Gas equipment operation data anomaly identification processing method and system

Abstract

The application provides a gas equipment operation data anomaly identification processing method and system, relates to the technical field of data processing, and comprises the following steps: receiving, in real time, equipment operation data streams uploaded by cell gas Internet of Things terminals, performing time stamp alignment, noise filtering and missing value interpolation processing on the equipment operation data streams, and obtaining a standardized time series data set; extracting synchronous parameter sequences of a pressure regulating station outlet convergence node, a supporting public building pipe network convergence node and a high-rise residential end table front node from the standardized time series data set, and constructing a three-dimensional working condition response surface; performing curvature differentiation and topological equivalent domain division on the three-dimensional working condition response surface, and obtaining a pipe network hydraulic and thermal coupling compensation amount. The application improves the accuracy, real-time performance and fine level of gas equipment operation data anomaly identification.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method and system for identifying and processing abnormal operating data of gas equipment. Background Technology

[0002] In the scenario of centralized gas supply in urban residential communities, the safe monitoring of gas equipment operation data is crucial to ensuring the gas supply for people's livelihood. Taking a typical comprehensive community as an example, the area includes multiple high-rise residential buildings, supporting commercial and catering buildings and centralized pressure regulating stations. The overlap of the peak gas consumption of residents in the morning and evening with the cooking time of merchants causes the pipeline flow and pressure to fluctuate drastically in a periodic manner. Moreover, the low temperature in winter can easily cause local freezing and blockage of pipelines and abnormal ignition of terminal equipment.

[0003] Existing anomaly detection technologies mostly rely on a single fixed threshold for exceeding limits, which makes it difficult to dynamically adapt to the coupled changes in gas usage time, seasonal temperature differences, and equipment load. During peak hours, false alarms occur frequently due to normal surges in flow. Furthermore, in typical leakage complex conditions such as the initial stage of pipeline freezing or a sudden drop in pressure accompanied by a sudden increase in flow, false alarms are seriously missed because the threshold is not triggered.

[0004] Existing monitoring systems mostly adopt independent parameter analysis architectures, lacking the ability to analyze the spatial topology correlation and dynamic coupling of multi-dimensional time-series data such as pressure, flow, and temperature. This makes it difficult to accurately separate normal operating condition fluctuations from real abnormal characteristics in massive high-frequency IoT data streams, resulting in delayed early warnings and frequent misjudgments. This hinders the intelligent upgrade of gas safety management, and existing solutions are unable to meet the urgent need for real-time control of gas safety under complex dynamic operating conditions. Summary of the Invention

[0005] This invention provides a method and system for identifying and processing abnormal operating data of gas equipment, which improves the accuracy, real-time performance and precision of identifying abnormal operating data of gas equipment.

[0006] To solve the above-mentioned technical problems, the technical solution of the present invention is as follows:

[0007] A first aspect is a method for identifying and processing abnormal operating data of gas equipment, the method comprising:

[0008] The system receives real-time data streams of equipment operation from the community gas IoT terminal, performs timestamp alignment, noise filtering, and missing value interpolation on the data streams to obtain a standardized time-series dataset.

[0009] Synchronous parameter sequences of the outlet confluence node of the pressure regulating station, the inflow node of the supporting public building pipeline network, and the terminal meter node of the high-rise residential building are extracted from the standardized time-series dataset to construct a three-dimensional operating condition response surface. The curvature differential and topological equal domain division of the three-dimensional operating condition response surface are performed to obtain the hydraulic-thermal coupling compensation amount of the pipeline network. The standardized time-series dataset is subjected to dynamic baseline generation processing to obtain the initial baseline data sequence. The initial baseline data sequence is superimposed with the hydraulic-thermal coupling compensation amount of the pipeline network through error compensation to obtain the corrected baseline sequence.

[0010] A multi-dimensional context feature matrix is ​​constructed by combining the corrected baseline sequence with the characteristics of the current gas consumption period, seasonal temperature characteristics, and equipment type characteristics. Nonlinear adaptive fitting is then performed on the multi-dimensional context feature matrix to obtain a multi-parameter dynamic normal interval baseline.

[0011] Using the multi-parameter dynamic normal interval baseline as a dynamic judgment benchmark, the standardized time series dataset is screened for exceeding limits to extract exceeding data segments; multi-parameter correlation analysis is performed on the exceeding data segments to obtain abnormal feature vectors;

[0012] The abnormal feature vector is matched with a preset gas anomaly pattern library to obtain anomaly matching results; the anomaly matching results are combined with historical evolution trends to calculate risk confidence, and anomaly identification results including anomaly type, associated equipment location and risk level are obtained.

[0013] Furthermore, the system receives real-time data streams of equipment operation from the community gas IoT terminal, performs timestamp alignment, noise filtering, and missing value interpolation on the data streams to obtain a standardized time-series dataset, including:

[0014] Receive the device operation data stream, extract the timestamp information of each data node, resample and sort the timestamp information according to a unified sampling period, and obtain a time-aligned data sequence.

[0015] Sliding window denoising is applied to the time-aligned data sequence to filter out high-frequency abrupt noise points, resulting in a smoothed data sequence.

[0016] By detecting data breakpoints in the smoothed data sequence, and performing interpolation and filling based on the changing trends of adjacent valid data segments before and after the breakpoint, a continuous and complete data sequence is obtained, which is then used as a standardized time series dataset.

[0017] Furthermore, synchronous parameter sequences of the surge tank outlet junction node, the supporting public building pipeline junction node, and the high-rise residential building terminal meter node are extracted from the standardized time-series dataset to construct a three-dimensional operating condition response surface. Curvature differentiation and topological equal domain partitioning are performed on the three-dimensional operating condition response surface to obtain the hydraulic-thermal coupling compensation quantities of the pipeline network, including:

[0018] The real-time collection points of the outlet junction node of the pressure regulating station, the junction node of the supporting public building pipeline network, and the terminal meter node of the high-rise residential building are located from the standardized time series dataset. The pressure value, flow value and temperature value of each node in the same sampling period are extracted to obtain the synchronous parameter sequence.

[0019] The flow rate and temperature values ​​in the synchronous parameter sequence are used as spatial horizontal and vertical coordinates, and the pressure value is used as elevation coordinates for spatial mapping to construct a three-dimensional working condition response surface.

[0020] Curvature differential calculation is performed on the three-dimensional working condition response surface to extract the surface deformation gradient features; using the surface deformation gradient features as the basis for topological partitioning, the three-dimensional working condition response surface is divided into equal-value domains to obtain multiple local response sub-regions.

[0021] The pressure gradient integral and temperature attenuation coefficient are calculated separately for each local response sub-region. The pressure gradient integral and temperature attenuation coefficient are then weighted and mapped to obtain the hydraulic-thermal coupling compensation amount of the pipe network.

[0022] Furthermore, the standardized time-series dataset undergoes dynamic baseline generation processing to obtain an initial baseline data sequence; the initial baseline data sequence is then superimposed with the network hydraulic-thermal coupling compensation amount to obtain a corrected baseline sequence, including:

[0023] Perform sliding window statistical calculations on the standardized time series dataset, extract the central trend value of each running parameter within the sliding window, and obtain the initial trend data sequence;

[0024] Boundary constraint filtering is performed on the initial trend data sequence to remove extreme points that exceed the preset physical reasonable range, thus obtaining the initial baseline data sequence; the initial baseline data sequence is then aligned with the hydraulic-thermal coupling compensation amount of the pipeline network on the time axis to obtain the time synchronization reference sequence.

[0025] Based on the time synchronization reference sequence and the hydraulic and thermal coupling compensation amount of the pipeline network, parameter dimension matching is performed to construct the node compensation mapping set corresponding to each sampling node.

[0026] The corrected baseline sequence is obtained by performing a deviation compensation superposition operation between the node compensation mapping set and the reference value of the corresponding sampling node in the time synchronization reference sequence.

[0027] Furthermore, a multi-dimensional context feature matrix is ​​constructed by combining the corrected baseline sequence with current gas consumption period characteristics, seasonal temperature characteristics, and equipment type characteristics. Nonlinear adaptive fitting is then performed on this multi-dimensional context feature matrix to obtain a multi-parameter dynamic normal interval baseline, including:

[0028] The corrected baseline sequence is sliced ​​into time windows to extract the baseline parameter values ​​within each time window, resulting in a baseline feature slice set. The current gas consumption period features, seasonal temperature features, and equipment type features are numerically encoded and normalized to obtain the period encoding vector, temperature encoding vector, and equipment type encoding vector, respectively.

[0029] The time period encoding vector, temperature encoding vector and device type encoding vector are concatenated to obtain the context feature vector set; the baseline feature slice set and the context feature vector set are cross-aligned and concatenated according to timestamp and dimension to construct a multi-dimensional context feature matrix.

[0030] Iterative weight optimization is performed on the multidimensional context feature matrix to dynamically adjust the baseline parameter dimension of the corresponding baseline feature slice set mapping in the multidimensional context feature matrix, as well as the fitting coefficients of the time period encoding dimension, temperature encoding dimension and device type encoding dimension of the corresponding context feature vector set mapping, to obtain the optimized fitting coefficients.

[0031] Based on the optimized fitting coefficients, a nonlinear mapping operation is performed on the multidimensional context feature matrix to obtain a multi-parameter dynamic normal interval baseline.

[0032] Furthermore, using the multi-parameter dynamic normal interval baseline as a dynamic judgment benchmark, the standardized time-series dataset is subjected to limit-crossing screening to extract limit-crossing data segments; multi-parameter correlation analysis is performed on the limit-crossing data segments to obtain abnormal feature vectors, including:

[0033] The standardized time series dataset is aligned point by point with the multi-parameter dynamic normal interval baseline. The deviation between the actual operating parameter value of each sampling node and the upper and lower limits of the corresponding baseline interval is calculated. Continuous data segments with deviations exceeding the preset tolerance threshold are selected to obtain the over-limit data segments.

[0034] Multi-parameter time-series synchronous extraction is performed on the out-of-limit data segments to separate the corresponding pressure time-series subseries, flow time-series subseries, and temperature time-series subseries within the segments;

[0035] The pressure time series, flow rate time series, and temperature time series were subjected to pairwise dynamic cross-correlation calculations to obtain the calculation results; a set of parameter coupling deviation indices was constructed based on the calculation results.

[0036] The abnormal feature vector is obtained by performing feature dimensionality reduction and vectorization reorganization on the parameter coupling deviation index set and the duration and peak deviation of the out-of-limit data segment.

[0037] Furthermore, the abnormal feature vectors are matched with a pre-defined gas anomaly pattern library to obtain anomaly matching results. These matching results are then combined with historical evolution trends to calculate risk confidence, resulting in anomaly identification results that include anomaly type, associated equipment location, and risk level.

[0038] Extract the prototype feature vectors of each standard anomaly pattern from the preset gas anomaly pattern library, calculate the spatial similarity distance between the anomaly feature vector and each prototype feature vector, select the standard anomaly pattern with the spatial similarity distance as the target matching pattern, and obtain the anomaly matching result;

[0039] Based on the abnormal matching results, locate the associated gas equipment locations and retrieve the historical operation records corresponding to the gas equipment locations; extract the occurrence frequency and deterioration rate parameters of the target matching pattern in the historical operation records and construct a historical evolution trend vector;

[0040] The pattern matching degree and historical evolution trend vector in the abnormal matching results are input into the preset risk confidence assessment model, and weighted fusion calculation is performed to obtain the risk confidence value.

[0041] The risk level is determined by mapping the risk confidence score to a preset risk level threshold. The risk level determination is then combined with the anomaly category identifier and associated device location of the target matching pattern to obtain an anomaly identification result that includes anomaly type, associated device location, and risk level.

[0042] Secondly, the gas equipment operation data anomaly identification and processing system includes:

[0043] The acquisition module is used to receive the device operation data stream uploaded by the community gas IoT terminal in real time, and perform timestamp alignment, noise filtering and missing value interpolation on the device operation data stream to obtain a standardized time series dataset.

[0044] The module is used to extract the synchronous parameter sequences of the outlet confluence node of the pressure regulating station, the inflow node of the supporting public building pipeline network, and the terminal meter node of the high-rise residential building from the standardized time series dataset, and construct a three-dimensional operating condition response surface; perform curvature differentiation and topological equal domain division on the three-dimensional operating condition response surface to obtain the hydraulic-thermal coupling compensation amount of the pipeline network; perform dynamic baseline generation processing on the standardized time series dataset to obtain the initial baseline data sequence; and perform error compensation superposition on the initial baseline data sequence and the hydraulic-thermal coupling compensation amount of the pipeline network to obtain the corrected baseline sequence.

[0045] The fitting module is used to construct a multi-dimensional context feature matrix by combining the corrected baseline sequence with the current gas consumption period characteristics, seasonal temperature characteristics, and equipment type characteristics. Nonlinear adaptive fitting is performed on the multi-dimensional context feature matrix to obtain a multi-parameter dynamic normal interval baseline.

[0046] The analysis module is used to use the multi-parameter dynamic normal interval baseline as a dynamic judgment benchmark to perform limit screening on the standardized time series dataset to extract limit-exceeding data segments; and to perform multi-parameter correlation analysis on the limit-exceeding data segments to obtain abnormal feature vectors.

[0047] The mapping module is used to perform feature mapping matching between abnormal feature vectors and a preset gas anomaly pattern library to obtain anomaly matching results; the anomaly matching results are combined with historical evolution trends to calculate risk confidence, and anomaly identification results including anomaly type, associated equipment location and risk level are obtained.

[0048] Thirdly, a computing device, comprising:

[0049] One or more processors;

[0050] A storage device for storing one or more programs that, when executed by one or more processors, cause the one or more processors to implement the method.

[0051] Fourthly, a computer-readable storage medium storing a program that, when executed by a processor, implements the method.

[0052] The above-described solution of the present invention has at least the following beneficial effects:

[0053] This method overcomes the technical problems of existing technologies, such as false alarms and false negatives caused by the inability of a single fixed threshold to adapt to changes in operating conditions, lack of multi-parameter spatiotemporal coupling analysis capability, rigid baseline model, and difficulty in accurately identifying anomaly types and quantifying risk levels. It employs standardized preprocessing techniques including timestamp alignment, noise filtering, and missing value interpolation to improve the accuracy, real-time performance, and refinement of anomaly identification in gas equipment operation data. This enables the location of anomaly-related equipment, clarification of anomaly types and risk levels, effective prevention of gas safety accidents, protection of residential gas safety, and promotion of intelligent upgrades to gas safety management. Furthermore, it utilizes standardized time-series datasets for over-limit screening and multi-parameter correlation analysis to obtain anomaly feature vectors. These vectors are then matched with a pre-defined gas anomaly pattern library through feature mapping and risk confidence calculation based on historical evolution trends. Attached Figure Description

[0054] Figure 1This is a flowchart illustrating the method for identifying and processing abnormal operating data of gas equipment provided in an embodiment of the present invention.

[0055] Figure 2 This is a schematic diagram of a gas equipment operation data anomaly identification and processing system provided in an embodiment of the present invention.

[0056] Figure 3 It is a simulation diagram of the preprocessing effect of equipment operation data.

[0057] Figure 4 It is a simulation diagram of dynamic baseline generation and compensation effect.

[0058] Figure 5 It is a trend chart of multi-parameter dynamic baseline fitting.

[0059] Figure 6 This is a comparison chart of anomaly detection results. Detailed Implementation

[0060] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0061] like Figure 1 As shown, an embodiment of the present invention proposes a method for identifying and processing abnormal operating data of gas equipment, the method comprising the following steps:

[0062] Step 1: Receive the device operation data stream uploaded by the community gas IoT terminal in real time, and perform timestamp alignment, noise filtering and missing value interpolation on the device operation data stream to obtain a standardized time series dataset;

[0063] Step 2: Extract the synchronous parameter sequences of the outlet confluence node of the pressure regulating station, the inflow node of the supporting public building pipeline network, and the terminal meter node of the high-rise residential building from the standardized time series dataset to construct a three-dimensional operating condition response surface; perform curvature differentiation and topological equal domain division on the three-dimensional operating condition response surface to obtain the hydraulic-thermal coupling compensation amount of the pipeline network; perform dynamic baseline generation processing on the standardized time series dataset to obtain the initial baseline data sequence; and perform error compensation superposition on the initial baseline data sequence and the hydraulic-thermal coupling compensation amount of the pipeline network to obtain the corrected baseline sequence.

[0064] Step 3: Construct a multi-dimensional context feature matrix by combining the corrected baseline sequence with the current gas consumption period characteristics, seasonal temperature characteristics, and equipment type characteristics. Perform nonlinear adaptive fitting on the multi-dimensional context feature matrix to obtain the multi-parameter dynamic normal interval baseline.

[0065] Step 4: Using the multi-parameter dynamic normal interval baseline as the dynamic judgment benchmark, perform limit screening on the standardized time series dataset to extract limit-exceeding data segments; perform multi-parameter correlation analysis on the limit-exceeding data segments to obtain abnormal feature vectors;

[0066] Step 5: Perform feature mapping matching between the abnormal feature vector and the preset gas anomaly pattern library to obtain the anomaly matching result; combine the anomaly matching result with the historical evolution trend to calculate the risk confidence level, and obtain the anomaly identification result including the anomaly type, the location of the associated equipment and the risk level.

[0067] In this embodiment of the invention, the invention employs a standardized process of receiving real-time data streams of equipment operation uploaded by community gas IoT terminals, performing timestamp alignment, noise filtering, and missing value interpolation. It extracts synchronous parameter sequences from the outlet junction nodes of pressure regulating stations, the junction nodes of supporting public building pipelines, and the terminal meter nodes of high-rise residential buildings to construct a three-dimensional operating condition response surface and calculate the hydraulic-thermal coupling compensation of the pipeline network. A corrected baseline sequence is obtained through dynamic baseline generation and error compensation superposition. A multi-dimensional context feature matrix is ​​constructed by combining the current gas usage period, seasonal temperature, and equipment type characteristics, and nonlinear adaptive fitting is performed to obtain a multi-parameter dynamic normal interval baseline. Anomaly feature vectors are obtained by performing over-limit screening and multi-parameter correlation analysis on the standardized time-series dataset. The abnormal feature vectors are then matched with a preset gas anomaly pattern library through feature mapping and the risk confidence level is calculated by combining historical evolution trends.

[0068] This technology overcomes the technical problems of existing gas equipment anomaly identification technologies, such as fixed thresholds that cannot adapt to changes in operating conditions, lack of multi-parameter coupling analysis capabilities, rigid baseline models, and difficulty in accurately identifying anomaly types, locating related equipment, and quantifying risk levels. As a result, it improves the accuracy and real-time performance of anomaly identification in gas equipment operation data, effectively reduces false alarms and missed alarms, accurately identifies anomaly types, locations of related equipment, and risk levels, provides reliable support for gas safety management, ensures the safety of residential gas use, and promotes the upgrading of gas safety management towards intelligence and refinement.

[0069] In a preferred embodiment of the present invention, step 1 above may include:

[0070] Step 1.1: Receive the device operation data stream, extract the timestamp information of each data node, and resample and align the timestamp information according to a unified sampling period to obtain a time-aligned data sequence. Specifically, this includes: receiving the device operation data stream uploaded by all gas IoT terminals within the community in real time. This data stream contains operating parameters such as gas pressure, flow rate, and temperature collected by each terminal, as well as the corresponding collection timestamp information for each parameter. Extract the timestamp information of each data node from each data entry, determine the collection time corresponding to each timestamp, and simultaneously calculate the original sampling period of all data nodes. Determine the maximum and minimum values ​​of the sampling period for each node, and calculate the unified sampling period. The unified sampling period is calculated using the least common multiple of all original sampling periods. The calculation formula is:

[0071] ;

[0072] in To standardize the sampling period, These represent the original sampling periods for each data node. This indicates the least common multiple operation. After determining the unified sampling period, the timestamp information of all data nodes is resampled. For data nodes with a sampling period shorter than the unified sampling period, the nearest neighbor interpolation method is used to supplement the missing sampling time data. For data nodes with a sampling period longer than the unified sampling period, redundant sampling data exceeding the unified sampling period is removed. After resampling, all data are sorted in order from earliest to latest according to the collection time corresponding to the timestamp, ensuring that the operating parameters of all data nodes correspond in the same time dimension, and finally obtaining a time-aligned data sequence.

[0073] Step 1.2 involves performing sliding window denoising on the time-aligned data sequence to filter out high-frequency abrupt noise points, resulting in a smoothed data sequence. Specifically, this includes: performing sliding window denoising on the time-aligned data sequence, determining the size of the sliding window based on the uniform sampling period and data acquisition frequency, using the following formula:

[0074] ;

[0075] in To adjust the sliding window size, For window coefficients, To standardize the sampling period, after the sliding window is determined, a sliding window mean filtering method is used to denoise the time-aligned data sequence. The mean of all data points within the sliding window is used as the filtered value of the center data point of the window. The window is moved sequentially to traverse the entire time-aligned data sequence. The specific calculation formula is as follows:

[0076] ;

[0077] in For the first The filtered values ​​of each data point To adjust the sliding window size, For the first in the sliding window The original values ​​of each data point To perform an integer operation on half the window size, the above filtering process removes high-frequency abrupt noise points in the time-aligned data sequence, eliminates data fluctuations caused by environmental interference and equipment errors, and obtains a smoothed data sequence, ensuring data stability and reliability.

[0078] Step 1.3 involves detecting data breakpoints in the smoothed data sequence and interpolating and filling them based on the changing trends of adjacent valid data segments before and after the breakpoints to obtain a continuous and complete data sequence. This continuous and complete data sequence is then used as a standardized time-series dataset. Specifically, this includes: detecting data breakpoints in the smoothed data sequence. A data breakpoint refers to a missing data segment caused by IoT terminal communication interruption, equipment failure, or other reasons. It is characterized by a time interval between two adjacent valid data points that is greater than twice the uniform sampling period. By traversing the smoothed data sequence, the time interval between two adjacent valid data points is calculated. If the time interval is greater than twice the uniform sampling period, a data breakpoint is determined to exist at that location. The calculation formula is as follows:

[0079] ;

[0080] in For the first The data point and the first +1 data point time interval, For the first The timestamp of each data point For the first +1 data point timestamp; when When a data breakpoint is detected, interpolation is performed based on the changing trends of adjacent valid data segments before and after the breakpoint. Linear interpolation is used for filling, and the rate of change of adjacent valid data points before and after the breakpoint is calculated using the following formula:

[0081] ;

[0082] in For the rate of change of data, The values ​​of the adjacent valid data points before the breakpoint. The values ​​of the adjacent valid data points after the breakpoint. and Given the corresponding timestamps, and based on a uniform sampling period, calculate the number of missing data points at the breakpoint using the following formula:

[0083] ;

[0084] in This represents the number of missing data points. To perform integer operations, based on the rate of change and the number of missing data points Calculate the value for each missing data point sequentially, using the following formula:

[0085] ;

[0086] in The index of the missing data point. For the first The values ​​of missing data points are used to fill in all detected data gaps, resulting in a continuous and complete data sequence. This data sequence eliminates the problems of time misalignment, high-frequency noise, and missing data, and is used as a standardized time series dataset.

[0087] In this embodiment of the invention, a series of processing steps, including timestamp alignment, sliding window denoising, and missing value interpolation, effectively solve the problems of time misalignment, high-frequency noise interference, and data loss in the data uploaded by gas IoT terminals in the background technology. This overcomes the technical defects of the prior art, such as false alarms and missed alarms in subsequent anomaly identification caused by the disorder and poor integrity of the data. Through specific processing methods such as unified sampling period, mean filtering, and linear interpolation, the accuracy, continuity, and stability of the standardized time series dataset are ensured, providing reliable data support for subsequent three-dimensional operating condition response surface construction, baseline generation, and anomaly feature extraction. This effectively improves the accuracy and reliability of the entire gas equipment operation data anomaly identification and processing method.

[0088] In a preferred embodiment of the present invention, step 2 above may include:

[0089] Step 2.1: Locate the real-time acquisition points of the pressure regulating station outlet junction node, the supporting public building pipeline inlet node, and the high-rise residential building end-meter node from the standardized time-series dataset. Extract the pressure, flow, and temperature values ​​of each node within the same sampling period to obtain a synchronous parameter sequence. Specifically, based on the standardized time-series dataset, perform node identifier matching on all acquisition points in the dataset to locate the real-time acquisition points of three key nodes: the pressure regulating station outlet junction node, the supporting public building pipeline inlet node, and the high-rise residential building end-meter node. These three nodes are the core monitoring nodes of the community's gas pipeline network, and their operating parameters directly reflect... After locating the overall pipeline network, the core operating parameters of the three nodes within the same sampling period are extracted. Three key parameters are extracted for each node: gas pressure, gas flow rate, and gas temperature. This ensures that the parameter acquisition time of the three nodes strictly corresponds to the unified sampling period. For each unified sampling period, the pressure, flow rate, and temperature values ​​of the three nodes are arranged sequentially according to the node order to form a set of three-dimensional parameter data. By traversing all unified sampling periods, all three-dimensional parameter data are concatenated in chronological order to obtain a synchronous parameter sequence. This sequence completely contains the core operating parameters of the three key nodes at each sampling time.

[0090] Step 2.2 involves spatially mapping the flow rate and temperature values ​​in the synchronization parameter sequence as the horizontal and vertical coordinates, and the pressure value as the elevation coordinate, to construct a three-dimensional operating condition response surface. Specifically, this includes: constructing a three-dimensional operating condition response surface for the synchronization parameter sequence using spatial mapping to visually present the coupling relationship of the operating parameters of the three key nodes, determining the coordinate system of the three-dimensional space, using the flow rate value in the synchronization parameter sequence as the horizontal axis coordinate of the three-dimensional space, the temperature value as the vertical axis coordinate of the three-dimensional space, and the pressure value as the elevation coordinate of the three-dimensional space. The three coordinates are independent and corresponding to each other, ensuring that the three-dimensional parameter data of each sampling period can correspond to a unique spatial point in the three-dimensional space. Spatial coordinate mapping is performed on all the three-dimensional parameter data in the synchronization parameter sequence, mapping the flow rate value, temperature value, and pressure value corresponding to each sampling period to the preset three-dimensional spatial coordinate system to obtain several discrete spatial data points.

[0091] A multi-scale neighborhood weighted iterative fitting algorithm is used to smooth the fitting of all discrete spatial data points. This algorithm achieves accurate fitting of discrete points and control of surface smoothness by dynamically dividing multi-scale neighborhoods and weighted iterative correction. The specific implementation process is as follows: multi-scale neighborhoods are divided for all discrete spatial data points. Three spherical neighborhoods of different scales are constructed with each discrete point as the center. The neighborhood radius increases geometrically, and the calculation formula is as follows:

[0092] ;

[0093] in For the first A neighborhood radius of a certain scale. The initial neighborhood radius is set to 1 / 20 of the range of the discrete point's x-coordinate. 2 is a scale increment coefficient to ensure the neighborhood coverage gradually expands, taking into account both the coupling relationships of local points and the overall operating trend. The weighting coefficients for each discrete point within its neighborhood at each scale are calculated. These weighting coefficients are inversely proportional to the spatial distance between the discrete points in the neighborhood and the central discrete point; the closer the distance, the larger the weighting coefficient. The specific calculation formula is as follows:

[0094] ;

[0095] In the formula, For the first The central discrete point corresponds to its i.e., ... the nth discrete point The weighted coefficients of discrete points in the neighborhood For the first The central discrete point and the first The three-dimensional spatial distance between discrete points in the neighborhood The total number of discrete points in the neighborhood at this scale is given by the denominator, which is the sum of the reciprocals of the distances between all discrete points in the neighborhood and the central discrete point. The sum of the weighting coefficients of all neighborhood discrete points is ensured to be 1. Iterative fitting calculations are performed based on these weighting coefficients. The weighted mean of each central discrete point in its neighborhood at each scale is calculated as the initial fitted value for that point. The deviation between the fitted value and the original discrete point elevation value is calculated for each iteration. If the deviation exceeds a preset fitting accuracy threshold, the weighting coefficients are adjusted and recalculated. The iteration terminates when the deviation is less than or equal to the accuracy threshold or the number of iterations reaches a preset maximum value. The accuracy threshold is calculated using the following formula:

[0096] ;

[0097] In the formula, This is the fitting accuracy threshold. The total number of discrete points. Let i be the original elevation value of the i-th discrete point. Let be the fitted elevation value for the k-th iteration, when Alternatively, when the number of iterations reaches 10, the iteration is stopped, and the final fitted elevation value is obtained. After the iterative fitting is completed, adjacent fitted discrete points are connected and surface interpolation is performed to fill the blank areas between discrete points, forming a continuous and smooth three-dimensional surface, which is the three-dimensional working condition response surface.

[0098] Step 2.3 involves calculating the curvature differential on the 3D operating condition response surface to extract surface deformation gradient features. Using these features as the basis for topology partitioning, the 3D operating condition response surface is divided into equal-domain segments to obtain multiple local response sub-regions. Specifically, this includes calculating the curvature differential on the constructed 3D operating condition response surface. The curvature differential reflects the degree of curvature at various locations on the surface, thereby extracting surface deformation gradient features. These features can reflect the local variations in pipeline operating conditions. The curvature differential calculation uses the 3D surface curvature calculation formula. For each spatial point on the 3D operating condition response surface, its curvature value is calculated. The specific calculation formula is as follows:

[0099] ;

[0100] in The curvature value of a spatial point on a three-dimensional surface; Let z be the second partial derivative of the elevation coordinate z with respect to the horizontal axis. Elevation coordinates Second partial derivative with respect to the vertical axis; The partial derivative of the elevation coordinate z with respect to the horizontal and vertical axes is given by . This is the first partial derivative of the elevation coordinate z with respect to the horizontal axis. Elevation coordinates After calculating the curvature values ​​of all spatial points on the surface using the first partial derivatives with respect to the vertical axis, the surface deformation gradient characteristics are calculated based on these curvature values. The deformation gradient characteristics are calculated using the gradient calculation formula, specifically:

[0101] ;

[0102] in These are the eigenvalues ​​of the surface deformation gradient. The curvature value, The x-axis coordinate is... The vertical axis is the coordinate. Let be the partial derivative of the curvature value with respect to the horizontal axis. is the partial derivative of the curvature value with respect to the vertical axis coordinate.

[0103] After extracting the surface deformation gradient features, these features are used as the basis for topology partitioning to divide the three-dimensional operating condition response surface into equal-domain segments. A gradient threshold is set, which is determined based on the normal operating condition fluctuation range of the community gas pipeline network. The calculation formula is as follows:

[0104] ;

[0105] in For gradient threshold, This represents the average of the deformation gradient eigenvalues ​​at all spatial points. Let the standard deviation be the deformation gradient eigenvalues ​​of all spatial points. As a threshold correction coefficient, the region on the three-dimensional working condition response surface with deformation gradient feature value less than or equal to the gradient threshold is divided into a local response sub-region. The region with deformation gradient feature value greater than the gradient threshold is further divided according to the gradient value. A local response sub-region is divided every gradient interval to ensure that the surface deformation gradient feature in each local response sub-region tends to be consistent, and finally multiple local response sub-regions are obtained.

[0106] Step 2.4: Calculate the pressure gradient integral and temperature attenuation coefficient within each local response sub-region. Perform a weighted mapping operation on the pressure gradient integral and temperature attenuation coefficient to obtain the hydraulic-thermal coupling compensation amount for the pipe network. Specifically, this includes: For each local response sub-region, calculate the pressure gradient integral and temperature attenuation coefficient within the region, and obtain the pipe network hydraulic-thermal coupling compensation amount through a weighted mapping operation. This compensation amount is used to correct subsequent baseline data and eliminate the influence of hydraulic-thermal coupling. Calculate the pressure gradient integral within each local response sub-region. The pressure gradient integral reflects the overall pressure change trend within the sub-region. The calculation formula is:

[0107] ;

[0108] In the formula, This is the integral value of the pressure gradient, used to quantify the overall drastic change in the pressure field within a local response sub-region; This represents the pressure value at a spatial point within a local response sub-region. Let x be the horizontal axis coordinate of the surface. The vertical coordinate of the surface; , These are the lower and upper limits of the horizontal axis coordinates of the sub-region, respectively. , These are the lower and upper limits of the ordinate of the sub-region, respectively; Pressure value The first partial derivative with respect to the flow rate; Pressure value The first partial derivative with respect to the temperature value is used to calculate the temperature decay coefficient in each local response sub-region. The temperature decay coefficient reflects the decay law of temperature with flow rate change in the sub-region, and the calculation formula is as follows:

[0109] ;

[0110] in This is the temperature decay coefficient; This represents the highest temperature value measured within the local response sub-region; This represents the lowest temperature value measured within the local response sub-region; This represents the maximum flow rate detected within the local response sub-region. This represents the minimum flow rate detected within a local response sub-region.

[0111] After calculating the integral value of the pressure gradient and the temperature attenuation coefficient, a weighted mapping operation is performed on the two to obtain the hydraulic-thermal coupling compensation amount of the pipe network. The calculation formula for the weighted mapping operation is as follows:

[0112] ;

[0113] in, This is the hydraulic-thermal coupling compensation amount for the pipeline network. These are the weighting coefficients for the integral value of the pressure gradient. This is the weighting factor for the temperature decay coefficient. and The sum of is 1, which is set according to the hydraulic and thermal characteristics of the community's gas pipeline network. The value is 0.6. The value is set to 0.4. The average value of the coupling compensation amount of all local response sub-regions is calculated to obtain the hydraulic and thermal coupling compensation amount of the entire pipe network.

[0114] In this embodiment of the invention, by locating three core nodes, constructing a three-dimensional operating condition response surface, segmenting local response sub-regions, and calculating the hydraulic-thermal coupling compensation amount of the pipeline network, the technical defects of existing anomaly identification technologies in the background art—namely, the lack of multi-parameter spatiotemporal coupling analysis capabilities and the failure to consider the influence of hydraulic-thermal coupling in the pipeline network—are effectively solved. This overcomes the problem that existing solutions cannot separate normal operating condition fluctuations from real anomaly characteristics. By accurately extracting the synchronization parameters of key nodes, quantifying the surface deformation characteristics, and calculating the coupling compensation amount, the hydraulic-thermal coupling law of the pipeline network can be reflected, providing a scientific compensation basis for baseline correction, effectively improving the accuracy of baseline data, and thus providing support for the accuracy of anomaly identification. This avoids false alarms and missed alarms caused by hydraulic-thermal coupling interference, promotes the upgrading of gas anomaly identification towards refinement and intelligence, and better adapts to the complex and dynamic gas usage conditions of residential communities.

[0115] In a preferred embodiment of the present invention, step 2 above may include:

[0116] Step 2.5: Perform sliding window statistical calculations on the standardized time-series dataset to extract the central trend value of each operating parameter within the sliding window, obtaining the initial trend data sequence. Specifically, this includes: performing sliding window statistical calculations on the standardized time-series dataset; the core purpose of this is to extract the central trend value of each operating parameter, filter out short-term minor fluctuations, obtain the initial trend data sequence reflecting the long-term operating trend of the pipeline network, and determine the size of the sliding window. The sliding window size is determined by combining the unified sampling period and the fluctuation period of the community gas operating parameters. The calculation formula is as follows:

[0117] ;

[0118] in, To adjust the sliding window size, The number of sampling periods per day. For the preset time span, To establish a uniform sampling period and ensure the window size covers short-term fluctuations while capturing trend changes, the standardized time-series dataset is slid sequentially through the window after the sliding window is determined. Statistical calculations are performed on each operating parameter within each window to extract the central trend value. The central trend value is calculated using a weighted average to avoid interference from extreme values. The specific calculation formula is as follows:

[0119] , , ;

[0120] in, The central trend value of the pressure values ​​within the window. The central trend value of the flow rate within the window. This represents the central trend value of the temperature values ​​within the window. For the first in the window The weighting coefficients for each data point decrease linearly from the center of the window towards both ends, satisfying the following... , , , The first in the window The pressure, flow, and temperature values ​​of each data point are sequentially slid across the entire standardized time-series dataset. The central trend values ​​of each window are then concatenated in chronological order to obtain the initial trend data sequence.

[0121] Step 2.6: Perform boundary constraint filtering on the initial trend data sequence to remove extreme points exceeding the preset physical reasonable range, obtaining the initial baseline data sequence; align the initial baseline data sequence with the hydraulic-thermal coupling compensation amount of the pipeline network on the time axis to obtain the time synchronization reference sequence. Specifically, this includes: performing boundary constraint filtering on the initial trend data sequence to remove extreme points exceeding the preset physical reasonable range. The preset physical reasonable range is determined based on the rated parameters of the community gas equipment, pipeline design standards, and historical normal operation data. Each operating parameter corresponds to an independent reasonable range, and the formula for calculating the upper and lower limits of the range is:

[0122] ;

[0123] ;

[0124] ;

[0125] ;

[0126] , ;

[0127] in, This refers to the rated pressure value of the gas equipment. , These are the upper and lower limits of the reasonable pressure range, respectively. This is the rated flow rate of the pipeline network. , These are the upper and lower limits of the reasonable flow range, respectively; This represents the upper limit of the reasonable temperature range. This is the lower limit of the reasonable temperature range, which is set in conjunction with the ambient temperature range and is expressed in degrees Celsius, to ensure coverage of normal temperature fluctuations under extreme weather conditions.

[0128] The initial trend data sequence is iterated through, and the pressure, flow, and temperature values ​​of each data point are compared with their corresponding reasonable intervals. If a parameter value exceeds the corresponding interval, it is identified as an extreme point and removed. After removing extreme points, the remaining data points are smoothed to obtain the initial baseline data sequence. The initial baseline data sequence is then aligned with the hydraulic-thermal coupling compensation amount of the pipeline network on the time axis. Since the timestamp of the initial baseline data sequence is consistent with the unified sampling period, and the hydraulic-thermal coupling compensation amount of the pipeline network is a global compensation value, it needs to be split according to the unified sampling period. Each sampling period corresponds to a compensation component. The splitting calculation formula is as follows:

[0129] ;

[0130] in, For the first Coupling compensation component for each sampling period, This is the overall hydraulic-thermal coupling compensation amount for the pipeline network. For the first Temperature deviation per sampling period The total number of sampling periods throughout the day is used to ensure that the compensation components correspond to the sampling periods. After the split is completed, the initial baseline data sequence and the coupling compensation components of each sampling period are aligned by timestamp to obtain the time synchronization reference sequence.

[0131] Step 2.7: Based on the time-synchronized reference sequence and the hydraulic-thermal coupling compensation amount of the pipeline network, parameter dimension matching is performed to construct the node compensation mapping set corresponding to each sampling node. Specifically, this includes: according to the parameter dimensions of the three key sampling nodes, each node contains three parameter dimensions: pressure, flow rate, and temperature, consistent with the parameter dimensions of the time-synchronized reference sequence. The compensation weight of each parameter dimension of each sampling node is calculated. The compensation weight is determined according to the node's location, function, and degree of hydraulic and thermal influence in the pipeline network. The pressure regulating station outlet confluence node, as the source of the pipeline network, has the highest pressure compensation weight; the high-rise residential terminal meter node is most affected by temperature, so the temperature compensation weight is the highest; the supporting public building pipeline network inflow node has large flow fluctuations, so the flow compensation weight is the highest. The specific weight calculation formula is as follows:

[0132] Voltage regulating station outlet junction node: , , ;

[0133] Supporting public building pipeline network access nodes: , , ;

[0134] High-rise residential building end-meter front node: , , ;

[0135] in , , The compensation weights are pressure, flow, and temperature parameters, respectively. The sum of the three weights for each node is 1. The system is dynamically calibrated based on the actual operating characteristics of the pipeline network. Based on the compensation weights of each node, and combined with the parameter values ​​of each node in the time synchronization reference sequence and the coupled compensation components of the corresponding sampling period, a node compensation mapping set is constructed.

[0136] Step 2.8 involves performing a deviation compensation superposition operation between the node compensation mapping set and the baseline values ​​of the corresponding sampling nodes in the time synchronization baseline sequence to obtain the corrected baseline sequence. Specifically, this includes calculating the deviation compensation value for each parameter dimension of each sampling node. The deviation compensation value is the product of the compensation weight of that parameter and the corresponding sampling period coupling compensation component. The calculation formula is as follows:

[0137] ; ; ;

[0138] in, , , These are the deviation compensation values ​​for pressure, flow rate, and temperature parameters, respectively. , , The parameter compensation weights for the corresponding nodes. For the coupling compensation component of this sampling period, the deviation compensation value is superimposed with the reference parameter value of the corresponding node in the time synchronization reference sequence to obtain the corrected parameter value. The calculation formula is as follows:

[0139] ; ; ;

[0140] in , , These are the corrected values ​​for pressure, flow rate, and temperature parameters, respectively. , , These are the baseline parameter values ​​in the time synchronization baseline sequence. The above deviation compensation superposition operation is performed on all sampling nodes and all sampling periods. The corrected parameter values ​​are then concatenated in time order and node order to obtain the corrected baseline sequence.

[0141] In this embodiment of the invention, by using sliding window statistical extraction of initial trends, boundary constraint filtering of extreme values, time axis alignment, and deviation compensation superposition, the technical defects of existing anomaly identification technologies in the background technology, such as fixed and rigid baseline models, failure to consider the influence of hydraulic and thermal coupling of pipeline networks, and insufficient baseline accuracy, are effectively solved. This overcomes the problems of existing baselines being unable to adapt to complex dynamic operating conditions and being susceptible to interference from extreme values. Through accurate trend extraction, extreme value removal, and coupling compensation, the corrected baseline sequence can dynamically adapt to the coupling changes of gas usage time periods, seasonal temperatures, and equipment loads, accurately reflecting the normal operating status of the pipeline network. This provides high-precision benchmark support for subsequent multi-parameter dynamic normal interval baseline construction and anomaly identification, effectively reducing false alarms and missed alarms caused by baseline deviations, further improving the accuracy and reliability of anomaly identification of gas equipment operating data, and adapting to the complex and dynamic gas usage conditions of residential communities.

[0142] In a preferred embodiment of the present invention, step 3 above may include:

[0143] Step 3.1: Perform time window slicing on the corrected baseline sequence, extract baseline parameter values ​​within each time window to obtain a baseline feature slice set; numerically encode and normalize the current gas consumption period features, seasonal temperature features, and equipment type features respectively to obtain time period encoding vectors, temperature encoding vectors, and equipment type encoding vectors. Specifically, this includes: performing time window slicing on the corrected baseline sequence, extracting baseline parameter values ​​within each time window to obtain a baseline feature slice set; numerically encoding and normalizing the current gas consumption period features, seasonal temperature features, and equipment type features respectively to obtain time period encoding vectors, temperature encoding vectors, and equipment type encoding vectors. Specifically, this includes: numerically encoding and normalizing the current gas consumption period features, seasonal temperature features, and equipment type features respectively to obtain a time period encoding vector, temperature encoding vector, and equipment type encoding vector. The linear sequence is sliced ​​into time windows. Based on a unified sampling period, continuous time windows are divided according to a fixed duration. The baseline sequence is traversed and corrected, and the window segmentation is completed sequentially. The mean, gradient, and discrete features of three baseline parameters (pressure, flow rate, and temperature) are extracted within each time window. All baseline parameter features within the same time window are integrated to obtain a baseline feature slice set. The current gas consumption period features, seasonal temperature features, and equipment type features are encoded into ordered integers and mapped to their corresponding discrete values. The encoded features are then subjected to hyperbolic tangent nonlinear scaling normalization to nonlinearly compress the feature values ​​to a stable range while preserving the relative differences and gradients between features. The calculation formula is as follows:

[0144] ;

[0145] In the formula The original numerical value after feature encoding; The global mean of the numerical values ​​encoded for this type of feature; The global standard deviation of the numerical encoding for this type of feature; These are the characteristic values ​​after nonlinear standardization; through the above processing, the time period encoding vector, temperature encoding vector, and equipment type encoding vector are obtained respectively.

[0146] Step 3.2 involves concatenating the time period encoding vector, temperature encoding vector, and equipment type encoding vector to obtain a context feature vector set. The baseline feature slice set and the context feature vector set are then cross-aligned and concatenated along the timestamps and dimensions to construct a multi-dimensional context feature matrix. Specifically, this includes: horizontally concatenating the time period encoding vector, temperature encoding vector, and equipment type encoding vector corresponding to the same time point to obtain a single fused feature vector; traversing all sampling times to obtain a complete context feature vector set; cross-aligning the baseline feature slice set and the context feature vector set point-by-point according to the same timestamp to ensure strict correspondence between baseline features and context features at the same sampling time; and sequentially concatenating the aligned baseline features and context features along the dimensional direction, using the time series as the vertical axis and all feature dimensions as the horizontal axis to construct a multi-dimensional context feature matrix. This matrix simultaneously carries the corrected baseline parameter features, gas usage time period features, seasonal temperature features, and equipment type features.

[0147] Step 3.3 involves performing iterative weight optimization on the multidimensional context feature matrix. This dynamically adjusts the baseline parameter dimensions mapped to the corresponding baseline feature slice set, as well as the fitting coefficients for the time period encoding dimension, temperature encoding dimension, and device type encoding dimension mapped to the corresponding context feature vector set, to obtain optimized fitting coefficients. Specifically, this includes performing iterative weight optimization on the multidimensional context feature matrix, with the optimization objective of minimizing the sum of squared global feature fitting errors, and dynamically adjusting the fitting coefficients for each feature dimension. First, a global fitting error function is constructed, where the error is the sum of the squared differences between the fitted outputs of all feature dimensions within the matrix and the actual baseline features. The calculation formula is as follows:

[0148] ;

[0149] in The actual baseline characteristic value, These are the currently fitted feature values. To minimize the global fitting error, the fitting coefficients for the baseline parameter dimension, time period encoding dimension, temperature encoding dimension, and device type encoding dimension are iteratively updated round by round. In each round, the coefficient value is adjusted in reverse based on the current error. The calculation formula is as follows:

[0150] ;

[0151] in The fitting coefficients for the current round are... These are the fitting coefficients for the next round. The iteration step size, This is the gradient value of the global error with respect to the current coefficients. When the global fitting error is less than the preset convergence threshold or the number of iterations reaches the preset upper limit, the iteration stops and the optimized fitting coefficients are output.

[0152] Step 3.4 involves performing a nonlinear mapping operation on the multidimensional context feature matrix based on the optimized fitting coefficients to obtain a multi-parameter dynamic normal interval baseline. Specifically, this includes performing a second-order cross-coupling nonlinear mapping operation on the multidimensional context feature matrix based on the optimized fitting coefficients, reflecting the coupling influence and nonlinear change patterns among multiple features. The calculation formula is as follows:

[0153] ;

[0154] in, For dynamic center baseline value, to These are the optimized fitting coefficients. , For the feature values ​​of different dimensions in the multidimensional context feature matrix, the above nonlinear mapping is performed on the three types of operating parameters, namely pressure, flow rate and temperature, to obtain their respective dynamic center baselines. Based on this, the dynamic floating threshold is calculated in combination with the feature dispersion. The dynamic center baseline and the floating threshold are added to obtain the upper limit of the normal interval, and subtracted to obtain the lower limit of the normal interval. The center baseline, upper limit and lower limit of the interval at the same time constitute the dynamic normal interval at that time. The dynamic normal intervals at all times are arranged in chronological order and fitted to obtain the multi-parameter dynamic normal interval baseline.

[0155] In this embodiment of the invention, because the invention employs time window slicing, hyperbolic tangent nonlinear standardization, multi-feature splicing to construct a multi-dimensional context feature matrix, iterative weight optimization based on the sum of squared errors, and secondary cross-coupling nonlinear mapping, it overcomes the technical problems of rigid and fixed baselines in existing technologies, inability to adapt to time periods, coupling changes between seasons and equipment types, and low operating condition matching due to overly simplistic linear processing. This enables dynamic adaptive adjustment of the normal operating range with the external environment, significantly improving baseline accuracy and operating condition adaptability, and effectively reducing false alarms during peak periods and missed alarms for hidden anomalies.

[0156] In a preferred embodiment of the present invention, step 4 above may include:

[0157] Step 4.1: Align the standardized time-series dataset with the multi-parameter dynamic normal range baseline point by point. Calculate the deviation between the actual operating parameter values ​​at each sampling node and the upper and lower limits of the corresponding baseline range. Filter out continuous data segments whose deviation exceeds the preset tolerance threshold to obtain the out-of-limit data segments. Specifically, this includes: aligning the standardized time-series dataset with the multi-parameter dynamic normal range baseline point by point at the same sampling time and the same sampling node to ensure that the actual operating parameters correspond to the dynamic baseline range. Calculate the relative deviation of the three types of actual operating parameters (pressure, flow rate, and temperature) relative to the corresponding dynamic baseline range. The relative deviation reflects the degree to which the actual parameters deviate from the normal range. The calculation formula is:

[0158] ;

[0159] in This is the relative deviation. These are the actual operating parameter values. The center value of the dynamic normal interval. This is the upper limit of the dynamic normal range. The lower limit of the dynamic normal range is set, and a relative deviation tolerance threshold is preset. All sampling points are traversed, and sampling points with relative deviations greater than the tolerance threshold or less than the negative tolerance threshold are marked as out-of-limit points. The continuity of out-of-limit points is determined. If multiple out-of-limit points are continuously distributed on the time axis and the number of continuous points is not less than the preset minimum number of continuous points, then the continuous data segment is determined as a valid out-of-limit segment. All valid out-of-limit segments together constitute an out-of-limit data segment.

[0160] Step 4.2 involves performing multi-parameter time-series synchronous extraction on the out-of-limit data segments, separating the corresponding pressure time-series subseries, flow time-series subseries, and temperature time-series subseries within the segments. Specifically, this includes: performing multi-parameter time-series synchronous extraction on the out-of-limit data segments according to a unified time axis, ensuring that the sampling times of the three types of parameters (pressure, flow, and temperature) are completely consistent and that the time-series lengths are the same; separating the pressure parameter sequence of the corresponding sampling node from the out-of-limit data segments time-by-time to form a pressure time-series subseries; separating the flow parameter sequence to form a flow time-series subseries; and separating the temperature parameter sequence to form a temperature time-series subseries. The three sets of subseries have equal lengths and completely aligned timestamps.

[0161] Step 4.3: Perform pairwise dynamic cross-correlation calculations on the pressure time series, flow rate time series, and temperature time series to obtain the calculation results. Based on the calculation results, construct a parameter coupling deviation index set, specifically including: pairwise combinations of the pressure time series, flow rate time series, and temperature time series, namely pressure and flow rate combination, pressure and temperature combination, and flow rate and temperature combination, and perform dynamic time series cross-correlation calculations on each combination sequence. The cross-correlation calculation considers small time series offsets and can truly reflect the hysteresis coupling characteristics between multiple parameters of the gas pipeline network. The formula for calculating the cross-correlation coefficient of a single group is:

[0162] ;

[0163] in, The timing offset is Cross-correlation coefficients at time , For two sets of time series subsequences at time... The value, , The mean of the two subsequences is given. Given the time series offset, iterate through all reasonable offsets, take the maximum cross-correlation coefficient for each combination, and calculate the parameter coupling deviation based on the maximum cross-correlation coefficient. The calculation formula is as follows:

[0164] ;

[0165] By combining the pressure-flow coupling deviation, pressure-temperature coupling deviation, and flow-temperature coupling deviation, a parameter coupling deviation index set is constructed.

[0166] Step 4.4 involves performing feature dimensionality reduction and vectorization reorganization on the parameter coupling deviation index set and the duration and peak deviation amplitude of the out-of-limit data segments to obtain anomaly feature vectors. Specifically, this includes calculating the duration and peak deviation amplitude of the out-of-limit data segments, where the duration is the total time span of the out-of-limit segment and the peak deviation amplitude is the absolute value of the maximum relative deviation within the segment. Weighted feature fusion dimensionality reduction is then performed on the parameter coupling deviation index set, duration, and peak deviation amplitude. Each feature is assigned a corresponding weight and linearly fused, compressing the multi-dimensional features into a single-dimensional feature sequence. The calculation formula is as follows:

[0167] ;

[0168] in, To fuse feature values, , , There are three sets of coupling deviations. For duration, This represents the deviation from the peak value. to To normalize the weight coefficients and ensure that the sum of all weights is 1, the fused feature values ​​at all times are arranged in chronological order and then vectorized and recombined to obtain anomaly feature vectors with unified dimensions and complete representation.

[0169] In this embodiment of the invention, because the invention employs techniques such as point-by-point relative deviation exceeding the limit judgment, elimination of isolated exceeding the limit points, multi-parameter time-series synchronous separation, dynamic time-series cross-correlation calculation of coupling deviation degree, and weighted feature fusion dimensionality reduction, it overcomes the technical problems of existing technologies that only use single-parameter exceeding the limit judgment, lack multi-parameter coupling correlation analysis, are susceptible to instantaneous interference and misjudgment, and cannot identify latent anomalies. Thus, it achieves accurate extraction and feature quantification of abnormal data, can fully characterize the magnitude, duration and coupling disruption degree of anomalies, and improves the distinguishability and reliability of abnormal features.

[0170] In a preferred embodiment of the present invention, step 5 above may include:

[0171] Step 5.1: Extract the prototype feature vectors of each standard anomaly mode in the preset gas anomaly mode library, and calculate the spatial similarity distance between the anomaly feature vector and each prototype feature vector; select the standard anomaly mode with the highest spatial similarity distance as the target matching mode to obtain the anomaly matching result. Specifically, this includes: extracting the prototype feature vectors corresponding to all standard anomaly modes in the mode library, calculating the spatial similarity distance between each one and the anomaly feature vector, and using the weighted Euclidean spatial similarity distance calculation method. This method assigns different weights to each feature component based on their importance to anomaly identification, and sets weight coefficients for each of the five feature components. The weight coefficients for pressure-flow coupling deviation are set to 0.3, pressure-temperature coupling deviation to 0.25, flow-temperature coupling deviation to 0.25, duration of exceeding limits to 0.1, and peak deviation amplitude to 0.1. The sum of all weight coefficients is 1. This can be dynamically calibrated according to the actual operating characteristics of the community gas pipeline network. The formula for calculating the weighted Euclidean spatial similarity distance is:

[0172] ;

[0173] In the formula, This represents the spatial similarity distance between the anomalous feature vector and the prototype feature vector. to These are the weight coefficients for the five feature components. to These are the values ​​of the five feature components of the abnormal feature vector. to These represent the five feature components of the prototype feature vector. The smaller the spatial similarity distance, the more similar the anomalous feature vector is to the standard anomalous pattern corresponding to the prototype feature vector, and the higher the degree of matching.

[0174] The spatial similarity distance between the abnormal feature vector and all prototype feature vectors in the pattern library is calculated sequentially to obtain a set of distance values. This set of distance values ​​is then iterated through, and the prototype feature vector corresponding to the smallest distance value is selected. The standard abnormal pattern corresponding to this prototype feature vector is determined as the target matching pattern. The pattern matching degree is calculated; it is negatively correlated with the spatial similarity distance. The calculation formula is as follows:

[0175] ;

[0176] In the formula, For pattern matching degree, The minimum spatial similarity distance. The maximum similarity distance threshold is preset, and the pattern matching score ranges from 0 to 1. The closer the value is to 1, the higher the matching accuracy. Finally, anomaly matching results are obtained, which include the target matching pattern identifier and the pattern matching score.

[0177] Step 5.2: Based on the anomaly matching results, locate the associated gas equipment locations and retrieve the historical operation records corresponding to the gas equipment locations; extract the occurrence frequency and deterioration rate parameters of the target matching pattern in the historical operation records, and construct a historical evolution trend vector. Specifically, based on the anomaly matching results, locate the associated gas equipment locations according to the anomaly type of the target matching pattern. Different types of anomalies correspond to different associated equipment. For example, pipeline leakage anomalies correspond to the inlet node of the supporting public building pipeline network or the end meter node of high-rise residential buildings; pressure drop anomalies correspond to the outlet inlet node of the pressure regulating station; and flow surge anomalies correspond to the inlet node of the supporting public building pipeline network. Through the correlation mapping relationship between anomaly type and equipment location, accurately locate the associated equipment location where the anomaly occurred, and clarify whether it is the outlet inlet node of the pressure regulating station, the inlet node of the supporting public building pipeline network, or the end meter node of high-rise residential buildings, as well as the specific equipment number, installation location, and other detailed information corresponding to the node.

[0178] After location is established, all historical operation records corresponding to the location of the associated device are retrieved. These records contain all operational data, anomaly records, and maintenance records since the device was put into use. The focus is on extracting two key parameters from the historical operation records: the frequency of occurrence and the rate of deterioration of the target matching pattern. These parameters are used to construct a historical evolution trend vector, reflecting the historical occurrence patterns and dynamic development trends of this type of anomaly on the device. The specific process for calculating the frequency is as follows: A historical statistical period is set, determined based on the device's operating years and anomaly occurrence patterns, typically set to 6 months. The total number of times the target matching pattern appears on the associated device within this statistical period is extracted. This total number is divided by the total historical statistical duration to obtain the anomaly occurrence frequency per unit time. If the anomaly does not occur within the historical statistical period, the frequency is set to 0. If the statistical period is less than 6 months, the frequency is calculated based on the actual statistical duration to ensure the rationality of the frequency calculation.

[0179] The specific process for calculating the deterioration rate is as follows: Retrieve the peak deviation and occurrence time of each occurrence of the target matching mode from the historical operation records of the associated device, select two adjacent abnormal records, and calculate using the following formula:

[0180] ;

[0181] in, For abnormal deterioration rate, This represents the deviation of the peak value from the previous anomaly. This represents the peak deviation magnitude of the subsequent anomaly. This refers to the time when the previous anomaly occurred. The deterioration rate is set to 0 if the anomaly occurs only once in the historical records. If the peak deviation of two adjacent anomalies shows a decreasing trend, it indicates that the anomaly is less severe and the deterioration rate is negative. If it shows an increasing trend, it indicates that the anomaly is more severe and the deterioration rate is positive. The calculated occurrence frequency and deterioration rate are combined in a fixed order with occurrence frequency first and deterioration rate second to obtain a two-dimensional historical evolution trend vector.

[0182] Step 5.3 involves inputting the pattern matching degree and historical evolution trend vector from the anomaly matching results into a preset risk confidence assessment model for weighted fusion calculation to obtain a risk confidence value. Specifically, this includes: a preset risk confidence assessment model, which is built based on gas anomaly risk assessment standards, historical anomaly handling data, and equipment operating characteristics. Its function is to perform weighted fusion calculation of the pattern matching degree and historical evolution trend vector from the anomaly matching results to obtain a risk confidence value that quantifies the degree of anomaly risk. Weight coefficients are assigned to each input parameter in the model, determined based on the degree of influence of each parameter on the risk assessment. Specifically, the pattern matching degree weight coefficient is set to 0.5, the occurrence frequency weight coefficient is set to 0.3, the deterioration rate weight coefficient is set to 0.2, and the sum of all weight coefficients is 1. The model is dynamically calibrated according to the community's gas safety management requirements to ensure that the weight allocation meets the actual risk assessment needs.

[0183] The pattern matching degree and historical evolution trend vector are input together into the preset risk confidence assessment model, and the calculation is performed according to the weighted fusion calculation method. The calculation formula is as follows:

[0184] ;

[0185] in This is the risk confidence level, ranging from 0 to 1. , , These are the weighting coefficients for pattern matching degree, occurrence frequency, and deterioration rate, respectively. For pattern matching degree, For frequency of occurrence, The risk confidence score directly reflects the degree of risk of the current anomaly. The closer the score is to 1, the higher the anomaly matching accuracy, the more frequent the recurrence, the faster the deterioration, and the higher the risk. The closer the score is to 0, the lower the anomaly matching accuracy, the lower the recurrence probability, the no obvious deterioration trend, and the lower the risk.

[0186] During the calculation process, range constraints are applied to each input parameter to ensure the reasonableness of the calculation results: the pattern matching degree is constrained to a range of 0 ≤ ≤1; if the calculation result exceeds this range, it will be automatically corrected to the nearest boundary value; the frequency range is constrained to 0≤ ≤5; if it exceeds this range, take the value of 5; the range of the deterioration rate is constrained to -0.5≤. If the value is ≤0.5, and it exceeds this range, the corresponding boundary value shall be used to ensure the stability and reliability of the risk confidence value.

[0187] Step 5.4: Based on the risk confidence value and the preset risk level classification threshold, perform interval mapping to obtain the risk level determination result; assemble the risk level determination result with the anomaly category identifier and associated equipment location of the target matching pattern to obtain the anomaly identification result containing anomaly type, associated equipment location, and risk level. Specifically, this includes: preset multi-level risk level classification thresholds, and combining gas safety management standards, anomaly handling priorities, and the operating characteristics of community gas equipment, dividing the risk confidence value into four continuous risk level intervals. The specific threshold settings are as follows: the low-risk threshold interval is 0 ≤ <0.3, the medium-risk threshold range is 0.3≤ <0.6, the high-risk threshold range is 0.6≤ <0.8, the critical risk threshold range is 0.8≤ ≤1.0, each risk level corresponds to a different priority for handling: low risk corresponds to routine inspection, medium risk corresponds to key monitoring, high risk corresponds to emergency investigation, and critical risk corresponds to immediate handling.

[0188] The risk confidence level is compared with the preset risk level classification thresholds one by one to determine the risk level corresponding to the current anomaly. The specific determination process is as follows: if the risk confidence level is in the low risk threshold range, it is determined to be low risk; if it is in the medium risk threshold range, it is determined to be medium risk; if it is in the high risk threshold range, it is determined to be high risk; if it is in the critical risk threshold range, it is determined to be critical risk. A clear risk level determination result is obtained. After the risk level determination is completed, the anomaly identification results are assembled. The assembled content includes three pieces of information: the anomaly category identifier corresponding to the target matching pattern, which clarifies the specific type of the current anomaly, such as pipeline leakage, pressure drop, abnormal flow surge, etc., and supplements a brief description of the anomaly, explaining the typical characteristics and potential hazards of this type of anomaly; the associated gas equipment location information, which clarifies the specific node where the anomaly occurred, equipment number, installation location, and other detailed information; and the risk level determination result, which clarifies the risk level of the current anomaly and the corresponding handling priority. During the data assembly process, the completeness and accuracy of all information are ensured, and the anomaly category identifier and equipment location information are standardized and coded. Finally, a complete anomaly identification result containing anomaly type, associated equipment location, and risk level is obtained.

[0189] In this embodiment of the invention, because the invention employs a pre-set complete gas anomaly pattern library, uses weighted Euclidean space similarity distance for anomaly feature matching, extracts occurrence frequency and deterioration rate from historical operation files to construct a historical evolution trend vector, calculates risk confidence based on weighted fusion, and determines risk level and completes data assembly through multi-level threshold interval mapping, it overcomes the technical defects of existing technologies that can only achieve simple limit violation prompts, are difficult to accurately identify anomaly types, cannot accurately locate related equipment, lack historical trend assessment and risk level quantification, and lack scientific basis for anomaly handling. This achieves accurate anomaly pattern matching, accurate fault equipment location, and scientific quantification of anomaly risks, comprehensively improving the intelligence, refinement, and reliability of gas equipment operation data anomaly identification. It provides comprehensive, accurate, and operable judgment basis for gas safety early warning, fault investigation, and emergency response, effectively reducing the probability of gas safety accidents and ensuring the safety of residential gas use.

[0190] like Figure 2 As shown, embodiments of the present invention also provide a gas equipment operation data anomaly identification and processing system, including:

[0191] The acquisition module is used to receive the device operation data stream uploaded by the community gas IoT terminal in real time, and perform timestamp alignment, noise filtering and missing value interpolation on the device operation data stream to obtain a standardized time series dataset.

[0192] The module is used to extract the synchronous parameter sequences of the outlet confluence node of the pressure regulating station, the inflow node of the supporting public building pipeline network, and the terminal meter node of the high-rise residential building from the standardized time series dataset, and construct a three-dimensional operating condition response surface; perform curvature differentiation and topological equal domain division on the three-dimensional operating condition response surface to obtain the hydraulic-thermal coupling compensation amount of the pipeline network; perform dynamic baseline generation processing on the standardized time series dataset to obtain the initial baseline data sequence; and perform error compensation superposition on the initial baseline data sequence and the hydraulic-thermal coupling compensation amount of the pipeline network to obtain the corrected baseline sequence.

[0193] The fitting module is used to construct a multi-dimensional context feature matrix by combining the corrected baseline sequence with the current gas consumption period characteristics, seasonal temperature characteristics, and equipment type characteristics. Nonlinear adaptive fitting is performed on the multi-dimensional context feature matrix to obtain a multi-parameter dynamic normal interval baseline.

[0194] The analysis module is used to use the multi-parameter dynamic normal interval baseline as a dynamic judgment benchmark to perform limit screening on the standardized time series dataset to extract limit-exceeding data segments; and to perform multi-parameter correlation analysis on the limit-exceeding data segments to obtain abnormal feature vectors.

[0195] The mapping module is used to perform feature mapping matching between abnormal feature vectors and a preset gas anomaly pattern library to obtain anomaly matching results; the anomaly matching results are combined with historical evolution trends to calculate risk confidence, and anomaly identification results including anomaly type, associated equipment location and risk level are obtained.

[0196] It should be noted that this system is a system corresponding to the above method. All implementation methods in the above method embodiments are applicable to this embodiment and can achieve the same technical effect.

[0197] Embodiments of the present invention also provide a computing device, including: a processor and a memory storing a computer program, wherein the computer program, when executed by the processor, performs the method described above. All implementations in the above method embodiments are applicable to this embodiment and can achieve the same technical effects.

[0198] Embodiments of the present invention also provide a computer-readable storage medium storing instructions that, when executed on a computer, cause the computer to perform the method described above. All implementations in the above method embodiments are applicable to this embodiment and can achieve the same technical effects.

[0199] Experimental example:

[0200] Example 1: Overall Implementation Process:

[0201] This embodiment was verified in the gas supply system of a comprehensive residential community in a city. The area covers 12 high-rise residential buildings, 3 supporting commercial and catering buildings and 1 centralized pressure regulating station, serving more than 2,500 users, with an average daily gas consumption of about 4,500 cubic meters.

[0202] like Figure 3 As shown: Equipment operation data preprocessing: Real-time reception of equipment operation data streams uploaded by community gas IoT terminals, including parameters such as pressure, flow rate, and temperature, with a sampling period of 1 minute. The data stream is processed by timestamp alignment, sliding window denoising, and missing value interpolation to obtain a standardized time-series dataset. Filtering effectively eliminates high-frequency noise in the original data.

[0203] like Figure 4 As shown: Dynamic baseline generation and compensation involves performing sliding window statistical calculations on a standardized time-series dataset, extracting the central trend value as the initial baseline, and combining it with the hydraulic and thermal coupling compensation of the pipeline network to perform error compensation superposition, resulting in a corrected baseline sequence. The corrected baseline can better track the dynamic changes of actual data.

[0204] like Figure 5As shown: Multi-parameter dynamic baseline fitting constructs a multi-dimensional contextual feature matrix by combining the corrected baseline sequence with current gas consumption period characteristics, seasonal temperature characteristics, and equipment type characteristics. Nonlinear adaptive fitting is then performed. Experimental results show that the baseline curves for the morning peak, noon peak, and evening peak periods exhibit significant differences.

[0205] like Figure 6 As shown: Abnormal pattern matching and risk assessment, the abnormal feature vector is matched with a preset gas abnormal pattern library, and the risk confidence is calculated by combining historical evolution trends. Compared with the traditional fixed threshold method, the method of the present invention has improved in terms of detection rate, accuracy and real-time performance.

[0206] Beneficial effects:

[0207] This embodiment verifies the effectiveness of the method for identifying and processing abnormal operating data of gas equipment. Through the synergistic effect of steps such as data preprocessing, three-dimensional operating condition response surface construction, dynamic baseline generation and compensation, multi-parameter dynamic baseline fitting, abnormal feature vector extraction, abnormal pattern matching and risk judgment, the method achieves intelligent and refined identification of abnormal gas equipment. Experimental results show that this method effectively improves the level of intelligence in gas safety management compared with the traditional fixed threshold method.

[0208] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications should also be considered within the scope of protection of the present invention.

Claims

1. A method for identifying and processing abnormal operating data of gas equipment, characterized in that, The method includes: The system receives real-time data streams of equipment operation from the community gas IoT terminal, performs timestamp alignment, noise filtering, and missing value interpolation on the data streams to obtain a standardized time-series dataset. Real-time data collection points were located at the outlet confluence node of the pressure regulating station, the inflow node of the supporting public building pipeline network, and the terminal meter node of the high-rise residential building from the standardized time-series dataset. Pressure, flow, and temperature values ​​of each node within the same sampling period were extracted to obtain a synchronous parameter sequence. The flow and temperature values ​​in the synchronous parameter sequence were used as spatial horizontal and vertical coordinates, and the pressure value was used as elevation coordinates for spatial mapping to construct a three-dimensional operating condition response surface. Curvature differential calculations were performed on the three-dimensional operating condition response surface to extract surface deformation gradient features. Using the surface deformation gradient features as the basis for topological partitioning, the three-dimensional operating condition response surface was divided into equal-domain segments to obtain multiple local response sub-regions. The pressure gradient integral and temperature attenuation coefficient within each local response sub-region were calculated, and a weighted mapping operation was performed on the pressure gradient integral and temperature attenuation coefficient to obtain the pipeline network hydraulic-thermal coupling compensation amount. The standardized time-series dataset was subjected to dynamic baseline generation processing to obtain an initial baseline data sequence. The initial baseline data sequence was superimposed with the pipeline network hydraulic-thermal coupling compensation amount to obtain a corrected baseline sequence. A multi-dimensional context feature matrix is ​​constructed by combining the corrected baseline sequence with the characteristics of the current gas consumption period, seasonal temperature characteristics, and equipment type characteristics. Nonlinear adaptive fitting is then performed on the multi-dimensional context feature matrix to obtain a multi-parameter dynamic normal interval baseline. Using the multi-parameter dynamic normal interval baseline as a dynamic judgment benchmark, the standardized time series dataset is screened for exceeding limits to extract exceeding data segments; multi-parameter correlation analysis is performed on the exceeding data segments to obtain abnormal feature vectors; The abnormal feature vectors are matched with a pre-defined gas anomaly pattern library to obtain anomaly matching results. The anomaly matching results are then combined with historical evolution trends to calculate risk confidence, resulting in anomaly identification results that include anomaly type, associated equipment location, and risk level. The formula for calculating the pressure gradient integral value is as follows: ； In the formula, This is the integral value of the pressure gradient; This represents the pressure value at a spatial point within a local response sub-region. Let x be the horizontal axis coordinate of the surface. The vertical coordinate of the surface; , These are the lower and upper limits of the horizontal axis coordinates of the sub-region, respectively. , These are the lower and upper limits of the ordinate of the sub-region, respectively; Pressure value The first partial derivative with respect to the flow rate; Pressure value The first partial derivative with respect to temperature; the formula for the temperature decay coefficient is: ,in This is the temperature decay coefficient; This represents the highest temperature value measured within the local response sub-region; This represents the lowest temperature value measured within the local response sub-region; This represents the maximum flow rate detected within the local response sub-region. This represents the minimum flow rate detected within a local response sub-region.

2. The method for identifying and processing abnormal operating data of gas equipment according to claim 1, characterized in that, The system receives real-time equipment operation data streams uploaded by the community gas IoT terminal, performs timestamp alignment, noise filtering, and missing value interpolation on the data streams to obtain a standardized time-series dataset, including: Receive the device operation data stream, extract the timestamp information of each data node, resample and sort the timestamp information according to a unified sampling period, and obtain a time-aligned data sequence. Sliding window denoising is applied to the time-aligned data sequence to filter out high-frequency abrupt noise points, resulting in a smoothed data sequence. By detecting data breakpoints in the smoothed data sequence, and performing interpolation and filling based on the changing trends of adjacent valid data segments before and after the breakpoint, a continuous and complete data sequence is obtained, which is then used as a standardized time series dataset.

3. The method for identifying and processing abnormal operating data of gas equipment according to claim 2, characterized in that, The standardized time-series dataset is subjected to dynamic baseline generation processing to obtain the initial baseline data sequence; The initial baseline data sequence is superimposed with the network hydraulic-thermal coupling compensation amount to obtain the corrected baseline sequence, including: Perform sliding window statistical calculations on the standardized time series dataset, extract the central trend value of each running parameter within the sliding window, and obtain the initial trend data sequence; Boundary constraint filtering is performed on the initial trend data sequence to remove extreme points that exceed the preset physical reasonable range, thus obtaining the initial baseline data sequence; the initial baseline data sequence is then aligned with the hydraulic-thermal coupling compensation amount of the pipeline network on the time axis to obtain the time synchronization reference sequence. Based on the time synchronization reference sequence and the hydraulic and thermal coupling compensation amount of the pipeline network, parameter dimension matching is performed to construct the node compensation mapping set corresponding to each sampling node. The corrected baseline sequence is obtained by performing a deviation compensation superposition operation between the node compensation mapping set and the reference value of the corresponding sampling node in the time synchronization reference sequence.

4. The method for identifying and processing abnormal operating data of gas equipment according to claim 3, characterized in that, A multi-dimensional context feature matrix is ​​constructed by combining the corrected baseline sequence with current gas consumption period characteristics, seasonal temperature characteristics, and equipment type characteristics. Nonlinear adaptive fitting is then performed on this multi-dimensional context feature matrix to obtain a multi-parameter dynamic normal interval baseline, including: The corrected baseline sequence is sliced ​​into time windows to extract the baseline parameter values ​​within each time window, resulting in a baseline feature slice set. The current gas consumption period features, seasonal temperature features, and equipment type features are numerically encoded and normalized to obtain the period encoding vector, temperature encoding vector, and equipment type encoding vector, respectively. The time period encoding vector, temperature encoding vector and device type encoding vector are concatenated to obtain the context feature vector set; the baseline feature slice set and the context feature vector set are cross-aligned and concatenated according to timestamp and dimension to construct a multi-dimensional context feature matrix. Iterative weight optimization is performed on the multidimensional context feature matrix to dynamically adjust the baseline parameter dimension of the corresponding baseline feature slice set mapping in the multidimensional context feature matrix, as well as the fitting coefficients of the time period encoding dimension, temperature encoding dimension and device type encoding dimension of the corresponding context feature vector set mapping, to obtain the optimized fitting coefficients. Based on the optimized fitting coefficients, a nonlinear mapping operation is performed on the multidimensional context feature matrix to obtain a multi-parameter dynamic normal interval baseline.

5. The method for identifying and processing abnormal operating data of gas equipment according to claim 4, characterized in that, Using the multi-parameter dynamic normal interval baseline as a dynamic judgment benchmark, the standardized time series dataset is screened for exceeding limits to extract data fragments that exceed limits. Multi-parameter correlation analysis was performed on the out-of-limit data segments to obtain anomaly feature vectors, including: The standardized time series dataset is aligned point by point with the multi-parameter dynamic normal interval baseline. The deviation between the actual operating parameter value of each sampling node and the upper and lower limits of the corresponding baseline interval is calculated. Continuous data segments with deviations exceeding the preset tolerance threshold are selected to obtain the over-limit data segments. Multi-parameter time-series synchronous extraction is performed on the out-of-limit data segments to separate the corresponding pressure time-series subseries, flow time-series subseries, and temperature time-series subseries within the segments; The pressure time series, flow rate time series, and temperature time series were subjected to pairwise dynamic cross-correlation calculations to obtain the calculation results; a set of parameter coupling deviation indices was constructed based on the calculation results. The abnormal feature vector is obtained by performing feature dimensionality reduction and vectorization reorganization on the parameter coupling deviation index set and the duration and peak deviation of the out-of-limit data segment.

6. The method for identifying and processing abnormal operating data of gas equipment according to claim 5, characterized in that, The abnormal feature vectors are matched with a preset gas anomaly pattern library to obtain the anomaly matching results. By combining the anomaly matching results with historical evolution trends to calculate the risk confidence level, anomaly identification results are obtained, including anomaly type, associated equipment location, and risk level, including: Extract the prototype feature vectors of each standard anomaly pattern from the preset gas anomaly pattern library, calculate the spatial similarity distance between the anomaly feature vector and each prototype feature vector, select the standard anomaly pattern with the spatial similarity distance as the target matching pattern, and obtain the anomaly matching result; Based on the abnormal matching results, locate the associated gas equipment locations and retrieve the historical operation records corresponding to the gas equipment locations; extract the occurrence frequency and deterioration rate parameters of the target matching pattern in the historical operation records and construct a historical evolution trend vector; The pattern matching degree and historical evolution trend vector in the abnormal matching results are input into the preset risk confidence assessment model, and weighted fusion calculation is performed to obtain the risk confidence value. The risk level is determined by mapping the risk confidence score to a preset risk level threshold. The risk level determination is then combined with the anomaly category identifier and associated device location of the target matching pattern to obtain an anomaly identification result that includes anomaly type, associated device location, and risk level.

7. A gas equipment operation data anomaly identification and processing system, wherein the system implements the method as described in any one of claims 1 to 6, characterized in that, include: The acquisition module is used to receive the device operation data stream uploaded by the community gas IoT terminal in real time, and perform timestamp alignment, noise filtering and missing value interpolation on the device operation data stream to obtain a standardized time series dataset. The module is used to extract the synchronous parameter sequences of the outlet confluence node of the pressure regulating station, the inflow node of the supporting public building pipeline network, and the terminal meter node of the high-rise residential building from the standardized time series dataset, and construct a three-dimensional operating condition response surface; the curvature differential and topological equal domain division are performed on the three-dimensional operating condition response surface to obtain the hydraulic and thermal coupling compensation of the pipeline network; the standardized time series dataset is subjected to dynamic baseline generation processing to obtain the initial baseline data sequence. The initial baseline data sequence is superimposed with the hydraulic-thermal coupling compensation amount of the pipe network to obtain the corrected baseline sequence; The fitting module is used to construct a multi-dimensional context feature matrix by combining the corrected baseline sequence with the current gas consumption period characteristics, seasonal temperature characteristics, and equipment type characteristics. Nonlinear adaptive fitting is performed on the multi-dimensional context feature matrix to obtain a multi-parameter dynamic normal interval baseline. The analysis module is used to use the multi-parameter dynamic normal interval baseline as a dynamic judgment benchmark to perform limit screening on the standardized time series dataset to extract limit-exceeding data segments; and to perform multi-parameter correlation analysis on the limit-exceeding data segments to obtain abnormal feature vectors. The mapping module is used to perform feature mapping matching between abnormal feature vectors and a preset gas anomaly pattern library to obtain anomaly matching results; the anomaly matching results are combined with historical evolution trends to calculate risk confidence, and anomaly identification results including anomaly type, associated equipment location and risk level are obtained.

8. A computing device, characterized in that, include: One or more processors; A storage device for storing one or more programs that, when executed by one or more processors, cause the one or more processors to implement the method as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a program that, when executed by a processor, implements the method as described in any one of claims 1 to 6.

Images