Artificial intelligence abnormal heating troubleshooting method and system

Abstract

The present invention provides an artificial intelligence-based abnormal heating troubleshooting method and system, applicable to the field of data processing technology. The method comprises: obtaining heating data; determining the high-order spatiotemporal characteristics of the heating data based on the heating data and a preset artificial intelligence algorithm; obtaining a sparse coupling factor matrix corresponding to the thermal expansion of the heating pipeline; obtaining an abnormality probability matrix and a set of propagation paths for the heating pipeline based on the high-order spatiotemporal characteristics of the heating data and the sparse coupling factor matrix corresponding to the thermal expansion of the heating pipeline; and obtaining the damage location of the heating pipeline based on the abnormality probability matrix and the set of propagation paths for the heating pipeline. The present invention improves the accuracy of abnormality location, shortens troubleshooting time, and reduces operation and maintenance costs, thereby improving troubleshooting efficiency and effectiveness.

Description

Technical Field

[0001] The present invention relates to data processing technology, and in particular to an artificial intelligence abnormal heating troubleshooting method and system. Background Art

[0002] During operation, centralized heating systems often experience heating anomalies such as thermal imbalance, pipe leaks, and air or sewage blockages due to factors such as complex pipe networks, environmental interference, and aging equipment. These issues not only lead to energy waste (such as hot water loss due to pipe leaks) and uneven room temperatures among users, but can also cause equipment failures and even safety incidents. For example, the "high hysteresis" characteristic of heating systems (where regulation lags behind actual demand changes) can lead to the normalization of data deviations, masking systemic anomalies and making it difficult for traditional detection methods to accurately identify the root cause of the problem. Furthermore, factors such as corrosion in aging pipe networks, unauthorized water discharge by users, and design and construction defects further exacerbate the complexity of heating anomalies.

[0003] At present, the existing abnormal heating troubleshooting methods are highly dependent on manual labor and have a low degree of automation, resulting in inaccurate positioning and low efficiency, which in turn affects the efficiency and effectiveness of the troubleshooting. Summary of the Invention

[0004] Based on the above problems, the present invention is proposed to provide an artificial intelligence abnormal heating troubleshooting method and system that overcomes the above problems or at least partially solves the above problems.

[0005] According to one aspect of the present invention, there is provided an artificial intelligence abnormal heating troubleshooting method, comprising the following steps:

[0006] Obtain heating data;

[0007] Based on heating data and preset artificial intelligence algorithms, determine the high-order spatiotemporal characteristics of heating data;

[0008] Obtain the sparse coupling factor matrix corresponding to the thermal expansion of the heating pipeline;

[0009] Based on the high-order spatiotemporal characteristics of the heating data and the sparse coupling factor matrix corresponding to the thermal expansion of the heating pipeline, the abnormal probability matrix of the heating pipeline and the propagation path set of the heating pipeline are obtained;

[0010] Based on the abnormal probability matrix of the heating pipeline and the propagation path set of the heating pipeline, the damage location of the heating pipeline is obtained.

[0011] Optionally, based on the heating data and a preset artificial intelligence algorithm, high-order spatiotemporal features of the heating data are determined, including:

[0012] Input the heating data into the preset artificial intelligence algorithm to obtain the spatiotemporal coding matrix and adjacency matrix;

[0013] Generate a dynamic spatiotemporal propagation graph sequence based on the spatiotemporal coding matrix and adjacency matrix;

[0014] Based on the dynamic spatiotemporal propagation graph sequence, the high-order spatiotemporal characteristics of heating data are determined.

[0015] Optionally, the heating data is input into a preset artificial intelligence algorithm to obtain a spatiotemporal coding matrix and an adjacency matrix, including:

[0016] Use the preset artificial intelligence algorithm to process the heating data, establish the heating network topology map, and generate the adjacency matrix based on the heating network topology map;

[0017] Perform time series alignment on the heating data according to the time window to obtain aligned data;

[0018] The historical fault labels are concatenated with the aligned data to obtain the spatiotemporal encoding matrix.

[0019] Optionally, based on the spatiotemporal coding matrix and the adjacency matrix, a dynamic spatiotemporal propagation graph sequence is generated, including:

[0020] Calculate the similarity between data in the spatiotemporal coding matrix;

[0021] Determine the fusion weight based on the similarity and historical propagation pattern between the data in the spatiotemporal coding matrix;

[0022] Determine the target edge set of the adjacency matrix based on the fusion weight and the preset threshold;

[0023] Based on the target edge set, a dynamic spatiotemporal propagation graph sequence is generated.

[0024] Optionally, based on the dynamic spatiotemporal propagation graph sequence, high-order spatiotemporal features of the heating data are determined, including:

[0025] Obtain spatial graph convolution and temporal convolution of dynamic spatiotemporal propagation graph sequences;

[0026] The spatial graph convolution and temporal convolution of the dynamic spatiotemporal propagation graph sequence are gated and fused to obtain high-order spatiotemporal features.

[0027] Optionally, obtaining a sparse coupling factor matrix corresponding to the thermal expansion of the heating pipeline includes:

[0028] The high-order spatiotemporal characteristics of the heating data are coupled and calculated to obtain the initial coupling factor matrix;

[0029] The initial coupling factor matrix is ​​topologically modified by the thermal expansion of the heating pipe to obtain a modified coupling factor matrix;

[0030] The corrected coupling factor matrix is ​​normalized to obtain a sparse coupling factor matrix.

[0031] or,

[0032] Calculate the first quantile corresponding to the covariance matrix of the high-order spatiotemporal characteristics of the heating data;

[0033] Extract the diagonal elements in the spatiotemporal coding matrix, and multiply the first quantile corresponding to the covariance matrix of the high-order spatiotemporal characteristics of the heating data by the diagonal elements in the spatiotemporal coding matrix to obtain the first quantile set of the covariance matrix of the high-order spatiotemporal characteristics of the heating data;

[0034] Calculate the second quantile corresponding to the covariance matrix corresponding to the thermal expansion of the heating pipe;

[0035] The target quantile is obtained by adding the first quantile corresponding to the covariance matrix of the high-order spatiotemporal characteristics of the heating data and the second quantile corresponding to the covariance matrix corresponding to the thermal expansion of the heating pipe;

[0036] Performing a dot product between the target quantile and each quantile in the first quantile set of the high-order spatiotemporal feature covariance matrix of the heating data, to obtain the second quantile set of the high-order spatiotemporal feature covariance matrix of the heating data;

[0037] The maximum quantile in the second quantile set is used to replace the central element in the transposed matrix of the space-time coding matrix and the central element in the covariance matrix corresponding to the thermal expansion of the heating pipe, and the transposed matrix of the replaced space-time coding matrix and the covariance matrix corresponding to the thermal expansion of the heating pipe are added to obtain a sparse coupling factor matrix.

[0038] Optionally, based on the high-order spatiotemporal characteristics of the heating data and the sparse coupling factor matrix corresponding to the thermal expansion of the heating pipeline, a target abnormality probability matrix of the heating pipeline and a propagation path set of the heating pipeline are obtained, including:

[0039] Perform time series anomaly detection on the high-order spatiotemporal features of the heating data to obtain the initial anomaly probability matrix of the heating pipeline;

[0040] The initial abnormality probability matrix of the heating pipeline is added to the sparse coupling factor matrix corresponding to the thermal expansion of the heating pipeline to obtain the target abnormality probability matrix of the heating pipeline;

[0041] The target anomaly probability matrix of the heating pipeline is subjected to spatial propagation enhancement and path backtracing to obtain the propagation path set of the heating pipeline.

[0042] or,

[0043] Extracting high-order spatiotemporal features of heating pipes from the high-order spatiotemporal features of heating data;

[0044] Obtaining a first covariance feature corresponding to the high-order spatiotemporal feature of the heating pipeline, and obtaining a second covariance feature corresponding to the sparse coupling factor matrix corresponding to the thermal expansion of the heating pipeline;

[0045] Merge the first covariance feature corresponding to the high-order spatiotemporal feature of the heating pipeline with the second covariance feature corresponding to the sparse coupling factor matrix corresponding to the thermal expansion of the heating pipeline to obtain a covariance feature set of the heating pipeline;

[0046] A set consisting of elements greater than the maximum value element in the transfer matrix of the high-order spatiotemporal characteristics of the heating pipeline is selected from the covariance feature set of the heating pipeline as the initial abnormality probability matrix of the heating pipeline;

[0047] The probability values ​​in the initial abnormal probability matrix of the heating pipeline are corrected by the target element in the covariance feature set of the heating pipeline to obtain the target abnormal probability matrix of the heating pipeline, wherein the target element in the covariance feature set of the heating pipeline refers to the element with the middle value in the covariance feature set of the heating pipeline;

[0048] The target anomaly probability matrix of the heating pipeline is subjected to spatial propagation enhancement and path backtracing to obtain the propagation path set of the heating pipeline.

[0049] Optionally, based on the target abnormality probability matrix of the heating pipeline and the propagation path set of the heating pipeline, obtaining the damage location of the heating pipeline includes:

[0050] The probability in the abnormal probability matrix of the heating pipeline is searched along the propagation path in the propagation path set of the heating pipeline. When the maximum probability is reached, the target node is obtained, and the damage position of the heating pipeline is determined by the target node.

[0051] According to another aspect of the present invention, there is provided an artificial intelligence abnormal heating troubleshooting system, comprising:

[0052] A first acquisition module is used to acquire heating data;

[0053] A feature determination module is used to determine the high-order spatiotemporal features of the heating data based on the heating data and a preset artificial intelligence algorithm;

[0054] The second acquisition module is used to obtain a sparse coupling factor matrix corresponding to the thermal expansion of the heating pipeline;

[0055] The third acquisition module is used to obtain the abnormal probability matrix of the heating pipeline and the propagation path set of the heating pipeline based on the high-order spatiotemporal characteristics of the heating data and the sparse coupling factor matrix corresponding to the heating pipeline;

[0056] The location determination module is used to obtain the damage location of the heating pipeline based on the abnormal probability matrix of the heating pipeline and the propagation path set of the heating pipeline.

[0057] According to the solution of the present invention, in this invention, heating data is first acquired. Then, based on the heating data and a preset artificial intelligence algorithm, the high-order spatiotemporal characteristics of the heating data are determined. Then, a sparse coupling factor matrix corresponding to the thermal expansion of the heating pipe is obtained. Based on the high-order spatiotemporal characteristics of the heating data and the sparse coupling factor matrix corresponding to the thermal expansion of the heating pipe, an abnormality probability matrix and a set of propagation paths of the heating pipe are obtained. Then, based on the abnormality probability matrix and the set of propagation paths of the heating pipe, the damage location of the heating pipe is determined. This invention improves the accuracy of abnormality location, shortens troubleshooting time, and reduces operation and maintenance costs, thereby improving the efficiency and effectiveness of troubleshooting. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] Figure 1 A flowchart of an artificial intelligence abnormal heating troubleshooting method according to an embodiment of the present invention is shown;

[0059] Figure 2 A structural block diagram of an artificial intelligence abnormal heating troubleshooting system according to an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0060] Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art.

[0061] In order to solve the above problems in the prior art, the present invention proposes a solution. One embodiment of the present invention provides a patent document management method and device based on artificial intelligence.

[0062] like Figure 1 As shown, the present embodiment proposes an artificial intelligence abnormal heating troubleshooting method, comprising the following steps:

[0063] Step S101: Acquire heating data.

[0064] Among them, heating data includes but is not limited to water supply temperature, return water pressure, flow, time, fault tags, etc.

[0065] Step S102: Based on the heating data and a preset artificial intelligence algorithm, determine the high-order spatiotemporal characteristics of the heating data.

[0066] Optionally, based on the heating data and a preset artificial intelligence algorithm, high-order spatiotemporal characteristics of the heating data are determined, including: inputting the heating data into a preset artificial intelligence algorithm to obtain a spatiotemporal coding matrix and an adjacency matrix; generating a dynamic spatiotemporal propagation graph sequence based on the spatiotemporal coding matrix and the adjacency matrix; and determining high-order spatiotemporal characteristics of the heating data based on the dynamic spatiotemporal propagation graph sequence.

[0067] The preset artificial intelligence algorithm is spatiotemporal propagation graph convolution.

[0068] Among them, the heating data is input into the preset artificial intelligence algorithm to obtain the spatiotemporal coding matrix and adjacency matrix, including: using the preset artificial intelligence algorithm to process the heating data, establish a heating pipe network topology map, and generate an adjacency matrix based on the heating pipe network topology map; align the heating data in time series according to the time window to obtain aligned data; splice the historical fault labels with the aligned data to obtain the spatiotemporal coding matrix.

[0069] As an example, we first create a heating network topology and adjacency matrix. For example, consider a residential heating network consisting of 100 heat exchanger stations. Using a geographic information system (GIS), we map the network connections to create a topology. If nodes are directly connected, the corresponding position in the adjacency matrix is ​​set to 1 (for example, if heat exchanger stations B and C are directly connected via a pipeline, the value in row p and column q of the adjacency matrix is ​​1). Otherwise, the value is 0.

[0070] The heating data is then time-series aligned according to a time window to generate aligned data. A 15-minute time window is set, and real-time data (such as supply water temperature, return water pressure, and flow rate) is collected from each node. These data are aligned by timestamp to form a time series, known as aligned data. For example, there are 96 time points between 0:00 AM and 12:00 PM on January 1, 2024. Each time point corresponds to 10-dimensional data (temperature, pressure, flow rate, etc.) from 100 nodes, resulting in a 96×100×10 three-dimensional dataset. Historical fault labels are then concatenated with the aligned data to generate a spatiotemporal encoding matrix. For example, if a pipeline leak occurs at 8:15 AM on January 1, 2024, at a heat exchange station, the fault label at that time point (e.g., "leak" corresponds to code 01) is concatenated with the time series data of the corresponding node, resulting in a spatiotemporal encoding matrix with dimensions of 96×100×(10+2) (the second dimension represents the fault label code).

[0071] The embodiment of the present application clearly presents the pipe network structure: a heating pipe network topology diagram and an adjacency matrix are established, which can intuitively display the connection relationship between the nodes of each heat exchange station in the heating pipe network, making it convenient for staff to quickly understand the pipe network layout, and providing basic network structure information for subsequent fault location and analysis. Unified data format and time standard: by setting a time window to align the heating data in time series, various heating data at different times and nodes (such as water supply temperature, return water pressure, flow, etc.) are organized into a unified three-dimensional data set format, so that the data is consistent in time and dimension, which is convenient for subsequent analysis and processing, and improves the availability and accuracy of the data. Fusion of fault information and data: splicing historical fault labels with aligned data to form a spatiotemporal coding matrix, closely combining the fault information with the heating data of the corresponding time points and nodes, providing rich feature information for fault prediction and diagnosis based on data mining and machine learning, helping to improve the accuracy of fault identification and location, and better ensure the stable operation of the heating system.

[0072] Among them, based on the space-time coding matrix and the adjacency matrix, a dynamic space-time propagation graph sequence is generated, including: calculating the similarity between the data in the space-time coding matrix; determining the fusion weight based on the similarity between the data in the space-time coding matrix and the historical propagation pattern; determining the target edge set of the adjacency matrix based on the fusion weight and the preset threshold; and generating a dynamic space-time propagation graph sequence based on the target edge set.

[0073] For example, the similarity between data in the spatiotemporal coding matrix is ​​calculated: for the node data at adjacent time points in the spatiotemporal coding matrix (such as the temperature and pressure data of a heat exchange station at time t and time t+1), the similarity is calculated using the Euclidean distance:

[0074]

[0075] in, are the feature vectors of nodes i and j respectively, and n is the feature dimension (e.g. 12 dimensions, including 10-dimensional operating data + 2-dimensional fault labels).

[0076] Then determine the fusion weight: Combined with historical fault propagation records (for example, in the past year, when heat exchange station A fails, heat exchange station B has an 80% probability of subsequently experiencing an abnormality), modify the similarity weight:

[0077] Fusion weight = ,

[0078] in, =0.6 is an empirical coefficient that balances the impact of real-time data and historical patterns.

[0079] Finally, the target edge set is filtered and a dynamic spatiotemporal propagation graph sequence is generated: the preset threshold is set to 0.7, and only edges with a fusion weight ≥ 0.7 are retained (for example, if the weight of heat exchange stations A and B is 0.8, then edge AB is retained). The dynamic spatiotemporal propagation graph sequence is generated according to the time window (15 minutes), and each time point corresponds to a spatiotemporal propagation graph containing the node connection relationship.

[0080] The embodiment of the present application accurately captures data associations: by calculating the similarity between data in the spatiotemporal coding matrix, it can accurately measure the similarity between data at different time points and different nodes, and dig out the potential connections in time and space between the operating states of each node in the heating system, providing quantitative correlation information for subsequent analysis. Combining historical and real-time data: Determine the fusion weight based on data similarity and historical propagation patterns, which not only takes into account the current system state reflected by real-time data, but also incorporates the empirical knowledge of historical fault propagation, making the analysis results more reliable and forward-looking, and can more comprehensively reflect the operating laws and fault propagation characteristics of the heating system. Effectively screen key connections: Determine the target edge set of the adjacency matrix based on the fusion weight and the preset threshold, and filter out those connections with weaker correlations and less impact on system operation and fault propagation, highlight the connections between key nodes, simplify the network structure, and focus more on the parts that have an important impact on the system, which helps to improve the efficiency of fault location and analysis. Intuitive display of system dynamics: The generated dynamic spatiotemporal propagation diagram sequence visually presents the node connection relationship and state changes of the heating system at different time points, allowing staff to intuitively observe the dynamic evolution process of the system, quickly discover abnormal changes and potential fault propagation paths, and provide a powerful decision-making support tool for the monitoring and management of the heating system, thereby improving monitoring efficiency and effectiveness.

[0081] Among them, based on the dynamic space-time propagation graph sequence, the high-order space-time characteristics of the heating data are determined, including: obtaining the spatial graph convolution and temporal convolution of the dynamic space-time propagation graph sequence; performing gated fusion on the spatial graph convolution and temporal convolution of the dynamic space-time propagation graph sequence to obtain high-order space-time characteristics.

[0082] For example, for the spatiotemporal propagation graph at each time point, a graph convolutional network is used to extract spatial features:

[0083] in, is the adjacency matrix + identity matrix, is the degree matrix, is the node feature of the lth layer, The weight matrix is ​​used to extract the spatial dependencies between nodes (such as the pressure correlation between adjacent heat exchange stations). The time series features of each node (such as the temperature series over the past 24 hours) are input into the time convolutional network to capture the dynamic changes in the time dimension and output a time feature vector (such as the temperature change trend of a node at time t).

[0084] Fusion of spatial and temporal features through gating mechanism:

[0085]

[0086]

[0087] in, is the GCN output, is the TCN output, for The weight of for The weight of is a constant, To multiply element by element, we finally get high-order features containing spatiotemporal coupling relationships.

[0088] The embodiments of the present application deeply capture spatiotemporal coupling relationships and enhance feature representation capabilities: Spatial Graph Convolution (GCN), based on the topological structure of a dynamic spatiotemporal propagation graph, can effectively extract spatial dependencies between nodes (such as the pressure and flow correlation between adjacent heat exchange stations), breaking the traditional method's reliance on "Euclidean spatial data," adapting to the physical connection characteristics of the irregular graph structure of the heating pipe network, and accurately modeling the direct and indirect impacts between nodes. Temporal Convolution (TCN) focuses on the dynamic changes of time series data (such as the fluctuation trends of temperature and pressure over time), capturing the periodicity, seasonality, and short-term abnormal fluctuations of the heating system operation, and avoiding the one-sidedness of relying solely on data at a single time point. The gated fusion mechanism dynamically adjusts the importance of spatiotemporal features through adaptive weighting (such as increasing the weight of temporal features when a fault occurs and focusing on spatial correlation when the pipe network is in a steady state), achieving the organic fusion of spatiotemporal information, generating high-order features containing complex spatiotemporal coupling relationships, and avoiding the insufficient feature representation caused by the fragmentation of spatiotemporal information.

[0089] Furthermore, it adapts to dynamic and complex systems, enhancing the ability to identify abnormal patterns. The operating state of a heating system is highly dynamic and nonlinear, influenced by multiple factors such as ambient temperature, user load, and equipment aging. High-order spatiotemporal features can capture the following key information: spatial propagation patterns, such as the path and intensity of the impact of a node's pressure anomaly on upstream and downstream nodes through the pipeline network; temporal evolution trends, such as the abnormal downward trend of the temperature at the same node within a continuous time window; and spatiotemporal interaction features, such as collective parameter shifts within a specific spatial region (e.g., a cluster of heat exchange stations in a certain area) within the same time window (which may indicate regional leaks or equipment failures). These features provide richer input for subsequent anomaly detection and fault location, significantly improving the recognition accuracy of complex scenarios such as "multi-node coordinated anomalies" and "progressive fault development," while avoiding missed detections or misjudgments caused by single-dimensional features (e.g., only space or time).

[0090] Furthermore, it supports multi-scale analysis to improve system monitoring and decision-making efficiency: Spatial Multi-Scale: Graph convolution operations aggregate neighbor node information layer by layer through a multi-layer network (e.g., direct connections between first-order neighbors and indirect connections between second-order neighbors), enabling multi-scale extraction from "local features of a single node" to "global structural features of the pipeline network." This is suitable for comprehensive analysis, from micro-level node failures to macro-level pipeline network operational status. Temporal Multi-Scale: Temporal convolution, through varying dilation rates (dilated convolution) or window sizes, can capture both short-term fluctuations (minute-level) and long-term trends (hourly / day-level), meeting the dual needs of heating systems for real-time monitoring (such as instant leak detection) and trend prediction (such as equipment aging warnings). The multi-scale nature of high-order features enables the system to quickly locate the specific node where an anomaly occurs (micro-level) and simultaneously analyze the anomaly's propagation and potential risks within the pipeline network (macro-level), providing precise guidance for dispatching maintenance resources and developing emergency response plans.

[0091] Furthermore, it provides high-quality input for intelligent algorithms, enabling precise fault location: the extracted high-order spatiotemporal features contain clear "fault-sensitive information" (such as the spatiotemporal characteristic patterns of historical fault nodes) and can be directly used as input for: anomaly probability matrix calculation: by comparing the similarity of current features with historical fault features, the anomaly probability of each node is quantified; propagation path search: combining the edge weights (fusion weights) of the dynamic spatiotemporal propagation graph, the key paths of anomaly propagation (such as pipelines corresponding to high-weight edges) are identified, implementing the precise positioning logic of "searching for the maximum probability along the propagation path" (such as the damage location determination method initially mentioned by the user). Compared to raw data or low-order features, high-order spatiotemporal features can significantly reduce data noise interference and improve the algorithm's sensitivity to "weak anomaly signals" (such as the subtle parameter changes that indicate early minor pipeline leaks), achieving "early detection and early location" of faults.

[0092] In summary, by extracting high-order spatiotemporal features, heating system monitoring has achieved an upgrade from "single-node data monitoring" to "spatiotemporal correlation pattern analysis", providing more comprehensive and accurate information support for anomaly detection, fault location, and system optimization, ultimately improving the reliability, operational efficiency, and intelligent operation and maintenance level of the heating system.

[0093] Step S103: Obtain a sparse coupling factor matrix corresponding to the thermal expansion of the heating pipeline.

[0094] In one embodiment, a sparse coupling factor matrix corresponding to the thermal expansion of a heating pipe is obtained, including: performing coupling calculation on high-order spatiotemporal features of heating data to obtain an initial coupling factor matrix; performing topological correction on the initial coupling factor matrix according to the thermal expansion of the heating pipe to obtain a corrected coupling factor matrix; and normalizing the corrected coupling factor matrix to obtain a sparse coupling factor matrix.

[0095] For example, for the high-order spatiotemporal feature matrix (N is the number of nodes, F is the feature dimension), calculate the coupling degree between nodes i and j:

[0096]

[0097] Get the initial coupling factor matrix , represents the feature similarity between nodes.

[0098] Then, the thermal expansion A of the heating pipes is used to filter the coupling of non-adjacent nodes (only the node pairs that are actually connected in the pipe network are retained):

[0099] like , ;otherwise, .

[0100] Finally, the corrected matrix is ​​normalized row by row to obtain the sparse coupling factor matrix S:

[0101]

[0102] Ensure that the sum of elements in each row is 1, highlighting the influence weight between adjacent nodes.

[0103] The embodiment of the present application integrates spatiotemporal features with physical topology to improve the accuracy of the coupling relationship: data-driven and physical constraints are combined: the initial coupling factor matrix calculates the feature similarity between nodes based on high-order spatiotemporal features (integrating the real-time operating data of the nodes, historical fault modes, and spatiotemporal dynamic associations), reflecting the actual dependency of the nodes in terms of operating status; non-adjacent nodes are filtered through the adjacency matrix to ensure that the coupling relationship only exists between physically connected nodes in the pipe network, avoiding non-real associations of nodes without actual connections, so that the coupling relationship conforms to the dynamic characteristics of data-driven and strictly follows the physical topology of the heating pipe network, thereby improving the physical interpretability and reliability of the model. Suppressing noise interference: eliminating invalid coupling of non-adjacent nodes, reducing the interference of irrelevant nodes on the propagation path analysis, focusing on the feature propagation on the real pipeline connection, making subsequent abnormality probability calculation and damage location positioning more accurate.

[0104] In addition, the model emphasizes the influence of key nodes to enhance model robustness. Normalization optimizes weight distribution: By normalizing rows so that the sum of each row is 1, the relative influence between adjacent nodes is clearly quantified (for example, when a node fails, the coupling weight ratio of its directly connected nodes is standardized). This avoids weight bias caused by differences in feature dimensions and facilitates subsequent propagation path searches based on probability matrices (for example, prioritizing abnormal propagation of high-weight connections). Sparsification reduces computational complexity: By retaining only the coupling relationships of physically connected nodes, the number of non-zero elements in the matrix is ​​significantly reduced. This reduces data dimensionality while maintaining key connection information, improving the efficiency of subsequent algorithms (such as path search and probability propagation calculations) and meeting the efficiency requirements of real-time monitoring scenarios.

[0105] Furthermore, it provides structured input for abnormal propagation analysis and supports precise positioning: clarifying the influence intensity between nodes: the sparse coupling factor matrix clearly depicts the correlation intensity of adjacent nodes in the pipeline network in terms of spatiotemporal characteristics (such as the transmission efficiency of temperature and pressure changes), and provides a quantitative basis for modeling the propagation path of abnormal probability in the pipeline network. For example, when an abnormality occurs at a node, the matrix weights can be used to quickly locate the highly correlated adjacent nodes, and the maximum probability can be searched preferentially along the propagation path to shorten the positioning time of the damaged location. Support dynamic propagation modeling: combining the dynamic spatiotemporal propagation graph sequence and high-order spatiotemporal characteristics, the sparse coupling factor matrix provides key parameters for constructing a dynamic model of pipeline network abnormal propagation, enabling the model to capture the spatiotemporal evolution of faults in physical connections (such as the transmission of pressure and flow changes from upstream to downstream nodes of leakage faults), thereby more accurately locking the source of the abnormality or the location of the damage.

[0106] In summary, the sparse coupling factor matrix provides a key intermediate representation for the precise positioning of damaged heating pipelines by integrating spatiotemporal features with physical topology, optimizing weight distribution, and reducing computational complexity. It supports the effective implementation from data-driven analysis to practical engineering applications and improves positioning accuracy.

[0107] In another embodiment, a sparse coupling factor matrix corresponding to the thermal expansion of the heating pipe is obtained, including: calculating the first quantile corresponding to the covariance matrix of the high-order spatiotemporal characteristics of the heating data; extracting the diagonal elements in the spatiotemporal coding matrix, and multiplying the first quantile corresponding to the covariance matrix of the high-order spatiotemporal characteristics of the heating data with the diagonal elements in the spatiotemporal coding matrix to obtain a first quantile set of the covariance matrix of the high-order spatiotemporal characteristics of the heating data; calculating the second quantile corresponding to the covariance matrix corresponding to the thermal expansion of the heating pipe; multiplying the first quantile corresponding to the covariance matrix of the high-order spatiotemporal characteristics of the heating data with the diagonal elements in the spatiotemporal coding matrix to obtain a first quantile set of the covariance matrix of the high-order spatiotemporal characteristics of the heating data; The second quantiles corresponding to the covariance matrix corresponding to the expansion are added to obtain the target quantile; the target quantile is point-multiplied with each quantile in the first quantile set of the high-order spatiotemporal feature covariance matrix of the heating data to obtain the second quantile set of the high-order spatiotemporal feature covariance matrix of the heating data; the maximum quantile in the second quantile set is used to replace the central element in the transposed matrix of the spatiotemporal coding matrix and the central element in the covariance matrix corresponding to the thermal expansion of the heating pipe, and the transposed matrix of the replaced spatiotemporal coding matrix is ​​added to the covariance matrix corresponding to the thermal expansion of the heating pipe to obtain the sparse coupling factor matrix.

[0108] For example, the high-order feature matrix of the heating data is first Calculate the covariance matrix ΣH = Cov(H), where ΣH[i,j] represents the covariance between the i-th and j-th spatiotemporal features, reflecting the correlation between their fluctuations. Calculate the first quantile q1 (e.g., the 0.25 quantile, representing the critical value for 25% of the data) for all nonzero elements (or global elements) in ΣH to mark the boundary between "weak correlation" and "strong correlation."

[0109] Then, extract the diagonal elements {E11, E22, …, Emm} of the spatiotemporal encoding matrix E, whose physical meaning is the self-attention weight of each spatiotemporal point (for example, the central node has a higher weight). Multiply q1 with each diagonal element to obtain the first quantile set .

[0110] The covariance matrix ΣY = Cov(Y) is then calculated for the thermal expansion data Y, reflecting the correlation between thermal expansion at different points in time and space. The second quantile q2 of ΣY (e.g., the 0.75 quantile, representing the critical value for 75% of the data) is calculated to mark the boundary of strong thermal expansion correlation.

[0111] Directly add q1 and q2 to obtain qtarget=q1+q2. Through linear superposition, qtarget simultaneously contains the "weak correlation intensity of heating characteristics" and the "strong correlation intensity of thermal expansion", providing a fused threshold for subsequent dot product.

[0112] After that, dot multiplication (element-wise multiplication) of qtarget with the obtained Q1 to obtain .

[0113] Finally, position the center element:

[0114] Assume that the space-time encoding transposed matrix E T The central element of the thermal expansion covariance matrix ΣY is E T [c,c] and ΣY[c,c] (corresponding to core space-time nodes, such as the center of the heat source).

[0115] Element replacement:

[0116] Replace the two central elements mentioned above with the maximum value max(Q2) in Q2 to highlight the leading role of the core node, and obtain the replaced spatiotemporal coding transposed matrix E T 1 and the thermal expansion covariance matrix ΣY1.

[0117] Matrix addition:

[0118] Calculate S = E T 1 +ΣY1, and the final sparse coupling factor matrix is ​​obtained through sparse processing (such as setting a threshold to filter elements below a certain value).

[0119] The embodiment of the present application accurately captures the coupling characteristics of spatiotemporal correlation and thermal expansion: by calculating the first quantile of the high-order spatiotemporal feature covariance matrix and the second quantile of the thermal expansion covariance matrix, the low-frequency key correlations of the heating data in the spatiotemporal dimension (such as long-term temperature change trends, regional thermal transmission laws) and the core influencing factors of the thermal expansion phenomenon (such as the deformation sensitivity threshold of key pipeline nodes) are extracted respectively. The use of quantiles can filter noise, retain statistically significant features in the data distribution, and avoid redundant information interference. The diagonal elements of the spatiotemporal coding matrix (representing the characteristics of each node itself) are multiplied by the first quantile, and then the spatiotemporal features and the key threshold of thermal expansion are fused through the target quantile, so that the coupling matrix simultaneously contains the node's own attributes (such as pipeline material and diameter) and spatiotemporal interaction relationships (such as the thermal conduction effect of adjacent pipelines). This design strengthens the combination of "local features" and "global correlations", avoiding the limitations of traditional methods that only consider a single dimension.

[0120] In addition, this application also constructs a sparse matrix to improve computational efficiency and model generalization: by replacing the central elements of the matrix and adding them together to obtain a sparse coupling factor matrix, only the key connections (non-zero elements) that have a significant impact on thermal expansion are retained, and irrelevant or weakly correlated node couplings are eliminated. This operation greatly reduces the matrix dimension and computational complexity, making it suitable for anomaly detection in large-scale heating networks, avoiding the "curse of dimensionality," and improving the accuracy of anomaly location. By replacing the "central elements" (usually corresponding to key pipeline nodes or hub locations) of the spatiotemporal coding matrix and the thermal expansion covariance matrix, it becomes more sensitive to the thermal expansion effects of core nodes. For example, in a heating network, the deformation of the trunk pipeline has a greater impact on the stability of the overall system. By strengthening the weight of the central elements, the prediction accuracy of such key nodes can be improved, and the noise interference of edge nodes can be avoided, thereby improving the accuracy and precision of anomaly location.

[0121] Furthermore, this application enhances adaptability to complex scenarios: through covariance matrix and quantile operations, it is naturally compatible with the spatiotemporal dimensions of heating data (such as temperature and flow in different time periods and regions) and the physical properties of thermal expansion (such as temperature-deformation function and material thermal expansion coefficient), and can automatically extract cross-modal correlations without manual feature engineering. This has stronger adaptability to nonlinear and time-varying complex coupling relationships in heating systems (such as periodic expansion caused by day and night temperature differences, and changes in material properties in different seasons), and can accurately locate anomalies and avoid positioning errors. Quantile operations are essentially robust statistics on data distribution. Compared with directly using the original covariance or correlation matrix, the introduction of quantiles reduces the interference of extreme values ​​on coupling relationships, improves the stability of the model in noisy environments, and improves the accuracy of the sparse coupling factor matrix, providing reliable support for anomaly detection and improving the accuracy and precision of anomaly detection.

[0122] Therefore, this embodiment uses three core mechanisms: quantile statistics to filter noise, sparse matrix to reduce complexity, and deep coupling of spatiotemporal-physical characteristics. While ensuring computational efficiency, it enhances the robustness to thermal expansion phenomena in heating pipes, provides efficient and reliable technical support for troubleshooting of intelligent heating systems, and thus improves the accuracy and precision of abnormality detection.

[0123] Step S104: Based on the high-order spatiotemporal characteristics of the heating data and the sparse coupling factor matrix corresponding to the thermal expansion of the heating pipeline, the abnormal probability matrix of the heating pipeline and the propagation path set of the heating pipeline are obtained.

[0124] In one embodiment, based on the high-order spatiotemporal characteristics of the heating data and the sparse coupling factor matrix corresponding to the thermal expansion of the heating pipe, the abnormality probability matrix of the heating pipe and the propagation path set of the heating pipe are obtained, including: performing time series anomaly detection on the high-order spatiotemporal characteristics of the heating data to obtain the abnormality probability matrix of the heating pipe; adding the initial abnormality probability matrix of the heating pipe and the sparse coupling factor matrix corresponding to the thermal expansion of the heating pipe to obtain the target abnormality probability matrix of the heating pipe; performing spatial propagation enhancement and path backtracing on the abnormality probability matrix of the heating pipe to obtain the propagation path set of the heating pipe.

[0125] For example, the isolation forest algorithm is first used to detect the high-order spatiotemporal feature sequence of each node, and the abnormal probability of node i at time t is output. (values ​​range from 0 to 1, such as 0.9 indicates a high probability of abnormality), forming an abnormality probability matrix (T is the number of time points).

[0126] The abnormal probability is then diffused (added) based on the sparse coupling factor matrix S:

[0127]

[0128] Reflects the impact of adjacent node anomalies on the current node.

[0129] Finally, from the node with abnormal probability mutation (such as Starting from (greater than 0.8), the historical propagation paths are traced back according to the adjacency matrix (for example, in the past three time windows, nodes A→B→C have abnormalities in sequence), forming a propagation path set containing nodes and timestamps (such as {(A, t-2), (B, t-1), (C, t)}).

[0130] The present embodiment enhances the spatiotemporal correlation of anomaly detection, improving detection accuracy. By integrating high-order spatiotemporal features (deep features that integrate spatial graph convolution and temporal convolution), this method performs time-series anomaly detection on high-order spatiotemporal features. This method can capture complex spatiotemporal coupling patterns in heating pipeline operation (e.g., pressure linkage between adjacent nodes and time series trends in temperature changes), avoiding the one-sidedness of single-dimensional detection. For example, the isolation forest algorithm, based on the distribution patterns of node historical feature sequences, can more accurately identify abnormal states that deviate from normal patterns (e.g., a sudden temperature drop at a node and simultaneous abnormal pressures at adjacent nodes). Spatial propagation and diffusion of anomaly probabilities: A sparse coupling factor matrix is ​​used to diffuse the initial anomaly probability. This essentially combines the physical connectivity of the pipeline network with the similarity of node features, ensuring that the calculation of anomaly probabilities not only relies on the state of a single node but also considers the impact of abnormalities on adjacent nodes (e.g., the abnormal probability of node B in the example is enhanced by the abnormal state of adjacent node A). This conforms to the physical laws of fault propagation in heating pipeline networks (e.g., a pipeline leak can cause chain changes in upstream and downstream pressure and flow), reduces false positives in isolated detection, and improves the reliability of the anomaly probability matrix.

[0131] Furthermore, the physical interpretability and fault location efficiency of propagation path backtracking are enhanced by network topology-based path modeling. This approach uses an adjacency matrix to backtrack propagation paths, ensuring that the path set strictly adheres to the actual connections between heating pipelines (e.g., only directly connected nodes in the network are included), avoiding the generation of physically meaningless "virtual paths." For example, starting from a node with a sudden abnormal probability increase (e.g., above a threshold of 0.8), the sequence of abnormal nodes within three historical time windows (e.g., A→B→C) is traced back based on timestamps. This directly corresponds to the actual fault propagation path in the network, providing maintenance personnel with clear spatial location clues. Path integrity in the dynamic time dimension: The propagation path, combined with timestamps (e.g., {(A, t-2), (B, t-1), (C, t)}), clearly illustrates the temporal evolution of the fault from the initial node to downstream nodes, helping to analyze the speed and scope of fault propagation (e.g., determining whether it is a rapidly propagating pressure anomaly or a slowly developing pipeline aging issue), and providing data support for emergency response strategies (e.g., prioritizing isolation of the source node).

[0132] Furthermore, sparsity and normalization improve computational efficiency and model robustness. The sparse coupling factor matrix offers the advantage of dimensionality reduction: the adjacency matrix filters out coupling relationships between non-adjacent nodes (retaining only the weights of actually connected nodes) and normalizes these weights (summing to 1 across each row), eliminating interference from irrelevant nodes and reducing computational complexity. For example, in a pipeline network consisting of 100 nodes, the sparse matrix retains only the actual pipe connections, reducing inefficient computations while highlighting the direct influence of adjacent nodes, allowing the anomaly diffusion process to be more focused on the actual pipeline network structure. Enhanced noise immunity: Normalization ensures that the diffusion intensity of anomaly probabilities remains within a reasonable range (weights sum to 1), preventing over-amplification or attenuation of anomaly probabilities due to random fluctuations at individual nodes, thereby improving the model's robustness to data noise. For example, if a node exhibits an artificially high anomaly probability due to a brief sensor failure, normalization of the weights of adjacent nodes will effectively dilute the anomaly's impact, preventing it from being misclassified as a true fault.

[0133] Furthermore, it provides precise spatial positioning and propagation chains for maintenance decisions: Rapid target node locking: By searching for the maximum value in the anomaly probability matrix to determine the target node (such as the node with a probability of 0.9 in the example), combined with the node with the earliest abnormality in the propagation path set (such as node A at timestamp t-2), the initial location of the fault can be quickly located, rather than just detecting the affected downstream nodes, reducing the scope of investigation (such as from 100 nodes to 3-5 nodes on the path). Full-chain fault analysis: The propagation path set not only provides the current abnormal node, but also records the historical trajectory of fault propagation, making it easier for operation and maintenance personnel to analyze the fault type (such as whether it is a single-point sudden leak or systematic aging of the pipeline network) and assess the scope of the fault impact (such as nodes that may be affected later). This provides a basis for preventive maintenance (such as checking weak nodes on the path in advance) and reduces operation and maintenance costs.

[0134] In summary, through the deep fusion of spatiotemporal features, anomaly diffusion modeling under the constraints of pipeline network topology, and the joint backtracking of time series and spatial paths, high-precision, strong interpretability and efficient positioning of heating pipeline anomaly detection are achieved. This solves the problems of traditional single-point detection ignoring spatial correlation and fault tracing relying on manual experience, and provides a scientific and feasible technical solution for real-time monitoring and precise operation and maintenance of smart heating pipeline networks.

[0135] In another embodiment, based on the high-order spatiotemporal features of the heating data and the sparse coupling factor matrix corresponding to the thermal expansion of the heating pipe, the abnormal probability matrix of the heating pipe and the propagation path set of the heating pipe are obtained, including: extracting the high-order spatiotemporal features of the heating pipe from the high-order spatiotemporal features of the heating data; obtaining the first covariance features corresponding to the high-order spatiotemporal features of the heating pipe, and obtaining the second covariance features corresponding to the sparse coupling factor matrix corresponding to the thermal expansion of the heating pipe; merging the first covariance features corresponding to the high-order spatiotemporal features of the heating pipe with the second covariance features corresponding to the sparse coupling factor matrix corresponding to the thermal expansion of the heating pipe to obtain the abnormal probability matrix of the heating pipe. Covariance feature set; screening out a set consisting of elements greater than the maximum value element in the transfer matrix of the high-order spatiotemporal features of the heating pipeline from the covariance feature set of the heating pipeline as the initial abnormality probability matrix of the heating pipeline; correcting the probability value in the initial abnormality probability matrix of the heating pipeline by the target element in the covariance feature set of the heating pipeline to obtain the target abnormality probability matrix of the heating pipeline, wherein the target element in the covariance feature set of the heating pipeline refers to the element with an intermediate value in the covariance feature set of the heating pipeline; performing spatial propagation enhancement and path backtracing on the target abnormality probability matrix of the heating pipeline to obtain a propagation path set of the heating pipeline.

[0136] For example, the high-order spatiotemporal features of heating data contain a wealth of information about the heating system across both time and space, such as temperature, pressure, and flow rate at various locations over different time periods. Using specific feature selection methods, we filter out high-order spatiotemporal features specific to heating pipes based on their physical location, function, and other attributes. For example, using information such as the pipe number and location, we can extract features related to that pipe, such as temperature trends and pressure fluctuations, from the multidimensional feature data.

[0137] Covariance calculations were performed on the extracted high-order spatiotemporal features of the heating pipeline. Covariance measures the overall error between two variables. In the case of multiple features, the calculated first covariance matrix F reflects the correlation between the features. For example, if the covariance between two features is positive and large, it indicates that the two features tend to increase or decrease together; if the covariance is negative, it means that when one feature increases, the other tends to decrease.

[0138] The covariance of the sparse coupling factor matrix corresponding to the thermal expansion of the heating pipe is calculated to obtain the second covariance matrix G. The sparse coupling factor matrix describes the degree of coupling between thermal expansion and other related factors. Calculating its covariance can further analyze the stability and changing trends of these coupling relationships.

[0139] Merge the first covariance feature with the second covariance feature. This can be done by simple concatenation, where the two covariance matrices are combined into a larger matrix according to certain rules, or by weighted merging, where the covariance features are weighted according to their importance before merging.

[0140] Obtain the transfer matrix J of the heating pipeline's high-order spatiotemporal features. This matrix describes the transitions between different states, with the maximum value representing the maximum degree of a particular feature transition. Then, select elements greater than this maximum value from the covariance feature set of the heating pipeline. This set of elements constitutes the initial abnormality probability matrix for the heating pipeline.

[0141] Find the element with the middle value in the covariance feature set of the heating pipes as the target element. The middle value element can reflect a typical characteristic of the data. Use the target element to correct the probability values ​​in the initial anomaly probability matrix. This correction can be achieved through some functional relationship, such as linear adjustment or nonlinear transformation, to make the initial probability values ​​more reasonable.

[0142] Spatial propagation enhancement: Considering the potential spatial diffusion of anomalies, algorithms are used to expand the spatial impact of anomaly information. For example, based on the physical connections and heat conduction characteristics of pipelines, anomaly information is propagated to adjacent pipeline nodes, enhancing the spatial coverage of anomaly information.

[0143] Path backtracking: Based on the enhanced anomaly information, reverse the path of the anomaly's occurrence and propagation. This can be accomplished by using graph theory algorithms, such as depth-first search or breadth-first search, starting from nodes with a high probability of anomaly and tracing back to the possible source of the anomaly.

[0144] This embodiment of the application integrates multi-source data fusion and precise modeling. By extracting high-order spatiotemporal characteristics of heating pipelines (such as the spatiotemporal fluctuations of temperature, pressure, and flow) and a thermal expansion sparse coupling factor matrix (reflecting the relationship between pipeline physical deformation and environmental factors), the system's operating status data and physical property data are combined. The combined covariance features of these two data provide dual constraints on anomaly triggers—capturing abnormal fluctuations in operating parameters while also considering the coupled effects of pipeline physical properties such as thermal expansion and contraction. This avoids the one-sidedness of single-dimensional analysis (such as relying solely on temperature data and ignoring structural changes). By screening the initial anomaly probability matrix and using the maximum value of the high-order spatiotemporal feature transfer matrix as a threshold, features that significantly deviate from normal conditions are prioritized. Further, the probability is corrected using the median covariance feature value, balancing the interference of extreme values ​​with the representativeness of typical features. This mechanism of "first screening for extreme anomalies, then refining and correcting based on data distribution characteristics" avoids the poor adaptability of traditional thresholding methods to data fluctuations and improves the robustness of anomaly probability assessment.

[0145] Furthermore, this application efficiently locates the source and propagation path of anomalies. Based on a target anomaly probability matrix, spatial propagation enhancement (taking into account the physical connections and heat conduction patterns of the pipelines) and a path tracing algorithm clearly visualize the anomaly's propagation path within the pipeline network. For example, when a minor leak occurs in a pipeline segment due to thermal expansion, the system not only detects the temperature anomaly at that node but also uses propagation analysis to predict the potential downstream impact of the anomaly or trace back to the source of the upstream pressure anomaly. This provides operations and maintenance personnel with a comprehensive view from a single point of anomaly to its global impact, significantly reducing troubleshooting time. The target anomaly probability matrix numerically quantifies the anomaly risk at each pipeline node. Combined with a collection of propagation paths, an "anomaly risk map" for the heating pipeline can be constructed. Operations and maintenance departments can use this information to formulate differentiated maintenance strategies: prioritizing pipelines with high-probability anomalies and located at critical nodes along the propagation path, while continuously monitoring nodes with low probability but located at the end of the propagation path. This shift from post-fault handling to pre-emptive risk management reduces the probability of unexpected incidents.

[0146] Therefore, through the closed loop of "data feature fusion → abnormality probability quantification → propagation path deduction", the "precision, efficiency and visualization" of heating pipeline abnormality detection is achieved. Its core value lies in the deep combination of the operating data and physical characteristics of the heating system. It not only solves the problem of insufficient adaptability of traditional methods to complex working conditions, but also provides a feasible technical path for intelligent operation and maintenance, and ultimately achieves the multiple goals of "reducing costs, increasing efficiency and maintaining stability".

[0147] Step S105: Based on the abnormal probability matrix of the heating pipeline and the propagation path set of the heating pipeline, the damage location of the heating pipeline is obtained.

[0148] Among them, based on the abnormal probability matrix of the heating pipe and the propagation path set of the heating pipe, the damage location of the heating pipe is obtained, including: searching the probability in the abnormal probability matrix of the heating pipe along the propagation path in the propagation path set of the heating pipe, and when the maximum probability is reached, obtaining the target node, so that the damage location of the heating pipe is determined by the target node.

[0149] For example, for each propagation path (such as node sequence → ...→ ), search the anomaly probability matrix in reverse chronological order (from the current anomaly node back to the historical node), and record the anomaly probability of each node at the corresponding time point.

[0150] If a node in the path In time The abnormal probability is the maximum value in the path (and greater than 0.9), and the node triggers an abnormality in the downstream node at a subsequent time point, then it is determined is the root node.

[0151] For example, the path ← ← middle, The abnormal probability of node is 0.95, the maximum value in the path, and the abnormality occurs before nodes B and C. Therefore, the target node is A. Therefore, it is determined that the damage location of the heating pipe is likely at node A. Subsequent maintenance personnel can focus on inspecting and repairing the heating pipes in and around node A to resolve possible damage to the heating pipes.

[0152] The embodiment of the present application accurately locks the source of the fault and avoids misjudging the spreading node: Time reverse search + probability maximization: trace back in reverse time along the propagation path (from the current abnormal node to the historical node), and prioritize identifying the root node where the abnormality occurs earliest and has the highest probability (such as the abnormal probability of node A in the example is 0.95, which is the maximum value of the path, and triggers the abnormality before B and C), rather than focusing only on the downstream nodes with a high probability at the moment. This is in line with the physical laws of fault propagation in the heating pipeline network (such as leaks usually start from a certain node and then spread to adjacent nodes), avoiding misjudging the "downstream nodes affected by the fault" as damaged locations, greatly improving positioning accuracy. Double constraint noise filtering: Target nodes are screened through the dual conditions of "abnormal probability threshold (such as ≥0.9)" and "triggering downstream abnormalities", effectively eliminating non-real abnormalities caused by accidental fluctuations or sensor noise, and ensuring that the positioning results are based on significant abnormal signals and conform to the causal logic of fault propagation (upstream node abnormalities appear before downstream nodes).

[0153] In addition, focusing on the physical propagation path improves positioning efficiency and interpretability: Path constraints narrow the search scope: the target node is searched only within the set of identified propagation paths (such as the node sequence A→B→C within three historical time windows), rather than traversing the entire pipeline network. This narrows the investigation scope from "all network nodes" to "key nodes on the path" (such as the three nodes in the example), significantly reducing manual inspection costs. Physical interpretability supports operation and maintenance decisions: the propagation path is strictly based on the pipeline network adjacency matrix (the actual pipeline connection relationship). The positioning results of the target node can be directly mapped to the specific location in the geographic information system (GIS) (such as the pipeline connection of heat exchange station A). Maintenance personnel do not need to rely on complex algorithm derivation and can quickly lock the inspection area through the pipeline network map, improving operation and maintenance efficiency.

[0154] Furthermore, it dynamically adapts to fault propagation patterns, enhancing robustness. Compatible with both gradual and sudden faults: For gradual faults (such as increasing leaks due to pipeline aging), a reverse-time search can be used to capture the consistently rising trend in early anomaly probabilities (e.g., node A has a probability of 0.8 at time t-2, 0.9 at time t-1, and 0.95 at time t), promptly locating potential hazards that have not yet fully erupted but are already affecting downstream areas. For sudden faults (such as sudden pipeline ruptures), the "maximum probability + triggering downstream anomalies" condition allows rapid identification of the source of sudden anomalies (even if anomalies occur simultaneously at downstream nodes, the originating node can be distinguished by timestamp sequence). Strong anti-interference capabilities: The anomaly probability matrix has been processed through spatiotemporal feature fusion and sparse coupling factor diffusion, filtering out interference from irrelevant nodes. The search process is performed only along valid propagation paths, reducing location ambiguity when multiple node anomalies occur in complex pipeline networks (e.g., distinguishing independent faults from cascading faults).

[0155] Furthermore, a closed-loop O&M process is formed to support preventive maintenance: fault location and propagation analysis are linked. Identifying the target node not only provides the damage location but also reveals the fault's propagation speed and impact range through the time series of the propagation path (such as an abnormal sequence from A to B to C). This helps O&M personnel determine the fault type (such as a single-point leak or a pressure imbalance in the pipeline network) and formulate targeted measures (such as preemptively reinforcing weak nodes along the path). Data-driven maintenance strategy optimization: By accumulating historical location results (such as multiple faults clustered near node A), network design or O&M plans can be optimized (such as increasing sensor density in that area and shortening inspection cycles), achieving a shift from "reactive maintenance" to "proactive prevention."

[0156] In summary, through the core logic of "searching for the maximum anomaly probability along the actual propagation path", abstract data analysis is transformed into specific physical node positioning, which has both The three major advantages of high precision (source identification), high efficiency (path constraint), and high reliability (causal verification) effectively solve the problems of "reliance on manual experience, wide investigation scope, and high misjudgment rate" in traditional heating pipeline fault location, and provide key technical support for the intelligent operation and maintenance of smart heating systems.

[0157] In summary, in this embodiment, heating data is first acquired. Based on the heating data and a preset artificial intelligence algorithm, the high-order spatiotemporal features of the heating data are determined. A sparse coupling factor matrix corresponding to the thermal expansion of the heating pipe is then obtained. Based on the high-order spatiotemporal features of the heating data and the sparse coupling factor matrix corresponding to the thermal expansion of the heating pipe, an abnormality probability matrix and a set of propagation paths for the heating pipe are obtained. Furthermore, based on the abnormality probability matrix and the set of propagation paths for the heating pipe, the damaged location of the heating pipe is determined. This invention improves the accuracy of abnormality location, shortens troubleshooting time, and reduces operation and maintenance costs, thereby improving troubleshooting efficiency and effectiveness.

[0158] Figure 2 An embodiment of the present invention provides an artificial intelligence abnormal heating troubleshooting system. Figure 2 As shown, the system includes:

[0159] The first acquisition module 201 is used to acquire heating data;

[0160] A feature determination module 202 is used to determine high-order spatiotemporal features of the heating data based on the heating data and a preset artificial intelligence algorithm;

[0161] The second acquisition module 203 is used to obtain a sparse coupling factor matrix corresponding to the thermal expansion of the heating pipeline;

[0162] The third acquisition module 204 is used to obtain the abnormal probability matrix of the heating pipeline and the propagation path set of the heating pipeline based on the high-order spatiotemporal characteristics of the heating data and the sparse coupling factor matrix corresponding to the heating pipeline;

[0163] The location determination module 205 is configured to obtain the damage location of the heating pipeline based on the abnormality probability matrix of the heating pipeline and the propagation path set of the heating pipeline.

[0164] Optionally, the feature determination module 202 is further configured to input the heating data into a preset artificial intelligence algorithm to obtain a spatiotemporal coding matrix and an adjacency matrix;

[0165] Generate a dynamic spatiotemporal propagation graph sequence based on the spatiotemporal coding matrix and adjacency matrix;

[0166] Based on the dynamic spatiotemporal propagation graph sequence, the high-order spatiotemporal characteristics of heating data are determined.

[0167] Optionally, the feature determination module 202 is further configured to process the heating data using a preset artificial intelligence algorithm, establish a heating network topology map, and generate an adjacency matrix based on the heating network topology map;

[0168] Perform time series alignment on the heating data according to the time window to obtain aligned data;

[0169] The historical fault labels are concatenated with the aligned data to obtain the spatiotemporal encoding matrix.

[0170] Optionally, based on the spatiotemporal coding matrix and the adjacency matrix, a dynamic spatiotemporal propagation graph sequence is generated, including:

[0171] Calculate the similarity between data in the spatiotemporal coding matrix;

[0172] Determine the fusion weight based on the similarity and historical propagation pattern between the data in the spatiotemporal coding matrix;

[0173] Determine the target edge set of the adjacency matrix based on the fusion weight and the preset threshold;

[0174] Based on the target edge set, a dynamic spatiotemporal propagation graph sequence is generated.

[0175] Optionally, the feature determination module 202 is further configured to obtain spatial graph convolution and temporal convolution of the dynamic spatiotemporal propagation graph sequence;

[0176] The spatial graph convolution and temporal convolution of the dynamic spatiotemporal propagation graph sequence are gated and fused to obtain high-order spatiotemporal features.

[0177] Optionally, the second acquisition module 203 is further configured to perform coupling calculation on the high-order spatiotemporal characteristics of the heating data to obtain an initial coupling factor matrix;

[0178] The initial coupling factor matrix is ​​topologically modified by the thermal expansion of the heating pipe to obtain a modified coupling factor matrix;

[0179] The corrected coupling factor matrix is ​​normalized to obtain a sparse coupling factor matrix.

[0180] or,

[0181] Calculate the first quantile corresponding to the covariance matrix of the high-order spatiotemporal characteristics of the heating data;

[0182] Extract the diagonal elements in the spatiotemporal coding matrix, and multiply the first quantile corresponding to the covariance matrix of the high-order spatiotemporal characteristics of the heating data by the diagonal elements in the spatiotemporal coding matrix to obtain the first quantile set of the covariance matrix of the high-order spatiotemporal characteristics of the heating data;

[0183] Calculate the second quantile corresponding to the covariance matrix corresponding to the thermal expansion of the heating pipe;

[0184] The target quantile is obtained by adding the first quantile corresponding to the covariance matrix of the high-order spatiotemporal characteristics of the heating data and the second quantile corresponding to the covariance matrix corresponding to the thermal expansion of the heating pipe;

[0185] Performing a dot product between the target quantile and each quantile in the first quantile set of the high-order spatiotemporal feature covariance matrix of the heating data, to obtain the second quantile set of the high-order spatiotemporal feature covariance matrix of the heating data;

[0186] The maximum quantile in the second quantile set is used to replace the central element in the transposed matrix of the space-time coding matrix and the central element in the covariance matrix corresponding to the thermal expansion of the heating pipe, and the transposed matrix of the replaced space-time coding matrix and the covariance matrix corresponding to the thermal expansion of the heating pipe are added to obtain a sparse coupling factor matrix.

[0187] Optionally, the third acquisition module 204 is further configured to perform time series anomaly detection on the high-order spatiotemporal features of the heating data to obtain an initial anomaly probability matrix of the heating pipeline;

[0188] The initial abnormality probability matrix of the heating pipeline is added to the sparse coupling factor matrix corresponding to the thermal expansion of the heating pipeline to obtain the target abnormality probability matrix of the heating pipeline;

[0189] The target anomaly probability matrix of the heating pipeline is subjected to spatial propagation enhancement and path backtracing to obtain the propagation path set of the heating pipeline.

[0190] or,

[0191] Extracting high-order spatiotemporal features of heating pipes from the high-order spatiotemporal features of heating data;

[0192] Obtaining a first covariance feature corresponding to the high-order spatiotemporal feature of the heating pipeline, and obtaining a second covariance feature corresponding to the sparse coupling factor matrix corresponding to the thermal expansion of the heating pipeline;

[0193] Merge the first covariance feature corresponding to the high-order spatiotemporal feature of the heating pipeline with the second covariance feature corresponding to the sparse coupling factor matrix corresponding to the thermal expansion of the heating pipeline to obtain a covariance feature set of the heating pipeline;

[0194] A set consisting of elements greater than the maximum value element in the transfer matrix of the high-order spatiotemporal characteristics of the heating pipeline is selected from the covariance feature set of the heating pipeline as the initial abnormality probability matrix of the heating pipeline;

[0195] The probability values ​​in the initial abnormal probability matrix of the heating pipeline are corrected by the target element in the covariance feature set of the heating pipeline to obtain the target abnormal probability matrix of the heating pipeline, wherein the target element in the covariance feature set of the heating pipeline refers to the element with the middle value in the covariance feature set of the heating pipeline;

[0196] The target anomaly probability matrix of the heating pipeline is subjected to spatial propagation enhancement and path backtracing to obtain the propagation path set of the heating pipeline.

[0197] Optionally, the location determination module 205 is also used to search for the probability in the abnormal probability matrix of the heating pipeline along the propagation path in the propagation path set of the heating pipeline. When the maximum probability is reached, the target node is obtained, and the damaged location of the heating pipeline is determined by the target node.

[0198] In the description provided herein, the algorithms and displays are not inherently related to any particular computer, virtual system, or other device. Various general-purpose systems may also be used in conjunction with the examples of the present invention. Based on the above description, it is apparent that the structure required for constructing such systems is well understood. In addition, the present invention is not directed to any specific programming language. It should be understood that various programming languages ​​may be utilized to implement the present invention described herein, and the description of specific languages ​​above is provided for the purpose of disclosing preferred embodiments of the present invention.

[0199] In the description provided herein, a large number of specific details are described. However, it is understood that embodiments of the present invention can be practiced without these specific details. In some instances, well-known methods, structures, and techniques are not shown in detail so as not to obscure the understanding of this description.

[0200] Those skilled in the art will appreciate that the modules, units, or components of the devices in the examples disclosed herein may be arranged in the device described in the embodiment, or alternatively may be located in one or more devices different from the devices in the examples. The modules in the foregoing examples may be combined into one module or further divided into multiple submodules.

[0201] Those skilled in the art will appreciate that the modules in the devices in the embodiments can be adaptively changed and arranged in one or more devices different from the embodiments. The modules or units or components in the embodiments can be combined into one module or unit or component, and in addition they can be divided into multiple submodules or subunits or subassemblies. Except that at least some of such features and / or processes or units are mutually exclusive, all features disclosed in this specification and all processes or units of any method or device disclosed in this manner can be combined in any combination. Unless otherwise expressly stated, each feature disclosed in this specification can be replaced by an alternative feature providing the same, equivalent or similar purpose.

[0202] In addition, some of the embodiments are described herein as methods or combinations of method elements that can be implemented by a processor of a computer system or by other devices that perform the functions described. Thus, a processor having the necessary instructions for implementing the method or method element forms a device for implementing the method or method element. Furthermore, the elements described herein of the device embodiments are examples of devices for implementing the functions performed by the elements for the purpose of implementing the invention.

[0203] As used herein, unless otherwise specified, the use of ordinal numbers "first," "second," "third," etc. to describe common objects merely indicates that different instances of similar objects are involved and are not intended to imply that the objects so described must have a given order in time, space, ranking, or in any other manner.

[0204] Although the present invention has been described with respect to a limited number of embodiments, it will be apparent to those skilled in the art, having benefit of the foregoing description, that other embodiments are contemplated within the scope of the invention thus described. Furthermore, it should be noted that the language used in this specification has been selected primarily for readability and didactic purposes, rather than for the purpose of explaining or defining the subject matter of the present invention. Consequently, many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the accompanying description. The disclosure of the present invention is intended to be illustrative and not restrictive of the scope of the invention, which is defined by the accompanying description.

Claims

1. An artificial intelligence abnormal heating troubleshooting method, characterized in that: The following steps are involved: Obtain heating data; Determining high-order spatiotemporal features of the heating data based on the heating data and a preset artificial intelligence algorithm; Obtain the sparse coupling factor matrix corresponding to the thermal expansion of the heating pipeline; Based on the high-order spatiotemporal characteristics of the heating data and the sparse coupling factor matrix corresponding to the thermal expansion of the heating pipeline, obtaining an abnormality probability matrix of the heating pipeline and a propagation path set of the heating pipeline; Obtaining a damage location of the heating pipeline based on an abnormality probability matrix of the heating pipeline and a propagation path set of the heating pipeline; The determining of high-order spatiotemporal features of the heating data based on the heating data and a preset artificial intelligence algorithm includes: Inputting the heating data into the preset artificial intelligence algorithm to obtain a spatiotemporal coding matrix and an adjacency matrix, wherein the preset artificial intelligence algorithm is a spatiotemporal propagation graph convolution; generating a dynamic spatiotemporal propagation graph sequence based on the spatiotemporal coding matrix and the adjacency matrix; Determining high-order spatiotemporal features of the heating data based on the dynamic spatiotemporal propagation graph sequence; The obtaining of a sparse coupling factor matrix corresponding to the thermal expansion of the heating pipeline includes: performing coupling calculation on the high-order spatiotemporal characteristics of the heating data to obtain an initial coupling factor matrix; Performing topological correction on the initial coupling factor matrix according to the thermal expansion of the heating pipe to obtain a corrected coupling factor matrix; Normalizing the corrected coupling factor matrix to obtain the sparse coupling factor matrix; or, Calculating the first quantile corresponding to the covariance matrix of the high-order spatiotemporal characteristics of the heating data; Extracting the diagonal elements in the spatiotemporal coding matrix, and multiplying the first quantile corresponding to the covariance matrix of the high-order spatiotemporal features of the heating data by the diagonal elements in the spatiotemporal coding matrix to obtain a first quantile set of the covariance matrix of the high-order spatiotemporal features of the heating data; Calculating the second quantile corresponding to the covariance matrix corresponding to the thermal expansion of the heating pipeline; Adding a first quantile corresponding to the covariance matrix of the high-order spatiotemporal features of the heating data and a second quantile corresponding to the covariance matrix corresponding to the thermal expansion of the heating pipe to obtain a target quantile; Performing a dot product between the target quantile and each quantile in the first quantile set of the high-order spatiotemporal feature covariance matrix of the heating data to obtain a second quantile set of the high-order spatiotemporal feature covariance matrix of the heating data; The maximum quantile in the second quantile set is used to replace the central element in the transposed matrix of the space-time coding matrix and the central element in the covariance matrix corresponding to the thermal expansion of the heating pipe, and the transposed matrix of the replaced space-time coding matrix and the covariance matrix corresponding to the thermal expansion of the heating pipe are added to obtain the sparse coupling factor matrix.

2. The artificial intelligence abnormal heating troubleshooting method according to claim 1 is characterized in that: The step of inputting the heating data into the preset artificial intelligence algorithm to obtain a spatiotemporal coding matrix and an adjacency matrix includes: Processing the heating data using the preset artificial intelligence algorithm to establish a heating network topology map, and generating an adjacency matrix based on the heating network topology map; Performing time series alignment on the heating data according to a time window to obtain aligned data; The historical fault labels are concatenated with the aligned data to obtain the spatiotemporal coding matrix.

3. The abnormal heating troubleshooting method based on artificial intelligence according to claim 1 is characterized in that: The generating a dynamic spatiotemporal propagation graph sequence based on the spatiotemporal coding matrix and the adjacency matrix includes: Calculating the similarity between the data in the spatiotemporal coding matrix; Determining a fusion weight based on the similarity and historical propagation pattern between the data in the spatiotemporal coding matrix; Determining a target edge set of the adjacency matrix based on a fusion weight and a preset threshold; Based on the target edge set, the dynamic spatiotemporal propagation graph sequence is generated.

4. The abnormal heating troubleshooting method based on artificial intelligence according to claim 1 is characterized in that: The determining of the high-order spatiotemporal features of the heating data based on the dynamic spatiotemporal propagation graph sequence includes: Obtaining spatial graph convolution and temporal convolution of the dynamic spatiotemporal propagation graph sequence; The spatial graph convolution and temporal convolution of the dynamic spatiotemporal propagation graph sequence are gated and fused to obtain the high-order spatiotemporal features.

5. The artificial intelligence abnormal heating troubleshooting method according to claim 1 is characterized in that: The step of obtaining a target abnormality probability matrix of the heating pipeline and a propagation path set of the heating pipeline based on the high-order spatiotemporal characteristics of the heating data and a sparse coupling factor matrix corresponding to the thermal expansion of the heating pipeline includes: Performing time series anomaly detection on the high-order spatiotemporal features of the heating data to obtain an initial anomaly probability matrix of the heating pipeline; Adding an initial abnormality probability matrix of the heating pipeline and a sparse coupling factor matrix corresponding to thermal expansion of the heating pipeline to obtain a target abnormality probability matrix of the heating pipeline; Performing spatial propagation enhancement and path backtracing on the target abnormality probability matrix of the heating pipeline to obtain a propagation path set of the heating pipeline; or, Extracting high-order spatiotemporal features of the heating pipeline from the high-order spatiotemporal features of the heating data; Obtaining a first covariance feature corresponding to the high-order spatiotemporal feature of the heating pipeline, and obtaining a second covariance feature corresponding to the sparse coupling factor matrix corresponding to the thermal expansion of the heating pipeline; Merging a first covariance feature corresponding to the high-order spatiotemporal feature of the heating pipeline with a second covariance feature corresponding to the sparse coupling factor matrix corresponding to the thermal expansion of the heating pipeline to obtain a covariance feature set of the heating pipeline; Screening out, from the covariance feature set of the heating pipeline, a set consisting of elements that are greater than the maximum value element in the transfer matrix of the high-order spatiotemporal feature of the heating pipeline as an initial abnormality probability matrix of the heating pipeline; Correcting the probability values ​​in the initial abnormality probability matrix of the heating pipeline by using the target element in the covariance feature set of the heating pipeline to obtain the target abnormality probability matrix of the heating pipeline, wherein the target element in the covariance feature set of the heating pipeline refers to the element with an intermediate value in the covariance feature set of the heating pipeline; The target abnormality probability matrix of the heating pipeline is subjected to spatial propagation enhancement and path backtracing to obtain a propagation path set of the heating pipeline.

6. The abnormal heating troubleshooting method based on artificial intelligence according to claim 1 is characterized in that: The obtaining of the damage location of the heating pipeline based on the target abnormality probability matrix of the heating pipeline and the propagation path set of the heating pipeline includes: The probability in the abnormal probability matrix of the heating pipeline is searched along the propagation path in the propagation path set of the heating pipeline. When the maximum probability is reached, the target node is obtained, so that the damage position of the heating pipeline is determined by the target node.

7. An artificial intelligence abnormal heating troubleshooting system, characterized in that: include: A first acquisition module is used to acquire heating data; a feature determination module, configured to determine high-order spatiotemporal features of the heating data based on the heating data and a preset artificial intelligence algorithm; The second acquisition module is used to obtain a sparse coupling factor matrix corresponding to the thermal expansion of the heating pipeline; a third acquisition module, configured to acquire an abnormality probability matrix of the heating pipeline and a propagation path set of the heating pipeline based on the high-order spatiotemporal characteristics of the heating data and a sparse coupling factor matrix corresponding to the heating pipeline; a location determination module, configured to obtain a damage location of the heating pipeline based on an abnormality probability matrix of the heating pipeline and a set of propagation paths of the heating pipeline; The determining of high-order spatiotemporal features of the heating data based on the heating data and a preset artificial intelligence algorithm includes: Inputting the heating data into the preset artificial intelligence algorithm to obtain a spatiotemporal coding matrix and an adjacency matrix, wherein the preset artificial intelligence algorithm is a spatiotemporal propagation graph convolution; generating a dynamic spatiotemporal propagation graph sequence based on the spatiotemporal coding matrix and the adjacency matrix; Determining high-order spatiotemporal features of the heating data based on the dynamic spatiotemporal propagation graph sequence; The obtaining of a sparse coupling factor matrix corresponding to the thermal expansion of the heating pipeline includes: performing coupling calculation on the high-order spatiotemporal characteristics of the heating data to obtain an initial coupling factor matrix; Performing topological correction on the initial coupling factor matrix according to the thermal expansion of the heating pipe to obtain a corrected coupling factor matrix; Normalizing the corrected coupling factor matrix to obtain the sparse coupling factor matrix; or, Calculating the first quantile corresponding to the covariance matrix of the high-order spatiotemporal characteristics of the heating data; Extracting the diagonal elements in the spatiotemporal coding matrix, and multiplying the first quantile corresponding to the covariance matrix of the high-order spatiotemporal features of the heating data by the diagonal elements in the spatiotemporal coding matrix to obtain a first quantile set of the covariance matrix of the high-order spatiotemporal features of the heating data; Calculating the second quantile corresponding to the covariance matrix corresponding to the thermal expansion of the heating pipeline; Adding a first quantile corresponding to the covariance matrix of the high-order spatiotemporal features of the heating data and a second quantile corresponding to the covariance matrix corresponding to the thermal expansion of the heating pipe to obtain a target quantile; Performing a dot product between the target quantile and each quantile in the first quantile set of the high-order spatiotemporal feature covariance matrix of the heating data to obtain a second quantile set of the high-order spatiotemporal feature covariance matrix of the heating data; The maximum quantile in the second quantile set is used to replace the central element in the transposed matrix of the space-time coding matrix and the central element in the covariance matrix corresponding to the thermal expansion of the heating pipe, and the transposed matrix of the replaced space-time coding matrix and the covariance matrix corresponding to the thermal expansion of the heating pipe are added to obtain the sparse coupling factor matrix.

Images