensors under both leak and no-leak conditions and used directly as time-series inputs to the model. Using a CNN eliminates manual feature extraction, as the network automatically learns relevant features from raw time-series data, while the LSTM component captures temporal dependencies. The proposed model performs binary classification of leak and no-leak conditions and localizes the leak position along the pipeline with a spatial resolution of ۱-km. Simulation results show that the CNN–LSTM model achieves over ۹۹% accuracy for both leak detection and leak localization, demonstrating its effectiveness for practical pipeline monitoring applications.

کلیدواژه ها:

Leak Detection ، Leak Localization ، Deep Learning ، Time-Series Data ، Classification