Cloud computing-based distributed full-scene intelligent operation and maintenance dynamic evaluation method

Abstract

This invention discloses a distributed, full-scenario intelligent operation and maintenance dynamic evaluation method based on cloud computing, belonging to the field of cloud computing and distributed system operation and maintenance technology. It includes the following six steps: S101, deploying sensors at monitoring nodes of the main pipeline structure and auxiliary facilities; S102, deploying edge gateways in partitions based on the monitoring nodes; S103, the cloud computing platform assigning key features to multiple preset evaluation dimensions; S104, the cloud computing platform obtaining preset baseline early warning thresholds; S105, based on risk levels and abnormal events; S106, the cloud computing platform generating early warning information based on the operation and maintenance health evaluation results. This invention, through edge-side preprocessing and anomaly detection, cloud-side multi-dimensional dynamic weight adaptation, and spatial topology propagation modeling, achieves dynamic early warning threshold linkage correction and multi-dimensional cross-quantitative evaluation, improving the spatial accuracy and risk prevention and control capabilities of distributed pipeline network operation and maintenance evaluation.

Description

Technical Field

[0001] This invention belongs to the field of cloud computing and distributed system operation and maintenance technology. Specifically, it is a dynamic evaluation method for distributed full-scenario intelligent operation and maintenance based on cloud computing. Background Technology

[0002] With the continuous expansion of distributed infrastructure such as urban underground integrated pipe corridors and heating pipe networks, the number of front-end sensing nodes has increased dramatically. Massive amounts of monitoring data converge on the central platform, putting severe pressure on network transmission and remote computing resources. The lack of effective local filtering and real-time identification mechanisms at the near end makes it difficult for the system's response time to meet the needs of safe operation and maintenance. Existing methods generally use fixed ratios to simply weight and integrate the system's operating conditions, surrounding factors, and facility performance. This fails to adaptively adjust coefficient allocation based on the fluctuation trends and abrupt changes of each monitoring indicator over time, resulting in judgments lagging behind actual situational evolution and making it difficult to accurately capture critical safety transition points. Furthermore, conventional alarm thresholds are usually set as uniform fixed values ​​across the entire network or static values ​​at a single point, failing to fully consider the constraints of the physical properties of pipe segments in the network topology on the anomaly propagation pattern. This lack of ability to coordinate corrections for related areas easily leads to missed and false alarms along the propagation path. Simultaneously, existing technologies are mostly limited to independent analysis of single spatial levels or single monitoring indicators, lacking effective means to quantify spatial distribution and multi-indicator layer-by-layer coupling, and thus failing to generate comprehensive judgment results reflecting the overall situation. Therefore, how to construct a cloud-based distributed full-scenario intelligent operation and maintenance dynamic evaluation method to achieve efficient preprocessing at the near end, adaptive allocation at the far end, spatial propagation modeling based on physical attributes, and hierarchical and domain-coupled quantitative evaluation has become a key technical problem that urgently needs to be solved in this field. Summary of the Invention

[0003] The purpose of this invention is to provide a cloud-based distributed, full-scenario intelligent operation and maintenance dynamic evaluation method to solve the above-mentioned technical problems. The technical solution adopted by this invention is as follows: The cloud-based distributed, full-scenario intelligent operation and maintenance dynamic evaluation method includes the following steps: S101, deploy various types of sensors at the monitoring nodes of the main structure and ancillary facilities of the pipeline network to synchronously collect raw data, including structural status parameters, environmental parameters and safety status data; S102 deploys edge gateways in partitions based on monitoring nodes, filters the raw data, then parses and extracts features to obtain key features, and obtains abnormal events based on the key features, and uploads the key features and abnormal events to the cloud computing platform. S103, the cloud computing platform assigns key features to multiple preset evaluation dimensions, uses preset models in parallel to analyze the evaluation dimensions, and dynamically adapts the weights of the evaluation dimensions according to the dynamic changes of key features in the time series to generate risk quantification indicators. S104, the cloud computing platform obtains the preset baseline warning threshold, corrects the baseline warning threshold based on key features and abnormal events to obtain the dynamic warning threshold, compares the risk quantification indicators with the dynamic warning threshold, and classifies the risk level according to the degree of exceedance of the risk quantification indicators and the number of nodes affected by the abnormal event. S105, based on risk level and abnormal events, quantitative scores are given according to the set spatial hierarchy and assessment dimensions to generate operation and maintenance health assessment results; S106, the cloud computing platform generates early warning information based on the operation and maintenance health assessment results and pushes the early warning information to the operation and maintenance terminal.

[0004] The above solution significantly reduces the transmission load and computing pressure on the cloud computing platform by employing a distributed, full-scenario intelligent operation and maintenance dynamic evaluation process and using edge gateway partitioning deployment and local caching mechanisms. It dynamically adapts the weights of evaluation dimensions based on the dynamic changes of key features over time, enabling risk quantification indicators to reflect the comprehensive risk status of the pipeline network structure, environmental parameters, and equipment operation in real time. Based on the linkage mechanism between dynamic early warning thresholds and risk levels, it achieves precise quantification of the scope and severity of abnormal events. Finally, it generates operation and maintenance health assessment results through a cross-evaluation matrix of spatial hierarchy and evaluation dimensions, improving the proactive early warning and risk prevention capabilities of the distributed pipeline network system.

[0005] S102 includes the following steps: S201, based on the spatial distribution density of monitoring nodes and the pipeline topology, divides the monitoring nodes into multiple acquisition zones, and deploys an edge gateway in each acquisition zone; S202, The edge gateway performs format verification and invalid value removal on the raw data in the collection partition to obtain standardized data; S203, the edge gateway performs mean or extreme value calculations on standardized data based on a preset time window to obtain key features that reflect the operating status of the pipeline network. S204, the edge gateway determines whether the key feature exceeds the edge judgment threshold. If it does not exceed the threshold, the key feature is cached locally, and multiple cached key features are uploaded to the cloud computing platform when the preset reporting period is reached. S205 If the key features exceed the edge judgment threshold, the edge gateway generates an abnormal event. The abnormal event includes the identifier of the abnormal source monitoring node, and the abnormal event and key features are uploaded to the cloud computing platform.

[0006] The above solution, by deploying edge gateways in partitions at monitoring nodes, achieves localized preprocessing of raw data and real-time detection of abnormal events. The edge gateways perform mean or extreme value calculations based on a set time window to extract key features, effectively filtering out sensor noise and communication interference. By determining whether key features exceed the edge detection threshold, local instant identification of abnormal events is achieved, significantly reducing invalid data transmission. Abnormal events and their corresponding key features are uploaded to the cloud computing platform in real time, ensuring that the cloud can promptly obtain high-value abnormal information. This edge computing architecture, while ensuring data quality, reduces the network bandwidth consumption and real-time computing pressure on the cloud computing platform, improving the response speed and resource utilization efficiency of the entire distributed monitoring system.

[0007] S103 includes the following steps: S301, the cloud computing platform assigns key features to multiple preset evaluation dimensions, including pipeline structure dimension, environmental parameter dimension and equipment operation dimension; S302, the preset models include a pipeline degradation model, a spatial risk assessment model, and an equipment deterioration prediction model; the pipeline degradation model, spatial risk assessment model, and equipment deterioration prediction model respectively analyze the pipeline structure dimension, environmental parameter dimension, and equipment operation dimension to obtain structural health sub-indicators, environmental safety sub-indicators, and equipment status sub-indicators; the preset models are called and executed in parallel by the cloud computing platform; S303, the cloud computing platform obtains key features from multiple consecutive preset time windows to form a historical key feature sequence, extracts dynamic change features of each evaluation dimension based on the historical key feature sequence, uses preset weights of each evaluation dimension as initial weights, and dynamically adapts the initial weights to obtain the target weights corresponding to each evaluation dimension. S304, based on the target weights corresponding to each assessment dimension, weighted and integrated the structural health sub-indicator, environmental safety sub-indicator, and equipment status sub-indicator to obtain the risk quantification index.

[0008] The aforementioned scheme achieves simultaneous quantitative assessment of the multi-dimensional risk status of the pipeline network by assigning key features to the pipeline structure dimension, environmental parameter dimension, and equipment operation dimension, and analyzing them through a pre-set model. The cloud computing platform extracts dynamic change features of each assessment dimension based on historical key feature sequences, and dynamically adapts the target weights using the weights of each assessment dimension as initial weights. This allows the weight allocation to adaptively adjust with the evolution of risk over time. Finally, the structural health sub-indicator, environmental safety sub-indicator, and equipment status sub-indicator are weighted and fused using the target weights to obtain a comprehensive risk quantification index. This multi-dimensional parallel analysis and dynamic weight adaptation mechanism improves the timeliness, accuracy, and multi-dimensional collaborative assessment capability of the risk quantification results.

[0009] S303 includes: extraction of key feature time-series changes. The cloud computing platform extracts dynamic change features based on the historical key feature sequences corresponding to each evaluation dimension. The dynamic change features include data fluctuation rate, trend change intensity, and correlation deviation index. Among them, the data fluctuation rate reflects the time-domain dispersion of the historical key feature sequence, the trend change intensity reflects the time-domain jump of the historical key feature sequence, and the correlation deviation index reflects the degree of deviation of the correlation between the key features of each evaluation dimension from the historical stable state. The correlation deviation index is determined by the deviation between the real-time correlation degree and the historical stable correlation degree between the key features of each evaluation dimension and the norm quantification.

[0010] The weight adjustment benchmark value is calculated by the cloud computing platform based on the data fluctuation rate. The weight adjustment benchmark value is positively correlated with the data fluctuation rate. The weight adjustment coefficient is amplified by the cloud computing platform based on the intensity of trend change. The greater the intensity of trend change, the larger the weight adjustment coefficient is relative to the weight adjustment benchmark value. Cross-validation correction: The cloud computing platform performs cross-validation correction on the weight adjustment coefficients of each evaluation dimension based on the correlation deviation index. The correlation deviation index of each evaluation dimension is sorted from largest to smallest. The weight adjustment coefficients of the evaluation dimensions with higher ranking are reduced, while the weight adjustment coefficients of the remaining evaluation dimensions remain unchanged. The target weight fusion process involves the cloud computing platform combining the corrected weight adjustment coefficients with the initial weights of the corresponding evaluation dimensions to obtain the target weights for each evaluation dimension.

[0011] The aforementioned scheme extracts three dynamic change features—data volatility rate, trend abrupt change intensity, and correlation deviation index—to construct a complete dynamic weight adaptation chain, encompassing temporal dispersion, temporal jump intensity, and inter-dimensional correlation deviation. A baseline weight adjustment value is calculated based on the data volatility rate, ensuring that the baseline weight reflects the historical volatility level of each assessment dimension. The baseline weight adjustment value is amplified by the trend abrupt change intensity, granting higher weight adjustment coefficients to dimensions with more dramatic changes. Furthermore, the weight adjustment coefficients are cross-validated and corrected based on the correlation deviation index, reducing the weight of top-ranked abnormal correlation dimensions, effectively suppressing weight allocation bias caused by distortion due to abnormal correlations in a single dimension. Finally, the corrected weight adjustment coefficients are merged with the initial weights to calculate the target weight. This mechanism ensures that weight allocation simultaneously considers historical volatility baselines, current abrupt change trends, and inter-dimensional correlation consistency, improving the dynamic response accuracy of risk quantification indicators to complex risk situations.

[0012] Cross-validation corrections include: The correlation matrix is ​​constructed by the cloud computing platform based on the key features corresponding to each evaluation dimension in the current time window in the historical key feature sequence, calculating the real-time correlation between the key features corresponding to any two evaluation dimensions, and constructing the correlation matrix. The correlation deviation index is calculated by the cloud computing platform based on the correlation degree matrix and the historical stable correlation degree matrix. The correlation deviation index is the norm of the row vector of the corresponding evaluation dimension in the difference matrix between the current correlation degree matrix and the historical stable correlation matrix. Once the abnormal correlation dimensions are determined, the cloud computing platform sorts the correlation deviation indices of each evaluation dimension from largest to smallest, and determines the evaluation dimensions with the highest ranking as the abnormal correlation dimensions. The cloud computing platform applies a decay factor to the weight adjustment coefficient of the abnormal correlation dimension. The decay factor is negatively correlated with the correlation deviation index of the abnormal correlation dimension, while the weight adjustment coefficients of the other evaluation dimensions remain unchanged. The cloud computing platform will output the corrected weight adjustment coefficient after applying the attenuation factor.

[0013] The above scheme achieves accurate identification of deviations between evaluation dimensions from their historical stable state by constructing a correlation matrix and calculating a correlation deviation index. The cloud computing platform obtains the correlation deviation index based on the norm of the row vector corresponding to the evaluation dimension in the difference matrix between the current correlation matrix and the historical stable correlation matrix, and identifies the top-ranked evaluation dimensions as anomalous correlation dimensions. Furthermore, a decay factor is applied to the weight adjustment coefficients of these anomalous correlation dimensions, making the decay more significant for dimensions with larger correlation deviations. This cross-validation correction mechanism effectively identifies and suppresses weight allocation distortion caused by anomalous correlations between dimensions, ensuring the robustness of weight allocation for each evaluation dimension during the target weight fusion process, and improving the reliability of risk quantification indicators in complex anomalous correlation scenarios.

[0014] S104 includes the following steps: S401, the cloud computing platform obtains the spatial topology relationship of each monitoring node and establishes a node association graph. The edges of the node association graph have edge weights that represent the pipe segment connection attributes. S402, based on the distribution statistics of key features within the historical security period, generates node-level benchmark early warning thresholds for each monitoring node; S403, when an abnormal event is detected, the monitoring node where the abnormal event occurred is identified as the abnormal source node, the set of associated nodes of the abnormal source node is determined based on the node association diagram, and the node-level baseline early warning threshold of each node in the set of associated nodes is adjusted and corrected in a coordinated manner according to the propagation characteristics of the abnormality in the pipeline network, so as to obtain the dynamic early warning threshold. S404 compares the risk quantification indicators of each monitoring node with its dynamic early warning threshold to identify nodes that exceed the limit; S405 uses the combined risk quantification indicators of the out-of-limit node and the coverage ratio of abnormal events in the associated node set to classify risk levels.

[0015] The above scheme achieves deep fusion modeling of pipeline network spatial topology and historical safety status by establishing a node association graph and generating node-level baseline early warning thresholds based on the distribution statistics of key features within historical safety periods. When an abnormal event is detected, the baseline early warning thresholds of each node in the associated node set of the abnormal source node are adjusted downwards in a coordinated manner to obtain a dynamic early warning threshold, which can adaptively adjust the threshold according to the characteristics of abnormal propagation. By comparing risk quantification indicators with the dynamic early warning threshold to identify nodes exceeding limits, and comprehensively classifying risk levels by combining the magnitude of exceeding limits with the coverage ratio of abnormal events in the associated node set, the scheme achieves accurate quantitative classification of the scope and severity of abnormal propagation. This improves the sensitivity and accuracy of risk early warning, providing quantitative basis for operation and maintenance decisions that combines spatial accuracy and risk level information.

[0016] S403 includes the following steps: S501, the cloud computing platform calculates the association influence coefficient from the abnormal source node to each associated node based on the edge weight of each edge in the node association graph. The edge weight is negatively correlated with the actual length of the pipeline between nodes. S502, Based on the correlation influence coefficient, determine the threshold reduction range for each correlated node; the larger the correlation influence coefficient, the larger the threshold reduction range. S503, subtract the corresponding threshold reduction amount from the node-level baseline early warning threshold of each associated node to obtain the dynamic early warning threshold of each associated node, and the threshold reduction amount shall not exceed the node-level baseline early warning threshold.

[0017] The above scheme calculates the correlation influence coefficient from the anomaly source node to each associated node based on the edge weights of each edge in the node association graph, achieving a spatial quantitative representation of the anomaly propagation intensity in the pipeline network. The edge weights are negatively correlated with the actual pipeline length between nodes, reflecting the constraint effect of pipe segment connection attributes on anomaly propagation. A larger correlation influence coefficient corresponds to a greater reduction in the threshold of the associated node, enabling the dynamic warning threshold to accurately reflect the spatial attenuation characteristics of anomaly propagation. Finally, the dynamic warning threshold is obtained by subtracting the corresponding threshold reduction from the node-level baseline warning threshold, and the threshold reduction does not exceed the baseline warning threshold, ensuring the rationality and safety of threshold correction. This dynamic threshold linkage reduction mechanism based on spatial topology association solves the problem that fixed thresholds cannot respond to anomaly spatial propagation, enabling the warning threshold to adaptively adjust according to the anomaly propagation path and connection strength, improving the spatial targeting and timeliness of risk warnings.

[0018] The edge weights in S501 are determined through the following steps: S601, the cloud computing platform obtains the physical attribute parameters of the pipe segment connecting the two monitoring nodes; S602, based on physical property parameters, calculates the structural fragility coefficient of the pipe segment. The structural fragility coefficient is negatively correlated with the pipe diameter and positively correlated with the material aging rate. S603, based on physical property parameters, calculates the environmental erosion coefficient of the pipe section; the environmental erosion coefficient is negatively correlated with burial depth and positively correlated with service life; S604 combines the structural fragility coefficient with the environmental erosion coefficient to obtain the comprehensive fragility coefficient of the pipe section; S605 calculates edge weights based on the comprehensive vulnerability coefficient of the pipe segment and the actual length of the pipeline network between nodes. The edge weights are positively correlated with the comprehensive vulnerability coefficient of the pipe segment and negatively correlated with the actual length of the pipeline network.

[0019] The above scheme achieves a scientific quantification of the comprehensive vulnerability of pipe segments by acquiring their physical property parameters and calculating structural vulnerability coefficients and environmental erosion coefficients. The structural vulnerability coefficient and environmental erosion coefficient are fused to obtain the comprehensive vulnerability coefficient of the pipe segment. Edge weights are then calculated based on this comprehensive vulnerability coefficient and the actual length of the pipeline network between nodes, ensuring that the edge weights are positively correlated with the comprehensive vulnerability coefficient and negatively correlated with the actual length of the pipeline network. This edge weight calculation mechanism deeply integrates physical attributes such as pipe diameter, material, burial depth, and service life with the spatial topology of the pipeline network, effectively quantifying the connection strength and vulnerability of pipe segments during anomaly propagation. This provides a basis for accurate modeling of anomaly propagation paths in the node association graph and improves the spatial physical rationality of dynamic early warning threshold correction.

[0020] S105 includes quantitative scoring based on the defined spatial hierarchy and evaluation dimensions, including: The cross-evaluation matrix is ​​constructed using a cloud computing platform, which generates a cross-evaluation matrix that represents spatial levels and evaluation dimensions. The rows of the cross-evaluation matrix correspond to spatial levels, including pipeline area levels, caisson levels, and pipeline levels. The columns of the cross-evaluation matrix correspond to evaluation dimensions. Initial value assignment for the matrix: The cloud computing platform assigns initial values ​​to each element in the cross-evaluation matrix based on the risk level; the higher the risk level, the larger the initial value of the corresponding matrix element. Abnormal increase correction: When an abnormal event occurs, the cloud computing platform determines the position of the matrix element affected by the abnormal event based on the location of the abnormal source monitoring node corresponding to the abnormal event. The abnormal source monitoring node is the monitoring node where the abnormal event occurred. The initial value of the matrix element located at the matrix element position is corrected for abnormal increase. The magnitude of the abnormal increase correction is positively correlated with the severity of the abnormal event. Hierarchical normalization processing: The cloud computing platform performs hierarchical normalization processing on different row elements under the same evaluation dimension in the cross-evaluation matrix based on the number and weight of monitoring nodes in each spatial level; the more monitoring nodes a spatial level has, the higher the weight ratio of its row elements. Once the evaluation results are generated, the cloud computing platform performs a summation operation on the cross-evaluation matrix after hierarchical normalization to generate the operation and maintenance health evaluation results.

[0021] The correction for the abnormal increase includes: The cloud computing platform determines the location of matrix elements affected by the abnormal event based on the monitoring node of the abnormal source corresponding to the abnormal event; the cloud computing platform obtains the values ​​of key features at the time of the abnormal event and calculates the deviation of the values ​​from the dynamic early warning threshold; based on the deviation, the cloud computing platform corrects the initial value of the matrix element position for abnormal increase. The magnitude of the abnormal increase correction is positively correlated with the deviation. The larger the deviation, the larger the magnitude of the abnormal increase correction.

[0022] In summary, this invention achieves closed-loop management of the entire process of a distributed pipeline network system, from data acquisition, feature extraction, risk quantification, dynamic threshold correction to health assessment and early warning push, through edge gateway partition preprocessing and local anomaly detection, cloud-side multi-dimensional parallel parsing and dynamic weight adaptation, spatial topology propagation modeling based on pipeline segment physical attributes, dynamic early warning threshold linkage adjustment and correction, and cross-quantitative evaluation of spatial hierarchy and evaluation dimensions. The dynamic weight adaptation mechanism solves the problem that static weights cannot respond to changes in risk status; the dynamic threshold correction based on spatial topology association solves the problem that fixed thresholds cannot adapt to anomaly propagation; and the cross-evaluation matrix achieves global quantitative representation of multi-dimensional risks, improving the spatial accuracy and risk prevention capabilities of distributed pipeline network operation and maintenance assessment. Attached Figure Description

[0023] The present invention can be further illustrated by the non-limiting embodiments given in the accompanying drawings; Figure 1 A three-dimensional deployment and real-time monitoring diagram of a cloud-based distributed full-scenario intelligent operation and maintenance dynamic evaluation method; Figure 2 This is a schematic diagram of the micro-strain monitoring of the strain sensor of the present invention; Figure 3 This is an overall flowchart of the present invention; Figure 4 This is a flowchart of the edge gateway data processing and anomaly detection process of the present invention; Figure 5 This is a flowchart illustrating the dynamic adaptation of evaluation dimension weights in this invention. Figure 6 This is a flowchart of the dynamic early warning threshold correction and risk level classification of the present invention. Detailed Implementation

[0024] To enable those skilled in the art to better understand the present invention, the technical solution of the present invention will be further described below in conjunction with the accompanying drawings and embodiments.

[0025] This embodiment takes the application scenario of urban underground integrated pipe gallery heating pipe network as an example, and this embodiment takes the eastern section of the pipe network as the specific subject of explanation.

[0026] like Figures 1 to 6 As shown in S101, various types of sensors are deployed at monitoring nodes along the main structure and ancillary facilities of the pipeline network to synchronously collect raw data, including structural status parameters, environmental parameters, and safety status data. Specifically, a monitoring node is set at regular intervals along the jacking pipe body. Each node is equipped with strain sensors (in this embodiment, fiber Bragg grating sensors, numbered FBG01 to FBG16), laser displacement monitors, and triaxial vibration sensors. Temperature and humidity sensors, liquid level sensors, gas monitors, and manhole cover monitoring sensors are deployed inside and at the manhole opening of each caisson. Equipment operation status sensors and on-site video monitoring equipment are deployed at pipeline ancillary facilities. All sensing devices are connected through a distributed wired and wireless hybrid network, and time synchronization is achieved by BeiDou. The collected raw data is indexed by timestamps and encapsulated according to monitoring node numbers to form raw data. Structural state parameters include pipe wall stress and strain values, pipe axial and radial displacement values, and triaxial vibration acceleration values. Environmental parameters include the internal temperature, humidity, liquid level, and underground gas concentration of the caisson. Safety status data includes the open / closed status of the manhole cover, vibration signals from illegal intrusions, on-site video image frames, and equipment operating current and switch status. For example... Figure 1 As shown, the fiber Bragg grating patch strain sensor at the monitoring node at pile number D9071 of the pipe jacking project is deployed in three dimensions. The strain sensor is numbered FBG12 and collects the actual micro-strain values ​​in real time.

[0027] S102 deploys edge gateways in partitions based on monitoring nodes. It filters raw data before performing real-time analysis and feature extraction to obtain key features, and identifies abnormal events based on these features. These key features and abnormal events are then uploaded to the cloud computing platform. For example... Figure 4 As shown, S102 includes the following steps: S201, based on the spatial distribution density of monitoring nodes and the pipeline topology, divides the monitoring nodes into multiple acquisition zones, with one edge gateway deployed in each acquisition zone. Specifically, taking 120 monitoring nodes along the entire underground integrated pipe gallery heating pipeline network as an example, the geographical coordinates of each monitoring node and the pipeline topology connection relationship are obtained through a GIS system. The actual pipeline distance between adjacent monitoring nodes is calculated, and a clustering algorithm based on the actual pipeline distance is used to divide the monitoring nodes into 8 acquisition zones. An edge gateway is deployed at the geometric center of each acquisition zone, and the edge gateway establishes a communication connection with each monitoring node within the zone.

[0028] S202, the edge gateway performs format verification and invalid value removal on the raw data within the acquisition zone to obtain standardized data. Specifically, the edge gateway checks whether the frame header, frame trailer, checksum, and timestamp format of the raw data frames meet the requirements, discards data frames with incorrect formats, removes invalid values ​​due to sensor offline or communication interruption, and performs time alignment on data collected at the same time to form standardized data.

[0029] S203, the edge gateway performs mean or extreme value calculations on standardized data based on a preset time window to obtain key features reflecting the pipeline network's operating status. The edge gateway processes data in units of a preset time window; in this embodiment, the preset time window is set to 10 seconds, but the specific time can be set according to the performance of the selected equipment. It performs time-domain statistical processing on the standardized data, calculates the arithmetic mean and maximum / minimum extreme values ​​of continuous sampled values ​​within the preset time window, and obtains trend characteristic values ​​reflecting the pipeline network's operating status. These trend characteristic values ​​are then used as key features.

[0030] S204, the edge gateway determines whether the key feature exceeds the edge judgment threshold. If it does not exceed the threshold, the key feature is cached locally, and multiple cached key features are uploaded to the cloud computing platform when the preset reporting period is reached. The edge judgment threshold is set as follows: the edge gateway automatically retrieves all data records of the same key feature of the monitoring node within the most recent 7-day historical security period, sorts them by value from smallest to largest, and takes the 95th percentile as the edge judgment threshold. That is, about 95% of the historical normal data is below this threshold, and 5% of normal fluctuations may approach or touch this line. If the key feature in the current 10-second time window does not exceed the edge judgment threshold, it is determined to be a normal working condition, and the key feature enters the local cache queue until the cache queue reaches the preset reporting period. In this embodiment, the preset reporting period is 5 minutes, that is, when 30 10-second windows of key features are accumulated, the edge gateway uploads the cached key features in batches to the cloud computing platform.

[0031] S205, if the key features exceed the edge detection threshold, the edge gateway generates an abnormal event. The abnormal event includes the identifier of the abnormal source monitoring node, and uploads the abnormal event and key features to the cloud computing platform. If the key features in the current 10-second time window exceed the edge detection threshold, the edge gateway immediately generates an abnormal event. The abnormal event includes an abnormality type identifier, an abnormality occurrence timestamp, an abnormal source monitoring node number, and abnormal key feature values, and uploads the abnormal event and corresponding key features to the cloud computing platform in real time. The cloud computing platform stores the received key features and abnormal events in the historical database according to the monitoring node number and timestamp as a historical data source.

[0032] S103, the cloud computing platform assigns key features to multiple preset evaluation dimensions, uses preset models in parallel to analyze the evaluation dimensions, and dynamically adapts the weights of the evaluation dimensions according to the dynamic changes of key features over time, generating risk quantification indicators. For example... Figure 5 As shown, S103 includes the following steps:

[0033] S301, the cloud computing platform assigns key features to multiple preset evaluation dimensions, including pipeline structure dimension, environmental parameter dimension and equipment operation dimension. Specifically, the cloud computing platform receives key features uploaded by the edge gateway and assigns these key features to three evaluation dimensions according to their data sources: key features derived from structural state parameters are classified into the pipeline structure dimension, including the average pipe wall stress calculated from the pipe wall stress and strain values ​​collected by strain sensors, the extreme values ​​of pipe displacement calculated from the axial and radial displacement values ​​collected by laser displacement monitors, and the average triaxial vibration acceleration calculated from the triaxial vibration acceleration values ​​collected by triaxial vibration sensors; key features derived from environmental parameters are classified into the environmental parameter dimension, including the average caisson temperature calculated from the temperature channel of the temperature and humidity sensor, the average caisson humidity calculated from the humidity channel of the temperature and humidity sensor, the average liquid level height calculated from the liquid level sensor, and the peak value of downhole gas concentration calculated from the downhole gas concentration collected by the gas monitor; and key features derived from equipment operation-related data in the safety status data are classified into the equipment operation dimension. Key features derived from the safety status data, such as the opening and closing status of manhole covers, vibration signals from illegal intrusions, and on-site video image frames, directly participate in the determination of abnormal events. When such key features exceed the corresponding edge determination threshold, the edge gateway generates an abnormal event and reports it in real time.

[0034] S302, the preset models include a pipeline degradation model, a spatial risk assessment model, and an equipment deterioration prediction model. These models analyze the pipeline structure dimension, environmental parameter dimension, and equipment operation dimension, respectively, to obtain structural health sub-indicators, environmental safety sub-indicators, and equipment status sub-indicators. The preset models are executed in parallel by the cloud computing platform. For example... Figure 2 As shown, the normalized microstrain index distribution of fiber Bragg grating strain sensors at each monitoring node along the pipe body is illustrated. The normalized microstrain index ranges from 0 to 100 (dimensionless), where 0 corresponds to the stress-free baseline state of the pipe wall, and 100 corresponds to the maximum allowable microstrain value. The cloud computing platform concurrently calls three preset models for analysis, and the outputs of each model are as follows:

[0035] The pipeline degradation model calculates the structural health sub-indicator using the following formula: ; Where σ is the mean stress of the pipe wall, in MPa, representing the average stress level of the pipe wall, which is calculated from the actual micro-strain value collected by the fiber Bragg grating strain sensor at the corresponding monitoring node; δ is the extreme value of the pipe displacement, in mm, representing the maximum displacement of the pipe. This is the average triaxial vibration acceleration, expressed in m / s², representing the average vibration intensity of the pipeline. This represents the material's fatigue limit, expressed in MPa. These are displacement safety limits, in mm. Vibration safety limits, in m / s²; This is the stress weighting coefficient. For displacement weighting coefficients, This is the vibration weighting coefficient, and .

[0036] The cloud computing platform extracts the material parameters of the pipe section from the pipeline asset management system that stores the pipe section material parameters and equipment files, and determines them according to the material mechanical property manual corresponding to the material type. (200MPa); The maximum allowable displacement of this pipe section was extracted from the pipeline network completion acceptance data as... (50mm), extract the maximum vibration acceleration allowed by the design as (10 m / s²). The cloud computing platform retrieved abnormal event records from the historical database of this pipe section and counted the frequency of each key feature independently triggering abnormal events: the average pipe wall stress triggered anomalies 15 times, the extreme value of pipe displacement triggered anomalies 9 times, and the average triaxial vibration acceleration triggered anomalies 6 times; the total trigger frequency was 30 times. The trigger frequency of each key feature was normalized by dividing by the total frequency, resulting in... =15 / 30=0.5, =9 / 30=0.3, =6 / 30=0.2.

[0037] Taking caisson J3 as an example: σ = 120 MPa, δ = 15 mm, a = 2.5 m / s². Substitute these values ​​into the calculation: ; The higher the value of the structural health sub-index, the greater the degree of structural degradation of the pipe section. The cloud computing platform outputs a structural health sub-index of 0.44.

[0038] The space risk assessment model calculates environmental safety sub-indicators using the following formula: ; In the formula, T is the average temperature of the caisson, in °C, representing the average temperature of the caisson; H is the average humidity of the caisson, in %, representing the average relative humidity of the caisson; L is the average liquid level height, in m, representing the average liquid level height of the caisson; and C is the peak concentration of the downhole gas, in ppm, representing the highest concentration of harmful gas during the monitoring period. For reference temperature, For reference humidity, For reference liquid level, Reference gas concentration; Temperature weighting coefficient, Humidity weighting coefficient This is the liquid level weighting coefficient. This is the gas concentration weighting coefficient, and .

[0039] The cloud computing platform extracts from the design specifications of the caisson (40℃) (80%) (4.0m) (100ppm); The cloud computing platform retrieved abnormal event records from the historical database of the caisson and counted the frequency of each key feature independently triggering the abnormal event: the average caisson temperature triggered an anomaly 8 times, the average caisson humidity triggered an anomaly 8 times, the average liquid level height triggered an anomaly 8 times, and the peak downhole gas concentration triggered an anomaly 8 times; the total trigger frequency was 32 times. The trigger frequency of each key feature was normalized by dividing by the total frequency, resulting in... =8 / 32=0.25, =8 / 32=0.25, =8 / 32=0.25, =8 / 32=0.25.

[0040] Taking caisson J3 as an example: T=35℃, H=75%, L=2.5m, C=50ppm. Substitute into the calculation: ;

[0041] The higher the value of the environmental safety sub-indicator, the higher the environmental safety risk of the well. The cloud computing platform outputs the environmental safety sub-indicator as 0.734375.

[0042] The equipment degradation prediction model calculates equipment condition sub-indicators using the following formula: ; In the formula, I is the average operating current of the equipment, in A, representing the average operating current of the equipment; S is the average switching state, a dimensionless ratio, representing the average consistency of the switching states of the equipment. The original switching state data are discrete values ​​of 0 / 1, where 0 represents open and 1 represents closed. After averaging, the average switching state between 0 and 1 is obtained. Rated current, in amperes (A); The historical steady-state switching baseline is represented by a dimensionless ratio. This is the current load weighting factor. For the switch deviation weighting coefficient, and .

[0043] The cloud computing platform extracts parameters from equipment nameplates and equipment files in the pipeline asset management system. (60A); retrieve the average switch status statistics of the device during the most recent 30 days of normal operation (i.e., the key features have not exceeded the edge detection threshold for 30 consecutive days) as the basis for determination. (0.80); The cloud computing platform retrieved abnormal event records from the device's historical database and counted the frequency of each key feature independently triggering abnormal events: the average operating current of the device triggered abnormal events 14 times, and the average switching status triggered abnormal events 6 times; the total trigger frequency was 20 times. The trigger frequency of each key feature was normalized by dividing by the total frequency, resulting in... =14 / 20=0.7, =6 / 20=0.3.

[0044] Taking the auxiliary equipment of caisson J3 as an example: I=45A, S=0.85. Substitute into the calculation: ; The higher the value of the device status sub-index, the greater the degree of device degradation. The cloud computing platform outputs a device status sub-index of 0.54375.

[0045] The three models are executed in parallel by the cloud computing platform. The above-mentioned structural health sub-indicator 0.44, environmental safety sub-indicator 0.734375, and equipment status sub-indicator 0.54375 are simultaneously output to S304 for weighted fusion.

[0046] S303, the cloud computing platform acquires key features from multiple consecutive preset time windows to form a historical key feature sequence. Based on the historical key feature sequence, it extracts dynamic change features for each evaluation dimension, uses preset weights for each evaluation dimension as initial weights, and dynamically adapts the initial weights to obtain the target weights corresponding to each evaluation dimension. Further, S303 includes: Key feature temporal change extraction: The cloud computing platform extracts dynamic change features based on the historical key feature sequences corresponding to each evaluation dimension. These dynamic change features include data fluctuation rate, trend abrupt change intensity, and correlation deviation index. Specifically, the data fluctuation rate reflects the temporal dispersion of the historical key feature sequence; the trend abrupt change intensity reflects the temporal jump of the historical key feature sequence; and the correlation deviation index reflects the degree to which the correlation between key features of each evaluation dimension deviates from its historical stable state. The correlation deviation index is quantified by the deviation between the real-time correlation and the historical stable correlation between key features of each evaluation dimension. Specifically, taking the J3 well as an example, the cloud computing platform constructs a historical key feature sequence based on the key features of the most recent 30 consecutive preset time windows. The standard deviation of the mean wall stress for 30 pipe network structural dimensions (fluctuating around the mean wall stress of 120 MPa at the monitoring node in S302, with a maximum value of 132 MPa and a minimum value of 108 MPa) was calculated, with a standard deviation ≈ 8.98 MPa. The data fluctuation rate after rounding was 9.0 MPa. The absolute value of the first-order difference of the mean wall stress for adjacent preset time windows was calculated, with the largest difference of 6.0 MPa between window 15 (126 MPa) and window 16 (120 MPa), which was used as the trend abrupt change intensity. The standard deviation of the mean well temperature for 30 environmental parameters (fluctuating around 35℃, with a maximum value of 41℃ and a minimum value of 29℃) was calculated, with a standard deviation = 4.0℃ and a data fluctuation rate of 4.0℃. The absolute value of the first-order difference of the mean well temperature for adjacent preset time windows was calculated, with the largest difference of 4.0℃ between window 8 (37℃) and window 9 (33℃), which was used as the trend abrupt change intensity. In this embodiment, the average temperature of the caisson is selected as a representative key feature of the environmental parameter dimension to extract dynamic change features. The average humidity of the caisson is used as another key feature of the environmental parameter dimension, participating in the calculation of sub-indicators of the spatial risk assessment model, but not as a representative feature of this dimension for dynamic change feature extraction. The standard deviation of the average operating current of 30 devices (fluctuating around 45A, with a maximum value of 51A and a minimum value of 39A) is calculated, with a standard deviation of 3.0A and a data fluctuation rate of 3.0A. The absolute value of the first-order difference of the average operating current of devices in adjacent preset time windows is calculated. The largest difference, 3.0A, is found between the 22nd window (48A) and the 23rd window (45A), and is used as the intensity of trend change.

[0047] Simultaneously, the cloud computing platform extracts representative key features from three evaluation dimensions across the aforementioned 30 windows: the average pipe wall stress in the pipeline structure dimension, the average well temperature in the environmental parameter dimension, and the average equipment operating current in the equipment operation dimension. Pearson correlation coefficients are calculated pairwise to obtain the current correlation matrix: structure-environment = 0.40, structure-equipment = 0.55, and environment-equipment = 0.45. The Pearson correlation coefficient matrix for the past 30 days without anomalies is retrieved from the historical database as the historical stable correlation matrix: structure-environment = 0.70, structure-equipment = 0.60, and environment-equipment = 0.50. The calculations of the Pearson correlation coefficients, difference matrix, and Euclidean norm are all existing technologies and will not be elaborated further. The difference matrix is ​​obtained by subtracting corresponding elements from the current correlation matrix and the historical stable correlation matrix. The Euclidean norm of each of the three row vectors is calculated and rounded to two decimal places to obtain the correlation deviation index: 0.30 for the pipeline structure dimension, 0.30 for the environmental parameter dimension, and 0.07 for the equipment operation dimension, all rounded to two decimal places.

[0048] The weight adjustment benchmark value is calculated by the cloud computing platform based on the data fluctuation rate. The weight adjustment benchmark value is positively correlated with the data fluctuation rate. Specifically, the input is the data fluctuation rate of the three evaluation dimensions extracted earlier: pipeline structure dimension (9.0 MPa), environmental parameters dimension (4.0℃), and equipment operation dimension (3.0A). The cloud computing platform retrieves historical abnormal event records for the caisson and counts the frequency of abnormal events triggered independently by key features of each evaluation dimension: pipeline structure dimension triggered 45 times, environmental parameters dimension triggered 25 times, and equipment operation dimension triggered 30 times, for a total trigger frequency of 100 times. The trigger frequency of each dimension is divided by the total frequency and normalized to obtain the risk contribution ratio: pipeline structure dimension 45%, environmental parameters dimension 25%, and equipment operation dimension 30%. The aforementioned risk contribution ratio is positively correlated with the data fluctuation rate (45% for 9.0MPa, 25% for 4.0℃, and 30% for 3.0A). This risk contribution ratio is directly mapped to the weight adjustment benchmark value, and the output is: 0.45 for pipeline structure dimension, 0.25 for environmental parameter dimension, and 0.30 for equipment operation dimension.

[0049] The weight adjustment coefficient is amplified. The cloud computing platform amplifies the weight adjustment benchmark value based on the intensity of trend mutation to obtain the weight adjustment coefficient. The greater the intensity of trend mutation, the larger the weight adjustment coefficient is relative to the weight adjustment benchmark value. Specifically, the inputs are: weight adjustment benchmark values ​​(pipeline structure dimension 0.45, environmental parameter dimension 0.25, equipment operation dimension 0.30), and the current trend mutation intensity extracted above (pipeline structure dimension 6.0MPa, environmental parameter dimension 4.0℃, equipment operation dimension 3.0A). The cloud computing platform retrieves the statistical average of the trend mutation intensity of this evaluation dimension during past anomaly-free periods from the historical database as the amplification benchmark: pipeline structure dimension 2.0MPa, environmental parameter dimension 1.33℃ is the statistical average of the trend mutation intensity during the historical 30-day anomaly-free period, and equipment operation dimension 1.0A. The amplification factor is calculated: the current trend mutation intensity is divided by the amplification benchmark, structure 6.0 / 2.0=3.0, environment 4.0 / 1.33≈3.0, equipment 3.0 / 1.0=3.0. Multiply the weight adjustment benchmark value by the amplification factor to obtain the weight adjustment coefficient, and the output is: pipeline structure dimension 0.45×3.0=1.35, environmental parameter dimension 0.25×3.0=0.75, equipment operation dimension 0.30×3.0=0.90.

[0050] Cross-validation correction: The cloud computing platform performs cross-validation correction on the weight adjustment coefficients of each evaluation dimension based on the correlation deviation index. The correlation deviation indices of each evaluation dimension are sorted from largest to smallest. The weight adjustment coefficients of the top-ranked evaluation dimensions are reduced, while the weight adjustment coefficients of the remaining evaluation dimensions remain unchanged. Specifically, cross-validation correction includes:

[0051] The correlation matrix is ​​constructed by the cloud computing platform based on the key features corresponding to each evaluation dimension within the current time window from the historical key feature sequence. It calculates the real-time correlation between any two key features corresponding to the evaluation dimensions and constructs a correlation matrix. Specifically, the cloud computing platform extracts representative key features from the historical key feature sequence for the three evaluation dimensions within the current time window: the average pipe wall stress (pipeline structure dimension), the average well temperature (environmental parameters dimension), and the average equipment operating current (equipment operation dimension). It calculates the Pearson correlation coefficient between any two dimensions and constructs a 3x3 symmetric correlation matrix, where the diagonal elements are all one, and the off-diagonal elements represent the real-time correlation coefficients between the two dimensions.

[0052] The correlation deviation index is calculated by the cloud computing platform based on the correlation degree matrix and the historical stable correlation degree matrix. The correlation deviation index is the norm of the row vector corresponding to the evaluation dimension in the difference matrix between the current correlation degree matrix and the historical stable correlation degree matrix. The cloud computing platform retrieves the correlation degree matrix from the past 30 days without anomalies as the historical stable correlation degree matrix. The corresponding elements of the current correlation degree matrix and the historical stable correlation matrix are subtracted to obtain the difference matrix. The Euclidean norm of the three row vectors of the difference matrix is ​​then calculated as the correlation deviation index for each of the three evaluation dimensions. Based on the calculation results above, the correlation deviation indices are: 0.30 for the pipeline structure dimension, 0.30 for the environmental parameter dimension, and 0.07 for the equipment operation dimension.

[0053] The abnormal correlation dimensions were determined by the cloud computing platform, which sorted the correlation deviation indices of each evaluation dimension from largest to smallest, identifying the top-ranked dimensions as the abnormal correlation dimensions. Specifically, the correlation deviation indices were sorted as follows: pipeline structure dimension 0.30, environmental parameters dimension 0.30, and equipment operation dimension 0.07.

[0054] A decay factor is applied. The cloud computing platform applies a decay factor to the weight adjustment coefficients of the abnormal correlation dimensions. The decay factor is negatively correlated with the correlation deviation index of the abnormal correlation dimensions, while the weight adjustment coefficients of the other evaluation dimensions remain unchanged. Specifically, the inputs are: weight adjustment coefficients (pipeline structure dimension 1.35, environmental parameter dimension 0.75, equipment operation dimension 0.90) and correlation deviation indexes (pipeline structure dimension 0.30, environmental parameter dimension 0.30, equipment operation dimension 0.07). The pipeline structure dimension and the environmental parameter dimension, which are ranked first and tied for first, are identified as abnormal correlation dimensions, and a decay factor is applied to them; the equipment operation dimension is used as one of the other evaluation dimensions, and its weight adjustment coefficient remains unchanged at 0.90. The attenuation factor is calculated as follows: Attenuation factor = 1 − k × Correlation deviation index, where k is the attenuation coefficient, calibrated through regression analysis of the correlation deviation index and the correction magnitude of the weight adjustment coefficient in historical abnormal events of the pipeline network. In this embodiment, k = 0.05 is used. For the pipeline network structure dimension: Attenuation factor = 1 − 0.05 × 0.30 = 0.985, and the corrected weight adjustment coefficient = 1.35 × 0.985 = 1.32975; For the environmental parameter dimension: Attenuation factor = 1 − 0.05 × 0.30 = 0.985, and the corrected weight adjustment coefficient = 0.75 × 0.985 = 0.73875. After applying the attenuation factor, the weight adjustment coefficients for the pipeline network structure dimension and the environmental parameter dimension decrease, while the equipment operation dimension remains at 0.9.

[0055] The cloud computing platform outputs the corrected weight adjustment coefficients after applying a decay factor. After cross-validation correction, the corrected weight adjustment coefficients are: 1.32975 for pipeline structure dimension, 0.73875 for environmental parameter dimension, and 0.90 for equipment operation dimension.

[0056] The target weight fusion process involves the cloud computing platform fusing the corrected weight adjustment coefficients with the initial weights of the corresponding evaluation dimensions to obtain the target weights for each evaluation dimension. The fusion calculation couples the initial weights with the dynamic adjustment coefficients, ensuring the target weights simultaneously reflect both the baseline allocation and the adjustment magnitude. Specifically, this embodiment uses linear weighted normalization for fusion; other embodiments may use nonlinear mapping or fuzzy weighting. Using an initial weight of 1 / 3 for each evaluation dimension as the baseline, the corrected weight adjustment coefficients are used as the adjustment magnitude, and the two are multiplied to obtain the preliminary target weights. These preliminary target weights are then normalized so that the sum of the three is 1.0. The inputs are: corrected weight adjustment coefficients (pipeline structure dimension 1.32975, environmental parameter dimension 0.73875, equipment operation dimension 0.90), and initial weights of 1 / 3. The processing involves multiplying these to obtain the preliminary target weights: pipeline structure dimension 0.44325, environmental parameter dimension 0.24625, and equipment operation dimension 0.30. The sum of the three is 0.9895, which is normalized and rounded to two decimal places: Pipeline structure dimension 0.45, Environmental parameter dimension 0.25, Equipment operation dimension 0.30. Output: Target weights for pipeline structure dimension 0.45, Environmental parameter dimension 0.25, and Equipment operation dimension 0.30.

[0057] S304, based on the target weights corresponding to each assessment dimension, weighted fusion of the structural health sub-indicator, environmental safety sub-indicator, and equipment status sub-indicator yields a risk quantification index. Specifically, the cloud computing platform weights and fuses the structural health sub-indicator (0.44), environmental safety sub-indicator (0.734375), and equipment status sub-indicator (0.54375) according to their target weights: the structural health sub-indicator (0.44) is multiplied by the target weight of the pipeline structure dimension (0.45), the environmental safety sub-indicator (0.734375) is multiplied by the target weight of the environmental parameter dimension (0.25), and the equipment status sub-indicator (0.54375) is multiplied by the target weight of the equipment operation dimension (0.30). The sum of these three values ​​yields the risk quantification index. Taking caisson J3 as an example, the calculation is as follows: Risk quantification index = 0.44 × 0.45 + 0.734375 × 0.25 + 0.54375 × 0.30 = 0.198 + 0.18359375 + 0.163125 = 0.54471875. The cloud computing platform outputs this risk quantification index of 0.5447 (rounded to four decimal places) and compares it with the dynamic early warning threshold in S104.

[0058] S104, the cloud computing platform obtains a preset baseline warning threshold, corrects the baseline warning threshold based on key features and abnormal events to obtain a dynamic warning threshold, compares the risk quantification indicators with the dynamic warning threshold, and classifies the risk level according to the degree of exceedance of the risk quantification indicators and the number of nodes affected by the abnormal event. For example... Figure 6 As shown, S104 includes the following steps:

[0059] S401, the cloud computing platform acquires the spatial topology relationships of each monitoring node and establishes a node association graph. The edges of the node association graph have edge weights that characterize the pipe segment connection attributes. Specifically, the cloud computing platform imports spatial topology relationship data of the monitoring nodes from the pipeline GIS management system, including the latitude and longitude coordinates, node number, node type of each monitoring node, and the pipe segment connection relationships in the pipeline CAD drawings. Based on the above information, the cloud computing platform establishes a node association graph: with monitoring nodes as vertices and pipe segments connecting two monitoring nodes as edges, forming an undirected graph structure. The node association graph is stored using an adjacency list, facilitating quick lookup of the direct neighbor nodes of any monitoring node. Each edge of the node association graph has an edge weight that characterizes the pipe segment connection attributes. This edge weight is a numerical attribute field attached to the edge, used to characterize the connection strength or vulnerability of the pipe segment during anomaly propagation. The initial value of this field is empty and will be calculated and filled in subsequent steps based on the physical attribute parameters of the pipe segment. For example, the cloud computing platform imports spatial topology data of 42 monitoring nodes. Taking caisson J3 as an example, its coordinates are 118.796° east longitude and 32.058° north latitude. The node type is inspection well. Its adjacency table record in the node association graph shows that its directly connected neighbor nodes are caisson J2, caisson J4 and caisson J5, with corresponding actual pipe lengths of 300 meters, 200 meters and 400 meters, respectively. All three edges have edge weight attribute fields.

[0060] S402. Based on the statistical distribution of key features within the historical safety period, a node-level baseline early warning threshold is generated for each monitoring node. Specifically, the cloud computing platform first determines the historical safety period. This historical safety period is defined as the continuous period within which the key features of the monitoring node have not exceeded the edge judgment threshold within the most recent seven consecutive days. Taking caissons J1 to J6 as an example, the cloud computing platform retrieves the key features of each monitoring node within this historical safety period from the database. Each monitoring node generates 8640 key feature values ​​per day, totaling 60480 values ​​over seven days. The cloud computing platform performs statistical distribution analysis on the 60480 historical key feature values ​​of the average pipe wall stress for each monitoring node: sorted from smallest to largest, the 95th percentile is taken as the node-level baseline early warning threshold. The statistical meaning of this node-level baseline early warning threshold is that under normal operating conditions, 95% of the historical key feature values ​​are below this line, and only 5% of normal fluctuations may approach this threshold. Similarly, statistical distribution analysis is performed on other key features such as the average caisson temperature and average equipment operating current of each monitoring node to generate corresponding node-level baseline early warning thresholds.

[0061] S403, when an abnormal event is detected, the monitoring node where the abnormal event occurred is identified as the abnormal source node. Based on the node association graph, a set of associated nodes of the abnormal source node is determined. According to the propagation characteristics of the abnormality in the pipeline network, the node-level baseline early warning threshold of each node in the associated node set is adjusted downwards in a coordinated manner to obtain a dynamic early warning threshold. S403 includes the following steps:

[0062] S501, the cloud computing platform calculates the association influence coefficient from the anomaly source node to each associated node based on the edge weights of each edge in the node association graph. The edge weights are negatively correlated with the actual length of the network between nodes. The association influence coefficient is determined by rounding the edge weights to one decimal place.

[0063] The edge weights are determined through the following steps: S601, the cloud computing platform obtains the physical attribute parameters of the pipe segment connecting two monitoring nodes. These parameters include pipe diameter, material, burial depth, and service life. For example, taking the pipe segment connecting caissons J3 and J4 as an example, the cloud computing platform extracts the physical attribute parameters of this pipe segment from the pipeline asset management system that stores the pipe segment's material parameters and equipment files: pipe diameter is DN800mm, material is Q235B carbon structural steel, burial depth is 3.5m below ground level, and service life is 12 years from the date of completion and acceptance.

[0064] S602, based on physical property parameters, calculates the structural fragility coefficient of the pipe segment. The structural fragility coefficient is negatively correlated with the pipe diameter and positively correlated with the material aging rate. The structural fragility coefficient reflects the pipe segment's resistance to failure at the structural mechanical level. It is negatively correlated with the pipe diameter (i.e., the larger the pipe diameter, the higher the structural strength) and positively correlated with the material aging rate (i.e., the more severe the aging, the more fragile the structure). The cloud computing platform, based on a pipe diameter of DN800mm and a material aging rate of 0.2 corresponding to 12 years of use of Q235B steel, combined with empirical coefficients derived from historical failure data regression calibration, calculates a structural fragility coefficient of 0.403.

[0065] S603, based on physical property parameters, calculates the environmental erosion coefficient of the pipe section. The environmental erosion coefficient is negatively correlated with burial depth and positively correlated with service life. The environmental erosion coefficient reflects the degree of corrosion damage to the pipe section from the external environment. It is negatively correlated with burial depth, meaning that the deeper the burial, the weaker the soil corrosion, and positively correlated with service life, meaning that the longer the service life, the more severe the cumulative erosion. Based on a burial depth of 3.5m and a service life of 12 years, combined with empirical coefficients calibrated from soil corrosion experimental data, the cloud computing platform calculates an environmental erosion coefficient of 0.299.

[0066] S604 combines the structural vulnerability coefficient and the environmental erosion coefficient to obtain the comprehensive vulnerability coefficient of the pipe section. This formula assigns equal weight to both types of coefficients. The cloud computing platform combines the structural vulnerability coefficient (0.403) and the environmental erosion coefficient (0.299) with equal weights to obtain a comprehensive vulnerability coefficient of 0.351 for the pipe section.

[0067] S605 calculates edge weights based on the comprehensive vulnerability coefficient of the pipe segment and the actual length of the pipeline network between nodes. The edge weights are positively correlated with the comprehensive vulnerability coefficient of the pipe segment and negatively correlated with the actual length of the pipeline network.

[0068] Edge weight calculation formula: ; in, The overall vulnerability coefficient of the pipeline segment. This represents the actual length of the pipeline (m). This is a reference length (1000m in this embodiment). The attenuation index is 1.0 in this embodiment. The normalization coefficient (0.343 in this example) is used to map the edge weights to the interval between 0 and 1.

[0069] In step S502, based on the correlation influence coefficient, the threshold reduction range for each associated node is determined; the larger the correlation influence coefficient, the larger the threshold reduction range. Specifically, the edge weights obtained in S501 are 0.602 for J4, 0.401 for J2, and 0.301 for J5. After rounding to one decimal place, the correlation influence coefficients are obtained as follows: J4 = 0.6, J2 = 0.4, and J5 = 0.3. Next, the cloud computing platform retrieves historical anomaly propagation event records for the pipeline network, calculates the ratio of the correlation influence coefficient of each associated node in historical anomaly events to its actual threshold reduction range, and uses the average value to calibrate the linear mapping coefficient as 0.5. The threshold reduction range is calculated as: Correlation Influence Coefficient × 0.5. Substituting the values ​​into the calculation: J4 threshold reduction = 0.6 × 0.5 = 0.3 (i.e., a reduction of 30%); J2 threshold reduction = 0.4 × 0.5 = 0.2 (i.e., a reduction of 20%); J5 threshold reduction = 0.3 × 0.5 = 0.15 (i.e., a reduction of 15%).

[0070] S503: Subtract the corresponding threshold reduction amount from the node-level baseline early warning threshold of each associated node to obtain the dynamic early warning threshold of each associated node. The threshold reduction amount shall not exceed the node-level baseline early warning threshold. Specifically, the inputs are: the node-level baseline early warning thresholds of each associated node generated by S402 (J4 is 80, J2 is 80, J5 is 80, these thresholds are generated by S402 by taking the 95th percentile of the historical key features of each monitoring node), and the threshold reduction amounts obtained by S502 (J4 is 0.3, J2 is 0.2, J5 is 0.15). Processing: Subtract the corresponding threshold reduction amount from the node-level baseline early warning thresholds: J4 dynamic early warning threshold = 80 − 80 × 0.3 = 56; J2 dynamic early warning threshold = 80 − 80 × 0.2 = 64; J5 dynamic early warning threshold = 80 − 80 × 0.15 = 68. Check the constraints: J4's downward adjustment of 24 does not exceed the baseline threshold of 80; J2's downward adjustment of 16 does not exceed 80; J5's downward adjustment of 12 does not exceed 80. Output: The dynamic warning thresholds for J4, J2, and J5 are 56, 64, and 68, respectively.

[0071] S404 compares the risk quantification indicators of each monitoring node with its dynamic early warning threshold to identify nodes exceeding the limit. Specifically, taking J2, J4, and J5 in the associated node set as examples, their risk quantification indicators, obtained through parallel analysis of the three models in S302 and weighted fusion calculation in S304, are 70, 75, and 70 respectively (expressed on a percentage basis for easy matching with the magnitude of the early warning threshold). The cloud computing platform compares the risk quantification indicators of each monitoring node with its dynamic early warning threshold: J4's risk quantification indicator of 75 is greater than the dynamic early warning threshold of 56, so J4 is determined to be an node exceeding the limit; J2's risk quantification indicator of 70 is greater than the dynamic early warning threshold of 64, so J2 is determined to be an node exceeding the limit; J5's risk quantification indicator of 70 is greater than the dynamic early warning threshold of 68, so J5 is determined to be an node exceeding the limit. All three nodes J2, J4, and J5 in the associated node set are identified as nodes exceeding the limit.

[0072] S405, the risk level is determined by combining the risk quantification indicators of the out-of-limit nodes and the coverage ratio of abnormal events in the associated node set. Specifically, based on the retrospective statistics of historical risk events in the pipeline network, the out-of-limit percentage is determined to be 60%, the coverage ratio is determined to be 40%, and the comprehensive risk index = 0.6 × (total out-of-limit magnitude / total dynamic warning threshold) + 0.4 × coverage ratio. In this embodiment, the out-of-limit magnitude of J4 is 75 − 56 = 19, the out-of-limit magnitude of J2 is 70 − 64 = 6, the out-of-limit magnitude of J5 is 70 − 68 = 2, and the total out-of-limit magnitude is 27; the total dynamic warning threshold is 56 + 64 + 68 = 188; and the coverage ratio is 3 / 3 = 100%. Substituting these values, the comprehensive risk index = 0.6 × (27 / 188) + 0.4 × 1.0 = 0.0862 + 0.4 = 0.4862. The risk level mapping rule is based on the correspondence between the historical risk level and emergency response level of the pipeline network: a comprehensive risk index greater than or equal to 0.4 is classified as Level 1 risk, 0.2 to 0.4 as Level 2 risk, and less than 0.2 as Level 3 risk. Since 0.4862 is greater than 0.4, this abnormal event is classified as Level 1 risk.

[0073] S105, based on risk level and abnormal events, quantifies and scores the data according to the set spatial levels and assessment dimensions to generate an operational health assessment result. Specifically, the cloud computing platform constructs a cross-assessment matrix, a 3x3 matrix, where the rows correspond to three spatial levels (first row: pipeline area level, second row: well level, third row: pipeline level), and the columns correspond to three assessment dimensions (first column: pipeline structure dimension, second column: environmental parameter dimension, third column: equipment operation dimension). The cloud computing platform sums all nine elements of the cross-assessment matrix after hierarchical normalization to obtain the comprehensive operational health assessment value for the current time segment. The operational health assessment result further includes: spatial level health index, assessment dimension health index, risk status evolution label, and abnormal event correlation graph. Among them, the spatial level health index is the arithmetic mean of all three matrix elements in the row of the spatial level after level normalization; the assessment dimension health index is the arithmetic mean of all three matrix elements in the column of the assessment dimension after level normalization; the risk situation evolution label is generated by comparing the current risk level with the historical risk level of the same period in the past 24 hours. If the current level is higher than the highest historical level, the label is deteriorating; if it is equal, it is continuing; if it is lower, it is mitigating; the abnormal event association graph is generated by constructing an abnormal propagation subgraph containing J2, J4, and J5 with the abnormal source node J3 as the center and the association influence coefficient in S501 as the edge weight, reflecting the propagation path of the abnormal from J3 to the surrounding nodes.

[0074] The quantitative scoring, based on the established spatial hierarchy and evaluation dimensions, includes: The cloud computing platform assigns initial values ​​to each element of the cross-assessment matrix based on the risk level; the higher the risk level, the larger the initial value of the corresponding matrix element. Specifically, the cloud computing platform assigns initial values ​​according to the risk levels determined by S104: Level 1 risk corresponds to an initial value of 80, Level 2 risk corresponds to an initial value of 50, and Level 3 risk corresponds to an initial value of 20. Taking the area where the caisson J3 is located as currently classified as Level 1 risk as an example, all nine elements of the cross-assessment matrix are initially assigned a value of 80. This initial value assignment method transforms the qualitative judgment of the risk level into quantitative base values ​​for the matrix elements, providing a unified numerical starting point for subsequent differentiated adjustments.

[0075] Anomaly increase correction: When an anomaly event occurs, the cloud computing platform determines the affected matrix element positions based on the location of the anomaly source monitoring node (the node where the anomaly event occurred). The initial value of the matrix element at the corresponding position is corrected for anomaly increase, and the magnitude of this correction is positively correlated with the severity of the anomaly event. Specifically, anomaly increase correction includes:

[0076] The cloud computing platform determines the location of matrix elements affected by the abnormal event based on the monitoring node of the abnormal source corresponding to the abnormal event; the cloud computing platform obtains the values ​​of key features at the time of the abnormal event and calculates the deviation of the values ​​from the dynamic early warning threshold; based on the deviation, the cloud computing platform corrects the initial value of the matrix element position for abnormal increase. The magnitude of the abnormal increase correction is positively correlated with the deviation. The larger the deviation, the larger the magnitude of the abnormal increase correction.

[0077] Specifically, for the anomaly source monitoring node J3, the key characteristic values ​​at the time of the anomaly event (average pipe wall stress 126MPa, average well temperature 41℃, where the average pipe wall stress 126MPa comes from window 15 in the key feature time series change extraction of S303, and the average well temperature 41℃ is the historical maximum value of the average well temperature within 30 preset time windows in the key feature time series change extraction of S303, appearing in window 8), and the node-level benchmark early warning threshold of the anomaly source node J3 generated by S402 (average pipe wall stress threshold 80MPa, average well temperature threshold 34℃, the anomaly source node is not included in the S403 associated node set, so it does not participate in the threshold linkage downgrade of S503, and its dynamic early warning threshold remains at the node-level benchmark early warning threshold). Processing: The deviation is calculated as follows: the average deviation of pipe wall stress = 126 − 80 = 46 MPa, and the deviation ratio is calculated as 46 / 80 = 57.5% based on the ratio of the deviation to the dynamic early warning threshold; the average deviation of well temperature = 41 − 34 = 7℃, and the deviation ratio is calculated as 7 / 34 ≈ 20.6% based on the ratio of the deviation to the dynamic early warning threshold. The cloud computing platform uses the deviation ratio as a quantitative intermediary to perform a stepped mapping to determine the correction range for abnormal increases: deviation ratio ≥ 50% is mapped to 30%; deviation ratio ≥ 20% and ≤ 50% is mapped to 20%; and deviation ratio ≤ 20% is mapped to 10%. The deviation ratio for the pipeline structure dimension is 57.5%, mapped to 30%; the deviation ratio for the environmental parameter dimension is 20.6%, mapped to 20%. A 30% anomalous increase correction was applied to the initial value of 80 for the location affected by the pipeline structure dimension (second row, first column, caisson level, pipeline structure dimension), resulting in a corrected value of 104; a 20% anomalous increase correction was applied to the initial value of 80 for the location affected by the environmental parameter dimension (second row, second column, caisson level, environmental parameter dimension), resulting in a corrected value of 96.

[0078] For hierarchical normalization, the cloud computing platform performs hierarchical normalization on different row elements in the same evaluation dimension of the cross-evaluation matrix based on the number and weight of monitoring nodes within each spatial level. Spatial levels with more monitoring nodes have higher weight proportions for their row elements. Specifically, the pipeline area level contains 24 monitoring nodes, the well level contains 6, and the pipeline level contains 12, with monitoring node weights of 24 / 42, 6 / 42, and 12 / 42 for the three spatial levels, respectively. The cloud computing platform performs hierarchical normalization by multiplying the three row elements in each evaluation dimension of the cross-evaluation matrix by their corresponding weights, ensuring that spatial levels with denser monitoring node distribution contribute more to the final evaluation results.

[0079] The evaluation results are generated by summing the hierarchically normalized cross-evaluation matrix on the cloud computing platform to produce the operational health assessment result. Specifically, the cloud computing platform sums all nine elements of the hierarchically normalized cross-evaluation matrix to obtain the comprehensive value of the operational health assessment result at the current moment. This comprehensive value quantitatively represents the overall health status of the spatial region under multi-dimensional evaluation in a single numerical form. The higher the value, the worse the health status and the more concentrated the risks, making it easier for operation and maintenance managers to quickly grasp the overall situation.

[0080] S106, the cloud computing platform generates early warning information based on the comprehensive value and risk level of the operation and maintenance health assessment and pushes the early warning information to the operation and maintenance terminal. Specifically, based on the comprehensive value and risk level of the operation and maintenance health assessment results generated in S105, the cloud computing platform generates structured early warning information, which includes risk level identifiers, a list of nodes exceeding limits, risk quantification indicators, and recommended handling measures. The cloud computing platform pushes the early warning information to the terminal of the person in charge of operation and maintenance and simultaneously pushes the abnormal event correlation graph to the emergency command platform screen, enabling operation and maintenance management personnel to quickly locate the abnormal propagation path.

[0081] The present invention has been described in detail above. The specific embodiments are provided only to help understand the method and core ideas of the present invention. It should be noted that those skilled in the art can make various improvements and modifications to the present invention without departing from its principles, and these improvements and modifications also fall within the protection scope of the claims of the present invention.

Claims

1. A distributed, full-scenario intelligent operation and maintenance dynamic evaluation method based on cloud computing, characterized in that, Includes the following steps: S101, various types of sensors are deployed at the monitoring nodes of the main structure and ancillary facilities of the pipeline network to synchronously collect raw data, including structural status parameters, environmental parameters and safety status data; S102, deploy edge gateways in partitions based on monitoring nodes, filter the raw data and then parse and extract key features to obtain key features, obtain abnormal events based on the key features, and upload the key features and abnormal events to the cloud computing platform. S103, the cloud computing platform assigns key features to multiple preset evaluation dimensions, uses preset models in parallel to analyze the evaluation dimensions, and dynamically adapts the weights of the evaluation dimensions according to the dynamic change patterns of key features in the time series to generate risk quantification indicators. S104, the cloud computing platform obtains a preset baseline warning threshold, corrects the baseline warning threshold based on the key features and the abnormal event to obtain a dynamic warning threshold, compares the risk quantification index with the dynamic warning threshold, and classifies the risk level according to the degree of exceedance of the risk quantification index and the number of nodes affected by the abnormal event. S105, Based on the risk level and the abnormal event, quantitative scoring is performed according to the set spatial hierarchy and evaluation dimensions to generate an operation and maintenance health assessment result; S106, the cloud computing platform generates early warning information based on the operation and maintenance health assessment results, and pushes the early warning information to the operation and maintenance terminal.

2. The cloud-based distributed full-scenario intelligent operation and maintenance dynamic evaluation method according to claim 1, characterized in that, S102 includes the following steps: S201, based on the spatial distribution density of monitoring nodes and the pipeline topology, divides the monitoring nodes into multiple acquisition zones, and deploys an edge gateway in each acquisition zone; S202, The edge gateway performs format verification and invalid value removal on the raw data in the collection partition to obtain standardized data; S203, the edge gateway performs mean or extreme value calculations on standardized data based on a preset time window to obtain key features that reflect the operating status of the pipeline network. S204, the edge gateway determines whether the key feature exceeds the edge judgment threshold. If it does not exceed the threshold, the key feature is cached locally, and multiple cached key features are uploaded to the cloud computing platform when the preset reporting period is reached. S205 If the key features exceed the edge judgment threshold, the edge gateway generates an abnormal event. The abnormal event includes the identifier of the abnormal source monitoring node, and the abnormal event and key features are uploaded to the cloud computing platform.

3. The cloud computing-based distributed full-scenario intelligent operation and maintenance dynamic evaluation method according to claim 1, characterized in that, S103 includes the following steps: S301, the cloud computing platform assigns key features to multiple preset evaluation dimensions, including pipeline structure dimension, environmental parameter dimension and equipment operation dimension; S302, the preset models include a pipeline degradation model, a spatial risk assessment model, and an equipment deterioration prediction model; the pipeline degradation model, spatial risk assessment model, and equipment deterioration prediction model respectively analyze the pipeline structure dimension, environmental parameter dimension, and equipment operation dimension to obtain structural health sub-indicators, environmental safety sub-indicators, and equipment status sub-indicators; the preset models are called and executed in parallel by the cloud computing platform; S303, the cloud computing platform obtains key features from multiple consecutive preset time windows to form a historical key feature sequence, extracts dynamic change features of each evaluation dimension based on the historical key feature sequence, uses preset weights of each evaluation dimension as initial weights, and dynamically adapts the initial weights to obtain the target weights corresponding to each evaluation dimension. S304, based on the target weights corresponding to each assessment dimension, weighted and integrated the structural health sub-indicator, environmental safety sub-indicator, and equipment status sub-indicator to obtain the risk quantification index.

4. The cloud computing-based distributed full-scenario intelligent operation and maintenance dynamic evaluation method according to claim 3, characterized in that, S303 includes: Key feature temporal change extraction: The cloud computing platform extracts dynamic change features based on the historical key feature sequences corresponding to each evaluation dimension. The dynamic change features include data fluctuation rate, trend change intensity, and correlation deviation index. Among them, data fluctuation rate reflects the temporal dispersion of the historical key feature sequence, trend change intensity reflects the temporal jump of the historical key feature sequence, and correlation deviation index reflects the degree of deviation of the correlation between key features of each evaluation dimension from the historical stable state. The correlation deviation index is determined by the deviation between the real-time correlation degree and the historical stable correlation degree between key features of each evaluation dimension through quantification. The weight adjustment benchmark value is calculated by the cloud computing platform based on the data fluctuation rate. The weight adjustment benchmark value is positively correlated with the data fluctuation rate. The weight adjustment coefficient is amplified by the cloud computing platform based on the intensity of trend change. The greater the intensity of trend change, the larger the weight adjustment coefficient is relative to the weight adjustment benchmark value. Cross-validation correction: The cloud computing platform performs cross-validation correction on the weight adjustment coefficients of each evaluation dimension based on the correlation deviation index. The correlation deviation index of each evaluation dimension is sorted from largest to smallest. The weight adjustment coefficients of the evaluation dimensions with higher ranking are reduced, while the weight adjustment coefficients of the remaining evaluation dimensions remain unchanged. The target weight fusion process involves the cloud computing platform combining the corrected weight adjustment coefficients with the initial weights of the corresponding evaluation dimensions to obtain the target weights for each evaluation dimension.

5. The cloud computing-based distributed full-scenario intelligent operation and maintenance dynamic evaluation method according to claim 4, characterized in that, The cross-validation correction includes: The correlation matrix is ​​constructed by the cloud computing platform based on the key features corresponding to each evaluation dimension in the current time window in the historical key feature sequence, calculating the real-time correlation between the key features corresponding to any two evaluation dimensions, and constructing the correlation matrix. The correlation deviation index is calculated by the cloud computing platform based on the correlation degree matrix and the historical stable correlation degree matrix. The correlation deviation index is the norm of the row vector of the corresponding evaluation dimension in the difference matrix between the current correlation degree matrix and the historical stable correlation matrix. Once the abnormal correlation dimensions are determined, the cloud computing platform sorts the correlation deviation indices of each evaluation dimension from largest to smallest, and determines the evaluation dimensions with the highest ranking as the abnormal correlation dimensions. The cloud computing platform applies a decay factor to the weight adjustment coefficient of the abnormal correlation dimension. The decay factor is negatively correlated with the correlation deviation index of the abnormal correlation dimension, while the weight adjustment coefficients of the other evaluation dimensions remain unchanged. The cloud computing platform will output the corrected weight adjustment coefficient after applying the attenuation factor.

6. The cloud computing-based distributed full-scenario intelligent operation and maintenance dynamic evaluation method according to claim 1, characterized in that, S104 includes the following steps: S401, the cloud computing platform obtains the spatial topology relationship of each monitoring node and establishes a node association graph. The edges of the node association graph have edge weights that represent the pipe segment connection attributes. S402, based on the distribution statistics of key features within the historical security period, generates node-level benchmark early warning thresholds for each monitoring node; S403, when an abnormal event is detected, the monitoring node where the abnormal event occurred is identified as the abnormal source node, the set of associated nodes of the abnormal source node is determined based on the node association diagram, and the node-level baseline early warning threshold of each node in the set of associated nodes is adjusted and corrected in a coordinated manner according to the propagation characteristics of the abnormality in the pipeline network, so as to obtain the dynamic early warning threshold. S404 compares the risk quantification indicators of each monitoring node with its dynamic early warning threshold to identify nodes that exceed the limit; S405 uses the combined risk quantification indicators of the out-of-limit node and the coverage ratio of abnormal events in the associated node set to classify risk levels.

7. The cloud computing-based distributed full-scenario intelligent operation and maintenance dynamic evaluation method according to claim 6, characterized in that, S403 includes the following steps: S501, the cloud computing platform calculates the association influence coefficient from the abnormal source node to each associated node based on the edge weight of each edge in the node association graph. The edge weight is negatively correlated with the actual length of the pipeline between nodes. S502, Based on the correlation influence coefficient, determine the threshold reduction range for each correlated node; the larger the correlation influence coefficient, the larger the threshold reduction range. S503, subtract the corresponding threshold reduction amount from the node-level baseline early warning threshold of each associated node to obtain the dynamic early warning threshold of each associated node, and the threshold reduction amount shall not exceed the node-level baseline early warning threshold.

8. The cloud computing-based distributed full-scenario intelligent operation and maintenance dynamic evaluation method according to claim 7, characterized in that, The edge weights in S501 are determined through the following steps: S601, the cloud computing platform obtains the physical attribute parameters of the pipe segment connecting the two monitoring nodes; S602, based on physical property parameters, calculates the structural fragility coefficient of the pipe segment. The structural fragility coefficient is negatively correlated with the pipe diameter and positively correlated with the material aging rate. S603, calculates the environmental erosion coefficient of the pipe section based on physical property parameters; The environmental erosion coefficient is negatively correlated with burial depth and positively correlated with service life. S604 combines the structural fragility coefficient with the environmental erosion coefficient to obtain the comprehensive fragility coefficient of the pipe section; S605 calculates edge weights based on the comprehensive vulnerability coefficient of the pipe segment and the actual length of the pipeline network between nodes. The edge weights are positively correlated with the comprehensive vulnerability coefficient of the pipe segment and negatively correlated with the actual length of the pipeline network.

9. The cloud computing-based distributed full-scenario intelligent operation and maintenance dynamic evaluation method according to claim 1, characterized in that, The quantitative scoring in S105, based on the set spatial hierarchy and evaluation dimensions, includes: The cross-evaluation matrix is ​​constructed using a cloud computing platform, which generates a cross-evaluation matrix that represents spatial levels and evaluation dimensions. The rows of the cross-evaluation matrix correspond to spatial levels, including pipeline area levels, caisson levels, and pipeline levels. The columns of the cross-evaluation matrix correspond to evaluation dimensions. Initial value assignment for the matrix: The cloud computing platform assigns initial values ​​to each element in the cross-evaluation matrix based on the risk level; the higher the risk level, the larger the initial value of the corresponding matrix element. Abnormal increase correction: When an abnormal event occurs, the cloud computing platform determines the position of the matrix element affected by the abnormal event based on the location of the abnormal source monitoring node corresponding to the abnormal event. The abnormal source monitoring node is the monitoring node where the abnormal event occurred. The initial value of the matrix element located at the matrix element position is corrected for abnormal increase. The magnitude of the abnormal increase correction is positively correlated with the severity of the abnormal event. Hierarchical normalization processing: The cloud computing platform performs hierarchical normalization processing on different row elements under the same evaluation dimension in the cross-evaluation matrix based on the number and weight of monitoring nodes in each spatial level; the more monitoring nodes a spatial level has, the higher the weight ratio of its row elements. Once the evaluation results are generated, the cloud computing platform performs a summation operation on the cross-evaluation matrix after hierarchical normalization to generate the operation and maintenance health evaluation results.

10. The cloud computing-based distributed full-scenario intelligent operation and maintenance dynamic evaluation method according to claim 9, characterized in that, The correction for the abnormal increase includes: The cloud computing platform determines the location of matrix elements affected by the abnormal event based on the monitoring node of the abnormal source corresponding to the abnormal event; the cloud computing platform obtains the values ​​of key features at the time of the abnormal event and calculates the deviation of the values ​​from the dynamic early warning threshold; based on the deviation, the cloud computing platform corrects the initial value of the matrix element position for abnormal increase. The magnitude of the abnormal increase correction is positively correlated with the deviation. The larger the deviation, the larger the magnitude of the abnormal increase correction.

Images