An early warning event processing system for an oil and gas pipeline optical fiber early warning system

Abstract

The present application relates to a kind of oil and gas pipeline optical fiber early warning system early warning event processing system, comprising: preprocessing module, for receiving the real-time early warning data of multiple early warning events and pre-processing, obtain the real-time early warning data of pre-processing;First merging module, for determining the first similar early warning event in the multiple early warning events according to the real-time early warning data of pre-processing, and the real-time early warning data of first similar early warning event is first merged, obtain merged early warning event;Dynamic upgrade processing module, for obtaining the multi-dimensional feature vector of merged early warning event;And using deep learning model, according to multi-dimensional feature vector, the initial early warning level of merged early warning event is adjusted, obtain adjusted early warning level;Output module, for output merged early warning event and its adjusted early warning level.The system of the present application can reduce repetition and redundant information, improve the efficiency, response speed and dynamic flexibility of early warning event processing.

Description

Technical Field

[0001] This invention relates to the field of oil and gas pipeline safety monitoring and information technology, and in particular to an early warning event processing system for an oil and gas pipeline fiber optic early warning system. Background Technology

[0002] In the oil and gas pipeline industry, ensuring the safe operation of pipelines is of paramount importance. Fiber optic early warning systems, due to their high sensitivity, long-distance monitoring capabilities, and resistance to electromagnetic interference, have become an important technical means for oil and gas pipeline safety monitoring. However, with the increasing complexity of actual application scenarios, higher demands are being placed on the data processing capabilities and intelligence level of early warning systems. Currently, various fiber optic early warning systems exist on the market, but most focus on signal identification and extraction, and the classification of early warning events, paying less attention to the secondary processing of early warning information. Faced with various complex situations, they rely solely on template configurations and cannot flexibly and dynamically optimize, resulting in a large number of duplicate and invalid alarms.

[0003] Disadvantages of existing technology:

[0004] Insufficient dynamism in event handling: Existing systems lack the ability to dynamically adjust event priorities based on event severity and real-time environmental factors, which may result in critical events not being handled in a timely manner.

[0005] Low efficiency of event merging: When processing a large number of early warning events, existing systems may fail to effectively merge events with similar characteristics or time correlations, resulting in a large amount of duplicate and redundant information.

[0006] The deduplication algorithm has limited accuracy: the deduplication algorithm of the existing system may not be accurate enough, resulting in duplicate items in the warning event data, which affects the accuracy and consistency of the data.

[0007] Lack of automated response: Some systems may rely on human intervention to handle early warning events and lack the ability to automatically execute pre-set response measures, which may lead to response delays. Summary of the Invention

[0008] (a) Technical problems to be solved

[0009] In view of the above-mentioned shortcomings and deficiencies of the prior art, the present invention provides an early warning event processing system for an oil and gas pipeline fiber optic early warning system, which solves the technical problem that the prior art relies solely on template configuration when facing various complex situations, and cannot flexibly and dynamically optimize, resulting in a large number of duplicate alarms and invalid alarms.

[0010] (II) Technical Solution

[0011] To achieve the above objectives, the main technical solutions adopted by the present invention include:

[0012] This invention provides an early warning event processing system for an oil and gas pipeline fiber optic early warning system, comprising:

[0013] The preprocessing module is used to receive real-time early warning data of multiple early warning events collected by the fiber optic early warning system of oil and gas pipelines and to preprocess the data to obtain preprocessed real-time early warning data.

[0014] The initial merging module is used to calculate the similarity of multiple early warning events based on the preprocessed real-time early warning data, determine the first similar early warning event among the multiple early warning events, and perform an initial merging of the real-time early warning data of the first similar early warning event to obtain one or more merged early warning events;

[0015] The dynamic upgrade / downgrade processing module is used to obtain a multi-dimensional feature vector of the merged warning events based on the real-time warning data of the merged warning events; and to use a deep learning model to adjust the initial warning level of the merged warning events according to the multi-dimensional feature vector to obtain the adjusted warning level.

[0016] The output module is used to output the merged early warning events and their adjusted early warning levels.

[0017] Optionally, the real-time early warning data includes: the event type of the early warning event, the initial early warning level, the time of occurrence, the coordinates, and the fiber optic number of the early warning event.

[0018] Optionally, the initial merging module includes:

[0019] The first vector construction unit is used to construct a first feature vector for each early warning event based on the associated preprocessed real-time early warning data; the first feature vector includes: the relative time of the early warning event, the event type, and the coordinates of the early warning event;

[0020] The first clustering analysis unit is used to calculate the similarity between any two early warning events based on the first feature vector, and to divide the multiple early warning events into multiple clusters according to the similarity of the early warning events;

[0021] The first merging unit identifies early warning events under the same cluster as first similar early warning events, and performs an initial merging of the real-time early warning data of the first similar early warning events to obtain merged early warning events.

[0022] Optionally, in the first vector construction unit, the construction of the first feature vector based on the preprocessed real-time early warning data related to the association includes: for the i-th early warning event, the constructed first feature vector is represented as: V i =[t i ,e i ,x i ,y i ]; where ti e represents the relative time of the i-th warning event. i Let x represent the event type of the i-th warning event. i Let y represent the longitude of the i-th warning event. i Represents the dimension of the i-th warning event;

[0023] In the first clustering analysis unit, the calculation of the similarity between any two early warning events based on the first feature vector includes: for the i-th early warning event and the j-th early warning event, the similarity between the two is calculated based on formula (1);

[0024] (1);

[0025] Where, d ij This represents the similarity between the i-th warning event and the j-th warning event.

[0026] Optionally, in the first merging unit, the initial merging of the real-time early warning data of the first similar early warning event includes:

[0027] For all first-similar warning events under a cluster, the average or median of the data under each entry of the associated real-time warning data is taken as the real-time warning data associated with the merged warning events.

[0028] Optionally, the multidimensional feature vector includes: time feature data, location feature data, event type feature data, environmental feature data, and historical feature data; wherein,

[0029] The time characteristic data includes: whether the occurrence time of the warning event falls within a preset sensitive time period;

[0030] The location feature data includes: topographic data, geological data, or hydrological data obtained based on GIS and the coordinates of the early warning event;

[0031] The event type characteristic data includes: the frequency of occurrence of the event type to which the warning event belongs, and / or the duration;

[0032] The environmental characteristic data includes: weather data, traffic conditions, or equipment status near the coordinates of the early warning event; the equipment status includes: pipeline flow rate and pipeline status of oil and gas pipelines;

[0033] The historical feature data includes: the occurrence pattern or trend of the warning event at the coordinates of the warning event.

[0034] Optionally, in the dynamic upgrade / downgrade processing module, the use of a deep learning model includes:

[0035] The deep learning model is: an AELDR model with adapted model parameters obtained based on a pre-training process;

[0036] The AELDR model includes:

[0037] A first LSTM model is used to predict a first result regarding the trend of the warning event based on the multidimensional feature vector;

[0038] The second LSTM model is used to predict a second result about the environmental pattern at the coordinates of the warning event based on the multidimensional feature vector.

[0039] A CNN model is used to predict a third result regarding the spatial pattern and local features of the warning event based on the multidimensional feature vector;

[0040] A random forest model is used to predict a fourth result based on the multidimensional feature vector, which shows the nonlinear relationship and complex patterns of the warning event.

[0041] An SVM model is used to predict a fifth result regarding the decision boundary of the warning event based on the high-dimensional features of the multi-dimensional feature vector.

[0042] A fusion model is used to integrate the first, second, third, fourth, and fifth results to predict the adjustment level of the warning event and make a recommended decision.

[0043] Optionally, the fusion model is a stacked meta-model.

[0044] Optionally, the pre-training process includes: acquiring historical data of multiple historical warning events, extracting historical multidimensional feature vectors of the historical data, and warning level adjustment records associated with the historical warning events, as a dataset for training;

[0045] The historical data includes: historical time characteristic data, historical location characteristic data, historical event type characteristic data, historical environmental characteristic data, and historical decision characteristic data.

[0046] Optionally, the dynamic upgrade / downgrade processing module and the output module are connected via a secondary merging module;

[0047] The secondary merging module is used to update the real-time warning data of the associated merged warning events according to the adjusted warning level output by the dynamic upgrade and downgrade processing module; and to recalculate the similarity of multiple merged warning events based on the updated real-time warning data to determine whether there is a second similar warning event among the multiple merged warning events. If so, the real-time warning data associated with the second similar warning event is merged again to obtain one or more re-merged warning events.

[0048] (III) Beneficial Effects

[0049] The early warning event processing system proposed in this invention includes a preprocessing module, an initial merging module, a dynamic escalation / de-escalation processing module, and an output module. The preprocessing module receives and preprocesses real-time early warning data from multiple early warning events collected by an oil and gas pipeline fiber optic early warning system to obtain preprocessed real-time early warning data. The initial merging module calculates the similarity of multiple early warning events based on the preprocessed real-time early warning data, determines the first similar early warning event among the multiple early warning events, and initially merges the real-time early warning data of the similar early warning events to obtain one or more merged early warning events. The dynamic escalation / de-escalation processing module obtains a multi-dimensional feature vector of the merged early warning events based on the real-time early warning data of the merged early warning events; and uses a deep learning model to adjust the initial early warning level of the merged early warning events based on the multi-dimensional feature vector to obtain an adjusted early warning level. The output module outputs the merged early warning events and their adjusted early warning levels.

[0050] The early warning event processing system provided in this embodiment is based on the aforementioned preprocessing module, initial merging module, dynamic escalation / de-escalation processing module, and output module. The initial merging module can merge early warning events with high similarity or correlation to reduce duplicate and redundant information, ensure the accuracy and consistency of event data, and improve the efficiency of early warning event processing and the system's response speed. The dynamic escalation / de-escalation module adjusts the priority of events based on multi-dimensional feature vectors reflecting the severity of the early warning event, historical data, and real-time environmental factors, generating merged early warning events and their adjusted early warning levels. This ensures that critical events receive timely and appropriate attention, better protecting the safety of oil and gas pipelines and reducing potential safety risks and losses. Attached Figure Description

[0051] Figure 1 A schematic diagram of the architecture of an early warning event processing system for an oil and gas pipeline fiber optic early warning system is provided for an embodiment.

[0052] Figure 2 This is a schematic diagram of the architecture of an early warning event processing system for another oil and gas pipeline fiber optic early warning system provided in an embodiment. Detailed Implementation

[0053] To better explain and facilitate understanding of the present invention, the present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0054] To better understand the specific implementation methods described below, some of the terms or nouns used will be explained:

[0055] Fiber Optic Early Warning: A system that uses fiber optic sensors to monitor changes in the environment around oil and gas pipelines. When an abnormal event is detected, the system issues an early warning signal.

[0056] Early warning event processing: A series of processing steps for receiving, analyzing, optimizing, and responding to early warning events detected by the fiber optic early warning system.

[0057] Dynamic escalation and escalation processing: Based on the severity of the warning event, historical data, and real-time environmental factors, the priority of the event is dynamically adjusted to ensure that critical events are handled in a timely manner.

[0058] Event merging: Merging warning events with similar characteristics or time correlation into one event to reduce duplicate and redundant information and improve processing efficiency.

[0059] Priority Adjustment: Adjust the order of event handling based on the urgency and importance of the event.

[0060] Redundant information reduction: By merging and deduplication, unnecessary information in early warning event data is reduced, improving the efficiency and accuracy of data processing.

[0061] Example 1

[0062] like Figure 1 As shown, this embodiment provides an early warning event processing system for an oil and gas pipeline fiber optic early warning system, including:

[0063] The preprocessing module is used to receive real-time early warning data of multiple early warning events collected by the fiber optic early warning system of oil and gas pipelines and to preprocess the data to obtain preprocessed real-time early warning data.

[0064] Specifically, the real-time early warning data includes: the event type of the early warning event, the initial early warning level, the time of occurrence, the coordinates, and the fiber optic number of the early warning event.

[0065] The initial merging module is used to calculate the similarity of multiple early warning events based on the preprocessed real-time early warning data, determine the first similar early warning event among the multiple early warning events, and perform an initial merging of the real-time early warning data of the first similar early warning event to obtain one or more merged early warning events.

[0066] The initial merging module can eliminate duplicate warning events, ensuring the accuracy and consistency of real-time warning data.

[0067] The dynamic upgrade / downgrade processing module is used to obtain a multi-dimensional feature vector of the merged warning events based on the real-time warning data of the merged warning events; and to use a deep learning model to adjust the initial warning level of the merged warning events according to the multi-dimensional feature vector, so as to obtain the adjusted warning level.

[0068] The dynamic escalation and escalation processing module can dynamically adjust the priority of early warning events, thereby ensuring the efficiency and accuracy of processing early warning events.

[0069] The output module is used to output the merged early warning events and their adjusted early warning levels.

[0070] The early warning event processing system provided in this embodiment, based on the aforementioned preprocessing module, initial merging module, dynamic escalation / de-escalation processing module, and output module, allows for the initial merging of early warning events with high similarity or correlation. This reduces duplicate and redundant information, ensures the accuracy and consistency of event data, and improves the efficiency of early warning event processing and the system's response speed. The dynamic escalation / de-escalation module adjusts the priority of events based on a multi-dimensional feature vector reflecting the severity of the early warning event, historical data, and real-time environmental factors. This generates merged early warning events and their adjusted early warning levels, ensuring that critical events receive timely and appropriate attention to better protect the safety of oil and gas pipelines and reduce potential safety risks and losses. The early warning event processing system provided in this embodiment enables real-time analysis, optimization, and response to oil and gas pipeline early warning events, thereby improving the intelligent management level of the early warning system.

[0071] Example 2

[0072] To better understand the technical solution provided in Embodiment 1, this embodiment provides a detailed description of each functional module involved in Embodiment 1.

[0073] In a first specific implementation, the real-time early warning data in the preprocessing module includes: the event type of the early warning event, the initial early warning level, the occurrence time, the coordinates, and the fiber optic number of the early warning event.

[0074] Specifically, the real-time early warning data can be structured in the following form:

[0075] {

[0076] "alarmLevel": 2, / / Initial warning level

[0077] "endOffset": 551.5, / / The offset of the warning event's coordinates to the end point

[0078] "endPileUUID": "4385ffaf-c577-4464-af43-26397598beff", / / Coordinate offset of the warning event, the UUID (fiber optic number) of the fiber optic cable to which the endpoint belongs.

[0079] "eventUUID":"ac7e6a4d-3d4e-4adc-9e3d-0903539e310e", / / Warning event UUID

[0080] "firstReportTime": 1724206165110, / / The first reporting time of the warning event

[0081] "latestReportTime": 1724206405110, / / Latest reporting time of the warning event

[0082] "latitude": 35.681786, / / Coordinates of the warning event

[0083] "longitude": 119.73313, / / Longitude coordinates of the warning event

[0084] "offset": 148.5, / / Coordinate offset starting point of the warning event

[0085] "riskPointDistance": 13463.5, / / Distance from the risk point

[0086] "startPileUUID": "6589e96f-aba7-4f79-a03a-01d0660dac08" / / The UUID (fiber optic number) of the starting fiber of the coordinate offset of the warning event.

[0087] }

[0088] The preprocessing module provided in this embodiment collects and preprocesses real-time early warning data from different fiber optic sensors. The preprocessing methods mainly include: data cleaning and input data standardization.

[0089] Data cleaning mainly involves removing missing values, outliers, and normalization. Assuming the initial real-time early warning dataset is D, and the cleaned real-time early warning dataset is D′, then...

[0090]

[0091] in, Let represent the i-th type of data in the initial real-time early warning dataset. Q represents the timestamp of the i-th type of data. 0.01 and Q 0.99 These represent the 1st percentile and 99th percentile of the data value of the i-th type of data in the initial real-time early warning dataset, respectively.

[0092] Data standardization mainly involves unifying the time format of real-time early warning data, the accuracy of coordinate data, the standards for early warning event types, the standards for early warning levels, and the numerical range of data.

[0093] In one specific implementation, the above data standardization includes adjusting the numerical range of the data using the following formula.

[0094]

[0095] x′ represents the standardized data value, and x represents the initial data value.

[0096] For real-time early warning data, such as time format, coordinate data accuracy, early warning event type standards, and early warning level, one-hot encoding can be performed to convert the initial data values ​​into binary codes.

[0097] In a second specific implementation, the initial merging module aims to merge early warning events based on their occurrence time, event type, and location. This embodiment employs a clustering algorithm to merge events by calculating the similarity between them. Specifically, the initial merging module includes: a first vector construction unit, a first clustering analysis unit, and a first merging unit.

[0098] The first vector construction unit is used to construct a first feature vector for each early warning event based on associated preprocessed real-time early warning data. The first feature vector includes: the relative time of the early warning event, the event type, and the coordinates of the early warning event. The relative time refers to the time from the occurrence to the end of an early warning event, i.e., the duration of the early warning event.

[0099] Specifically, for the i-th warning event, the constructed first feature vector is represented as:

[0100] V i =[t i ,e i ,x i ,y i ];

[0101] Among them, t i e represents the relative time of the i-th warning event. i Let x represent the event type of the i-th warning event. iLet y represent the longitude of the i-th warning event. i This represents the dimension of the i-th warning event.

[0102] The first clustering analysis unit is used to calculate the similarity between any two early warning events based on the first feature vector, and to divide the multiple early warning events into multiple clusters according to the similarity of the early warning events.

[0103] Specifically, the method for calculating the similarity between any two warning events based on the first feature vector is as follows: for the i-th warning event and the j-th warning event, the similarity between the two is calculated based on formula (1);

[0104] (1);

[0105] Where, d ij This represents the similarity between the i-th warning event and the j-th warning event.

[0106] The first merging unit identifies early warning events under the same cluster as first similar early warning events, and performs an initial merging of the real-time early warning data of the first similar early warning events to obtain merged early warning events.

[0107] Specifically, the initial merger described above was conducted in the following manner:

[0108] For all first-similar warning events under a cluster, the average or median of the data under each entry of the associated real-time warning data is taken as the real-time warning data associated with the merged warning events.

[0109] In the third specific implementation, for the dynamic upgrade and downgrade processing module, this embodiment uses a dynamic upgrade and downgrade algorithm based on adaptive ensemble learning (AELDR) to dynamically adjust the warning level of the warning event.

[0110] AELDR (Adaptive Ensemble Learning-based Dynamic Rating) is a dynamic rating algorithm based on adaptive ensemble learning, designed to dynamically adjust the priority of early warning events according to environmental factors, historical factors, historical feedback results, and expert database information. This algorithm improves processing efficiency and accuracy by integrating multiple machine learning models and dynamically adjusting event priorities based on real-time data and model predictions.

[0111] Specifically, the multidimensional feature vector includes: time feature data, location feature data, event type feature data, environmental feature data, and historical feature data; wherein,

[0112] The time feature data includes whether the occurrence time of the warning event falls within a preset sensitive time period. The time feature data converts the timestamp of the warning event into a relative time and extracts the time period characteristics of the event, such as daytime, nighttime, weekday, weekend, etc.

[0113] The location feature data includes: topographic data, geological data, or hydrological data obtained based on GIS and the coordinates of the early warning event. The coordinates of the location where the early warning event occurred can be converted into a relative position, such as using the distance relative to the pipeline starting point. Combined with GIS data, environmental features around the location, such as topography, geology, and hydrology, can be extracted.

[0114] The event type characteristic data includes: the frequency of occurrence of the event type to which the warning event belongs, and / or the duration.

[0115] The environmental feature data includes: weather data, traffic conditions, or equipment status near the coordinates of the early warning event; the equipment status includes: pipeline flow rate and pipeline status of oil and gas pipelines.

[0116] The historical feature data includes: the occurrence pattern or trend of warning events at the coordinates of the warning event. It can also extract features from historical feedback results such as user feedback and expert evaluations, and extract keywords and sentiment tendencies from the feedback results.

[0117] It also allows setting up an expert database, providing access to expert knowledge, experience rules, and other expert database information.

[0118] To facilitate the dynamic upgrade and downgrade processing module in handling multi-dimensional feature vectors, information gain can also be used as a partitioning criterion to divide the multi-dimensional feature vectors of the warning event into multiple data subsets.

[0119]

[0120] Where D is the dataset, A is the feature, and H(D) is the entropy of the dataset. It is a certain value of feature A |D V | The corresponding subset of data.

[0121] In multidimensional feature vectors, each type of data has relatively uniform processing characteristics, such as time, coordinates, type, and warning level. By dividing the data into subsets, we can obtain data with more clustered features, which allows the AELDR model to process each type of data subset.

[0122] In the dynamic upgrade / downgrade processing module, the deep learning model is an AELDR model with adapted model parameters obtained based on a pre-training process.

[0123] The AELDR model includes:

[0124] A first LSTM model is used to predict a first result regarding the trend of the warning event based on the multidimensional feature vector.

[0125] The second LSTM model is used to predict a second result about the environmental pattern at the coordinates of the warning event based on the multidimensional feature vector.

[0126] A CNN model is used to predict a third result regarding the spatial pattern and local features of the warning event based on the multidimensional feature vector.

[0127] A random forest model is used to predict a fourth result based on the multidimensional feature vectors, which yields nonlinear relationships and complex patterns related to the warning events.

[0128] An SVM model is used to predict a fifth result regarding the decision boundary of the warning event based on the high-dimensional features of the multidimensional feature vector.

[0129] A fusion model is used to integrate the first, second, third, fourth, and fifth results to predict the adjustment level of the warning event and make a recommended decision.

[0130] Specifically, the fusion model is a stacked meta-model with appropriate model parameters obtained through a pre-training process.

[0131] Optionally, the pre-training process includes: acquiring historical data of multiple historical warning events, extracting historical multidimensional feature vectors of the historical data, and warning level adjustment records associated with the historical warning events, as a dataset for training.

[0132] The historical data includes: historical time characteristic data, historical location characteristic data, historical event type characteristic data, historical environmental characteristic data, and historical decision characteristic data.

[0133] Based on the AELDR model with adapted model parameters obtained through the pre-training process, it can extract the temporal, spatial, level, and type features from the real-time warning data corresponding to each warning event, and make comprehensive decisions based on geographical environment, historical factors, time, and spatial information. For different task types, the learning mechanism is improved by collecting event processing feedback. Manual tasks generally require delayed processing and short-term intervention is unnecessary, while mechanical tasks require short-term intervention. If the warning event is reported at night, due to infrequent activity and a higher probability of false alarms, it can be downgraded. If the warning event occurs during busy farming seasons, which can easily cause interference, it can be downgraded. If the warning event occurs during severe weather, such as heavy rain, which can easily cause interference, it can be downgraded. If the warning event occurs near highways or waterways, it can be downgraded. If there are frequent alarms within a short period, but the environmental factors are normal, the warning can be escalated. If analysis of historical processing results reveals too many alarms requiring unnecessary processing, the event can be downgraded. For alarms of the same type, within the same range, and occurring frequently within the same time period, they can be merged and escalated. If a manual alarm occurs at location A, and after it is confirmed and cleared, X minutes later, manual alarms of different levels continue to occur within 100 meters upstream and downstream of location A. At this time, the alarms can be merged and processed, and the alarm level can be escalated.

[0134] In addition, the dynamic escalation and escalation process takes into account the above factors. If an alarm exists in a location and the first alarm is confirmed, subsequent alarms will continue to occur. Based on the historical alarm frequency and alarm confirmation results, alarms of the same type will be dynamically merged and the alarm level will be adjusted. The merging scope and merging time will be dynamically adjusted. Multiple frequent alarms will be merged and the alarm level will be downgraded or increased. If there are geographical and weather factors at the location, further judgment will be made as to whether it is severe weather or a difficult terrain such as a highway, construction site, or water flow. Noise reduction will also be performed based on these factors. Further judgment will be made based on the time of alarm occurrence, such as daytime or nighttime, spring, summer, autumn, and winter, and the alarm level will be lowered or increased accordingly. Finally, the decision will be made automatically based on the historical alarm handling results at this location and the frequency and time of the current warning event.

[0135] In the fourth specific implementation, the dynamic upgrade / downgrade processing module and the output module are connected through a secondary merging module.

[0136] The secondary merging module is used to update the real-time warning data of the associated merged warning events according to the adjusted warning level output by the dynamic upgrade and downgrade processing module; and to recalculate the similarity of multiple merged warning events based on the updated real-time warning data to determine whether there is a second similar warning event among the multiple merged warning events. If so, the real-time warning data associated with the second similar warning event is merged again to obtain one or more re-merged warning events.

[0137] The secondary merging module deduplicates duplicate push notifications and optimized duplicate alarms to avoid data corruption.

[0138] The output module outputs optimized early warning events to trigger corresponding response measures, and sends the optimized early warning events to operators to provide real-time monitoring and decision support.

[0139] This embodiment proposes a highly efficient early warning event processing system capable of identifying and merging early warning events with similar characteristics or temporal correlations, reducing duplicate and redundant information and improving data processing efficiency. Furthermore, this embodiment employs an AELDR model for dynamic escalation and escalation of early warning events, flexibly adjusting event priorities based on severity, historical data, and real-time environmental factors. This dynamism is relatively lacking in existing technologies, ensuring that critical events receive timely and appropriate attention. This provides a more efficient, accurate, and automated fiber optic early warning system to better protect the safety of oil and gas pipelines and reduce potential safety risks and losses.

[0140] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, EEPROM, optical storage, etc.) containing computer-usable program code.

[0141] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, as well as combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions.

[0142] It should be noted that any reference numerals placed between parentheses in the claims should not be construed as limiting the claims. The word "comprising" does not exclude the presence of components or steps not listed in the claims. The word "a" or "an" preceding a component does not exclude the presence of a plurality of such components. The invention can be implemented by means of hardware comprising several different components and by means of a suitably programmed computer. In claims that enumerate several means, several of these means may be embodied by the same hardware. The use of the terms first, second, third, etc., is merely for convenience of expression and does not indicate any order. These terms can be understood as part of the component names.

[0143] Furthermore, it should be noted that in the description of this specification, the terms "one embodiment," "some embodiments," "embodiment," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Furthermore, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0144] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the claims should be interpreted to include both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0145] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, then this invention should also include these modifications and variations.

Claims

1. A warning event processing system for an oil and gas pipeline fiber optic early warning system, characterized in that, include: The preprocessing module is used to receive real-time early warning data of multiple early warning events collected by the fiber optic early warning system of oil and gas pipelines and to preprocess the data to obtain preprocessed real-time early warning data. The initial merging module is used to calculate the similarity of multiple early warning events based on the preprocessed real-time early warning data, determine the first similar early warning event among the multiple early warning events, and perform an initial merging of the real-time early warning data of the first similar early warning event to obtain one or more merged early warning events. The dynamic upgrade / downgrade processing module is used to obtain the multi-dimensional feature vector of the merged early warning events based on the real-time early warning data of the merged early warning events. Then, using a deep learning model, the initial warning level of the merged warning events is adjusted based on multidimensional feature vectors to obtain the adjusted warning level; The output module is used to output the merged early warning events and their adjusted early warning levels; The initial merging module includes: The first vector construction unit is used to construct a first feature vector for each early warning event based on the associated preprocessed real-time early warning data; the first feature vector includes: the relative time of the early warning event, the event type, and the coordinates of the early warning event; The first clustering analysis unit is used to calculate the similarity between any two early warning events based on the first feature vector, and to divide the multiple early warning events into multiple clusters according to the similarity of the early warning events; The first merging unit identifies early warning events under the same cluster as first similar early warning events, and performs an initial merging of the real-time early warning data of the first similar early warning events to obtain merged early warning events.

2. The early warning event processing system according to claim 1, characterized in that, The real-time early warning data includes: the event type of the early warning event, the initial early warning level, the time of occurrence, the coordinates, and the fiber optic number of the early warning event.

3. The early warning event processing system according to claim 1, characterized in that, In the first vector construction unit, the construction of the first feature vector based on the preprocessed real-time early warning data related to the association includes: for the i-th early warning event, the constructed first feature vector is represented as: V i =[t i ,e i ,x i ,y i ]; where t i e represents the relative time of the i-th warning event. i Let x represent the event type of the i-th warning event. i Let y represent the longitude of the i-th warning event. i Represents the dimension of the i-th warning event; In the first clustering analysis unit, the calculation of the similarity between any two early warning events based on the first feature vector includes: for the i-th early warning event and the j-th early warning event, the similarity between the two is calculated based on formula (1); （1）； Where, d ij This represents the similarity between the i-th warning event and the j-th warning event.

4. The early warning event processing system according to claim 1, characterized in that, In the first merging unit, the initial merging of real-time early warning data for the first similar early warning event includes: For all first-similar warning events under a cluster, the average or median of the data under each entry of the associated real-time warning data is taken as the real-time warning data associated with the merged warning events.

5. The early warning event processing system according to claim 1, characterized in that, The multidimensional feature vector includes: time feature data, location feature data, event type feature data, environmental feature data, and historical feature data; wherein, The time characteristic data includes: whether the occurrence time of the warning event falls within a preset sensitive time period; The location feature data includes: topographic data, geological data, or hydrological data obtained based on GIS and the coordinates of the early warning event; The event type characteristic data includes: the frequency of occurrence of the event type to which the warning event belongs, and / or the duration; The environmental characteristic data includes: weather data, traffic conditions, or equipment status near the coordinates of the early warning event; the equipment status includes: pipeline flow rate and pipeline status of oil and gas pipelines; The historical feature data includes: the occurrence pattern or trend of the warning event at the coordinates of the warning event.

6. The early warning event processing system according to claim 5, characterized in that, In the dynamic upgrade / downgrade processing module, the use of a deep learning model includes: The deep learning model is: an AELDR model with adapted model parameters obtained based on a pre-training process; The AELDR model includes: A first LSTM model is used to predict a first result regarding the trend of the warning event based on the multidimensional feature vector; The second LSTM model is used to predict a second result about the environmental pattern at the coordinates of the warning event based on the multidimensional feature vector. A CNN model is used to predict a third result regarding the spatial pattern and local features of the warning event based on the multidimensional feature vector; A random forest model is used to predict a fourth result based on the multidimensional feature vector, which shows the nonlinear relationship and complex patterns of the warning event. An SVM model is used to predict a fifth result regarding the decision boundary of the warning event based on the high-dimensional features of the multi-dimensional feature vector. A fusion model is used to integrate the first, second, third, fourth, and fifth results to predict the adjustment level of the warning event and make a recommended decision.

7. The early warning event processing system according to claim 6, characterized in that, The fusion model is a stacked meta-model.

8. The early warning event processing system according to claim 6, characterized in that, The pre-training process includes: acquiring historical data of multiple historical warning events, extracting historical multidimensional feature vectors of the historical data, and warning level adjustment records associated with the historical warning events, as a dataset for training; The historical data includes: historical time characteristic data, historical location characteristic data, historical event type characteristic data, historical environmental characteristic data, and historical decision characteristic data.

9. The early warning event processing system according to claim 1, characterized in that, The dynamic upgrade / downgrade processing module and the output module are connected through a secondary merging module; The secondary merging module is used to update the real-time warning data of the associated merged warning events based on the adjusted warning level output by the dynamic upgrade and downgrade processing module. Based on the updated real-time early warning data, the similarity of multiple merged early warning events is recalculated to determine whether there is a second similar early warning event among the multiple merged early warning events. If so, the real-time early warning data associated with the second similar early warning event is merged again to obtain one or more re-merged early warning events.

Images