Steam pipe network transmission and distribution loss location method and device

Abstract

The present application provides a method and device for locating transmission and distribution losses in a steam pipeline network. The method includes: collecting historical time series data of the pipeline network and obtaining loss points and their occurrence times; analyzing the historical time series data of the pipeline network based on the occurrence times of the loss points to obtain parameter information of the loss points; using a neural network model to perform loss attribution analysis on the loss points based on the parameter information of the loss points to obtain attribution results; reviewing the attribution results based on a preset interval; obtaining location information of the pipeline network transmission and distribution losses based on the parameter information of the loss points and the occurrence times of the loss points; and storing the parameter information of the loss points, the attribution results, and the location information of the pipeline network transmission and distribution losses. This method integrates statistical analysis, time-frequency analysis, and causal inference to achieve real-time, accurate, and explainable detection of transmission and distribution losses in complex pipeline networks, providing effective technical support for energy conservation and consumption reduction and safe operation, and improving the stability and accuracy of pipeline network operation.

Description

Technical Field

[0001] The present application relates to the field of smart city operations, and in particular to a method and device for locating transmission and distribution losses in a steam pipeline network. Background Art

[0002] In modern industrial production, steam pipeline networks serve as critical infrastructure for transporting heat energy. Their stable and efficient operation is crucial for ensuring production continuity and reducing costs. However, steam pipelines inevitably experience some losses during transportation, a major challenge for businesses. These losses not only directly waste energy and increase operating costs, but also have profound impacts on their economic performance, environmental protection, and fulfillment of their social responsibilities. Detecting abnormal steam pipeline damage, locating it, and analyzing its cause have become key concerns for steam pipeline operators.

[0003] Currently, steam pipeline network operation monitoring still relies primarily on manual inspections and fixed-point instrument readings. Operations and maintenance personnel regularly visually inspect insulation, supports, and valves along the pipeline corridors, using portable devices such as temperature guns, pressure gauges, and listening rods to preliminarily determine if there are leaks, blockages, or equipment failures. Losses at key nodes are determined solely through set alarm thresholds. However, steam pipeline networks are typically strongly coupled, nonlinear, and time-varying systems: small pressure differences, subtle temperature drops, or weak energy peaks in the spectrum often indicate fault precursors. However, these weak signals are drowned out by load fluctuations, noise, and operating mode switching, making them indistinguishable using a single threshold alarm. Manual methods also fail to mine correlations and causal relationships across multiple devices and sources in high-dimensional data spaces, making it even more difficult to explain the causes of failures.

[0004] Therefore, real-time detection of abnormal pipeline losses has become an urgent problem to be solved.

[0005] Patent CN115713095A discloses a method for detecting anomalies in natural gas pipelines based on a hybrid deep neural network. The method comprises: normalizing the characteristic data of the natural gas pipeline to obtain input eigenvalues; constructing a stacked sparse denoising autoencoder deep neural network model based on the input eigenvalues ​​as a first hybrid deep neural network; constructing a cost function based on the input eigenvalues, and using the cost function to perform unsupervised feature learning on the first hybrid deep neural network to obtain a second hybrid deep neural network; adding a supervised classifier to the second hybrid deep neural network to obtain a third hybrid deep neural network; inputting the input eigenvalues ​​into the third hybrid deep neural network to obtain output eigenvalues; defining the output eigenvalues ​​based on the label value; wherein data with a label value of 0 in the output eigenvalues ​​is defined as normal data, and data with a label value of r is defined as anomaly data, where r is an integer greater than 0; calculating the maximum probability value of the output eigenvalue using the supervised classifier; and training the third hybrid deep neural network to minimize the difference between the maximum probability value of the output eigenvalue and the label. This method only provides anomaly classification, lacks cause analysis and location functions, and is insufficiently user-friendly.

[0006] Based on this, the present application provides a method and device for locating steam pipeline distribution losses to improve the existing technology. Summary of the Invention

[0007] The purpose of this application is to provide a method and device for locating transmission and distribution losses in a steam pipeline network. This method integrates statistical analysis, time-frequency analysis, and causal inference to achieve real-time, accurate, and explainable detection of transmission and distribution losses in complex pipeline networks, providing effective technical support for energy conservation and consumption reduction and safe operation, and improving the stability and accuracy of pipeline network operation.

[0008] The purpose of this application is achieved by the following technical solutions:

[0009] In a first aspect, the present application provides a method for locating steam pipe network transmission and distribution losses, which dynamically monitors, locates, and stores steam pipe network transmission and distribution losses at multiple time scales based on multi-dimensional data. The method comprises:

[0010] Collect historical time series data of the pipeline network and obtain loss points and their occurrence time;

[0011] Based on the occurrence time of the loss point, analyze the historical time series data of the pipeline network to obtain the parameter information of the loss point;

[0012] Based on the parameter information of the loss point, the neural network model is used to perform loss attribution analysis on the loss point to obtain the attribution result;

[0013] Review the attribution results based on the preset interval;

[0014] Based on the parameter information of the loss point and the occurrence time of the loss point, the location information of the pipeline network transmission and distribution loss is obtained;

[0015] The parameter information of the loss point, attribution results and location information of the pipeline network transmission and distribution loss are stored.

[0016] This technical solution offers the following benefits: by collecting historical time-series data from the pipeline network, loss points and their occurrence times are identified, and parametric information about the loss points is analyzed and extracted. Based on this parametric information, a variational Bayesian neural network model is used to perform loss attribution analysis and obtain attribution results. The attribution results are then reviewed within pre-set intervals. Combining the loss point parameters and occurrence times, the location of the pipeline network transmission and distribution loss is located and the relevant information is stored. This method significantly improves the accuracy and efficiency of steam pipeline transmission and distribution loss detection. Traditional pipeline network monitoring methods typically rely on single-time-scale analysis, which struggles to capture the dynamic characteristics of complex losses. However, multi-time-scale analysis can accurately extract the time-frequency characteristics of loss points, enhancing the robustness of loss detection. The application of a neural network model makes loss attribution analysis more intelligent, enabling accurate identification of loss causes and quantification of confidence levels, providing a reliable basis for subsequent processing. Furthermore, the precise acquisition of loss locations and structured storage of information not only facilitate real-time monitoring and historical data tracing, but also provide data support for pipeline network maintenance, reducing misjudgment rates and maintenance costs. Overall, this method integrates statistical analysis, time-frequency analysis and causal inference to achieve real-time, accurate and explainable detection of transmission and distribution losses in complex steam pipeline networks, providing effective technical support for energy conservation and consumption reduction and safe operation, and improving the stability and accuracy of steam pipeline network operation.

[0017] In some optional implementations, collecting historical time series data of the pipeline network and obtaining loss points and their occurrence times include:

[0018] Online access to the pipeline network online operation system and pipeline network database;

[0019] Pulling historical time series data of the pipeline network for a preset inspection task within a preset period from the pipeline network database, wherein the historical time series data is multi-dimensional data;

[0020] Perform differential processing on historical time series data;

[0021] Cluster analysis is performed on the historical time series data after difference processing to obtain the loss points and the time when the loss points occur.

[0022] The beneficial effects of this technical solution are: accessing the pipeline network's online operation system and database and pulling multi-dimensional historical time series data within a preset time period. Subsequently, the data is differentially processed to highlight changes in losses, and cluster analysis is used to identify loss points and their occurrence times. Utilizing an online standardized access method with strong compatibility, it can seamlessly connect to a variety of pipeline network systems, ensuring the stability and real-time nature of data collection. It allows users to customize the data pulling range according to the detection task, reducing redundant data processing and improving efficiency. The collection of multi-dimensional data retains comprehensive information on pipeline network operation, and differential processing effectively highlights changes in losses and reduces noise interference. The application of cluster analysis further improves the accuracy of loss point identification and can quickly extract loss characteristics and time information from complex data. Compared with traditional manual screening or simple threshold methods, this method has a higher degree of automation and a lower error rate, providing reliable data support for subsequent loss attribution and positioning, significantly shortening the response time of loss detection, and improving the intelligence level of pipeline network management.

[0023] In some optional implementations, the parameter information of the pipeline network transmission and distribution loss includes loss duration, loss pattern, loss trend and characteristic change frequency.

[0024] The beneficial effects of this technical solution are: the extraction of loss duration and loss trend helps to distinguish between instantaneous losses and continuous losses, thereby judging the severity and potential impact of the losses. The identification of loss patterns can reveal the regular characteristics of losses, such as periodic fluctuations or mutations, providing important clues for attribution analysis. The quantification of the frequency of characteristic changes further enhances the ability to capture the dynamic characteristics of losses, and is particularly suitable for multi-time scale loss detection in complex pipe network systems. Compared with traditional methods that only focus on a single indicator (such as pressure or flow), the multi-dimensional parameter system of this method can more comprehensively characterize loss characteristics and reduce the missed detection rate and misjudgment rate.

[0025] In some optional implementations, the loss attribution analysis of the loss points using a neural network model based on the parameter information of the loss points to obtain the attribution results includes:

[0026] Inputting the loss point to be measured into the neural network model to obtain an attribution result of the loss point to be measured, wherein the attribution result of the loss point to be measured is used to indicate the loss cause of the loss point to be measured and the confidence level corresponding to the loss cause, wherein the neural network model is a variational Bayesian neural network model;

[0027] The training process of the variational Bayesian neural network model includes:

[0028] Acquire a training set, the training set including a plurality of training data, each of the training data including parameter information of a sample pipeline network transmission and distribution loss and labeled data of attribution results of the sample loss point;

[0029] For each training data in the training set, perform the following processing:

[0030] Inputting parameter information of sample pipeline network transmission and distribution losses in the training data into a preset deep learning model to obtain predicted data of attribution results of the sample loss points;

[0031] Updating the model parameters of the deep learning model based on the predicted data and the labeled data of the attribution result of the sample loss point;

[0032] Check whether the preset training end condition is met; if so, use the trained deep learning model as the variational Bayesian neural network model.

[0033] This technical solution offers the following benefits: It utilizes a variational Bayesian neural network model to perform attribution analysis on loss points. The model inputs parameter information for a loss point and outputs the cause of the loss and a confidence level. The model undergoes deep learning training using a training set (consisting of sample loss parameters and annotated attribution results), iteratively updating model parameters until training termination criteria are met. The variational Bayesian neural network model effectively handles uncertainty in data through probabilistic reasoning, resulting in more reliable and interpretable attribution results than traditional rule-based or simple machine learning approaches. The confidence level output by the model quantifies the credibility of the loss cause, providing a scientific basis for subsequent review and decision-making. The training process relies on sample data with multi-dimensional parameters, ensuring the model's adaptability to complex loss scenarios and reducing the risk of overfitting. Furthermore, this method's automated attribution analysis significantly reduces manual intervention and improves processing efficiency. This allows for rapid identification of loss causes and shortens response time, particularly when working with large-scale pipeline network data. This approach, leveraging deep learning and probabilistic modeling, significantly enhances the intelligence and real-time nature of pipeline transmission and distribution loss management.

[0034] In some optional implementations, reviewing the attribution result based on the preset interval includes:

[0035] Push loss causes with confidence levels higher than the preset range to the system platform and automatically match the corresponding preset emergency strategies;

[0036] Mark the loss causes with confidence levels within a preset range, and record the characteristic snapshots and waveform trend graphs of the loss points corresponding to the loss causes for manual review;

[0037] Push loss causes with confidence levels lower than the preset range to the system front end.

[0038] The beneficial effects of this technical solution are: based on the preset interval, the loss attribution results are reviewed, and the loss causes with high confidence are pushed to the system platform and matched with the emergency strategy. The characteristic snapshots and waveform trend charts of the causes with moderate confidence are recorded for manual review, and the low-confidence causes are pushed to the front end. The automatic push of high-confidence loss causes and the matching of emergency strategies achieve rapid response and reduce the impact of losses on pipeline network operations. The characteristic snapshots and waveform trend chart records of moderate-confidence causes provide intuitive data support for manual review and improve the accuracy and traceability of the review. The front-end push of low-confidence causes facilitates further monitoring and analysis and avoids the interference of invalid information. The hierarchical review mechanism of this method effectively balances automation and manual intervention, reduces the risk of misjudgment, and improves the transparency and operability of the system. Compared with the traditional fully manual review method, this method significantly shortens the processing time, improves the real-time and reliability of pipeline transmission and distribution loss management, and provides strong technical support for ensuring the safe operation of the pipeline network.

[0039] In some optional implementations, obtaining location information of pipeline network transmission and distribution loss based on parameter information of the loss point and the occurrence time of the loss point includes:

[0040] Based on the parameter information of the loss points, a preset number of loss points with relatively strong feature change frequencies are set as a loss point set;

[0041] For each set of loss points, do the following:

[0042] Detect whether there is a proximity relationship between the loss points in the loss point cluster;

[0043] If there is a proximity relationship, the two loss points with the earliest occurrence time are selected as the main loss collection points;

[0044] If there is no adjacent relationship, a loss subgraph is generated;

[0045] Sort the loss subgraphs by position, and set the loss point corresponding to the upstream loss subgraph as the main loss collection point;

[0046] Connect to the main loss collection points to obtain the location information of pipeline network transmission and distribution losses.

[0047] The beneficial effects of this technical solution are: based on the parameter information and occurrence time of the loss points, a loss point set is constructed by screening loss points with strong feature change frequencies. The point set is detected for proximity, and the earliest loss point is selected to generate a loss subgraph. After sorting, the main loss point is determined and connected to obtain the loss location information. By screening loss points with strong feature change frequencies, the method focuses on key loss features and reduces noise interference. The proximity relationship detection and subgraph generation strategy fully utilizes the spatiotemporal correlation of loss points and can accurately restore the spatial distribution of loss occurrence. The selection and connection of the main loss point further simplifies the positioning process, ensures the consistency and reliability of the location information, and realizes the precise positioning of the loss location. Compared with traditional methods that rely on manual inspections or simple threshold positioning, this method has a high degree of automation and a fast positioning speed, and is suitable for large-scale loss detection in complex pipe network systems.

[0048] In a second aspect, the present application provides a pipeline network transmission and distribution loss detection device, the device comprising a processor, the processor being configured to implement the following steps:

[0049] Data acquisition module, used to collect historical time series data of the pipeline network and obtain loss points and their occurrence time;

[0050] The parameter information acquisition module is used to analyze the historical time series data of the pipeline network based on the occurrence time of the loss point to obtain the parameter information of the loss point;

[0051] The attribution result acquisition module is used to perform loss attribution analysis on the loss points using a neural network model based on the parameter information of the loss points to obtain the attribution results;

[0052] An attribution result review module is used to review the attribution results based on a preset interval;

[0053] The loss location module is used to obtain the location information of the pipeline network transmission and distribution loss based on the parameter information of the loss point and the occurrence time of the loss point;

[0054] The loss storage module is used to store the parameter information of the loss point, the attribution results and the location information of the pipeline network transmission and distribution loss.

[0055] In some optional implementations, the processor is configured to collect historical time series data of the pipeline network and obtain loss points and their occurrence times in the following manner:

[0056] Online access to the pipeline network online operation system and pipeline network database;

[0057] Pulling historical time series data of the pipeline network for a preset inspection task within a preset period from the pipeline network database, wherein the historical time series data is multi-dimensional data;

[0058] Perform differential processing on historical time series data;

[0059] Cluster analysis is performed on the historical time series data after difference processing to obtain the loss points and the time when the loss points occur.

[0060] In some optional implementations, the parameter information of the pipeline network transmission and distribution loss includes loss duration, loss pattern, loss trend and characteristic change frequency.

[0061] In some optional embodiments, the processor is configured to perform loss attribution analysis on the loss point using a neural network model based on parameter information of the loss point in the following manner to obtain an attribution result:

[0062] Inputting the loss point to be measured into the neural network model to obtain an attribution result of the loss point to be measured, wherein the attribution result of the loss point to be measured is used to indicate the loss cause of the loss point to be measured and the confidence level corresponding to the loss cause, wherein the neural network model is a variational Bayesian neural network model;

[0063] The training process of the variational Bayesian neural network model includes:

[0064] Acquire a training set, the training set including a plurality of training data, each of the training data including parameter information of a sample pipeline network transmission and distribution loss and labeled data of attribution results of the sample loss point;

[0065] For each training data in the training set, perform the following processing:

[0066] Inputting parameter information of sample pipeline network transmission and distribution losses in the training data into a preset deep learning model to obtain predicted data of attribution results of the sample loss points;

[0067] Updating the model parameters of the deep learning model based on the predicted data and the labeled data of the attribution result of the sample loss point;

[0068] Check whether the preset training end condition is met; if so, use the trained deep learning model as the variational Bayesian neural network model.

[0069] In some optional implementations, the processor is configured to review the attribution result based on a preset interval in the following manner:

[0070] Push loss causes with confidence levels higher than the preset range to the system platform and automatically match the corresponding preset emergency strategies;

[0071] Mark the loss causes with confidence levels within a preset range, and record the characteristic snapshots and waveform trend graphs of the loss points corresponding to the loss causes for manual review;

[0072] Push loss causes with confidence levels lower than the preset range to the system front end.

[0073] In some optional embodiments, the processor is configured to obtain the location information of the pipeline network transmission and distribution loss based on the parameter information of the loss point and the occurrence time of the loss point in the following manner:

[0074] Based on the parameter information of the loss points, a preset number of loss points with relatively strong feature change frequencies are set as a loss point set;

[0075] For each set of loss points, do the following:

[0076] Detect whether there is a proximity relationship between the loss points in the loss point cluster;

[0077] If there is a proximity relationship, the two loss points with the earliest occurrence time are selected as the main loss collection points;

[0078] If there is no adjacent relationship, a loss subgraph is generated;

[0079] Sort the loss subgraphs by position, and set the loss point corresponding to the upstream loss subgraph as the main loss collection point;

[0080] Connect to the main loss collection points to obtain the location information of pipeline network transmission and distribution losses.

[0081] In a third aspect, the present application provides an electronic device comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements the steps of any of the above methods when executing the computer program.

[0082] In a fourth aspect, the present application provides a steam pipe network transmission and distribution loss location system, the system comprising:

[0083] The above electronic equipment.

[0084] In a fifth aspect, the present application provides a chip, wherein the chip stores a computer program, and when the computer program is executed by a processor, the steps of any of the above methods are implemented. BRIEF DESCRIPTION OF THE DRAWINGS

[0085] The present application is further described below with reference to the accompanying drawings and implementation methods.

[0086] Figure 1 A schematic diagram of a process for locating steam network transmission and distribution losses provided in an embodiment of the present application is shown.

[0087] Figure 2 A flow chart of a method for obtaining loss points and their time provided in an embodiment of the present application is shown.

[0088] Figure 3 A flow chart of a method for obtaining location information of pipeline network transmission and distribution losses provided in an embodiment of the present application is shown.

[0089] Figure 4 A structural block diagram of a pipeline network transmission and distribution loss detection device provided in an embodiment of the present application is shown.

[0090] Figure 5 A structural framework diagram of an electronic device provided in an embodiment of the present application is shown.

[0091] Figure 6 A structural schematic diagram of a steam pipe network transmission and distribution loss locating system provided in an embodiment of the present application is shown.

[0092] Figure 7 A schematic diagram of the structure of a program product provided in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0093] Below, the embodiments of the present application are further described in conjunction with the accompanying drawings and specific implementation methods. It should be noted that, under the premise of no conflict, the various embodiments or technical features described below can be arbitrarily combined to form a new implementation method.

[0094] In the embodiments of the present application, "at least one" refers to one or more, and "more" refers to two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent: the existence of A alone, the existence of A and B at the same time, and the existence of B alone, where A and B can be singular or plural. The character " / " generally indicates that the objects associated before and after are in an "or" relationship. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b or c can represent: a, b, c, a and b, a and c, b and c, a and b and c, where a, b and c can be single or multiple. It is worth noting that "at least one" can also be interpreted as "one or more items".

[0095] It should also be noted that in the embodiments of this application, words such as "exemplary" or "for example" are used to indicate examples, illustrations, or descriptions. Any implementation or design scheme described as "exemplary" or "for example" in the embodiments of this application should not be interpreted as being preferred or advantageous over other implementations or designs. Rather, the use of words such as "exemplary" or "for example" is intended to present the relevant concepts in a concrete manner.

[0096] Method Example

[0097] See also Figure 1 , Figure 1 A schematic diagram of a process for locating steam network transmission and distribution losses provided in an embodiment of the present application is shown.

[0098] The present application provides a method for locating steam pipe network transmission and distribution losses, which dynamically monitors, locates, and stores steam pipe network transmission and distribution losses at multiple time scales based on multi-dimensional data. The method includes:

[0099] Step S101: Collect historical time series data of the pipeline network and obtain loss points and their occurrence times;

[0100] Step S102: Based on the occurrence time of the loss point, analyze the historical time series data of the pipeline network to obtain parameter information of the loss point;

[0101] Step S103: Based on the parameter information of the loss point, a neural network model is used to perform loss attribution analysis on the loss point to obtain an attribution result;

[0102] Step S104: reviewing the attribution results based on the preset interval;

[0103] Step S105: obtaining location information of the pipeline network transmission and distribution loss based on parameter information of the loss point and the occurrence time of the loss point;

[0104] Step S106: storing the parameter information of the loss point, the attribution result and the location information of the pipeline network transmission and distribution loss.

[0105] Therefore, by collecting historical time-series data from the pipeline network, we can identify loss points and their occurrence times, and then analyze and extract parameter information about these loss points. Based on this parameter information, we use a variational Bayesian neural network model to perform loss attribution analysis and obtain attribution results. Subsequently, we review the attribution results within preset intervals, and by combining the parameter information and occurrence time of the loss points, we can locate the location of the pipeline network transmission and distribution loss, and store the relevant information.

[0106] This method significantly improves the accuracy and efficiency of steam pipeline transmission and distribution loss detection. Traditional pipeline network monitoring methods typically rely on single-time-scale analysis, which struggles to capture the dynamic characteristics of complex losses. However, multi-time-scale analysis can accurately extract the time-frequency characteristics of loss points, enhancing the robustness of loss detection.

[0107] The application of neural network models makes loss attribution analysis more intelligent, accurately identifying loss causes and quantifying confidence levels, providing a reliable basis for subsequent processing. Furthermore, the precise acquisition of loss locations and structured storage of information not only facilitate real-time monitoring and historical data tracing, but also provide data support for pipeline network maintenance, reducing misjudgment rates and maintenance costs.

[0108] Overall, this method integrates statistical analysis, time-frequency analysis and causal inference to achieve real-time, accurate and explainable detection of transmission and distribution losses in complex steam pipeline networks, providing effective technical support for energy conservation and consumption reduction and safe operation, and improving the stability and accuracy of steam pipeline network operation.

[0109] In the embodiment of the present application, time series data refers to data recorded over time, usually arranged in chronological order. The characteristics of time series data include: Timestamp: the specific time of each data record. Continuity: the time interval between data points can be uniform (such as every minute, every hour) or uneven. Trend: Time series data can be used to analyze patterns such as changing trends, periodicity, and seasonality. Time series data includes operating data, which is recorded regularly and changes over time. Operating data refers to data related to the operating status and performance of the pipeline system. These data typically include: Flow data: the flow rate of gas in the pipeline (for example, cubic meters / hour). Pressure data: the pressure of the gas in the pipeline (for example, MPa). Temperature data: the temperature of the gas in the pipeline. Valve status: the switch status of each valve. Equipment operating status: the operating status of equipment such as compressors and pressure regulators.

[0110] In the implementation of this application, the required data include: 1) instantaneous faults, such as leakage, valve failure, etc.; analysis is performed every 30 minutes, using the operating data of the most recent hour; 2) short-term slow faults, such as short-term blockages, slow failures caused by insulation water absorption; data is pulled and analyzed every 1 to 2 days to obtain the operating data within these 1 to 2 days; 3) long-term equipment and component aging, such as aging of insulation materials, increased resistance caused by long-term fouling in pipelines, etc.; data analysis is performed every year to obtain historical operating data within 3 years to avoid problems in subsequent analysis and processing caused by excessive data volume.

[0111] In addition to the network's own operational data, weather loads also significantly impact the amount of steam used by each user in the network. Therefore, these loads are recorded in the database as external variables to supplement pipeline operational monitoring data. Specific data includes hourly temperature, humidity, wind speed, and rain and snow conditions.

[0112] See also Figure 2 , Figure 2 A flow chart of a method for obtaining loss points and their time provided in an embodiment of the present application is shown.

[0113] In some optional implementations, collecting historical time series data of the pipeline network and obtaining loss points and their occurrence times include:

[0114] Step S201: Accessing the pipe network online operation system and the pipe network database online;

[0115] Step S202: Pulling historical time series data of the pipe network for a preset inspection task within a preset time period from the pipe network database, wherein the historical time series data is multi-dimensional data;

[0116] Step S203: performing differential processing on the historical time series data;

[0117] Step S204: performing cluster analysis on the historical time series data after differential processing to obtain loss points and the time when the loss points occur.

[0118] This approach connects to the network's online operation system and database, extracting multi-dimensional historical time series data for a preset period. This data is then differentially processed to highlight changes in losses, and cluster analysis is used to identify loss points and their occurrence times.

[0119] Utilizing a standardized online access method with strong compatibility, it seamlessly connects to various pipeline network systems, ensuring stable and real-time data collection. Users can customize the data extraction scope based on the inspection task, reducing redundant data processing and improving efficiency. Multidimensional data collection preserves comprehensive information about pipeline network operations, while differential processing effectively highlights loss changes and reduces noise interference. The application of cluster analysis further improves the accuracy of loss point identification, enabling rapid extraction of loss characteristics and temporal information from complex data.

[0120] Compared with traditional manual screening or simple threshold methods, this method has a higher degree of automation and a lower error rate. It provides reliable data support for subsequent loss attribution and positioning, significantly shortens the response time of loss detection, and improves the intelligence level of pipeline network management.

[0121] In some optional embodiments, the OBDC interface can be used to access the pipeline network online operation system and the pipeline network database, and SQL statements can be used to pull the historical time series data of the pipeline network for the preset inspection task within a preset time period.

[0122] In some optional embodiments, it is used to collect real-time operating data of key nodes in the pipeline network, including but not limited to parameters such as steam temperature, flow, pressure, humidity, insulation layer surface temperature, external ambient temperature, condensate level and discharge flow; at the same time, it can also access GIS pipeline network structure information and historical operation records.

[0123] In some optional implementations, the data extracted is multi-dimensional time series data. Time series data can be differentiated to eliminate trend and periodic components of the data, making the data more stable and facilitating further analysis and detection of loss values. This application uses a combination of first-order and second-order differences to combine the pressure and temperature of all points in the pipeline network at each moment, and the first-order and second-order differences of the flow rate of all pipelines as features for that moment. This momentary data is then classified using a clustering algorithm and the loss points (categorized as loss points) are output.

[0124] Considering the need to regularly update the clustering model after pipe loss data is stored, we chose to use the COP-Kmeans model for clustering. COP-Kmeans, based on traditional Kmeans, requires specifying which points must be in one class or not. During the iteration process, if any points do not match the specified values, the number of clusters is reallocated and the cluster center is updated until the stopping number is reached or convergence is achieved.

[0125] Since there is an obvious quantitative mismatch in the first-order differences of the parameters of the pipe network during clustering, the Mahalanobis distance is used as the basis for Kmeans clustering. The calculation formula of the Mahalanobis distance is as follows:

[0126]

[0127] In some optional implementations, the parameter information of the pipeline network transmission and distribution loss includes loss duration, loss pattern, loss trend and characteristic change frequency.

[0128] Therefore, extracting the duration and trend of losses helps distinguish between instantaneous losses and sustained losses, thereby determining the severity and potential impact of the losses. Identifying loss patterns can reveal regular characteristics of losses, such as periodic fluctuations or sudden changes, providing important clues for attribution analysis. Quantifying the frequency of characteristic changes further enhances the ability to capture the dynamic characteristics of losses, making it particularly suitable for multi-timescale loss detection in complex pipe network systems. Compared to traditional methods that focus only on a single indicator (such as pressure or flow), the multidimensional parameter system of this method can more comprehensively characterize loss characteristics, reducing missed detection rates and false positive rates.

[0129] In some optional embodiments, wavelet transform analysis is performed on the data of long-term changes or 1 hour before and after the initial screening to confirm the existence of pipeline loss fluctuations, including the flow rate of each pipeline, the pressure of each node and the temperature, to further extract the loss fluctuation pattern and trend of the data.

[0130] Wavelet transform analysis is a time-frequency localized, multi-resolution analysis tool suitable for processing non-stationary signals. It can separate signal details and overall trends at different scales. In loss detection, wavelet transforms can be used to distinguish the characteristics of sudden changes in signals (such as sudden pressure changes caused by leaks) from long-term trends (such as systemic losses). In practical applications, the discrete wavelet transform (DWT) is often used to decompose signals. At each level, it contains low-frequency (approximate) and high-frequency (detail) components, as shown in Equation (2).

[0131]

[0132] Daubechies wavelets (DB wavelets for short) are a family of wavelets with compact support, strong orthogonality, and good smoothness. They are widely used in signal denoising, compression, and loss detection. DB wavelets are named after their vanishing moments. For example, the db4 wavelet has four vanishing moments and can accurately represent cubic polynomials; the db6 wavelet has six vanishing moments and is suitable for extracting smooth trends in higher-order signals.

[0133] For long-term loss analysis, we used db6, a wavelet transform at level 6, to perform long-term trend analysis on the total daily pipe loss data. Level 6 decomposition was chosen to obtain lower-frequency coefficients while maintaining sufficient temporal resolution, thereby better reflecting the overall trend of pipeline loss. The low-frequency coefficients obtained from the level 6 decomposition were analyzed and plotted. This plot reflects the overall trend of pipeline loss. By observing the low-frequency coefficient plot, we can identify whether there is a long-term upward or downward trend in pipeline loss, as well as any cyclical fluctuations. By combining known pipeline parameters (such as pipe material, diameter, length, and conveying medium) with historical operating data, we can infer the insulation status of the pipeline. For example, a persistently rising low-frequency coefficient may indicate aging or damage of the insulation layer, leading to increased heat loss. The results of the wavelet transform analysis were processed, retaining the low-frequency component and reconstructing the signal. The reconstructed data better illustrates the long-term changes in pipe loss. Then, the ARIMA model is used to predict the pipe damage status in the next 2 to 3 months based on the reconstructed low-frequency signal. This is compared with historical data and combined with relevant parameters such as pipeline maintenance, renewal costs, and insulation layer replacement costs. The economic efficiency is calculated and compared with historical data, and economic analysis results and pipeline maintenance recommendations are given.

[0134] COP-Kmeans detects loss fluctuations at the current moment, indicating short-term or transient losses in the pipeline network. For each node with data collection, the db4 wavelet transform analysis parameters at level 4 are applied, focusing on short-term, sudden changes. The db4 wavelet also has a shorter support length and sharper time-domain characteristics, which better captures sudden changes in the signal. At level 4, the time window is smaller, allowing for more precise localization of the time point when the loss fluctuation occurred. Low-frequency components are removed from the signal, and high-frequency components are reconstructed, using coefficients from higher levels to constrain lower levels. If the high-frequency components exceed a pre-set threshold, a loss is considered to have occurred during this time period. The aforementioned wavelet transform analysis is also performed on parameters such as pressure, flow rate, and temperature to determine whether the duration of the loss exceeds 5%. If so, the loss is considered to be caused by a short-term loss. Otherwise, the loss is considered to be caused by transient leakage or insulation failure. The detail coefficients at levels 3 / 4 obtained from the wavelet transform analysis are normalized and then used as input to the Bayesian model for loss attribution to further determine the cause.

[0135] In some optional implementations, the loss attribution analysis of the loss points using a neural network model based on the parameter information of the loss points to obtain the attribution results includes:

[0136] Inputting the loss point to be measured into the neural network model to obtain an attribution result of the loss point to be measured, wherein the attribution result of the loss point to be measured is used to indicate the loss cause of the loss point to be measured and the confidence level corresponding to the loss cause, wherein the neural network model is a variational Bayesian neural network model;

[0137] The training process of the variational Bayesian neural network model includes:

[0138] Acquire a training set, the training set including a plurality of training data, each of the training data including parameter information of a sample pipeline network transmission and distribution loss and labeled data of attribution results of the sample loss point;

[0139] For each training data in the training set, perform the following processing:

[0140] Inputting parameter information of sample pipeline network transmission and distribution losses in the training data into a preset deep learning model to obtain predicted data of attribution results of the sample loss points;

[0141] Updating the model parameters of the deep learning model based on the predicted data and the labeled data of the attribution result of the sample loss point;

[0142] Check whether the preset training end condition is met; if so, use the trained deep learning model as the variational Bayesian neural network model.

[0143] Therefore, a variational Bayesian neural network model is used to perform attribution analysis on loss points. The model inputs parameter information for a loss point and outputs the loss cause and confidence level. The model undergoes deep learning training using a training set (consisting of sample loss parameters and annotated attribution results), iteratively updating model parameters until the training termination criteria are met. The variational Bayesian neural network model effectively handles uncertainty in data through probabilistic reasoning, resulting in more reliable and interpretable attribution results than traditional rule-based or simple machine learning methods. The confidence level output by the model quantifies the credibility of the loss cause, providing a scientific basis for subsequent review and decision-making. The training process relies on sample data with multi-dimensional parameters, ensuring the model's adaptability to complex loss scenarios and reducing the risk of overfitting. Furthermore, this method's automated attribution analysis significantly reduces manual intervention and improves processing efficiency. Especially when working with large-scale pipeline network data, it can quickly identify loss causes and shorten response time. This method, leveraging deep learning and probabilistic modeling, significantly enhances the intelligent and real-time management of pipeline transmission and distribution losses.

[0144] In an embodiment of the present application, variational Bayes is a method for inferring unknown parameters in a probability model, which approximates the posterior distribution by optimizing variational inference. Variational neural networks are different from general neural networks. They consider intermediate parameters to be latent variables and conform to a certain prior distribution. On this basis, the model is trained, and the minimum classification error and the minimum KL divergence between the prior distribution and the posterior distribution are used as the goals of model optimization. In addition to the classification category, during actual reasoning, the probability of the current data under each loss category can be obtained based on a sampling statistical analysis of the distribution of latent variables, and the maximum probability loss condition is selected as the attribution result of the current loss time. Considering that each variable is a superposition of random fluctuations in measurements, gas sources, and users, the present invention uses the standard normal distribution of N(0,1) as the prior distribution. The calculation method of KL divergence is:

[0145]

[0146] In the initial phase, the dependency structure between loss types (such as leakage and insulation failure) and features is established using existing historical loss data. Model training then forms an initial approximate posterior distribution. During operation, the model employs online variational inference, incrementally updating local latent variables and global parameters with each new sample, dynamically revising the posterior probability distribution and thereby continuously improving attribution accuracy.

[0147] In some optional implementations, reviewing the attribution result based on the preset interval includes:

[0148] Push loss causes with confidence levels higher than the preset range to the system platform and automatically match the corresponding preset emergency strategies;

[0149] Mark the loss causes with confidence levels within a preset range, and record the characteristic snapshots and waveform trend graphs of the loss points corresponding to the loss causes for manual review;

[0150] Push loss causes with confidence levels lower than the preset range to the system front end.

[0151] Therefore, the loss attribution results are reviewed based on the preset intervals. High-confidence loss causes are pushed to the system platform and matched with emergency strategies. Moderate-confidence causes record feature snapshots and waveform trend graphs for manual review, and low-confidence causes are pushed to the front end. The automatic push of high-confidence loss causes and the matching of emergency strategies achieve a rapid response and reduce the impact of losses on pipeline network operations. The feature snapshots and waveform trend graph records of moderate-confidence causes provide intuitive data support for manual review, improving the accuracy and traceability of the review. The front-end push of low-confidence causes facilitates further monitoring and analysis, avoiding interference from invalid information. The hierarchical review mechanism of this method effectively balances automation and manual intervention, reduces the risk of misjudgment, and improves the transparency and operability of the system. Compared with the traditional fully manual review method, this method significantly shortens processing time, improves the real-time and reliability of pipeline transmission and distribution loss management, and provides strong technical support for ensuring the safe operation of the pipeline network.

[0152] In some optional embodiments, after the attribution results are output, the system directly pushes loss types with high confidence (e.g., attribution probability ≥ 85%) to the system platform and automatically matches the corresponding emergency response strategy. For example, a red alert appears for leak attribution, while a yellow alert appears for short-term insulation failure, informing operations and maintenance personnel of the cause of the loss and the corresponding solution. For losses with medium confidence levels (e.g., 50%–85%), the diagnosis results are marked as "pending review," and feature snapshots and waveform trend graphs are recorded for manual review and confirmation. Low confidence levels (<50%) are displayed on the front end. Throughout the operation process, manual review feedback is used as model label sample back-injection to periodically update the model structure and parameters, achieving self-evolution and accuracy improvement.

[0153] See also Figure 3 , Figure 3 A flow chart of a method for obtaining location information of pipeline network transmission and distribution losses provided in an embodiment of the present application is shown.

[0154] In some optional implementations, obtaining location information of pipeline network transmission and distribution loss based on parameter information of the loss point and the occurrence time of the loss point includes:

[0155] Step S301: Based on the parameter information of the loss points, a preset number of loss points with relatively strong feature change frequencies are set as a loss point set;

[0156] For each set of loss points, do the following:

[0157] Step S302: Detect whether there is a proximity relationship between the loss points in the loss point set;

[0158] Step S303: If there is a proximity relationship, the two loss points with the earliest occurrence time are selected as the main loss collection points;

[0159] Step S304: if there is no proximity relationship, generate a loss subgraph;

[0160] Step S305: sort the loss subgraphs by position, and set the loss point corresponding to the most upstream loss subgraph as the main loss collection point;

[0161] Step S306: Connect to the main loss collection point to obtain the location information of the pipeline network transmission and distribution loss.

[0162] Therefore, based on the parameter information and occurrence time of the loss points, a loss point set is constructed by screening loss points with strong feature change frequencies. The point set is tested for proximity, and the earliest loss point is selected to generate a loss subgraph. After sorting, the main loss point is determined and connected to obtain the loss location information. By screening loss points with strong feature change frequencies, the method focuses on key loss features and reduces noise interference. The proximity detection and subgraph generation strategy fully utilizes the spatiotemporal correlation of loss points and can accurately restore the spatial distribution of loss occurrence. The selection and connection of the main loss point further simplifies the positioning process, ensures the consistency and reliability of the location information, and realizes the precise positioning of the loss location. Compared with traditional methods that rely on manual inspections or simple threshold positioning, this method has a high degree of automation and a fast positioning speed, and is suitable for large-scale loss detection in complex pipe network systems.

[0163] In some optional embodiments, when the high-frequency change characteristics of each node exceed the set threshold, it is considered that a loss situation has occurred, and the 5% of nodes with the strongest high-frequency changes are focused on and set as the loss point set. If there is an obvious neighboring relationship between these nodes in the pipeline network topology, the two nodes with the earliest loss time are marked as collection points. The remaining points in the pipeline network are eliminated to form a loss subgraph. The relative positions of each loss subgraph in the pipeline network are compared, and the loss subgraphs are sorted from upstream to downstream, where the collection points in the upstreammost loss subgraph are considered to be the main loss points. The pipelines connected between the loss points are located as the locations where the loss occurs, and are displayed as the main loss points to prompt inspection personnel to focus on the investigation. The remaining loss points also need to be highlighted on the front-end interface as warning and investigation prompts.

[0164] In one specific embodiment, an ODBC interface is used to connect the pipeline network's online operation system and database. SQL statements are used to retrieve pipeline network operation data from 00:00 to 23:59 on May 14, 2025, including multi-dimensional time series data such as pressure, flow, and temperature (sampling frequency is 1 minute). The data is differentially processed to highlight loss fluctuations, and then K-means cluster analysis is used to identify loss points. For example, a sudden drop in pressure in a certain section of pipeline is detected at 14:35 and marked as a loss point with an occurrence time of 14:35:20. A wavelet transform is applied to this loss point to extract parameter information: the loss duration is 6 minutes, the loss pattern is sudden, the loss trend is a continuous pressure drop, and the characteristic change frequency is high (multiple fluctuations per second). These parameters comprehensively characterize the loss characteristics and provide a data foundation for subsequent analysis. The loss point parameters are then input into a pre-trained variational Bayesian neural network model. This model is trained based on a historical dataset (including annotated samples of pipeline leaks, valve failures, and blockages). High accuracy is ensured through iterative optimization of deep learning model parameters. The model output attribution results: The cause of the loss was a "minor pipeline leak" with an 88% confidence level. The confidence level quantifies the confidence level of the result, facilitating subsequent decision-making. Since the confidence level of 88% exceeds the preset threshold (80%), the system automatically pushes the "minor pipeline leak" notification to the management platform and implements emergency response strategies (such as reducing the pressure in the pipeline section or activating a backup pipeline). Simultaneously, a snapshot of the loss point's characteristics (pressure waveform) and trend chart are recorded for manual review by engineers. Causes with a confidence level below 60% (such as equipment failure, with a confidence level of 45%) are pushed to the system frontend and marked as pending monitoring. Based on the loss point parameters, the five loss points with the highest frequency of characteristic changes are selected to form a point set. The proximity of the loss points within the point set was examined, revealing a spatiotemporal correlation between the loss points at 14:35:20 and 14:35:25. The two earliest loss points were selected to generate a loss subgraph. After position sorting, the primary loss point was determined to be located upstream of pipeline node C. By connecting the primary loss points, the loss was located in the pipeline section from node C to D, approximately 50 meters long. The parameter information of the loss point (duration 6 minutes, mutation pattern, etc.), attribution results (minor pipeline leak, confidence level 88%) and positioning information (nodes C to D) are stored in the database to generate a unique identification record to support subsequent tracing and maintenance analysis.

[0165] Device embodiment

[0166] See also Figure 4 , Figure 4 A structural block diagram of a pipeline network transmission and distribution loss detection device provided in an embodiment of the present application is shown.

[0167] The embodiment of the present application also provides a pipeline distribution loss detection device, the specific implementation of which is consistent with the implementation and technical effects recorded in the above method embodiment, and some contents will not be repeated here.

[0168] The present application provides a pipeline network transmission and distribution loss detection device, the device comprising a processor, the processor being configured to implement the following steps:

[0169] The data collection module 101 is used to collect historical time series data of the pipe network and obtain loss points and their occurrence time;

[0170] The parameter information acquisition module 102 is used to analyze the historical time series data of the pipeline network based on the occurrence time of the loss point to obtain the parameter information of the loss point;

[0171] Attribution result acquisition module 103, configured to perform loss attribution analysis on the loss points using a neural network model based on parameter information of the loss points to obtain attribution results;

[0172] The attribution result review module 104 is used to review the attribution result based on a preset interval;

[0173] The loss location module 105 is used to obtain the location information of the pipeline network transmission and distribution loss based on the parameter information of the loss point and the occurrence time of the loss point;

[0174] The loss storage module 106 is used to store parameter information of the loss point, attribution results and location information of the pipeline network transmission and distribution loss.

[0175] In some optional implementations, the processor is configured to collect historical time series data of the pipeline network and obtain loss points and their occurrence times in the following manner:

[0176] Online access to the pipeline network online operation system and pipeline network database;

[0177] Pulling historical time series data of the pipeline network for a preset inspection task within a preset period from the pipeline network database, wherein the historical time series data is multi-dimensional data;

[0178] Perform differential processing on historical time series data;

[0179] Cluster analysis is performed on the historical time series data after difference processing to obtain the loss points and the time when the loss points occur.

[0180] In some optional implementations, the parameter information of the pipeline network transmission and distribution loss includes loss duration, loss pattern, loss trend and characteristic change frequency.

[0181] In some optional embodiments, the processor is configured to perform loss attribution analysis on the loss point using a neural network model based on parameter information of the loss point in the following manner to obtain an attribution result:

[0182] Inputting the loss point to be measured into the neural network model to obtain an attribution result of the loss point to be measured, wherein the attribution result of the loss point to be measured is used to indicate the loss cause of the loss point to be measured and the confidence level corresponding to the loss cause, wherein the neural network model is a variational Bayesian neural network model;

[0183] The training process of the variational Bayesian neural network model includes:

[0184] Acquire a training set, the training set including a plurality of training data, each of the training data including parameter information of a sample pipeline network transmission and distribution loss and labeled data of attribution results of the sample loss point;

[0185] For each training data in the training set, perform the following processing:

[0186] Inputting parameter information of sample pipeline network transmission and distribution losses in the training data into a preset deep learning model to obtain predicted data of attribution results of the sample loss points;

[0187] Updating the model parameters of the deep learning model based on the predicted data and the labeled data of the attribution result of the sample loss point;

[0188] Check whether the preset training end condition is met; if so, use the trained deep learning model as the variational Bayesian neural network model.

[0189] In some optional implementations, the processor is configured to review the attribution result based on a preset interval in the following manner:

[0190] Push loss causes with confidence levels higher than the preset range to the system platform and automatically match the corresponding preset emergency strategies;

[0191] Mark the loss causes with confidence levels within a preset range, and record the characteristic snapshots and waveform trend graphs of the loss points corresponding to the loss causes for manual review;

[0192] Push loss causes with confidence levels lower than the preset range to the system front end.

[0193] In some optional embodiments, the processor is configured to obtain the location information of the pipeline network transmission and distribution loss based on the parameter information of the loss point and the occurrence time of the loss point in the following manner:

[0194] Based on the parameter information of the loss points, a preset number of loss points with relatively strong feature change frequencies are set as a loss point set;

[0195] For each set of loss points, do the following:

[0196] Detect whether there is a proximity relationship between the loss points in the loss point cluster;

[0197] If there is a proximity relationship, the two loss points with the earliest occurrence time are selected as the main loss collection points;

[0198] If there is no adjacent relationship, a loss subgraph is generated;

[0199] Sort the loss subgraphs by position, and set the loss point corresponding to the upstream loss subgraph as the main loss collection point;

[0200] Connect to the main loss collection points to obtain the location information of pipeline network transmission and distribution losses.

[0201] Device Example

[0202] An embodiment of the present application also provides an electronic device, which includes a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of any of the above methods are implemented. The specific implementation method is consistent with the implementation method and the technical effect achieved in the above method embodiment, and some contents will not be repeated here.

[0203] See also Figure 5 , Figure 5 A structural framework diagram of an electronic device provided in an embodiment of the present application is shown.

[0204] The electronic device includes at least one memory 210, at least one processor 220, and a bus 230 connecting different platform systems.

[0205] The memory 210 may include a readable medium in the form of a volatile memory, such as a random access memory (RAM) 211 and / or a cache memory 212 , and may further include a read-only memory (ROM) 213 .

[0206] The memory 210 also stores a computer program, which can be executed by the processor 220, so that the processor 220 implements the steps of any of the above methods.

[0207] The memory 210 may also include a utility 214 having at least one program module 215, such program module 215 including but not limited to: an operating system, one or more application programs, other program modules, and program data, each of which or some combination may include an implementation of a network environment.

[0208] Accordingly, the processor 220 may execute the aforementioned computer program and the utility 214 .

[0209] The processor 220 may be implemented as one or more application specific integrated circuits (ASICs), DSPs, programmable logic devices (PLDs), complex programmable logic devices (CPLDs), field programmable gate arrays (FPGAs), or other electronic components.

[0210] Bus 230 may represent one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, an accelerated graphics port, a processor, or a local bus using any of a variety of bus architectures.

[0211] The electronic device may also communicate with one or more external devices 240, such as a keyboard, pointing device, Bluetooth device, etc., and may also communicate with one or more devices capable of interacting with the electronic device, and / or any device that enables the electronic device to communicate with one or more other computing devices (e.g., a router, modem, etc.). Such communication may be performed via input / output interface 250. Furthermore, the electronic device may also communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) via network adapter 260. The network adapter 260 may communicate with other modules of the electronic device via bus 230. It should be understood that, although not shown in the figures, other hardware and / or software modules may be used in conjunction with the electronic device, including but not limited to microcode, device drivers, redundant processors, external disk drive arrays, RAID systems, tape drives, and data backup storage platforms.

[0212] System Example

[0213] See also Figure 6 , Figure 6 A structural schematic diagram of a steam pipe network transmission and distribution loss locating system provided in an embodiment of the present application is shown.

[0214] The present application also provides a steam pipe network transmission and distribution loss location system, the system comprising:

[0215] The above electronic equipment.

[0216] Media Examples

[0217] An embodiment of the present application also provides a chip, which stores a computer program. When the computer program is executed by a processor, the steps of any of the above methods are implemented. The specific implementation method is consistent with the implementation method and the technical effect achieved in the above method embodiment, and some contents will not be repeated here.

[0218] See also Figure 7 , Figure 7 A schematic diagram of the structure of a program product provided in an embodiment of the present application is shown.

[0219] The program product is used to implement any of the above methods. The program product can use a portable compact disc read-only memory (CD-ROM) and include program code, and can be run on a terminal device, such as a personal computer. However, the program product of the present invention is not limited to this. In the embodiment of the present application, the readable storage medium can be any tangible medium that contains or stores a program, and the program can be used by or in combination with an instruction execution system, device or device. The program product can use any combination of one or more readable media. The readable medium can be a readable signal medium or a readable storage medium. The readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of readable storage media (a non-exhaustive list) include: an electrical connection with one or more wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0220] The chip may include a data signal propagated in baseband or as part of a carrier wave, which carries readable program code. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the foregoing. The readable storage medium may also be any readable medium that can send, propagate, or transmit a program for use by or in conjunction with an instruction execution system, device, or component. The program code contained on the readable storage medium may be transmitted using any suitable medium, including but not limited to wireless, wired, optical cable, RF, or any suitable combination of the foregoing. The program code for performing the operations of the present invention may be written in any combination of one or more programming languages, including object-oriented programming languages ​​such as Java and C++, as well as conventional procedural programming languages ​​such as C, Python, or similar programming languages. The program code may be executed entirely on the user computing device, partially on the user computing device, as a standalone software package, partially on the user computing device and partially on a remote computing device, or entirely on a remote computing device or server. Where a remote computing device is involved, the remote computing device may be connected to the user computing device through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computing device (e.g., through the Internet using an Internet service provider).

[0221] This application is explained from the perspectives of purpose of use, effectiveness, progress and novelty, and has complied with the functional enhancement and use requirements emphasized by the Patent Law. The above description and drawings of this application are only preferred embodiments of this application and are not intended to limit this application. Therefore, all structures, devices, features, etc. that are similar or identical to those of this application, that is, all equivalent replacements or modifications made in accordance with the scope of the patent application of this application, should fall within the scope of protection of the patent application of this application.

Claims

1. A method for locating steam pipe network transmission and distribution losses, characterized in that: Dynamic monitoring, positioning, and storage of steam network transmission and distribution losses at multiple time scales based on multi-dimensional data include: Collect historical time series data of the pipeline network and obtain the loss points and their occurrence times. The historical time series data of the pipeline network includes: the flow rate of gas in the pipeline, the pressure of gas in the pipeline, the temperature of gas in the pipeline, and the switch status of each valve; Based on the occurrence time of the loss point, analyze the historical time series data of the pipeline network to obtain the parameter information of the loss point; Based on the parameter information of the loss point, the neural network model is used to perform loss attribution analysis on the loss point to obtain the attribution result; Review the attribution results based on the preset interval; Based on the parameter information of the loss point and the occurrence time of the loss point, the location information of the pipeline network transmission and distribution loss is obtained; Store the parameter information of the loss point, attribution results and location information of the pipeline network transmission and distribution loss; The parameter information of the loss point includes loss duration, loss pattern, loss trend and characteristic change frequency; The obtaining of location information of pipeline network transmission and distribution loss based on parameter information of the loss point and the occurrence time of the loss point includes: Based on the parameter information of the loss points, a preset number of loss points with relatively strong feature change frequencies are set as a loss point set; For each set of loss points, do the following: Detect whether there is a proximity relationship between the loss points in the loss point cluster; If there is a proximity relationship, the two loss points with the earliest occurrence time are selected as the main loss collection points; If there is no adjacent relationship, a loss subgraph is generated; Sort the loss subgraphs by position, and set the loss point corresponding to the upstream loss subgraph as the main loss collection point; Connect to the main loss collection points to obtain the location information of pipeline network transmission and distribution losses.

2. The method for locating steam pipe network transmission and distribution loss according to claim 1, characterized in that: The collection of historical time series data of the pipeline network and the acquisition of loss points and their occurrence times include: Online access to the pipeline network online operation system and pipeline network database; Pulling historical time series data of the pipeline network for a preset inspection task within a preset period from the pipeline network database, wherein the historical time series data is multi-dimensional data; Perform differential processing on historical time series data; Cluster analysis is performed on the historical time series data after difference processing to obtain the loss points and the time when the loss points occur.

3. The method for locating steam pipe network transmission and distribution loss according to claim 1, characterized in that: The loss attribution analysis of the loss points using a neural network model based on the parameter information of the loss points to obtain attribution results includes: Inputting the loss point to be measured into the neural network model to obtain an attribution result of the loss point to be measured, wherein the attribution result of the loss point to be measured is used to indicate the loss cause of the loss point to be measured and the confidence level corresponding to the loss cause, wherein the neural network model is a variational Bayesian neural network model; The training process of the variational Bayesian neural network model includes: Acquire a training set, the training set including a plurality of training data, each of the training data including parameter information of a sample pipeline network transmission and distribution loss and labeled data of attribution results of the sample loss point; For each training data in the training set, perform the following processing: Inputting parameter information of sample pipeline network transmission and distribution losses in the training data into a preset deep learning model to obtain predicted data of attribution results of the sample loss points; Updating the model parameters of the deep learning model based on the predicted data and the labeled data of the attribution result of the sample loss point; Check whether the preset training end condition is met; if so, use the trained deep learning model as the variational Bayesian neural network model.

4. The method for locating steam pipe network transmission and distribution loss according to claim 1, characterized in that: The attribution result is reviewed based on the preset interval, including: Push loss causes with confidence levels higher than the preset range to the system platform and automatically match the corresponding preset emergency strategies; Mark the loss causes with confidence levels within a preset range, and record the characteristic snapshots and waveform trend graphs of the loss points corresponding to the loss causes for manual review; Push loss causes with confidence levels lower than the preset range to the system front end.

5. A pipeline network transmission and distribution loss detection device, characterized in that: The pipeline network transmission and distribution loss detection device includes: The data acquisition module is used to collect historical time series data of the pipeline network and obtain the loss points and their occurrence time. The historical time series data of the pipeline network includes: the flow rate of gas in the pipeline, the pressure of gas in the pipeline, the temperature of gas in the pipeline and the switch status of each valve; The parameter information acquisition module is used to analyze the historical time series data of the pipeline network based on the occurrence time of the loss point to obtain the parameter information of the loss point; The attribution result acquisition module is used to perform loss attribution analysis on the loss points using a neural network model based on the parameter information of the loss points to obtain the attribution results; An attribution result review module is used to review the attribution results based on a preset interval; The loss location module is used to obtain the location information of the pipeline network transmission and distribution loss based on the parameter information of the loss point and the occurrence time of the loss point; The loss storage module is used to store the parameter information of the loss point, the attribution results and the location information of the pipeline network transmission and distribution loss; The parameter information of the loss point includes loss duration, loss pattern, loss trend and characteristic change frequency; The obtaining of location information of pipeline network transmission and distribution loss based on parameter information of the loss point and the occurrence time of the loss point includes: Based on the parameter information of the loss points, a preset number of loss points with relatively strong feature change frequencies are set as a loss point set; For each set of loss points, do the following: Detect whether there is a proximity relationship between the loss points in the loss point cluster; If there is a proximity relationship, the two loss points with the earliest occurrence time are selected as the main loss collection points; If there is no adjacent relationship, a loss subgraph is generated; Sort the loss subgraphs by position, and set the loss point corresponding to the upstream loss subgraph as the main loss collection point; Connect to the main loss collection points to obtain the location information of pipeline network transmission and distribution losses.

6. An electronic device, characterized in that: The electronic device includes a memory and a processor, the memory stores a computer program, and the processor is configured to execute the method according to any one of claims 1 to 4.

7. A steam pipe network transmission and distribution loss positioning system, characterized in that: The steam pipe network transmission and distribution loss positioning system includes: The electronic device according to claim 6.

8. A chip, characterized in that: The chip stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 4 is implemented.

Images