Intelligent identification and fault early warning method for abnormal fluctuation of hydraulic and thermal working conditions of heat supply network pipeline

Abstract

The invention discloses a heat supply network pipeline hydraulic and thermal working condition abnormal fluctuation intelligent identification and fault early warning method, and relates to the technical field of computer processing, and the method comprises the following steps: S01, collecting hydraulic and thermal parameters of a heat supply network pipeline in real time; s02, acquiring abnormal data of the time sequence sample set based on the thermodynamic mechanism model, detecting and repairing the abnormal data; s03, extracting characteristic parameters capable of reflecting abnormal fluctuation of the hydraulic and thermal working conditions of the heat supply network pipeline from the repaired time sequence sample set; s04, establishing an abnormal fluctuation identification model by utilizing an intelligent algorithm, and judging whether the hydraulic and thermal working conditions of the heat supply network pipeline have abnormal fluctuation or not; and S05, when the abnormal fluctuation is identified, giving a grading early warning mechanism and risk prevention and control suggestions of the damage degree grade of the associated pipeline. According to the invention, by fusing multi-source data and an intelligent algorithm, full-chain intelligent management of the hydraulic and thermal working conditions of the heat supply network pipeline from abnormal accurate sensing, fault intelligent diagnosis to risk active early warning is realized.

Description

Technical Field

[0001] This invention relates to the field of computer processing technology, and in particular to a method for intelligent identification and fault early warning of abnormal fluctuations in the hydraulic and thermal operating conditions of heating network pipelines. Background Technology

[0002] A heating network system mainly consists of heat sources, transmission pipelines, and user terminals. Heat sources generate heat through various methods such as combined heat and power (CHP), coal combustion, and gas combustion, and then transfer this heat to user terminals via the transmission pipelines. In actual operation, the hydraulic and thermal conditions of the heating network pipelines are highly susceptible to interference from various factors. From the perspective of the heat source, equipment malfunctions and changes in fuel supply can cause fluctuations in the supply water temperature and pressure, thereby affecting the overall operational stability of the heating network. For example, a boiler malfunction in a power plant can cause a drop in supply water temperature, directly reducing the heating quality for users.

[0003] Regarding pipeline networks, factors such as unreasonable network structure design, improper valve adjustment, pipe aging, construction and installation quality issues, and external damage can all lead to problems such as hydraulic imbalance and pipeline leaks. Hydraulic imbalance causes uneven pressure and flow distribution throughout the network, resulting in excess heat in some areas and insufficient heat in others, severely impacting user comfort. Pipeline leaks lead to heat and water loss, increasing energy consumption and operating costs.

[0004] There are also many factors affecting the operating conditions of the heating network at the user end. Faulty heating equipment or abnormal heating behavior at the user end, such as malfunctioning indoor thermostat valves or unauthorized modifications to heating equipment, can disrupt the balance and regulation of the heating network and increase management difficulty.

[0005] Traditional methods for diagnosing heating network faults have certain limitations. Expert rule-based methods are often used for applications with limited operational data and suffer from poor portability, making them difficult to adapt to the complex and ever-changing operating environments of heating networks. While theoretical modeling methods can achieve high accuracy, they require high sensor reliability and struggle to identify fault conditions not simulated. Data analysis methods rely solely on mathematical calculations; without incorporating expert rule-based methods, their portability decreases, and their excessive data requirements make them unsuitable for practical applications. Summary of the Invention

[0006] To address the aforementioned technical problems, the present invention provides a method for intelligent identification and fault early warning of abnormal fluctuations in the hydraulic and thermal operating conditions of heating network pipelines. The method includes the following steps: S01. Install sensors at key nodes of the heating network pipeline to collect the hydraulic and thermal parameters of the heating network pipeline in real time. S02. The hydro-thermal parameters are cleaned and fused preprocessed to obtain a time series sample set. Abnormal data detection of the time series sample set is obtained based on the thermodynamic mechanism model and repaired to ensure that the repaired time series sample set and multi-sensor data are consistent. S03. Extract feature parameters that can reflect abnormal fluctuations in the hydraulic and thermal operating conditions of the heating network pipeline from the repaired time series sample set. The feature parameters include residual features based on the mechanism model, operating condition coupling features, and frequency domain features based on signal processing. S04. Based on the extracted feature parameters, an abnormal fluctuation identification model is established using an intelligent algorithm to determine whether there are abnormal fluctuations in the hydraulic and thermal conditions of the heating network pipeline. S05. When abnormal fluctuations are identified, a graded early warning mechanism for the degree of hazard of the associated pipeline and risk prevention and control recommendations shall be given based on the type and degree of the abnormal fluctuations.

[0007] Preferably, the installation of sensors at key nodes of the heating network pipeline includes: It consists of multiple pressure and temperature sensors arranged on elbows, tees, valves and steam traps to obtain soft measurement data sets; A sensor simulating the mechanism of thermal pipeline operation characteristics is established. It is used to collect data on pipe sections not arranged on the heating network pipeline by the physical data acquisition module. It is combined with historical operation data to establish a data-driven model. The two are integrated to construct a soft measurement model of thermal pipeline pressure and temperature, and obtain soft measurement datasets that are not easy to measure.

[0008] Preferably, step S02, which involves cleaning and fusing the hydro-thermal parameters to obtain a time-series sample set, includes: S21. Substitute the real-time collected data into the thermodynamic mechanism model to obtain the model prediction value, and subtract the model prediction value from the actual measured value in the hydrothermal parameters to obtain the residual. S22. When the residual continuously exceeds the preset threshold and continues to exceed the first time period, the measurement data within that time period is determined to be potentially abnormal data. S23. For the potential abnormal data, forward extrapolation is performed using the mechanistic relationship model, or data repair is performed using a data prediction model based on long short-term memory networks to generate alternative values.

[0009] Preferably, the step of keeping the repaired time-series sample set and multi-sensor data consistent in step S02 includes: S24. For multiple sensors of the same type at the same monitoring point, or two sensors at adjacent points, fuzzy set theory is used to fuse the data collected by the multiple sensors to obtain the fusion result. S25. Calculate the confidence distance between each sensor data and the fusion result. When the confidence distance of a certain sensor data exceeds the dynamic threshold, the sensor data is determined to be unreliable and is then removed.

[0010] Preferably, step S03, which involves extracting characteristic parameters from the repaired time-series sample set that reflect abnormal fluctuations in the hydraulic and thermal operating conditions of the heating network pipeline, includes: The extraction of residual features based on the mechanism model includes: S31. Construct a dynamic hydraulic-thermal coupling model for the heating network pipe section. The dynamic hydraulic-thermal coupling model uses the pressure, temperature, and flow rate of the upstream node as input to predict the pressure, temperature, and flow rate of the downstream node. S32. Input the real-time data into the dynamic hydraulic-thermal coupling model and calculate the real-time residuals of the corresponding type between the actual measured values ​​of each parameter of the downstream node and the predicted pressure, temperature and flow rate of the downstream node. S33. Calculate the statistical characteristics of the real-time residual within the sliding time window, including the mean, variance, skewness, and cumulative sum of the residual, as a residual feature set reflecting the degree to which the system state deviates from the normal mechanism. The extraction of the operating condition coupling features includes: S34. The ratio of the pressure difference change to the flow rate change at adjacent sampling times yields the real-time slope of the pressure-flow characteristic curve, which serves as a key feature characterizing the instantaneous change in pipeline impedance. S35. Calculate the real-time efficiency ratio of heat power to pump power consumption by using the ratio of heat power calculated from flow rate and supply / return water temperature difference to the power consumption of the circulating pump calculated from current and voltage, as a characteristic characterizing the system's energy efficiency status. S36. Calculate the deviation rate between the supply and return water pressure difference of the same pipe section on the heating network pipeline and the theoretical design value, as a characteristic for judging the degree of hydraulic imbalance; The extraction of frequency domain features based on signal processing specifically includes: S37. Perform wavelet packet transform on the acquired pressure time-series signal and flow time-series signal and decompose them to a specific frequency band; S38. The energy ratio of the acquired pressure signal in the specific high-frequency band is used as an early feature characterizing the occurrence of water hammer or cavitation. S39. Calculate the moving average value of the main peak value of the power spectral density of the flow signal in the specific high frequency band, as a characterization of the slow-varying disturbance characteristics caused by heat source regulation or large user switching. S36. Calculate the deviation rate between the supply and return water pressure difference of the same pipe section on the heating network pipeline and the theoretical design value, as a characteristic for judging the degree of hydraulic imbalance.

[0011] Preferably, in step S04, the intelligent algorithm establishes the abnormal fluctuation identification model by combining two models deployed in parallel, wherein: A baseline isolation forest model trained on historical normal data is used to detect known normal pattern biases. An adaptive isolation forest model based on dynamically updated online data is used to capture new anomalous patterns generated during system evolution; In step S04, when determining whether there are abnormal fluctuations in the hydraulic and thermal conditions of the heating network pipeline, if either of the two models determines that the input comprehensive feature vector is abnormal, the subsequent fine classification process is triggered. S41. The comprehensive feature vector is split into a static feature stream and a dynamic temporal feature stream; S42. The static feature stream is embedded into a context vector through a fully connected network to obtain a static context vector, and the dynamic temporal feature stream is input into a multi-head self-attention encoder to obtain a dynamic temporal feature vector. S43. Aggregate the static context vector and the dynamic temporal feature vector to form the final working condition representation vector, and input it into the classifier; S44. Convert the heating network topology into graph data, wherein: nodes in the graph data represent key monitoring points, and edges represent pipeline connections; S45. When an anomaly is detected at a single node, the condition representation vector and the graph structure data are input into a pre-trained graph neural network. S46. The graph neural network simulates the propagation effect of faults along the pipeline network through a message passing mechanism, and outputs a joint judgment on the location of the fault root cause and the fault type by combining the abnormal characteristics of the current single node with the state of the adjacent nodes of the previous single node.

[0012] Preferably, step S05, when abnormal fluctuations are identified, includes a graded early warning mechanism for the severity of hazard to the associated pipeline and risk prevention and control recommendations based on the type and degree of the abnormal fluctuations, including: S51. Based on the identified fault type and location of the node, inject the corresponding fault parameters into the digital twin model of the heating network, and simulate the propagation path of the fault parameters and their impact on key system indicators within a future preset time window. S52. Based on the simulation results, calculate the scope of the fault's impact, the rate of parameter deterioration, and the potential impact level on heating supply security. S53. Based on the above calculation results, a dynamic and quantitative comprehensive risk index is generated, and the warning level is dynamically divided into any one of four levels: attention, warning, serious and emergency, according to the index. S54. Construct a knowledge graph that integrates historical fault cases, equipment maintenance manuals, and expert experience, so that data pop-ups with faults, symptoms, equipment, and handling measures are formed at the nodes, and edges represent the logical relationships between them, as well as trigger symbols to display the data pop-ups. S55. Match the feature vector of the current fault with the handling knowledge graph to retrieve the Top-K most similar historical cases and their successful handling solutions; S56. Generate a personalized handling suggestion list for the current fault based on graph reasoning. The personalized handling suggestion list includes key confirmation steps, priority operation sequence, required spare parts and tools, and a list of related interlocking equipment.

[0013] The present invention has at least the following beneficial effects: 1. By employing multi-sensor data fusion and anomaly data repair mechanisms, the accuracy and reliability of fault identification are significantly improved. Furthermore, a thermodynamic mechanism model is used to clean and preprocess the real-time acquired hydraulic and thermal parameters, and residual analysis is used to detect potential anomalies. These anomalies are then repaired using long short-term memory networks or mechanistic models, ensuring the consistency and integrity of the time-series sample set. Simultaneously, the extracted multiple feature parameters can comprehensively capture abnormal fluctuation patterns in the heating network pipelines, avoiding misjudgments or omissions caused by traditional methods relying on a single data source or model. This enables more accurate fault detection in complex operating environments.

[0014] 2. By adopting a baseline isolation forest model and an adaptive isolation forest model deployed in parallel, it can detect deviations from known normal patterns and capture new abnormal patterns generated during system evolution. This overcomes the shortcomings of traditional expert rule methods in terms of poor portability and theoretical model methods in terms of difficulty in identifying unsimulated faults.

[0015] 3. By combining graph neural networks to analyze the topology of the heating network and simulating the propagation effect of faults along the pipeline, the method achieves joint judgment of the location and type of fault root causes, enabling it to adapt to different heating network configurations and operating conditions, and reducing reliance on external expert knowledge or large amounts of historical data.

[0016] 4. Enables maintenance personnel to take timely and targeted measures based on the type and severity of abnormal fluctuations. Simultaneously, by utilizing a knowledge graph to match historical cases and expert experience, a personalized list of handling suggestions is generated, improving the efficiency and accuracy of fault response and reducing the risk of heating interruptions or equipment damage caused by abnormal hydraulic and thermal operating conditions, thereby optimizing the operation and maintenance management of the heating network system. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 The flowchart illustrates a method for intelligent identification and fault early warning of abnormal fluctuations in the hydraulic and thermal operating conditions of a heating network pipeline, as provided in Embodiment 1 of the present invention. Detailed Implementation

[0019] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0020] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0021] Example 1

[0022] This embodiment provides a method for intelligent identification and fault early warning of abnormal fluctuations in the hydraulic and thermal operating conditions of heating network pipelines. The method includes the following steps: Figure 1 As shown: S01. Install sensors at key nodes of the heating network pipeline to collect the hydraulic and thermal parameters of the heating network pipeline in real time. Specifically, the aforementioned installation of sensors at key nodes in the heating network pipeline includes: It consists of multiple pressure and temperature sensors arranged on elbows, tees, valves and steam traps to obtain soft measurement data sets; The mechanism simulation sensor for establishing the operating characteristics of thermal pipelines is used to collect physical data acquisition modules for pipe sections not deployed on the heating network pipeline. It is combined with historical operating data to establish a data-driven model. The two are integrated to construct a soft measurement model of thermal pipeline pressure and temperature, and obtain soft measurement datasets that are not easy to measure.

[0023] Secondly, the hydraulic and thermal parameters of the aforementioned heating network pipelines are packaged and uploaded according to a predetermined window length and cycle.

[0024] Pressure and temperature sensors are densely installed at critical nodes in the pipe network where the hydraulic and thermal states change drastically or are prone to failure (such as elbows, tees, valves, and steam traps). These physical sensors are responsible for acquiring directly measurable hard data.

[0025] Secondly, the mechanism simulation sensor is constructed based on the principles of thermodynamics and fluid mechanics to establish a mechanism model of the heating network pipeline. This model can simulate and calculate the pressure, temperature, and other data of the downstream pipe section without installed sensors based on the parameters of known points upstream (measured by physical sensors), which is equivalent to a virtual sensor.

[0026] Furthermore, by integrating historical operational data, a data-driven model is trained to correct calculation biases in the mechanistic model and adapt to dynamic changes in pipeline network operation. Finally, the mechanistic model and the data-driven model are merged to construct a soft-sensor model for pressure and temperature in thermal pipelines. This model can continuously output soft-sensor datasets for pipe sections that are difficult to measure directly, effectively filling monitoring gaps that physical sensors cannot cover.

[0027] Furthermore, the system combines all data points collected within a time window (which may include readings from multiple sensors) into a single data packet, and then periodically uploads this data packet to the data processing center. This processing method transforms the continuous data stream into a series of well-organized time-series sample sets, providing directly usable, structured data input for subsequent time-series feature extraction and anomaly detection algorithms.

[0028] S02. Clean and fuse the hydro-thermal parameters to obtain a time series sample set, and detect and repair abnormal data in the time series sample set based on the thermodynamic mechanism model, so that the repaired time series sample set and multi-sensor data are consistent. Specifically, the above-mentioned cleaning and fusion preprocessing of hydrodynamic and thermal parameters to obtain a time-series sample set includes: S21. Substitute the real-time collected data into the thermodynamic mechanism model to obtain the model prediction value, and calculate the difference between the model prediction value and the actual measured value in the hydraulic and thermodynamic parameters to obtain the residual. S22. When the residual continuously exceeds the preset threshold and continues to exceed the first time period, the measurement data within that time period is determined to be potentially abnormal data. S23. For potentially abnormal data, use a mechanistic relationship model for forward extrapolation, or use a data prediction model based on long short-term memory networks for data repair to generate alternative values.

[0029] Furthermore, ensuring consistency between the repaired time-series sample set and the multi-sensor data includes: S24. For multiple sensors of the same type at the same monitoring point, or two sensors at adjacent points, fuzzy set theory is used to fuse the data collected by the multiple sensors to obtain the fusion result. S25. Calculate the confidence distance between each sensor data and the fusion result. When the confidence distance of a certain sensor data exceeds the dynamic threshold, the sensor data is determined to be unreliable and is then removed.

[0030] Specifically, real-time collected data is input into a pre-established thermodynamic mechanism model, which calculates predicted parameter values ​​based on physical laws (such as energy conservation and momentum conservation). The reasonableness of the data is monitored by calculating the residual between the predicted and actual measured values. To distinguish between transient disturbances and genuine anomalies, a dual judgment condition is set: the residual must not only exceed a preset threshold but also persist for a certain period. This mechanism effectively avoids misjudgments caused by random noise and accurately identifies truly potential anomalies.

[0031] For data deemed anomalous, the system provides two repair paths. The first is to use a mechanistic relationship model for forward extrapolation, using data from the previous reliable time point to extrapolate the replacement value for the current time point based on the physical model, ensuring that the repaired value conforms to physical laws. The second is to use a data prediction model based on a Long Short-Term Memory (LSTM) network, which predicts the current normal value by learning the temporal patterns in historical data.

[0032] For multiple similar sensors deployed at the same monitoring point, or two physically adjacent sensors whose data should be related, the system employs fuzzy set theory for data fusion. This approach can handle the uncertainty and fuzziness of sensor information, and by calculating the membership degree of each data point, it obtains a more reliable fusion result that integrates all the information.

[0033] The confidence distance between the raw data from each sensor and the fusion result is calculated. This confidence distance refers to the degree of deviation between the sensor reading and the "collective consensus." The system sets a dynamic threshold; when the confidence distance of a sensor continuously exceeds the threshold, the sensor is deemed to be faulty or severely drifting, and its data is no longer reliable, thus being removed from the valid data source. This step ensures that the dataset participating in subsequent analysis is self-consistent and consistent, further improving the overall credibility of the data.

[0034] S03. Extract characteristic parameters that can reflect abnormal fluctuations in the hydraulic and thermal operating conditions of the heating network pipeline from the repaired time series sample set. The characteristic parameters include residual characteristics based on the mechanism model, operating condition coupling characteristics, and frequency domain characteristics based on signal processing. The aforementioned characteristic parameters that can reflect abnormal fluctuations in the hydraulic and thermal operating conditions of the heating network pipelines are extracted from the repaired time-series sample set, including: Extraction of residual features based on mechanistic models includes: S31. Construct a dynamic hydraulic-thermal coupling model for the heating network pipe section. The dynamic hydraulic-thermal coupling model uses the pressure, temperature, and flow rate of the upstream node as input to predict the pressure, temperature, and flow rate of the downstream node. S32. Input the real-time data into the dynamic hydraulic-thermal coupling model and calculate the corresponding type of real-time residual between the actual measured values ​​of each parameter of the downstream node and the predicted pressure, temperature and flow rate of the downstream node. S33. Calculate the statistical characteristics of the real-time residual within the sliding time window, including the mean, variance, skewness, and cumulative sum of the residual, as a residual feature set reflecting the degree to which the system state deviates from the normal mechanism. Extraction of operating condition coupling features includes: S34. The ratio of the pressure difference change to the flow rate change at adjacent sampling times yields the real-time slope of the pressure-flow characteristic curve, which serves as a key feature characterizing the instantaneous change in pipeline impedance. S35. Calculate the real-time efficiency ratio of heat power to pump power consumption by using the ratio of heat power calculated from flow rate and supply / return water temperature difference to the power consumption of the circulating pump calculated from current and voltage, as a characteristic characterizing the system's energy efficiency status. S36. Calculate the deviation rate between the supply and return water pressure difference of the same pipe section on the heating network pipeline and the theoretical design value, as a characteristic for judging the degree of hydraulic imbalance; The extraction of frequency domain features based on signal processing specifically includes: S37. Perform wavelet packet transform on the acquired pressure time-series signal and flow time-series signal and decompose them to a specific frequency band; S38. Based on the energy ratio of the acquired pressure signal in a specific high-frequency band, it serves as an early characteristic of water hammer or cavitation phenomena. S39. Calculate the moving average of the main peak value of the power spectral density of the flow signal in a specific high-frequency band, as a characterization of the slow-varying disturbance caused by heat source regulation or large user switching. S36. Calculate the deviation rate between the supply and return water pressure difference of the same pipe section on the heating network pipeline and the theoretical design value, as a characteristic for judging the degree of hydraulic imbalance.

[0035] Specifically, this method detects latent anomalies that violate physical laws. First, an accurate dynamic hydraulic-thermal coupling model is constructed. This model can predict the theoretical values ​​of downstream nodes based on the state of upstream nodes and physical laws. Then, real-time data is input into the model, and the real-time residuals between the measured values ​​of downstream nodes and the model's predictions are calculated. In a stable system, the residuals should fluctuate slightly around zero. Finally, by calculating the statistical characteristics of the residuals within a sliding time window (e.g., mean reflects systematic deviation, variance reflects the severity of fluctuations, skewness reflects the direction of deviation, and cumulative sum reflects the trend of deviation), the instantaneous residuals are transformed into a stable set of features that quantifies the degree to which the system deviates from its normal mechanism.

[0036] The pipeline impedance is monitored in real time by calculating the ratio of pressure difference to flow rate change between adjacent time points. Abrupt changes in this slope directly indicate pipeline blockage, valve malfunction, or leakage. The energy efficiency of the system is directly characterized by calculating the ratio of delivered heat power to circulating pump power consumption. A decrease in this ratio is a combined signal of pump performance degradation, pipeline fouling, or hydraulic imbalance. The degree of hydraulic imbalance in the current system is directly quantified by calculating the deviation between the actual supply and return water pressure difference and the theoretical design value; this is a crucial characteristic for determining whether the pipeline network is operating under optimal conditions.

[0037] It captures periodic or transient information related to faults that is difficult to detect in time-domain signals. This is achieved by transforming pressure and flow signals from the time domain to the frequency domain. First, wavelet packet transform is used to decompose the signal into different frequency bands without loss, so that it can be focused on specific frequency components.

[0038] Then, the energy proportion of the pressure signal in the high-frequency band is analyzed. Transient impact events such as water hammer and cavitation can excite high-frequency vibrations, so this characteristic can serve as an early warning of their occurrence.

[0039] Simultaneously, the main peak value of the power spectral density of the flow signal in a specific high-frequency band is analyzed. This feature can capture slow-varying disturbances with specific frequencies caused by operations such as heat source regulation or large user switching, thereby distinguishing operational disturbances from fault disturbances.

[0040] S04. Based on the extracted feature parameters, an abnormal fluctuation identification model is established using intelligent algorithms to determine whether there are abnormal fluctuations in the hydraulic and thermal conditions of the heating network pipeline. Specifically, the aforementioned intelligent algorithm establishes an abnormal fluctuation identification model by combining two models deployed in parallel, wherein: A baseline isolation forest model trained on historical normal data is used to detect known normal pattern biases. An adaptive isolation forest model based on dynamically updated online data is used to capture new anomalous patterns generated during system evolution; In step S04, when determining whether there are abnormal fluctuations in the hydraulic and thermal conditions of the heating network pipeline, if either of the two models determines that the input comprehensive feature vector is abnormal, the subsequent fine classification process is triggered. S41. Decompose the comprehensive feature vector into a static feature stream and a dynamic temporal feature stream; S42. The static feature stream is embedded into a context vector through a fully connected network to obtain a static context vector. The dynamic temporal feature stream is input into a multi-head self-attention encoder to obtain a dynamic temporal feature vector. S43. Aggregate the static context vector and the dynamic temporal feature vector to form the final working condition representation vector, and input it into the classifier. S44. Convert the heating network topology into graph data, where: nodes in the graph data represent key monitoring points, and edges represent pipeline connections; S45. When an anomaly is detected at a single node, the condition representation vector and graph structure data are input together into the pre-trained graph neural network. S46. Graph neural networks simulate the propagation effect of faults along the pipeline network through message passing mechanisms. They combine the abnormal characteristics of the current single node with the state of the adjacent nodes of the previous single node to output a joint judgment on the location of the fault root cause and the fault type.

[0041] Specifically, a hierarchical intelligent model architecture is used to bridge the gap from rapid anomaly detection to accurate fault diagnosis. This step first employs a parallel-deployed dual-isolation forest model as the trigger mechanism for anomaly detection: one is a baseline isolation forest model trained on historical normal data, used to stably detect anomalies deviating from known normal patterns; the other is an adaptive isolation forest model that can be dynamically updated using online data, used to capture new anomaly patterns generated during system evolution. This "dual-insurance" design ensures that the system can effectively control false alarms while flexibly identifying new faults. Whenever either model determines that the input comprehensive feature vector is anomaly, it immediately initiates a more refined and resource-intensive diagnostic process, thereby optimizing the allocation of computational resources while maintaining a high detection rate. Once an alarm is triggered, the system immediately enters a refined classification process. This process first splits the comprehensive feature vector into a static feature stream and a dynamic temporal feature stream, and then performs deep processing through a fully connected network and a multi-head self-attention encoder, respectively, to fuse steady-state context information and dynamic temporal dependencies, forming a highly condensed final condition representation vector.

[0042] To further achieve accurate fault location and type identification, the system transforms the physical heating network topology into graph structure data and uses a pre-trained graph neural network to simulate the propagation effect of faults in the pipeline network through its message passing mechanism. This makes the diagnostic results no longer dependent on the data of a single isolated node, but comprehensively considers the state correlation between the fault source and adjacent nodes, and finally outputs a joint judgment on the location and specific type of the fault root cause.

[0043] S05. When abnormal fluctuations are identified, a graded early warning mechanism and risk control recommendations are provided based on the type and severity of the abnormal fluctuations, specifying the level of hazard to the associated pipelines, including: S51. Based on the identified fault type and location of the node, inject the corresponding fault parameters into the digital twin model of the heating network, and simulate the propagation path of the fault parameters and their impact on key system indicators within a future preset time window. S52. Based on the simulation results, calculate the scope of the fault's impact, the rate of parameter deterioration, and the potential impact level on heating supply security. S53. Based on the above calculation results, a dynamic and quantitative comprehensive risk index is generated, and the warning level is dynamically divided into any one of four levels: attention, warning, serious and emergency, according to the index. S54. Construct a knowledge graph that integrates historical fault cases, equipment maintenance manuals, and expert experience, so that data pop-ups with faults, symptoms, equipment, and handling measures are formed at the nodes, and edges represent the logical relationships between them, as well as trigger symbols to display the data pop-ups. S55. Match the feature vector of the current fault with the handling knowledge graph to retrieve the Top-K most similar historical cases and their successful handling solutions; S56. Generate a personalized handling suggestion list for the current fault based on graph reasoning. The personalized handling suggestion list includes key confirmation steps, priority operation sequence, required spare parts and tools, and a list of related interlocking equipment.

[0044] Specifically, the abnormal signals identified in the preceding steps are transformed into operational decisions with clear guidance, thereby achieving an intelligent upgrade from "passive alarm" to "proactive prevention and control." This system is designed to address the disconnect between early warning and response in traditional monitoring systems. Through two core components—dynamic risk prediction and intelligent decision support—it provides operations personnel with a comprehensive solution that includes not only risk levels but also specific operational paths.

[0045] In terms of implementation, after obtaining the fault type and location, the system will inject them as parameters into the digital twin model of the heating network to simulate the propagation path and impact of the fault in the future. Then, it will quantify and calculate key indicators such as the scope of impact and the rate of deterioration, and finally integrate them to generate a dynamic comprehensive risk index. Based on this, the warning level will be scientifically divided into four levels: "attention, warning, serious, and emergency", thereby realizing the accurate assessment of risk from qualitative to quantitative and from the present to the future.

[0046] Subsequently, the system initiates an intelligent decision support process: through a pre-built knowledge graph that integrates historical cases, maintenance manuals, and expert experience, the characteristics of the current fault are matched with nodes in the graph to quickly retrieve the Top-K most similar historical cases and their handling solutions. Based on this, the system uses the logical relationships of the graph to reason and generate a highly personalized list of handling suggestions. This list not only includes priority operation sequences and key confirmation steps, but also clarifies the required spare parts, tools, and affected related equipment, thereby directly transforming the diagnostic results into an immediately executable and systematic maintenance work order, greatly improving the speed and accuracy of fault response.

[0047] This embodiment significantly improves the accuracy and reliability of fault identification through multi-sensor data fusion and anomaly data repair mechanisms. It utilizes a thermodynamic mechanism model to clean and preprocess real-time acquired hydraulic and thermal parameters, and detects potential anomalies through residual analysis. Repairing these anomalies is then performed using a long short-term memory network or mechanism model, ensuring the consistency and integrity of the time-series sample set. Simultaneously, the extracted feature parameters comprehensively capture abnormal fluctuation patterns in the heating network pipelines, avoiding misjudgments or omissions caused by traditional methods relying on a single data source or model, thus achieving more accurate fault detection in complex operating environments. The parallel deployment of baseline isolation forest models and adaptive isolation forest models detects deviations from known normal patterns and captures new anomaly patterns generated during system evolution, overcoming the shortcomings of poor portability in traditional expert rule methods and the difficulty of identifying unsimulated faults in theoretical model methods. By combining graph neural networks to analyze the heating network topology and simulate the propagation effect of faults along the pipeline, the method achieves joint judgment of fault root cause location and type, enabling it to adapt to different heating network configurations and operating conditions, reducing reliance on external expert knowledge or large amounts of historical data. This enables maintenance personnel to take timely and targeted measures based on the type and severity of abnormal fluctuations. Simultaneously, by utilizing a knowledge graph to match historical cases and expert experience, a personalized list of handling suggestions is generated, improving the efficiency and accuracy of fault response and reducing the risk of heating interruptions or equipment damage caused by abnormal hydraulic and thermal operating conditions, thereby optimizing the operation and maintenance management of the heating network system.

[0048] Example 2

[0049] This invention provides a non-transitory computer-readable storage medium storing at least one instruction or at least one program segment, which is loaded and executed by a processor to implement the following steps: Sensors are installed at key nodes of the heating network pipeline to collect the hydraulic and thermal parameters of the heating network pipeline in real time. The hydro-thermal parameters are cleaned and fused preprocessed to obtain a time series sample set. Abnormal data in the time series sample set are detected and repaired based on the thermodynamic mechanism model, so that the repaired time series sample set and multi-sensor data are consistent. Feature parameters that reflect abnormal fluctuations in the hydraulic and thermal operating conditions of the heating network pipeline are extracted from the restored time-series sample set. These feature parameters include residual features based on the mechanism model, operating condition coupling features, and frequency domain features based on signal processing. Based on the extracted feature parameters, an abnormal fluctuation identification model is established using intelligent algorithms to determine whether there are abnormal fluctuations in the hydraulic and thermal conditions of the heating network pipeline. When abnormal fluctuations are identified, a graded early warning mechanism and risk prevention and control recommendations are given based on the type and degree of the abnormal fluctuations, indicating the severity of the hazard to the associated pipeline.

[0050] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments of the above methods. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0051] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is used as an example. In practical applications, the above functions can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above.

[0052] Example 3

[0053] This invention provides an electronic device, including a processor and a memory, wherein the memory stores at least one instruction or at least one program segment, and the at least one instruction or the at least one program segment is loaded and executed by the processor to implement the following steps: Sensors are installed at key nodes of the heating network pipeline to collect the hydraulic and thermal parameters of the heating network pipeline in real time. The hydro-thermal parameters are cleaned and fused preprocessed to obtain a time series sample set. Abnormal data in the time series sample set are detected and repaired based on the thermodynamic mechanism model, so that the repaired time series sample set and multi-sensor data are consistent. Feature parameters that reflect abnormal fluctuations in the hydraulic and thermal operating conditions of the heating network pipeline are extracted from the restored time-series sample set. These feature parameters include residual features based on the mechanism model, operating condition coupling features, and frequency domain features based on signal processing. Based on the extracted feature parameters, an abnormal fluctuation identification model is established using intelligent algorithms to determine whether there are abnormal fluctuations in the hydraulic and thermal conditions of the heating network pipeline. When abnormal fluctuations are identified, a graded early warning mechanism and risk prevention and control recommendations are given based on the type and degree of the abnormal fluctuations, indicating the severity of the hazard to the associated pipeline.

[0054] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A method for intelligent identification and fault early warning of abnormal fluctuations in the hydraulic and thermal operating conditions of a heating network pipeline, characterized in that, The method includes the following steps: S01. Install sensors at key nodes of the heating network pipeline to collect the hydraulic and thermal parameters of the heating network pipeline in real time. S02. The hydro-thermal parameters are cleaned and fused preprocessed to obtain a time series sample set. Abnormal data detection of the time series sample set is obtained based on the thermodynamic mechanism model and repaired to ensure that the repaired time series sample set and multi-sensor data are consistent. S03. Extract feature parameters that can reflect abnormal fluctuations in the hydraulic and thermal operating conditions of the heating network pipeline from the repaired time series sample set. The feature parameters include residual features based on the mechanism model, operating condition coupling features, and frequency domain features based on signal processing. S04. Based on the extracted feature parameters, an abnormal fluctuation identification model is established using an intelligent algorithm to determine whether there are abnormal fluctuations in the hydraulic and thermal conditions of the heating network pipeline. S05. When abnormal fluctuations are identified, a graded early warning mechanism for the degree of hazard of the associated pipeline and risk prevention and control recommendations shall be given based on the type and degree of the abnormal fluctuations.

2. The method for intelligent identification and fault early warning of abnormal fluctuations in the hydraulic and thermal operating conditions of a heating network pipeline according to claim 1, characterized in that, The installation of sensors at key nodes in the heating network pipeline includes: It consists of multiple pressure and temperature sensors arranged on elbows, tees, valves and steam traps to obtain soft measurement data sets; A sensor simulating the mechanism of thermal pipeline operation characteristics is established. It is used to collect data on pipe sections not arranged on the heating network pipeline by the physical data acquisition module. It is combined with historical operation data to establish a data-driven model. The two are integrated to construct a soft measurement model of thermal pipeline pressure and temperature, and obtain soft measurement datasets that are not easy to measure.

3. The method for intelligent identification and fault early warning of abnormal fluctuations in the hydraulic and thermal operating conditions of a heating network pipeline according to claim 1, characterized in that, The step S02, which involves cleaning and fusing the hydrothermal parameters to obtain a time-series sample set, includes: S21. Substitute the real-time collected data into the thermodynamic mechanism model to obtain the model prediction value, and subtract the model prediction value from the actual measured value in the hydrothermal parameters to obtain the residual. S22. When the residual continuously exceeds the preset threshold and continues to exceed the first time period, the measurement data within that time period is determined to be potentially abnormal data. S23. For the potential abnormal data, forward extrapolation is performed using the mechanistic relationship model, or data repair is performed using a data prediction model based on long short-term memory networks to generate alternative values.

4. The method for intelligent identification and fault early warning of abnormal fluctuations in the hydraulic and thermal operating conditions of a heating network pipeline according to claim 1, characterized in that, The step of keeping the repaired time-series sample set and multi-sensor data consistent in S02 includes: S24. For multiple sensors of the same type at the same monitoring point, or two sensors at adjacent points, fuzzy set theory is used to fuse the data collected by the multiple sensors to obtain the fusion result. S25. Calculate the confidence distance between each sensor data and the fusion result. When the confidence distance of a certain sensor data exceeds the dynamic threshold, the sensor data is determined to be unreliable and is then removed.

5. The method for intelligent identification and fault early warning of abnormal fluctuations in the hydraulic and thermal operating conditions of a heating network pipeline according to claim 1, characterized in that, The step S03, which involves extracting characteristic parameters from the repaired time-series sample set that reflect abnormal fluctuations in the hydraulic and thermal operating conditions of the heating network pipeline, includes: The extraction of residual features based on the mechanism model includes: S31. Construct a dynamic hydraulic-thermal coupling model for the heating network pipe section. The dynamic hydraulic-thermal coupling model uses the pressure, temperature, and flow rate of the upstream node as input to predict the pressure, temperature, and flow rate of the downstream node. S32. Input the real-time data into the dynamic hydraulic-thermal coupling model and calculate the real-time residuals of the corresponding type between the actual measured values ​​of each parameter of the downstream node and the predicted pressure, temperature and flow rate of the downstream node. S33. Calculate the statistical characteristics of the real-time residual within the sliding time window, including the mean, variance, skewness, and cumulative sum of the residual, as a residual feature set reflecting the degree to which the system state deviates from the normal mechanism. The extraction of the operating condition coupling features includes: S34. The ratio of the pressure difference change to the flow rate change at adjacent sampling times yields the real-time slope of the pressure-flow characteristic curve, which serves as a key feature characterizing the instantaneous change in pipeline impedance. S35. Calculate the real-time efficiency ratio of heat power to pump power consumption by using the ratio of heat power calculated from flow rate and supply / return water temperature difference to the power consumption of the circulating pump calculated from current and voltage, as a characteristic characterizing the system's energy efficiency status. S36. Calculate the deviation rate between the supply and return water pressure difference of the same pipe section on the heating network pipeline and the theoretical design value, as a characteristic for judging the degree of hydraulic imbalance; The extraction of frequency domain features based on signal processing specifically includes: S37. Perform wavelet packet transform on the acquired pressure time-series signal and flow time-series signal and decompose them to a specific frequency band; S38. The energy ratio of the acquired pressure signal in the specific high-frequency band is used as an early feature characterizing the occurrence of water hammer or cavitation. S39. Calculate the moving average value of the main peak value of the power spectral density of the flow signal in the specific high frequency band, as a characterization of the slow-varying disturbance characteristics caused by heat source regulation or large user switching. S36. Calculate the deviation rate between the supply and return water pressure difference of the same pipe section on the heating network pipeline and the theoretical design value, as a characteristic for judging the degree of hydraulic imbalance.

6. The method for intelligent identification and fault early warning of abnormal fluctuations in the hydraulic and thermal operating conditions of a heating network pipeline according to claim 1, characterized in that, In step S04, the intelligent algorithm establishes an abnormal fluctuation identification model by combining two models deployed in parallel, wherein: A baseline isolation forest model trained on historical normal data is used to detect known normal pattern biases. An adaptive isolation forest model based on dynamically updated online data is used to capture new anomalous patterns generated during system evolution; In step S04, when determining whether there are abnormal fluctuations in the hydraulic and thermal conditions of the heating network pipeline, if either of the two models determines that the input comprehensive feature vector is abnormal, the subsequent fine classification process is triggered. S41. The comprehensive feature vector is split into a static feature stream and a dynamic temporal feature stream; S42. The static feature stream is embedded into a context vector through a fully connected network to obtain a static context vector, and the dynamic temporal feature stream is input into a multi-head self-attention encoder to obtain a dynamic temporal feature vector. S43. Aggregate the static context vector and the dynamic temporal feature vector to form the final working condition representation vector, and input it into the classifier; S44. Convert the heating network topology into graph data, wherein: nodes in the graph data represent key monitoring points, and edges represent pipeline connections; S45. When an anomaly is detected at a single node, the condition representation vector and the graph structure data are input into a pre-trained graph neural network. S46. The graph neural network simulates the propagation effect of faults along the pipeline network through a message passing mechanism, and outputs a joint judgment on the location of the fault root cause and the fault type by combining the abnormal characteristics of the current single node with the state of the adjacent nodes of the previous single node.

7. The method for intelligent identification and fault early warning of abnormal fluctuations in the hydraulic and thermal operating conditions of a heating network pipeline according to claim 1, characterized in that, Step S05, when abnormal fluctuations are identified, provides a graded early warning mechanism for the severity of hazard to the associated pipeline and risk prevention and control recommendations based on the type and degree of the abnormal fluctuations, including: S51. Based on the identified fault type and location of the node, inject the corresponding fault parameters into the digital twin model of the heating network, and simulate the propagation path of the fault parameters and their impact on key system indicators within a future preset time window. S52. Based on the simulation results, calculate the scope of the fault's impact, the rate of parameter deterioration, and the potential impact level on heating supply security. S53. Based on the above calculation results, a dynamic and quantitative comprehensive risk index is generated, and the warning level is dynamically divided into any one of four levels: attention, warning, serious and emergency, according to the index. S54. Construct a knowledge graph that integrates historical fault cases, equipment maintenance manuals, and expert experience, so that data pop-ups with faults, symptoms, equipment, and handling measures are formed at the nodes, and edges represent the logical relationships between them, as well as trigger symbols to display the data pop-ups. S55. Match the feature vector of the current fault with the handling knowledge graph to retrieve the Top-K most similar historical cases and their successful handling solutions; S56. Generate a personalized handling suggestion list for the current fault based on graph reasoning. The personalized handling suggestion list includes key confirmation steps, priority operation sequence, required spare parts and tools, and a list of related interlocking equipment.

8. A non-transitory computer-readable storage medium, wherein the non-transitory computer-readable storage medium stores at least one instruction or at least one program segment, characterized in that, The at least one instruction or the at least one program segment is loaded and executed by the processor to implement the intelligent identification and fault early warning method for abnormal fluctuations in the hydraulic and thermal operating conditions of heating network pipelines as described in any one of claims 1-7.

9. An electronic device, characterized in that, The method includes a processor and a memory, wherein the memory stores at least one instruction or at least one program, and the at least one instruction or at least one program is loaded and executed by the processor to implement the intelligent identification and fault early warning method for abnormal fluctuations in the hydraulic and thermal conditions of heating network pipelines as described in any one of claims 1-7.

Images