Oil and gas pipeline monitoring data processing method and system based on artificial intelligence

Abstract

The invention relates to the technical field of oil and gas pipeline monitoring, and discloses an oil and gas pipeline monitoring data processing method and system based on artificial intelligence, and the method comprises the steps: obtaining the distributed optical fiber sensing data of an oil and gas pipeline, and carrying out the noise reduction processing of the distributed optical fiber sensing data, and obtaining the processed pipeline state data; based on the processed pipeline state data and a preset abnormal event judgment model, whether an abnormal event exists or not is judged, and the preset abnormal event judgment model is used for conducting probabilistic abnormal event judgment on the pipeline state data; feedback information of the operation and maintenance personnel for the abnormal event is received, the feedback information comprises confirmation or rejection, and an abnormal early warning processing strategy is determined according to the feedback information. According to the invention, the abnormal event is effectively identified, strategy adjustment is carried out according to the feedback of the operation and maintenance personnel, and the intelligent level and response efficiency of monitoring are improved.

Description

Technical Field

[0001] This application relates to the field of oil and gas pipeline monitoring technology, and more specifically, to an artificial intelligence-based method and system for processing oil and gas pipeline monitoring data. Background Technology

[0002] As a core infrastructure of the national energy strategy, the safety monitoring of oil and gas pipelines is of paramount importance. Traditional methods of oil and gas pipeline safety monitoring, such as manual inspections, are inefficient and costly, failing to meet the real-time and accuracy requirements of modern oil and gas pipelines. While ultrasonic and negative pressure wave detection methods have improved monitoring capabilities to some extent, they still struggle to meet the real-time and accuracy demands when monitoring oil and gas pipelines over long distances in complex environments.

[0003] In recent years, distributed fiber optic sensing technology has been increasingly applied to the safety monitoring of oil and gas pipelines due to its unique advantages such as all-weather operation, long-distance transmission, and resistance to electromagnetic interference. However, existing oil and gas pipeline monitoring systems based on distributed fiber optic sensing technology still face many technical bottlenecks. For example, in terms of hardware performance, the sensing distance is often limited to within 70km, and the effective signal acquisition rate is less than 96%, which restricts its application in large-scale oil and gas pipeline networks. Regarding data processing and anomaly detection, existing systems generally suffer from poor anomaly location accuracy (approximately 10m) and a high false alarm rate (5%–10%). This is mainly because they rely excessively on traditional threshold judgment methods, making it difficult to effectively handle uncertainties in complex environments, leading to numerous false alarms or missed alarms, severely impacting the reliability and operational efficiency of the monitoring system. Furthermore, existing systems suffer from weak hardware and software integration, poor compatibility between management systems and industry platforms, and a lack of customization capabilities, resulting in low utilization of distributed fiber optic sensing data. This further exacerbates the workload of maintenance personnel and reduces overall operational efficiency. These problems collectively constrain the large-scale application and development of distributed fiber optic sensing technology in the field of oil and gas pipeline safety monitoring.

[0004] To address the aforementioned issues, existing technologies urgently need improvement. Summary of the Invention

[0005] To address the shortcomings of existing technologies, this application provides an artificial intelligence-based method and system for processing oil and gas pipeline monitoring data. This system addresses the deficiencies of existing oil and gas pipeline monitoring systems in terms of data processing, abnormal event judgment, positioning accuracy, and false alarm rate, as well as issues such as weak hardware and software integration, poor compatibility between the management system and industry platforms, and lack of customization capabilities.

[0006] In a first aspect, this application discloses an artificial intelligence-based method for processing oil and gas pipeline monitoring data, including: Acquire distributed fiber optic sensing data of oil and gas pipelines, perform noise reduction processing on the distributed fiber optic sensing data, and obtain processed pipeline status data. Based on the processed pipeline status data and the preset abnormal event judgment model, it is determined whether there are abnormal events. The preset abnormal event judgment model is used to perform probabilistic abnormal event judgment on the pipeline status data. Receive feedback from operations and maintenance personnel regarding abnormal events, including confirmation or rejection. Based on the feedback, determine the abnormal early warning and handling strategy.

[0007] This technical solution enables intelligent processing of oil and gas pipeline monitoring data, effectively identifies abnormal events, and adjusts strategies based on feedback from maintenance personnel, thereby improving the intelligence level and response efficiency of monitoring.

[0008] Furthermore, the steps for acquiring distributed fiber optic sensing data of oil and gas pipelines include: Vibration and acoustic monitoring data along the oil and gas pipelines are collected using distributed fiber optic sensors laid along the pipelines.

[0009] This technical solution ensures the comprehensiveness and accuracy of the raw data, laying the foundation for subsequent intelligent analysis.

[0010] Based on this, the steps for denoising the distributed fiber optic sensing data to obtain the processed pipeline status data include: The distributed optical fiber sensing data is decomposed into multiple intrinsic mode function components using the ensemble empirical mode decomposition method. The multiple intrinsic mode function components are then denoised using the wavelet thresholding method. Finally, the denoised multiple intrinsic mode function components are reconstructed to obtain the pipeline state data.

[0011] This technical solution can effectively remove noise from the original data, improve the purity and reliability of pipeline status data, and provide high-quality data for accurate judgment of abnormal events.

[0012] In some preferred embodiments, the step of determining whether an abnormal event exists, based on the processed pipeline status data and a preset abnormal event judgment model, includes: A Bayesian hypothesis testing model is used as the pre-defined abnormal event judgment model to calculate the posterior probability of the pipeline being in an abnormal state based on the pipeline status data. If the posterior probability is higher than the first preset threshold, an abnormal event is determined to exist, and a two-layer localization optimization is performed using the K-means clustering algorithm to obtain the location of the abnormal event.

[0013] Furthermore, the two-layer positioning optimization includes a first-layer positioning optimization and a second-layer positioning optimization. The first-layer positioning optimization includes determining a rough area where anomalies occur along the oil and gas pipeline based on the signal propagation characteristics of pipeline status data. The second-layer positioning optimization includes performing cluster analysis on the signal propagation characteristics of pipeline status data within the rough area to determine the location of the anomaly and improve the positioning accuracy of the anomaly to a preset range.

[0014] As a technical improvement, the following is included before performing two-layer positioning optimization: Multi-channel spatiotemporal feature extraction and correlation analysis are performed on pipeline status data. Based on the inherent consistency of pipeline status data in spatial, frequency and time dimensions, signal propagation characteristics of pipeline status data are extracted.

[0015] To enhance functionality, after locating the abnormal event, the following is also included: A one-dimensional convolutional neural network is used to extract the abnormal features of anomalous events, and the abnormal features are then input into a support vector machine for classification to identify the type of anomalous events.

[0016] To improve the plan, based on feedback information, the steps to determine the anomaly warning handling strategy include: Based on the historical accuracy of feedback from operations and maintenance personnel and the current workload assessment results, the feedback confidence level of the feedback information is calculated. When the feedback information is confirmed and the feedback confidence level is higher than the second preset threshold, the first processing strategy is executed. The first processing strategy includes generating a confirmed anomaly warning, marking the location of the anomaly on the map and displaying the type of the anomaly through a visualization operation platform. When the feedback information is confirmation and the feedback confidence level is not higher than the second preset threshold, or when the feedback information is rejection and the feedback confidence level is not higher than the second preset threshold, the second processing strategy is executed. The second processing strategy includes generating an anomaly warning to be reviewed and notifying the operation and maintenance personnel to conduct a second review of the anomaly warning to be reviewed through at least one of the following methods: highlighting prompts on the visual operation platform, voice reminders, and pushing messages to mobile terminals. When the feedback is a rejection and the feedback confidence level is higher than the second preset threshold, a third processing strategy is executed, which includes not generating an anomaly warning.

[0017] For specific situations, the process may include the following after executing the second processing strategy: If no secondary confirmation feedback is received from maintenance personnel regarding the pending anomaly warning within the preset time, the pipeline status data within the preset time window before the generation of the pending anomaly warning, as well as the auxiliary monitoring data sources associated with the pipeline status data, will be retrieved for data backtracking analysis.

[0018] Secondly, this application also discloses an artificial intelligence-based oil and gas pipeline monitoring data processing system for executing the aforementioned artificial intelligence-based oil and gas pipeline monitoring data processing method. The system includes: The data acquisition and processing module is used to acquire distributed optical fiber sensing data of oil and gas pipelines, perform noise reduction processing on the distributed optical fiber sensing data, and obtain processed pipeline status data. The abnormal event judgment module is used to determine whether there are abnormal events based on the processed pipeline status data and the preset abnormal event judgment model. The preset abnormal event judgment model is used to perform probabilistic abnormal event judgment on the pipeline status data. The anomaly warning module is used to receive feedback information from operation and maintenance personnel regarding abnormal events. The feedback information includes confirmation or rejection. Based on the feedback information, the anomaly warning handling strategy is determined.

[0019] This technical solution provides a system that integrates data acquisition, processing, anomaly detection, and early warning, enabling comprehensive and intelligent monitoring of oil and gas pipelines and improving overall operation and maintenance efficiency and safety.

[0020] In summary, this application provides an artificial intelligence-based method and system for processing oil and gas pipeline monitoring data. The method acquires distributed fiber optic sensor data and performs noise reduction processing to obtain high-quality pipeline status data. Based on this, a pre-defined anomaly judgment model is used to probabilistically judge anomalies in the pipeline status data, effectively reducing the high false alarm rate problem caused by traditional threshold judgment methods. Furthermore, this application introduces a feedback mechanism from maintenance personnel, determining anomaly early warning handling strategies based on feedback information, achieving intelligent decision-making through human-machine collaboration, and improving the accuracy and reliability of early warnings. Compared with existing technologies, this application effectively solves the problems of poor anomaly location accuracy, high false alarm rate, and low data utilization in existing systems, significantly improving the intelligence level and operational efficiency of oil and gas pipeline safety monitoring, and providing strong protection for the safe operation of oil and gas pipelines. Attached Figure Description

[0021] Figure 1 This is a flowchart illustrating an artificial intelligence-based oil and gas pipeline monitoring data processing method provided in an embodiment of this application.

[0022] Figure 2 This is a schematic diagram of the structure of an artificial intelligence-based oil and gas pipeline monitoring data processing system provided in an embodiment of this application.

[0023] Labeling Explanation: 210, Data Acquisition and Processing Module; 220, Abnormal Event Judgment Module; 230, Abnormal Early Warning Module. Detailed Implementation

[0024] The technical solutions of this application will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments. The components of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0025] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this application, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0026] Traditional methods of oil and gas pipeline safety monitoring, such as manual inspections, are inefficient and costly, failing to meet the real-time and accuracy requirements of modern oil and gas pipelines. While ultrasonic and negative pressure wave detection methods have improved monitoring capabilities to some extent, they still struggle to meet the real-time and accuracy demands of monitoring oil and gas pipelines over long distances in complex environments. Existing oil and gas pipeline monitoring systems based on distributed fiber optic sensing technology still face numerous technical bottlenecks. For example, in terms of hardware performance, the sensing distance is often limited to within 70km, and the effective signal acquisition rate is below 96%, restricting their application in large-scale oil and gas pipeline networks. Regarding data processing and anomaly detection, existing systems generally suffer from poor anomaly location accuracy and high false alarm rates. This is primarily because they rely excessively on traditional threshold-based methods, which struggle to effectively handle uncertainties in complex environments, leading to numerous false alarms or missed alarms, severely impacting the reliability and operational efficiency of the monitoring system. In addition, the existing system has poor hardware and software integration, poor compatibility between the management system and industry platforms, and a lack of customization capabilities, resulting in low utilization of distributed fiber optic sensing data, which further increases the workload of maintenance personnel and reduces overall maintenance efficiency.

[0027] In this regard, firstly, referring to Figure 1 This application proposes an artificial intelligence-based method for processing oil and gas pipeline monitoring data, including: S1. Acquire distributed optical fiber sensing data of oil and gas pipelines, perform noise reduction processing on the distributed optical fiber sensing data, and obtain the processed pipeline status data. S2. Based on the processed pipeline status data and the preset abnormal event judgment model, determine whether there is an abnormal event. The preset abnormal event judgment model is used to perform probabilistic abnormal event judgment on the pipeline status data. S3. Receive feedback information from operations and maintenance personnel regarding abnormal events. The feedback information includes confirmation or rejection. Based on the feedback information, determine the abnormal early warning handling strategy.

[0028] Distributed fiber optic sensing data refers to data collected by distributed fiber optic sensors laid along oil and gas pipelines, reflecting changes in physical quantities such as vibration and acoustics along the pipeline. This data is characterized by continuity, high density, and long-distance transmission, serving as the foundation for judging the condition of oil and gas pipelines. Pipeline status data refers to distributed fiber optic sensing data after noise reduction processing. It more accurately reflects the actual operating status of oil and gas pipelines, providing reliable input for subsequent anomaly event judgment. The anomaly event judgment model is a pre-defined model based on artificial intelligence algorithms. Its core function is to perform probabilistic anomaly event judgment on pipeline status data, i.e., assess the probability of the pipeline being in an abnormal state. Feedback information refers to the operational results of maintenance personnel confirming or rejecting anomaly event warnings based on the actual situation. This is crucial for the continuous optimization of the model and the adjustment of the warning strategy. The anomaly warning handling strategy consists of different response measures taken for anomalies based on feedback information and system evaluation results, aiming to achieve accurate warnings and efficient handling.

[0029] Specifically, there are several methods for acquiring distributed fiber optic sensing data from oil and gas pipelines. For example, distributed fiber optic sensors can be deployed along the pipeline route to collect real-time data on changes in physical quantities such as vibration, strain, and temperature, utilizing the Rayleigh scattering, Brillouin scattering, or Raman scattering effects of optical fibers. These sensors can be connected in series or parallel to form a sensing system covering the entire pipeline network. Another approach is to utilize existing fiber optic infrastructure by injecting probe light pulses into the fiber optic cable and analyzing the returned scattered light signals to acquire data along the route. This raw data typically contains a significant amount of noise and requires noise reduction processing.

[0030] Various noise reduction techniques can be employed to denoise distributed fiber optic sensing data and obtain processed pipeline status data. For example, digital filters, such as Butterworth or Chebyshev filters, can be used to filter the data in the frequency domain, removing high-frequency or low-frequency noise. Wavelet transform methods can also be used to decompose the signal into different scales, then thresholding or shrinking the scale containing noise before reconstructing the signal. Furthermore, adaptive signal processing methods such as Empirical Mode Decomposition (EMD) or Ensemble Empirical Mode Decomposition (EEMD) can be used to decompose the signal into a series of intrinsic mode functions (IMFs), then identify and remove noisy IMFs, and finally reconstruct the denoised signal. After noise reduction, the noise components in the original distributed fiber optic sensing data are effectively suppressed, resulting in cleaner pipeline status data that better reflects the true state of the pipeline.

[0031] In determining the existence of abnormal events based on processed pipeline state data and a pre-defined abnormal event judgment model, the pre-defined abnormal event judgment model is crucial. This model is used to probabilistically judge abnormal events in pipeline state data. For example, a Support Vector Machine (SVM) model can be used, which learns the decision boundary that distinguishes between normal and abnormal states by training on historical normal and abnormal pipeline state data. When new pipeline state data is input, the SVM model outputs the probability that the data belongs to an abnormal state. Another approach is to use neural network models, such as Recurrent Neural Networks (RNNs) or Long Short-Term Memory Networks (LSTMs). These models are good at processing time-series data, can learn the temporal characteristics of pipeline state data, and predict future states, thereby determining whether the current state deviates from the normal range. In addition, statistical methods, such as Gaussian Mixture Models (GMMs), can be used to model normal pipeline state data and calculate the probability that new data points fall outside the normal distribution, thereby judging abnormalities. The core of these models lies in their ability to extract effective features from complex pipeline state data and evaluate the probability of abnormal events in the form of probabilities, rather than simple binary judgments, which provides richer information for subsequent decision-making.

[0032] In receiving feedback from operations and maintenance (O&M) personnel regarding abnormal events and determining anomaly warning handling strategies based on this feedback, the feedback information includes confirmation or rejection. For example, after the system issues an anomaly warning, O&M personnel can view detailed information about the anomaly, such as its location and type, through a visual operation platform, and confirm or reject the warning based on on-site investigation or other auxiliary information. If O&M personnel confirm the anomaly, the system can execute a first handling strategy, such as generating a confirmed anomaly warning and marking the location and type of the anomaly on a map through the visual operation platform, so that O&M personnel can take timely measures. If O&M personnel reject the anomaly, the system can execute a third handling strategy, such as not generating an anomaly warning to avoid invalid warnings interfering with O&M personnel. Furthermore, if the feedback information is confirmation but with low confidence, or rejection but with low confidence, the system can execute a second handling strategy, such as generating an anomaly warning pending review and notifying O&M personnel to conduct a second confirmation through highlighting, voice reminders, or push notifications to mobile terminals to ensure the accuracy of the warning. This strategy adjustment mechanism based on feedback information enables the system to continuously learn and optimize, improving the accuracy and reliability of early warnings.

[0033] This application presents an AI-based oil and gas pipeline monitoring data processing method that integrates distributed fiber optic sensing technology with advanced artificial intelligence algorithms to form a highly efficient and intelligent oil and gas pipeline safety monitoring system. The method first acquires distributed fiber optic sensing data of the oil and gas pipeline, which includes key information such as vibration and acoustics along the pipeline route. Since the raw sensing data is often affected by environmental noise, this application further performs noise reduction processing on this data to remove irrelevant noise, thereby obtaining cleaner and more accurate pipeline status data. This step is the foundation for subsequent accurate judgment of abnormal events, ensuring the effectiveness of the input model.

[0034] Subsequently, the processed pipeline status data is input into a pre-defined anomaly detection model. This model is one of the core innovations of this application; unlike traditional threshold-based methods, it employs a probabilistic detection mechanism. This means that the model not only determines whether an anomaly exists but also assesses the probability of its occurrence, thus providing maintenance personnel with more refined decision-making support. For example, when the model calculates a high posterior probability that the pipeline is in an abnormal state, the system will issue a preliminary warning.

[0035] Upon receiving an initial alert, operations and maintenance (O&M) personnel will provide feedback on the anomaly based on the actual situation, including confirmation or rejection. This step incorporates a human-machine collaboration concept, fully leveraging the experience and expertise of O&M personnel. The system will determine the final anomaly alert handling strategy based on the feedback from O&M personnel and its internal evaluation mechanism. For example, if O&M personnel confirm the anomaly, the system will generate a confirmed anomaly alert, providing detailed location and type information; if O&M personnel reject the anomaly, the system may not generate an alert to avoid false alarms. This feedback mechanism allows the system to continuously learn and optimize, improving the accuracy of anomaly judgment and the intelligence level of the alert strategy.

[0036] Overall, this application forms a closed-loop intelligent monitoring and early warning system through data acquisition and noise reduction, probabilistic anomaly event judgment based on artificial intelligence, and human-machine collaborative feedback and strategy adjustment. The various technical features work together to solve the problems of high false alarm rates, poor positioning accuracy, and low operation and maintenance efficiency in traditional oil and gas pipeline monitoring, significantly improving the intelligence level and reliability of oil and gas pipeline safety monitoring.

[0037] In summary, this application overcomes the limitations of existing oil and gas pipeline monitoring technologies in data processing, anomaly detection, and early warning strategies by integrating advanced data processing technologies, artificial intelligence models, and human-machine collaboration mechanisms. It significantly improves the accuracy, reliability, and intelligence level of oil and gas pipeline safety monitoring, providing a more solid guarantee for the safe operation of oil and gas pipelines.

[0038] In some embodiments of this application, the step of acquiring distributed optical fiber sensing data of oil and gas pipelines can be understood as real-time and continuous sensing of the physical state along the oil and gas pipeline.

[0039] Specifically, the steps for acquiring distributed fiber optic sensing data from oil and gas pipelines include: Vibration and acoustic monitoring data along the oil and gas pipelines are collected using distributed fiber optic sensors laid along the pipelines.

[0040] Distributed fiber optic sensors are sensing devices capable of continuously measuring physical quantities (such as temperature, strain, and vibration) along the length of an optical fiber. Their working principle is typically based on phenomena such as Rayleigh scattering, Brillouin scattering, or Raman scattering in optical fibers. When the fiber is affected by external vibrations or sound waves, the characteristics of the optical signal (such as phase, frequency, and intensity) change. By analyzing these changes, vibration and acoustic information at various points along the pipeline can be deduced. Specifically, distributed fiber optic sensors can be laid outside or inside oil and gas pipelines, or laid parallel to the pipeline, to achieve full-line coverage monitoring. The collected vibration monitoring data can reflect information such as mechanical stress, deformation, and disturbances from third-party construction in the pipeline, while the acoustic monitoring data can capture sound wave signals generated by fluid anomalies, leaks, and foreign object impacts inside the pipeline.

[0041] The aforementioned technical solutions enable real-time, all-weather, and comprehensive monitoring of oil and gas pipelines, effectively overcoming the limitations of traditional point sensors, such as limited coverage and susceptibility to electromagnetic interference. Distributed fiber optic sensors offer advantages such as strong resistance to electromagnetic interference, inherent safety, and long transmission distances, making data acquisition more stable and reliable. Furthermore, by simultaneously acquiring both vibration and acoustic monitoring data, pipeline conditions can be cross-verified from different dimensions, improving the accuracy and robustness of anomaly detection and providing solid data support for the safe operation of oil and gas pipelines.

[0042] In some of the embodiments described above in this application, noise reduction processing of distributed fiber optic sensing data is a crucial step in obtaining accurate pipeline status data. However, distributed fiber optic sensing data from oil and gas pipelines is often affected by various complex noises, such as environmental noise and equipment noise. If the noise reduction processing is insufficient or inaccurate, it may lead to distortion of the pipeline status data, thereby affecting the accuracy of subsequent anomaly event assessment. Therefore, this application further proposes a more refined and effective noise reduction method to improve the quality and reliability of pipeline status data.

[0043] In this regard, this application further proposes the following steps for denoising distributed optical fiber sensing data to obtain processed pipeline status data: The distributed optical fiber sensing data is decomposed into multiple intrinsic mode function components using the ensemble empirical mode decomposition method. The multiple intrinsic mode function components are then denoised using the wavelet thresholding method. Finally, the denoised multiple intrinsic mode function components are reconstructed to obtain the pipeline state data.

[0044] Specifically, ensemble empirical mode decomposition (EMD) is an adaptive signal processing technique that aims to decompose complex nonlinear and nonstationary signals into a series of physically meaningful intrinsic mode function (EMF) components. Each EMF component represents the oscillation mode of the original signal at different time scales and satisfies specific conditions. For example, the number of local maxima and local minima differs from the number of zero-crossings by at most one across the entire data segment, and the mean of the upper and lower envelopes defined by the local maxima and local minima is zero. Through this decomposition, different frequency components and noise components in the original distributed fiber optic sensing data can be effectively separated.

[0045] Wavelet thresholding can be understood as a commonly used signal denoising technique. Its principle is to decompose the signal into wavelet domains of different scales using wavelet transform. Since noise in the wavelet domain typically manifests as small-amplitude coefficients, while signals manifest as large-amplitude coefficients, a threshold can be set to perform soft or hard thresholding on the wavelet coefficients, thereby removing noise. Specifically, for the multiple intrinsic mode function components obtained from the decomposition, wavelet thresholding can be applied to each component to remove the noise components it contains.

[0046] In practical applications, when reconstructing multiple denoised intrinsic mode function (IMF) components, all denoised IMF components are typically superimposed to recover the effective components of the original signal and remove noise. This yields cleaner and more accurate pipeline state data, providing a high-quality data foundation for subsequent anomaly event detection.

[0047] Through the above technical solution, this application can significantly improve the noise reduction effect of distributed optical fiber sensing data. The combination of empirical mode decomposition (EMD) and wavelet thresholding methods enables the system to more effectively handle the complex nonlinear and non-stationary noise in oil and gas pipeline monitoring data, avoiding the limitations of traditional single noise reduction methods. Therefore, the obtained pipeline status data has a higher signal-to-noise ratio and accuracy, and can more realistically reflect the actual operating status of oil and gas pipelines. This high-quality pipeline status data provides a solid foundation for subsequent anomaly event judgment based on an anomaly event judgment model, thereby effectively reducing the false alarm rate and missed alarm rate, and improving the reliability and early warning accuracy of the entire oil and gas pipeline monitoring system.

[0048] In some preferred embodiments, it is assumed that distributed fiber optic sensors laid along the oil and gas pipeline collect raw distributed fiber optic sensing data containing various signals such as environmental vibration, vehicle passage, and slight pipeline deformation. First, this raw data is processed using an ensemble empirical mode decomposition method, decomposing it into, for example, ten intrinsic mode function (EMF) components. Some of these components may primarily contain high-frequency noise, others may contain low-frequency environmental vibration, and still others may contain critical signals such as pipeline deformation. Next, for each EMF component, an appropriate wavelet basis function (e.g., Daubechies wavelet) and threshold type (e.g., soft thresholding) are selected, and wavelet thresholding is used for denoising. For example, for components primarily containing high-frequency noise, a higher threshold can be set to completely remove the noise; for components containing critical signals, a lower threshold is set to preserve signal details. Finally, all denoised EMF components are reconstructed to obtain pipeline state data that has significantly removed noise and clearly reflects the true state of the pipeline. This processed data can clearly display minor deformations or abnormal vibration patterns in the pipeline, thus providing accurate input for subsequent anomaly event detection.

[0049] Traditional oil and gas pipeline monitoring data processing methods typically provide only probabilistic judgments when determining the presence of anomalies. They lack mechanisms to translate these probabilistic judgments into definitive confirmations of anomalies, and also fail to provide precise location information. For example, assuming the system only outputs an 80% probability of pipeline anomaly, maintenance personnel may find it difficult to directly determine whether immediate action is necessary, or even after confirming the anomaly, they may be unable to quickly pinpoint the problem. Failure to address these issues could lead to delayed responses to anomalies, increased pipeline operational risks, and potentially unnecessary resource waste.

[0050] In response, this application proposes an artificial intelligence-based method for processing oil and gas pipeline monitoring data. The step of determining whether an abnormal event exists based on the processed pipeline status data and a pre-defined abnormal event judgment model includes: A Bayesian hypothesis testing model is used as the pre-defined abnormal event judgment model to calculate the posterior probability of the pipeline being in an abnormal state based on the pipeline status data. If the posterior probability is higher than the first preset threshold, an abnormal event is determined to exist, and a two-layer localization optimization is performed using the K-means clustering algorithm to obtain the location of the abnormal event.

[0051] Specifically, the aforementioned pre-defined abnormal event judgment model is implemented as a Bayesian hypothesis testing model. A Bayesian hypothesis testing model is a statistical inference method that aims to update beliefs about hypotheses by combining prior information and observed data. Here, this model is used to calculate the posterior probability that pipeline status data indicates the pipeline is in an abnormal state. The posterior probability can be understood as the probability that the pipeline is indeed in an abnormal state given the observed current pipeline status data.

[0052] Furthermore, to transform probabilistic judgments into explicit anomaly event judgments, a first preset threshold is set. When the calculated posterior probability is higher than the first preset threshold, the system determines that an anomaly event exists. The first preset threshold can be adjusted according to the needs of the actual application scenario and the tolerance for false positives and false negatives to balance the sensitivity and accuracy of detection.

[0053] Furthermore, once an anomaly is identified, this application employs a K-means clustering algorithm to perform a two-layer localization optimization to pinpoint the location of the anomaly. K-means clustering is an iterative clustering analysis algorithm that aims to divide data points into K clusters, ensuring that each data point belongs to the cluster corresponding to its nearest mean (centroid). Here, this algorithm is applied to pipeline status data, and through a two-layer optimization process, the range of anomalies is gradually narrowed down, ultimately determining their precise location.

[0054] Through the above technical solutions, this application can significantly improve the accuracy and reliability of anomaly detection in oil and gas pipelines. The use of the Bayesian hypothesis testing model makes anomaly judgment more intelligent and adaptive, effectively reducing false alarm and false negative rates. Simultaneously, by introducing a clear posterior probability threshold, probabilistic judgments are transformed into deterministic judgments, greatly simplifying the decision-making process for maintenance personnel. Most importantly, the two-layer localization optimization combined with the K-means clustering algorithm enables anomalies not only to be detected but also to be precisely located. This has irreplaceable value in guiding on-site inspections, maintenance, and emergency response, thereby significantly improving the safe operation level and emergency response efficiency of oil and gas pipelines.

[0055] In some preferred embodiments, it is assumed that the distributed fiber optic sensing data of the oil and gas pipeline, after noise reduction processing, yields processed pipeline status data. To determine whether an abnormal event exists, the system first employs a Bayesian hypothesis testing model as a preset abnormal event judgment model. This model calculates the posterior probability that the pipeline is in an abnormal state based on the current pipeline status data and a pre-set prior probability. For example, when a continuous, high-energy vibration signal appears in the pipeline status data, the Bayesian model calculates the posterior probability that the signal represents an abnormal pipeline state. If the calculated posterior probability, for example, is 0.95, and the preset first threshold is 0.8, since 0.95 is higher than 0.8, the system will determine that an abnormal event exists. Next, to determine the specific location of the abnormal event, the system will initiate a two-layer localization optimization using a K-means clustering algorithm. The first layer of localization optimization may, based on the signal propagation characteristics, preliminarily determine that the abnormal event occurs within a relatively long pipeline area, for example, between 10 and 12 kilometers in pipeline length. Subsequently, the second layer of positioning optimization performs K-means clustering analysis on more refined pipeline status data within this coarse area. For example, it clusters sensor data points within the area to identify the clusters with the highest concentration of abnormal signals, thereby improving the positioning accuracy of abnormal events to the meter or even sub-meter level, such as determining that an abnormal event occurred at a pipeline mileage of 10.5 kilometers. In this way, maintenance personnel can obtain precise location information of abnormal events for targeted on-site inspections and handling.

[0056] In some of the embodiments described above in this application, a two-layer localization optimization using the K-means clustering algorithm is proposed to obtain the location of abnormal events. However, in practical applications, localization optimization using only the K-means clustering algorithm may not meet the high requirements for the localization accuracy of abnormal events, especially in the case of complex oil and gas pipeline environments and variable signal propagation characteristics. This may result in inaccurate localization results and affect the efficiency of subsequent emergency response.

[0057] In response, this application further proposes a two-layer positioning optimization, including a first-layer positioning optimization and a second-layer positioning optimization. The first-layer positioning optimization includes determining a rough area along the oil and gas pipeline where anomalies occur based on the signal propagation characteristics of pipeline status data. The second-layer positioning optimization includes performing cluster analysis on the signal propagation characteristics of pipeline status data within the rough area to determine the location of the anomaly, thereby improving the positioning accuracy of the anomaly to a preset range.

[0058] Specifically, the first layer of location optimization aims to quickly identify the approximate range of anomalies by utilizing the signal propagation characteristics of pipeline status data. For example, by analyzing features such as signal attenuation, propagation speed, and time difference of arrival, the approximate location of the anomaly source can be preliminarily determined. This determination of a coarse area can effectively narrow the search range for subsequent precise location, improving processing efficiency. The second layer of location optimization then performs a more refined clustering analysis of the signal propagation characteristics of the pipeline status data within the coarse area determined by the first layer. For example, K-means clustering or other advanced clustering methods can be used to finely divide the signal features within the coarse area, identifying a more precise location of the anomaly source. This hierarchical, progressive optimization strategy effectively overcomes the problem of insufficient location accuracy of a single algorithm in complex environments. In practical applications, improving the location accuracy of anomalies to a preset range means that, through two layers of location optimization, the error between the final determined location of the anomaly and the actual location is controlled within an acceptable engineering range, for example, less than 1 meter or 5 meters, to meet the needs of oil and gas pipeline safety management and emergency repair.

[0059] Through the above technical solution, this application can significantly improve the location accuracy of abnormal events, enabling it to meet preset engineering requirements. This high-precision location capability helps maintenance personnel quickly and accurately pinpoint the source of anomalies, thereby shortening emergency response time and reducing accident losses. Furthermore, the hierarchical optimization strategy also improves the efficiency and robustness of the location algorithm, enabling it to provide reliable location results when facing abnormal events of different types and intensities, thus providing a more solid technical guarantee for the safe operation of oil and gas pipelines.

[0060] In some preferred embodiments, it is assumed that an anomaly occurs within a 100-kilometer monitoring section of the oil and gas pipeline. In the first-level positioning optimization, the system first analyzes the signal propagation speed and attenuation characteristics based on pipeline status data collected by distributed fiber optic sensors, initially determining that the anomaly may occur in a coarse area between 50 and 60 kilometers. Subsequently, in the second-level positioning optimization, the system focuses on analyzing the pipeline status data within this 10-kilometer coarse area, performing K-means clustering analysis on the signal propagation characteristics within this area. For example, by clustering features such as signal strength and frequency changes, the location of the anomaly can be further refined to 55.3 kilometers, and the positioning error can be controlled within 1 meter, thus meeting the preset positioning accuracy requirements. This hierarchical positioning strategy effectively avoids the complexity of performing high-precision calculations across the entire 100-kilometer pipeline, while ensuring the accuracy of the final positioning result.

[0061] In some embodiments described above, this application proposes a method to determine the existence of abnormal events based on pipeline status data and a pre-defined abnormal event judgment model, employing a two-layer positioning optimization to determine the location of the abnormal event. However, in practical applications, when directly using raw or pre-processed pipeline status data for positioning optimization, its signal propagation characteristics may be affected by environmental noise, data acquisition errors, or the superposition of multiple events, resulting in insufficient robustness or ambiguity of the extracted features, thereby affecting the accuracy and stability of the subsequent two-layer positioning optimization. If the above problems are not addressed, the positioning accuracy of abnormal events may not meet actual operation and maintenance needs, and may even lead to false alarms or missed alarms.

[0062] In response, this application further proposes a method for multi-channel spatiotemporal feature extraction and correlation analysis of pipeline status data before performing the above two-layer positioning optimization. The aim is to extract more accurate and robust signal propagation features based on the inherent consistency of pipeline status data in spatial, frequency and time dimensions, so as to provide high-quality input for subsequent positioning optimization.

[0063] Before performing the above two-layer positioning optimization, it also includes multi-channel spatiotemporal feature extraction and correlation analysis of pipeline status data. Based on the inherent consistency of pipeline status data in spatial, frequency and time dimensions, the signal propagation characteristics of pipeline status data are extracted.

[0064] Specifically, multi-channel spatiotemporal feature extraction and correlation analysis refers to the comprehensive processing of multi-channel vibration and acoustic monitoring data collected from distributed fiber optic sensors laid along oil and gas pipelines. "Multi-channel" refers to data streams collected at different locations along the pipeline; these streams are synchronous in time and continuous in space. Spatiotemporal feature extraction aims to identify signal features related to anomalous events and exhibiting specific patterns in time and space from these multi-channel data. For example, a sliding window technique can be used to segment the data in the time dimension and analyze data from different spatial locations within each time window. Correlation analysis establishes connections between the extracted spatiotemporal features to identify their inherent consistency in the spatial, frequency, and temporal dimensions. This consistency is an inherent property of anomalous event signal propagation; for example, vibration or sound waves caused by an anomalous event will propagate along the pipeline, and their arrival time, amplitude attenuation, and frequency variations at different sensors will exhibit specific spatiotemporal correlation patterns.

[0065] Signal propagation characteristics can be understood as physical quantities or mathematical models that characterize the propagation patterns of abnormal events in a pipeline. These characteristics may include, but are not limited to, signal propagation speed, attenuation rate, frequency shift, phase difference, and differences in signal arrival time between different channels. The aim is to transform the raw, complex pipeline state data into a simpler, more discriminative feature representation, enabling subsequent localization algorithms to more accurately identify the location of abnormal events. In practical applications, signal propagation characteristics are extracted based on the inherent consistency of pipeline state data in the spatial, frequency, and temporal dimensions. For example, cross-correlation functions, time-frequency analysis (such as short-time Fourier transform and wavelet transform), and array signal processing techniques (such as beamforming and MUSIC algorithms) can be used to analyze the time delay, phase, and amplitude relationships between multi-channel data, thereby accurately capturing the signal propagation characteristics.

[0066] By employing the aforementioned technical solution, before implementing two-layer positioning optimization, multi-channel spatiotemporal feature extraction and correlation analysis are performed on pipeline status data. Based on the inherent consistency of the data across spatial, frequency, and temporal dimensions, signal propagation features are extracted, effectively overcoming the impact of noise and interference in the original data on positioning accuracy. Compared to directly using pipeline status data without refined feature extraction for positioning, this solution obtains more discriminative and robust signal propagation features, significantly improving the accuracy and stability of anomaly location. Furthermore, by utilizing multi-dimensional inherent consistency, the positioning process becomes more adaptable to complex environments and various anomalies, reducing the risk of false alarms and missed alarms, and providing more reliable technical support for the safe operation and maintenance of oil and gas pipelines.

[0067] In some of the embodiments described above in this application, although it is possible to determine the existence and location of abnormal events based on the processed pipeline status data, it does not further provide specific information about the type of abnormal event. In practical applications, different types of abnormal events may require different response strategies, and the lack of type identification capability will affect the decision-making efficiency and response speed of operation and maintenance personnel. To address this, this application further proposes a method that can identify the type of abnormal event after its location has been obtained.

[0068] After locating the anomalous event, the process also includes: using a one-dimensional convolutional neural network to extract the anomalous features of the anomalous event, and inputting the anomalous features into a support vector machine for classification to identify the type of anomalous event.

[0069] Specifically, after the location of the aforementioned abnormal event is determined, in order to identify the type of abnormal event, this application employs a one-dimensional convolutional neural network (1D CNN) to extract the abnormal features of the abnormal event. A one-dimensional convolutional neural network is a deep learning model specifically designed for processing sequential data. It automatically learns and extracts local features and patterns in the data by sliding convolutional kernels along the time dimension. Here, the abnormal features of an abnormal event refer to unique data patterns or signal characteristics that can distinguish different types of abnormal events, such as the frequency distribution and amplitude variation trends of different types of vibration events. The extracted abnormal features are then input into a support vector machine (SVM) for classification. A support vector machine is a classic machine learning classification algorithm that constructs one or a set of hyperplanes to separate sample points of different categories, thereby achieving the identification of abnormal event types. For example, a support vector machine can be trained to distinguish between different types of abnormal events such as pipeline leaks, third-party construction, and equipment failures.

[0070] Through the above technical solution, this application overcomes the limitation of only providing abnormal event location information without specifying the type of event. When maintenance personnel receive an anomaly alert, they can not only know the location of the abnormal event but also obtain its specific type, such as pipeline leakage, third-party construction, or equipment failure. This detailed type information enables maintenance personnel to more quickly and accurately determine the nature and potential hazards of the event, thereby immediately activating targeted emergency plans and resource allocation, avoiding unnecessary on-site investigation and analysis time, and significantly improving the efficiency and accuracy of emergency response. Furthermore, identifying the type of abnormal event also helps accumulate richer historical data, providing a more accurate basis for subsequent risk assessment and preventative maintenance, further enhancing the safe operation and maintenance capabilities of oil and gas pipelines.

[0071] However, in practice, the feedback from maintenance personnel may be affected by a variety of factors, such as their experience level and current workload. If warnings are generated directly based solely on the surface content of the feedback information (confirmation or rejection), it may lead to false alarms or missed alarms, thereby affecting the accuracy and reliability of the warning system.

[0072] In response, this application further proposes a more intelligent and robust method for determining anomaly early warning handling strategies. This method comprehensively considers the feedback information and confidence level of operation and maintenance personnel to achieve more accurate and reliable early warning decisions.

[0073] The steps for determining the anomaly warning handling strategy based on the feedback information mentioned above include: Based on the historical accuracy of feedback from operations and maintenance personnel and the current workload assessment results, the feedback confidence level of the feedback information is calculated. When the feedback information is confirmed and the feedback confidence level is higher than the second preset threshold, the first processing strategy is executed. The first processing strategy includes generating a confirmed anomaly warning, marking the location of the anomaly on the map and displaying the type of the anomaly through a visualization operation platform. When the feedback information is confirmation and the feedback confidence level is not higher than the second preset threshold, or when the feedback information is rejection and the feedback confidence level is not higher than the second preset threshold, the second processing strategy is executed. The second processing strategy includes generating an anomaly warning to be reviewed and notifying the operation and maintenance personnel to conduct a second review of the anomaly warning to be reviewed through at least one of the following methods: highlighting prompts on the visual operation platform, voice reminders, and pushing messages to mobile terminals. When the feedback is a rejection and the feedback confidence level is higher than the second preset threshold, a third processing strategy is executed, which includes not generating an anomaly warning.

[0074] Specifically, feedback confidence refers to a quantitative assessment of the reliability of feedback information provided by maintenance personnel. This confidence is calculated by comprehensively analyzing the historical feedback accuracy rate of maintenance personnel and the current workload assessment results. The historical feedback accuracy rate can be understood as the proportion of times the maintenance personnel's feedback matches the actual situation when handling similar abnormal events in the past, reflecting the maintenance personnel's experience and judgment. The current workload assessment result refers to the busyness of maintenance personnel when receiving feedback requests for current abnormal events. This can be assessed based on indicators such as the number of events they are handling and response time; excessive workload may lead to misjudgment. The second preset threshold is a key parameter used to distinguish between high and low feedback confidence; its value can be adjusted according to the actual application scenario and the requirements for early warning accuracy. The visual operation platform refers to a graphical user interface used to display monitoring data of oil and gas pipelines, the location and type of abnormal events, and early warning information. Maintenance personnel can interact with this platform. Mobile terminals can be understood as smart devices carried by maintenance personnel, such as smartphones or tablets, used to receive system push notifications.

[0075] Through the above technical solution, this application can adaptively adjust the early warning strategy based on the reliability feedback from maintenance personnel, significantly improving the accuracy and intelligence level of early warning for abnormal events in oil and gas pipelines. This solution not only effectively reduces false alarms and missed alarms, lowering maintenance costs, but also improves the work efficiency of maintenance personnel, ensuring the safe and stable operation of oil and gas pipelines.

[0076] In some embodiments of this application, a second processing strategy is proposed to be executed when the feedback information is confirmation and the feedback confidence level is not higher than a second preset threshold, or when the feedback information is rejection and the feedback confidence level is not higher than the second preset threshold. The second processing strategy includes generating a pending review anomaly warning and notifying maintenance personnel to conduct a secondary confirmation of the warning through at least one of the following methods: highlighting on a visual operation platform, voice reminder, and pushing a message to a mobile terminal. However, in its implementation, if maintenance personnel fail to provide timely secondary confirmation feedback for the pending review anomaly warning, potential abnormal events may not be handled effectively and promptly, thereby affecting the safe operation of oil and gas pipelines.

[0077] In this regard, this application further proposes that after implementing the aforementioned second processing strategy, it also includes: If no secondary confirmation feedback from maintenance personnel is received within the preset time, the pipeline status data within the preset time window before the generated abnormality warning will be retrieved, along with the auxiliary monitoring data sources associated with the pipeline status data, for data backtracking analysis.

[0078] Specifically, the preset time refers to the maximum time the system waits for maintenance personnel to provide secondary confirmation feedback. This time can be flexibly configured based on actual business needs, the urgency of the anomaly, and the workload of maintenance personnel, for example, it can be set to 10 minutes, 30 minutes, or 1 hour. Its purpose is to provide maintenance personnel with a reasonable response window while avoiding delays in anomaly handling due to long waiting times. Secondary confirmation feedback refers to the operation by which maintenance personnel, after receiving an anomaly alert requiring review, confirm or reject the alert through a visual operation platform or other designated methods.

[0079] In practical applications, retrieving pipeline status data within a preset time window before the generation of the pending anomaly warning refers to the system automatically retracing back to the pipeline status data collected and processed within a period prior to the generation of the warning when no secondary confirmation feedback is received. This preset time window can be set according to the needs of data analysis; for example, it can trace back to data from minutes, hours, or even days before the anomaly occurred, allowing for a more comprehensive analysis of the anomaly's evolution. The auxiliary monitoring data sources associated with the pipeline status data refer to monitoring data that provides information on the operating status of oil and gas pipelines, in addition to distributed fiber optic sensor data. These include pressure sensor data, temperature sensor data, flow meter data, SCADA system data, video surveillance data, and environmental meteorological data. These auxiliary data sources can provide multi-dimensional and multi-faceted support information for the analysis of anomalies.

[0080] Furthermore, data backtracking analysis refers to the system's in-depth analysis of retrieved pipeline status data and auxiliary monitoring data sources to reassess the likelihood, nature, and impact of abnormal events. This may include, but is not limited to, rerunning the abnormal event judgment model, performing trend analysis, correlation analysis, and pattern recognition, aiming to find evidence of abnormal events or rule out the possibility of false alarms from a broader data context.

[0081] Through the above technical solution, this application significantly improves the robustness and reliability of oil and gas pipeline anomaly early warning processing. Even if maintenance personnel fail to conduct a secondary confirmation of the anomaly early warning in a timely manner due to various reasons (such as busy work, network failure, misoperation, etc.), the system can automatically initiate data backtracking analysis, avoiding the risk of overlooking potential anomalies. This proactive data backtracking mechanism not only provides the system with a basis for re-evaluating anomalies but also helps to discover deeper anomaly patterns or related information, thereby improving the accuracy of anomaly identification and processing efficiency. In addition, by combining multi-source auxiliary monitoring data for analysis, the comprehensiveness and reliability of anomaly judgment are further enhanced, effectively compensating for the limitations that may exist in a single data source and ensuring the intelligence and automation level of the oil and gas pipeline monitoring system.

[0082] In some preferred embodiments, suppose the oil and gas pipeline monitoring system detects an abnormal vibration signal on a section of the pipeline and generates a pending anomaly warning based on a preset anomaly judgment model. The system highlights the warning to maintenance personnel A through a visual operation platform and pushes a message to them via a mobile terminal, requiring them to confirm the warning again within 30 minutes. However, because maintenance personnel A is handling another emergency at the time, they fail to confirm or reject the warning within 30 minutes. At this point, the system will automatically trigger a data backtracking analysis process. The system will retrieve distributed fiber optic sensor data from the hour prior to the generation of the pending anomaly warning, and simultaneously acquire pressure, temperature, and flow data for that section of the pipeline during that time period. The system will comprehensively analyze this backtracking data, for example, by comparing the vibration spectrum changes before and after the anomaly occurred, and combining the fluctuations in pressure and temperature data, to reassess the nature of the anomaly event. If the backtracking analysis results further support the existence of the abnormal event, the system may automatically raise the priority of the warning, or notify another maintenance personnel B to handle it. In extreme cases, it may even automatically trigger a higher level of emergency response according to preset rules, such as remotely shutting down the relevant valves or dispatching on-site personnel for inspection, thereby avoiding serious consequences that may result from the lack of manual confirmation.

[0083] Regarding this, secondly, refer to Figure 2This application proposes an artificial intelligence-based oil and gas pipeline monitoring data processing system for executing the aforementioned artificial intelligence-based oil and gas pipeline monitoring data processing method. The system includes: The data acquisition and processing module 210 is used to acquire distributed optical fiber sensing data of oil and gas pipelines, perform noise reduction processing on the distributed optical fiber sensing data, and obtain processed pipeline status data. The abnormal event judgment module 220 is used to determine whether there is an abnormal event based on the processed pipeline status data and the preset abnormal event judgment model. The preset abnormal event judgment model is used to perform probabilistic abnormal event judgment on the pipeline status data. The anomaly warning module 230 is used to receive feedback information from operation and maintenance personnel regarding abnormal events. The feedback information includes confirmation or rejection. Based on the feedback information, the anomaly warning handling strategy is determined.

[0084] Specifically, the data acquisition and processing module 210 is configured to perform the data acquisition and noise reduction steps in the above method. This module can integrate the data interface of a distributed fiber optic sensor to receive vibration and acoustic monitoring data collected in real time by distributed fiber optic sensors laid along the oil and gas pipeline. Subsequently, the module preprocesses the received raw data, for example, by using ensemble empirical mode decomposition to decompose the distributed fiber optic sensor data into multiple intrinsic mode function components, and then using wavelet thresholding to denoise these components, finally reconstructing the processed pipeline status data. The purpose is to provide high-quality, low-noise pipeline status data, laying the foundation for subsequent anomaly event judgment.

[0085] The abnormal event judgment module 220 is configured to execute the abnormal event judgment step in the above method. This module receives the processed pipeline status data output by the data acquisition and processing module 210 and analyzes this data based on a preset abnormal event judgment model. For example, this module can use a Bayesian hypothesis testing model as the preset abnormal event judgment model to calculate the posterior probability that the pipeline status data represents an abnormal state. If the posterior probability is higher than a first preset threshold, an abnormal event is judged to exist, and a two-layer localization optimization can be further performed using a K-means clustering algorithm to determine the precise location of the abnormal event. Its purpose is to accurately and promptly identify potential abnormal situations and provide preliminary localization information.

[0086] In practical applications, the anomaly warning module 230 is configured to execute the anomaly warning handling strategy determination step in the above method. This module is responsible for receiving feedback information from operations and maintenance personnel regarding identified anomalies. This feedback information typically includes confirmation or rejection of the anomaly. Based on this feedback information, and combined with the historical accuracy rate of feedback from operations and maintenance personnel and the current workload assessment results, the feedback confidence level is calculated to determine the corresponding anomaly warning handling strategy. For example, depending on different combinations of feedback information and feedback confidence levels, a confirmed anomaly warning, a pending review anomaly warning, or no anomaly warning can be generated. The aim is to achieve intelligent management of anomaly warnings, improve the accuracy and reliability of warnings, and optimize the workflow of operations and maintenance personnel.

[0087] Through the aforementioned system solution, this application transforms the complex oil and gas pipeline monitoring data processing method into a deployable and operable physical system, significantly improving the method's practicality and automation level. The system, with its clear modular division, ensures clear data flow and well-defined responsibilities for each processing stage, thereby enhancing the efficiency and stability of the entire processing flow. Furthermore, the systematic implementation facilitates integration and expansion, better adapting to oil and gas pipeline monitoring scenarios of varying scales and needs. It reduces the need for manual intervention, lowers maintenance costs, and significantly improves the timeliness and accuracy of anomaly detection, location, and early warning, providing a solid guarantee for the safe operation of oil and gas pipelines.

[0088] The above are merely embodiments of this application and are not intended to limit the scope of protection of this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A data processing method for oil and gas pipeline monitoring based on artificial intelligence, characterized in that, include: Acquire distributed optical fiber sensing data of oil and gas pipelines, perform noise reduction processing on the distributed optical fiber sensing data, and obtain processed pipeline status data. Based on the processed pipeline status data and the preset abnormal event judgment model, it is determined whether there is an abnormal event. The preset abnormal event judgment model is used to perform probabilistic abnormal event judgment on the pipeline status data. Receive feedback information from operations and maintenance personnel regarding the abnormal event, wherein the feedback information includes confirmation or rejection, and determine the abnormal early warning handling strategy based on the feedback information.

2. The method for processing oil and gas pipeline monitoring data based on artificial intelligence according to claim 1, characterized in that, The steps for acquiring distributed fiber optic sensing data of oil and gas pipelines include: Vibration and acoustic monitoring data along the oil and gas pipelines are collected using distributed fiber optic sensors laid along the pipelines.

3. The method for processing oil and gas pipeline monitoring data based on artificial intelligence according to claim 1, characterized in that, The step of performing noise reduction processing on the distributed optical fiber sensing data to obtain processed pipeline status data includes: The distributed optical fiber sensing data is decomposed into multiple intrinsic mode function components using the ensemble empirical mode decomposition method. The multiple intrinsic mode function components are then denoised using the wavelet thresholding method. Finally, the denoised multiple intrinsic mode function components are reconstructed to obtain the pipeline state data.

4. The method for processing oil and gas pipeline monitoring data based on artificial intelligence according to claim 1, characterized in that, The step of determining whether an abnormal event exists based on the processed pipeline status data and a preset abnormal event judgment model includes: A Bayesian hypothesis testing model is used as the preset abnormal event judgment model to calculate the posterior probability that the pipeline is in an abnormal state based on the pipeline status data. If the posterior probability is higher than the first preset threshold, it is determined that the abnormal event exists, and a two-layer localization optimization is performed using the K-means clustering algorithm to obtain the localization of the abnormal event.

5. The method for processing oil and gas pipeline monitoring data based on artificial intelligence according to claim 4, characterized in that, The two-layer positioning optimization includes a first-layer positioning optimization and a second-layer positioning optimization. The first-layer positioning optimization includes determining a rough area along the oil and gas pipeline where the abnormal event occurs based on the signal propagation characteristics of the pipeline status data. The second-layer positioning optimization includes performing cluster analysis on the signal propagation characteristics of the pipeline status data within the rough area to determine the location of the abnormal event and improve the positioning accuracy of the abnormal event to a preset range.

6. The method for processing oil and gas pipeline monitoring data based on artificial intelligence according to claim 5, characterized in that, Before performing the two-layer localization optimization, the following is also included: Multi-channel spatiotemporal feature extraction and correlation analysis are performed on the pipeline status data. Based on the inherent consistency of the pipeline status data in the spatial, frequency and time dimensions, the signal propagation features of the pipeline status data are extracted.

7. The method for processing oil and gas pipeline monitoring data based on artificial intelligence according to claim 4, characterized in that, After locating the abnormal event, the process also includes: A one-dimensional convolutional neural network is used to extract the abnormal features of the abnormal event, and the abnormal features are input into a support vector machine for classification to identify the type of the abnormal event.

8. The method for processing oil and gas pipeline monitoring data based on artificial intelligence according to claim 7, characterized in that, The step of determining the anomaly warning handling strategy based on the feedback information includes: Based on the historical accuracy of feedback from operations and maintenance personnel and the current workload assessment results, the feedback confidence level of the feedback information is calculated. When the feedback information is confirmed and the confidence level of the feedback is higher than the second preset threshold, a first processing strategy is executed, wherein the first processing strategy includes generating a confirmed anomaly warning, marking the location of the anomaly on a map and displaying the type of the anomaly through a visualization operation platform; When the feedback information is confirmation but the feedback confidence level is not higher than the second preset threshold, or when the feedback information is rejection and the feedback confidence level is not higher than the second preset threshold, the second processing strategy is executed. The second processing strategy includes generating a pending review anomaly warning and notifying the operation and maintenance personnel to conduct a second review of the pending review anomaly warning through at least one of the following methods: highlighting prompts on the visual operation platform, voice reminders, and pushing messages to mobile terminals. When the feedback information is a rejection and the confidence level of the feedback is higher than the second preset threshold, a third processing strategy is executed, wherein the third processing strategy includes not generating an abnormal warning.

9. The method for processing oil and gas pipeline monitoring data based on artificial intelligence according to claim 8, characterized in that, Following the execution of the second processing strategy, the following is also included: If no secondary confirmation feedback from maintenance personnel is received regarding the pending anomaly warning within a preset time period, the pipeline status data within the preset time window prior to the generation of the pending anomaly warning, as well as the auxiliary monitoring data source associated with the pipeline status data, will be retrieved for data backtracking analysis.

10. An artificial intelligence-based oil and gas pipeline monitoring data processing system, used to execute the artificial intelligence-based oil and gas pipeline monitoring data processing method as described in any one of claims 1 to 9, characterized in that, The system includes: The data acquisition and processing module is used to acquire distributed optical fiber sensing data of oil and gas pipelines, and to perform noise reduction processing on the distributed optical fiber sensing data to obtain processed pipeline status data. An abnormal event judgment module is used to determine whether there is an abnormal event based on the processed pipeline status data and a preset abnormal event judgment model, wherein the preset abnormal event judgment model is used to perform probabilistic abnormal event judgment on the pipeline status data. An anomaly warning module is used to receive feedback information from operation and maintenance personnel regarding the anomaly event, wherein the feedback information includes confirmation or rejection, and an anomaly warning handling strategy is determined based on the feedback information.

Images