Search

CN-121977168-A - Pipeline leakage identification and control method and device based on artificial intelligence

CN121977168ACN 121977168 ACN121977168 ACN 121977168ACN-121977168-A

Abstract

A pipeline leakage identification and control method and device based on artificial intelligence relate to the field of pipeline monitoring. The method comprises the steps of collecting parameter signal sequences of pipelines at a plurality of monitoring points, analyzing the parameter signal sequences, identifying signal transient characteristics, marking the signal transient characteristics as disturbance events, determining characteristic time points of the disturbance events in the parameter signal sequences corresponding to the monitoring points, calculating propagation characteristic parameters of the disturbance events based on spatial separation degree among the monitoring points and the characteristic time points, training a pipeline leakage identification model based on historical pipeline operation data, inputting the propagation characteristic parameters into the pipeline leakage identification model to obtain classification results of the disturbance events, generating risk assessment grades if the classification results are determined to be leakage events, and activating a management and control execution strategy through a preset response logic tree based on the risk assessment grades. By implementing the technical scheme provided by the application, the accuracy of identifying the pipeline leakage is improved.

Inventors

  • LI GUANGXI

Assignees

  • 北京华科合创科技发展有限公司

Dates

Publication Date
20260505
Application Date
20260318

Claims (10)

  1. 1. An artificial intelligence-based pipeline leakage identification and control method, which is characterized by comprising the following steps: acquiring parameter signal sequences of the pipeline in a time synchronous mode at a plurality of monitoring points along a preset path of the pipeline, wherein the monitoring points are provided with space separation degrees; analyzing parameter signal sequences acquired at each monitoring point, identifying signal transient characteristics meeting preset conditions, and marking the signal transient characteristics as disturbance events; determining a characteristic time point of the disturbance event in a parameter signal sequence corresponding to each monitoring point; Calculating a propagation characteristic parameter of the disturbance event based on the spatial separation degree between the monitoring points and the characteristic time points; acquiring historical pipeline operation data, and training a pipeline leakage identification model based on the historical pipeline operation data; Inputting the propagation characteristic parameters into the pipeline leakage identification model to obtain a classification result of the disturbance event, wherein the classification result is a normal operation disturbance or leakage event; and if the classification result is determined to be a leakage event, generating a risk assessment grade, and activating a management and control execution strategy through a preset response logic tree based on the risk assessment grade.
  2. 2. The method of claim 1, wherein analyzing the parameter signal sequences collected at each monitoring point to identify signal transient characteristics that meet a preset condition specifically comprises: Performing multi-scale wavelet decomposition on the parameter signal sequences of the monitoring points to obtain wavelet coefficients of different frequency components; Calculating the local energy density distribution of each wavelet coefficient, and identifying energy density abrupt change points as potential transient characteristic points; Performing time-frequency domain joint analysis on the potential transient characteristic points, and extracting the rising time, peak amplitude and decay time constant of the transient characteristic; and determining the transient characteristics of which the rising time is smaller than the first threshold value, the peak amplitude is larger than the second threshold value and the decay time constant is in a preset interval as the signal transient characteristics meeting the preset condition.
  3. 3. The method according to claim 1, wherein the calculating the propagation characteristic parameter of the disturbance event based on the spatial separation between the monitoring points and the characteristic time points specifically includes: Constructing a monitoring point space topological graph, wherein nodes of the monitoring point space topological graph represent the monitoring points, the weight of the edges of the monitoring point space topological graph represents the pipeline path length between adjacent monitoring points, and the adjacent monitoring points are two monitoring points connected along the preset path; Fitting a disturbance propagation speed curve by using a least square method based on the time difference of the disturbance event detected by the adjacent monitoring points and the corresponding pipeline path length; and calculating the speed gradient change rate and the speed discrete coefficient of the disturbance propagation speed curve as the propagation characteristic parameters.
  4. 4. The method of claim 1, wherein the training a pipeline leak identification model based on the historical pipeline operation data comprises: preprocessing the historical pipeline operation data to generate a training data set, wherein the preprocessing comprises the steps of extracting pressure gradient characteristics by wavelet transformation, extracting sound source positioning characteristics by a beam forming technology, and eliminating environmental drift in temperature data by Kalman filtering; constructing the pipeline leakage identification model, wherein the pipeline leakage identification model adopts a framework which takes a residual network as a characteristic to extract backbone, and combines a time sequence dependency relationship in long-term and short-term memory network captured data to realize the self-adaptive fusion of multi-mode characteristics through an attention mechanism; And training the pipeline leakage recognition model by adopting a distributed training strategy based on the training data set.
  5. 5. The method according to claim 1, wherein the activating a management execution policy by a preset response logic tree based on the risk assessment level specifically comprises: dividing the risk level of the leakage event into a first risk level, a second risk level and a third risk level based on the position information, the leakage rate and the influence range of the leakage event; when the risk level is determined to be the first risk level, activating an audible and visual alarm tower, pushing alarm information to a monitoring center, and improving the monitoring frequency of the pipeline; when the risk level is determined to be the second risk level, partition isolation is performed through an electric ball valve group so as to close valves in a preset upstream and downstream range of the leakage event; When the risk level is determined to be the third risk level, activating a full-line emergency shutdown system, and closing a plurality of electric ball valves on the pipeline according to a preset sequence; And when the pipeline pressure is monitored to exceed the preset pressure threshold value and the preset time is continued, starting the emergency pressure relief device.
  6. 6. The method of claim 1, wherein after the determining the characteristic point in time of the disturbance event, the method further comprises: Carrying out waveform similarity analysis on the signal transient characteristics detected by each monitoring point, and calculating cross-correlation coefficients of the signal transient characteristics among different monitoring points; Constructing a disturbance event association matrix based on the cross-correlation coefficient, and grouping the disturbance events by adopting a spectral clustering algorithm; And carrying out space-time consistency check on disturbance events in the same group, determining an abnormal detection result and eliminating the abnormal detection result.
  7. 7. The method of claim 6, wherein the performing space-time consistency check on the disturbance events in the same group, determining an anomaly detection result and rejecting the anomaly detection result, specifically comprises: establishing a disturbance propagation theoretical model based on physical characteristics of a pipeline, and determining a theoretical propagation speed range; Calculating the actual propagation speed of the disturbance event detected by adjacent monitoring points in the same group, and comparing the actual propagation speed with the theoretical propagation speed range; And determining a disturbance event pair with the actual propagation speed exceeding the theoretical propagation speed range, marking the disturbance event pair as the abnormal detection result, and eliminating the abnormal detection result.
  8. 8. An artificial intelligence based pipe leak identification and management device comprising one or more processors and memory coupled to the one or more processors, the memory to store computer program code comprising computer instructions that the one or more processors invoke to cause the artificial intelligence based pipe leak identification and management device to perform the method of any of claims 1-7.
  9. 9. A computer readable storage medium comprising instructions that, when run on an artificial intelligence based pipe leak identification and management device, cause the artificial intelligence based pipe leak identification and management device to perform the method of any of claims 1-7.
  10. 10. A computer program product, characterized in that the computer program product, when run on an artificial intelligence based pipe leakage identification and management device, causes the artificial intelligence based pipe leakage identification and management device to perform the method according to any one of claims 1-7.

Description

Pipeline leakage identification and control method and device based on artificial intelligence Technical Field The application relates to the field of pipeline monitoring, in particular to an artificial intelligence-based pipeline leakage identification and control method and device. Background Pipelines are critical infrastructure for transporting important media such as oil, gas, etc., but the risk of leakage in their operation can lead to serious economic losses, environmental pollution and safety accidents. The traditional pipeline monitoring method, such as manual inspection or single parameter monitoring, has the defects of slow response, low positioning precision, high false alarm rate and the like, and is difficult to meet the requirement of modern pipelines on high safety. In order to improve the monitoring level, the prior art proposes a scheme based on multi-sensor fusion and intelligent analysis. According to the scheme, various sensors such as pressure, sound waves and temperature are deployed along a pipeline, multidimensional operation data are collected in real time, and leakage is judged by identifying composite characteristics such as pressure transient. The scheme improves the real-time performance and accuracy of leakage identification to a certain extent. However, the intelligent monitoring scheme described above has inherent drawbacks in distinguishing real leaks from certain specific normal operations, such as pigging operations. When the pipe cleaner runs at high speed in the pipeline, the generated composite signal characteristics of severe pressure fluctuation, strong acoustic signals, local temperature change and the like are highly similar to the signal characteristics of partial real leakage. Since the prior art is typically a single point detection method, normal pig operations are easily misjudged as leaking, triggering unnecessary downtime and emergency response. Frequent false alarms not only cause direct economic loss, but also seriously reduce the trust degree of operation and maintenance personnel on an automation system, once true leakage occurs, the handling is delayed possibly due to the reduced vigilance, and deeper potential safety hazards are formed. Therefore, how to improve the accuracy of interference and real leakage signals for pipe cleaning operation is a technical problem to be solved in the current pipeline safety monitoring field. Disclosure of Invention The application provides an artificial intelligence-based pipeline leakage identification and control method and device, which improve the accuracy of pipeline leakage identification. The application provides an artificial intelligence-based pipeline leakage identification and control method, which comprises the steps of acquiring parameter signal sequences of a pipeline in a time synchronization mode at a plurality of monitoring points along a preset path of the pipeline, analyzing the parameter signal sequences acquired at the monitoring points, identifying signal transient characteristics meeting preset conditions, marking the signal transient characteristics as disturbance events, determining characteristic time points of the disturbance events in the parameter signal sequences corresponding to the monitoring points, calculating propagation characteristic parameters of the disturbance events based on the spatial separation degree between the monitoring points and the characteristic time points, acquiring historical pipeline operation data, training a pipeline leakage identification model based on the historical pipeline operation data, inputting the propagation characteristic parameters into the pipeline leakage identification model to obtain classification results of the disturbance events, wherein the classification results are normal operation disturbance or leakage events, generating risk assessment grades if the classification results are leakage events, and executing a management and control strategy by activating a risk assessment logic tree based on the preset risk assessment grades. By adopting the technical scheme, the parameter signal sequences are acquired at a plurality of monitoring points in a time synchronous mode along the preset path of the pipeline, so that the omnibearing real-time monitoring of the running state of the pipeline is realized, and the accuracy of the positioning of disturbance events is ensured by the space separation degree among the monitoring points. And the parameter signal sequence is analyzed to identify signal transient characteristics and marked as a disturbance event, so that abnormal changes in the pipeline can be rapidly captured. The propagation rule of the disturbance in the pipeline is accurately obtained by determining the characteristic time point of the disturbance event and calculating the propagation characteristic parameter. The pipeline leakage recognition model trained based on historical pipeline operation data has strong mode recognition capabil