Long oil pipeline leakage real-time detection system based on transfer learning LSTM

Abstract

The invention belongs to the technical field of oil field production equipment, and particularly to a long oil pipeline leakage real-time detection system based on the transfer learning LSTM. The method of the system comprises the following steps that 1, collecting data including normal data and leakage data through a long oil pipeline leakage experiment platform, and the LSTM network detection accuracy reaches 95% or above through data training; 2, migrating the trained LSTM model to field data through transfer learning, and predicting the state of the pipeline in real time; 3, observing through an observer, comparing the relationship between r (k) and R (k), judging whether leakage occurs, and if yes, generating an alarm by the system. The system can diagnose a pipeline fault in real time through a mode recognition method, generates an alarm if the fault occurs, and has a certain application value in the crude oil transportation process. In order to solve the technical problems, theproblems of real-time performance, small samples, misinformation and missing information are solved.

Description

Technical field: [0001] The invention belongs to the technical field of oilfield production equipment, and in particular relates to a real-time detection system for leakage of long oil pipelines based on transfer learning LSTM. Background technique: [0002] Oil is an important strategic resource, which is related to the economic lifeline of the country. Pipelines are widely used in oil transportation due to their convenience and speed. However, since most of my country's pipelines were first built in the 1980s and have been in service for nearly 40 years, they are already seriously aging and prone to oil spill accidents. This will not only cause energy waste, pollute the environment and other issues, but even endanger the safety of human life and property. Therefore, it is very necessary to propose a real-time detection system for long oil pipelines. [0003] At present, for pipeline leakage detection, most of them adopt the negative pressure wave method, that is, instal...

AI Technical Summary

Problems solved by technology

[0004] However, doing so will face three disadvantages: 1. Real-time problem: A series of signal conversion processing will result in the real-time problem of detection and even delay the best time for maintenance.

2. Small sample problem: Since pipeline leakage is a small probability event, the leakage sample is relatively small, there is no reference in the signal processing process, and a large number of expert opinions need to be relied on, which is time-consuming and greatly increases the detection cost

3. It is prone to false positives and false positives: due to the environment where the sensor is located (vehicle driving, nearby construction, etc.), the collected pressure signal will contain a lot of noise, which further increases the difficulty of detection

Images