A valve hall equipment infrared temperature measurement early warning method and system based on a lightweight detection network

Abstract

This invention relates to the field of power equipment condition monitoring and fault early warning technology, specifically disclosing an infrared temperature measurement and early warning method and system for valve hall equipment based on a lightweight detection network. This invention constructs a spatial topology coupling model to achieve digital twins of physical nodes and thermal interaction paths of the equipment. It utilizes real-time spatial reference anchor point identification and geometric transformation to accurately deduce the location of key monitoring nodes. By estimating cross-thermal pollution components and separating and calculating the net calorific value of nodes, it effectively identifies the true source of equipment temperature rise. Simultaneously, the system evaluates the thermal imbalance index of related subsystems to identify potential systemic risks in advance. Finally, it generates a composite early warning decision that integrates node status and system status, providing maintenance personnel with comprehensive and objective early warning information, significantly improving the intelligent operation and maintenance level of valve halls, and compensating for the diagnostic blind spots of traditional methods that only focus on individuals while neglecting the overall picture.

Description

Technical Field

[0001] This invention belongs to the field of power equipment condition monitoring and fault early warning technology, and relates to an infrared temperature measurement early warning method and system for valve hall equipment based on a lightweight detection network. Background Technology

[0002] Valve halls, as the core locations of high-voltage direct current (HVDC) transmission systems, are characterized by dense internal equipment, complex electromagnetic environments, and stringent operating temperature requirements. Real-time and accurate monitoring of the equipment's thermal status is crucial for ensuring the safe and stable operation of the power grid. The current core challenge lies in accurately separating abnormal temperature rises caused by equipment defects from complex infrared thermal images, and effectively distinguishing between localized faults in a single component and functional thermal imbalances in the entire interconnected system. Furthermore, minute camera angle fluctuations, displacements caused by equipment thermal expansion and contraction, and cross-thermal radiation contamination from nearby heat sources all severely interfere with the accuracy of temperature measurements and the reliability of fault diagnosis.

[0003] Currently, the most common solutions in the industry rely on fixed infrared thermal imaging systems combined with regular manual inspections. Fixed systems typically pre-define several fixed regions of interest on the infrared image, continuously monitor the highest or average temperature of these regions, and set simple absolute temperature thresholds or temperature rise rate thresholds to trigger alarms. Manual inspections involve experienced maintenance personnel using handheld infrared thermal imagers to periodically scan critical equipment on-site according to procedures, combining this with subjective judgment to assess the equipment's thermal condition.

[0004] The drawbacks of traditional methods are obvious: automatic alarm systems based on fixed regions of interest and simple thresholds cannot effectively isolate the effects of ambient temperature changes, equipment load fluctuations, and heat radiation from nearby equipment, resulting in high false alarm and false negative rates and failing to reveal the true net heat generation status of the equipment. Furthermore, this method is ill-suited to addressing issues such as camera vibration causing the monitored target to deviate from the preset area. Manual inspection, on the other hand, is not only labor-intensive and inefficient, but also lacks real-time detection capabilities, making it difficult to capture sudden and rapidly developing thermal faults. Moreover, diagnostic results heavily rely on human experience and lack objective, quantifiable, and unified standards. Summary of the Invention

[0005] In view of this, in order to solve the problems mentioned in the background technology, an infrared temperature measurement and early warning method and system for valve hall equipment based on a lightweight detection network is proposed.

[0006] The objective of this invention can be achieved through the following technical solutions: The first aspect of this invention provides an infrared temperature measurement and early warning method for valve hall equipment based on a lightweight detection network, including: S1, spatial topology coupling model construction: generating a spatial topology coupling model, which establishes a digital twin of the physical nodes of the equipment and the thermal interaction path for the monitoring perspective of a fixed infrared camera.

[0007] S2. Real-time spatial reference anchor point identification: Acquire real-time infrared thermal images and identify multiple spatial reference anchor points in the real-time infrared thermal images to generate a real-time anchor point coordinate set.

[0008] S3. Coordinate derivation of key monitoring nodes: Input the real-time anchor point coordinate set and the spatial topology coupling model, and calculate the precise mapping position of each key monitoring node in the real-time infrared thermal image through geometric transformation to generate a node positioning coordinate map.

[0009] S4. Estimation of cross-thermal pollution components: Based on the node location coordinate map, extract the measured temperature values ​​of each key monitoring node from the real-time infrared thermal image, and record the nodes with abnormal measured temperature values ​​as abnormal nodes. Call the spatial topology coupling model to trace its neighboring heat source nodes in order to calculate and generate cross-thermal pollution components.

[0010] S5. Calculation of Net Heating Value of Nodes: Subtract the corresponding cross-heat contamination component from the measured temperature value of the abnormal node, and use the result as the net heating value that characterizes the degree of heating caused by the fault of the abnormal node.

[0011] S6. System-level thermal imbalance index assessment: The net calorific value of all nodes belonging to the same related subsystem is selected, and the model is compared with the ideal temperature gradient preset in the spatial topology coupling model to quantify and generate a thermal imbalance index to assess the overall health status of the same related subsystem.

[0012] S7. Composite Early Warning Decision Generation: Based on the preset alarm threshold and early warning threshold, the net calorific value and thermal imbalance index are judged respectively, and the judgment results are integrated to generate a composite early warning decision.

[0013] The second aspect of the present invention provides an infrared temperature measurement and early warning system for valve hall equipment based on a lightweight detection network, comprising: a spatial topology coupling model construction module, which generates a spatial topology coupling model, which establishes a digital twin of the physical nodes of the equipment and the thermal interaction path for the monitoring perspective of a fixed infrared camera.

[0014] The real-time spatial reference anchor point identification module acquires real-time infrared thermal images and identifies multiple spatial reference anchor points in the real-time infrared thermal images to generate a real-time anchor point coordinate set.

[0015] The key monitoring node coordinate derivation module takes as input a real-time anchor point coordinate set and a spatial topology coupling model, and calculates the precise mapping position of each key monitoring node in the real-time infrared thermal image through geometric transformation, so as to generate a node positioning coordinate map.

[0016] The cross-thermal pollution component estimation module extracts the measured temperature values ​​of each key monitoring node from the real-time infrared thermal image based on the node location coordinate map, and records nodes with abnormal measured temperature values ​​as abnormal nodes. It then calls the spatial topology coupling model to trace its neighboring heat source nodes in order to calculate and generate the cross-thermal pollution component.

[0017] The node net calorific value separation calculation module subtracts the corresponding cross-heat contamination component from the measured temperature value of the abnormal node, and the calculation result is used as the net calorific value characterizing the degree of fault heating of the abnormal node itself.

[0018] The system-level thermal imbalance index assessment module filters out the net calorific value of all nodes belonging to the same related subsystem, and performs pattern comparison based on the preset ideal temperature gradient in the spatial topology coupling model to quantify and generate a thermal imbalance index to assess the overall health status of the same related subsystem.

[0019] The composite early warning decision generation module determines the net calorific value and thermal imbalance index based on preset alarm thresholds and early warning thresholds, and integrates the determination results to generate a composite early warning decision.

[0020] Compared with the prior art, the embodiments of the present invention have at least the following advantages or beneficial effects: (1) The present invention constructs a spatial topological coupling model to digitally integrate the physical structure of the device, the heat transfer path and the camera visual space, thereby achieving accurate, dynamic and anti-interference positioning of the monitoring target. Even if the camera is slightly displaced or the device body is deformed due to thermal expansion and contraction, the system can still reverse-calculate the accurate image coordinates of the key monitoring nodes that are obscured or have indistinct features by identifying the spatial reference anchor point in real time and performing geometric transformation, thus ensuring the long-term consistency and accuracy of the temperature data acquisition points and fundamentally solving the monitoring failure problem caused by target mismatch in traditional methods.

[0021] (2) This invention proposes a quantitative estimation and separation method for cross-thermal contamination, which can effectively identify the true source of equipment temperature rise. Instead of simply judging the observed high temperature as a fault, it uses a pre-constructed equipment thermal coupling network diagram to trace and calculate the contribution of nearby heat sources to the "background" temperature rise of the target node through heat conduction and heat radiation. By subtracting this cross-thermal contamination component from the measured temperature, a "net calorific value" that can intuitively reflect the health status of the equipment is obtained, which greatly improves the accuracy of single-node fault diagnosis and significantly reduces false alarms caused by environmental factors.

[0022] (3) This invention compares the real-time temperature gradient of the cooling circuit and other related equipment groups with the ideal temperature gradient under the design conditions, thereby quantitatively assessing the degree of deviation of the entire subsystem's operating state. This method can detect potential systemic risks caused by factors such as decreased cooling efficiency and blockage of flow channels in advance. Such risks may not manifest as a sharp temperature rise at any single node in the early stages, thus making up for the diagnostic blind spot of traditional methods that only focus on individuals and ignore the whole.

[0023] (4) This invention provides maintenance personnel with more comprehensive and instructive early warning information by generating a composite early warning decision based on the fusion of node status and system status. It combines the net heat value alarm reflecting the actual fault of a single component with the thermal imbalance index early warning reflecting the overall health trend of the system. It can clearly distinguish between urgent equipment failures that need to be dealt with immediately and systemic operational anomalies that need to be monitored and investigated. This makes the formulation of fault response and maintenance strategies more targeted and predictive, and improves the intelligent operation and maintenance level of the entire valve hall. Attached Figure Description

[0024] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the 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.

[0025] Figure 1 This is a schematic diagram of the method steps of the present invention.

[0026] Figure 2 This is a schematic diagram of the system structure connection of the present invention. Detailed Implementation

[0027] 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.

[0028] Please see Figure 1 The first aspect of the present invention provides an infrared temperature measurement and early warning method for valve hall equipment based on a lightweight detection network, including: S1, spatial topology coupling model construction: generating a spatial topology coupling model, which establishes a digital twin of the physical nodes of the equipment and the thermal interaction path for the monitoring perspective of a fixed infrared camera.

[0029] In a specific embodiment of the present invention, the specific steps for generating the spatial topological coupling model include: fusing the three-dimensional design drawings of the valve hall equipment with the installation pose information of the fixed infrared camera, and calibrating the estimated imaging area of ​​each key monitoring node in the infrared image to generate a physical node position mapping library.

[0030] It should be noted that this step aims to construct a digital thermodynamic model of the valve hall equipment to simulate the heat distribution and transfer relationships of the equipment as observed from a fixed infrared camera perspective in the real world. The implementation process begins with fusing design data, loading the 3D design drawings of the valve hall equipment into computer-aided design software or a dedicated 3D modeling environment. Simultaneously, the precise 3D spatial coordinates and orientation angles of the fixed infrared camera installed on-site are acquired. Using this information, the 3D model of the equipment is aligned with the camera's virtual perspective through coordinate system transformation, thereby reproducing the actual monitoring image from the camera in the computer. Under this virtual perspective, based on engineering experience and the equipment operation manual, all key monitoring nodes requiring focused monitoring are marked on the 3D model. These key monitoring nodes are potential high-risk points for equipment malfunctions and overheating, such as electrical connection points like the thyristor module terminals of the converter valve, or fluid path inflection points like bends in cooling water pipes. Using projection transformation algorithms from computer graphics, the positions of these marked nodes in 3D space are projected onto the camera's 2D imaging plane, calculating the rectangular range that each key monitoring node might appear in the infrared image, i.e., the estimated imaging area. Each key monitoring node's unique identifier is associated with and stored along with its corresponding estimated imaging area coordinates, ultimately forming a physical node location mapping library.

[0031] Based on the physical connections and spatial adjacency between devices, the heat conduction and heat radiation paths between key monitoring nodes are analyzed and quantified to construct a device thermal coupling network diagram.

[0032] It should be noted that, next, based on the loaded 3D design drawings of the equipment, the physical connection relationships and non-contact but spatially adjacent relationships between equipment components are analyzed. For components connected by bolts or direct welding, heat conduction paths are defined between them; for component surfaces that are "visible" to each other in space without obstruction, heat radiation paths are defined between them. Through thermodynamic simulation, a heat transfer coefficient or view factor is quantified for each path to describe the efficiency of heat transfer. All key monitoring nodes and their heat conduction and heat radiation paths are organized into nodes and edges to construct a thermal coupling network diagram of the equipment. In this thermal coupling network diagram, nodes represent equipment components, and edges represent heat transfer channels and their strength.

[0033] Based on the design parameters of the associated subsystems, the ideal temperature gradient between key monitoring nodes is defined to construct the heat flow topology of the subsystems.

[0034] It should be noted that, going a step further, for equipment groups with clearly defined fluid flow directions, such as cooling systems in valve halls, the inlet and outlet temperatures, rated flow rates, and design temperature values ​​at various points along the flow path under standard operating conditions are extracted from the equipment's design parameters. Based on these design values, the normal temperature drop or rise from upstream to downstream nodes is calculated, forming a standardized temperature change sequence, which is the ideal temperature gradient. All relevant nodes within the same cooling loop are connected in sequence according to the fluid flow direction, and the ideal temperature gradients between nodes are labeled, thereby constructing a subsystem heat flow topology diagram to depict the heat distribution pattern that this cooling system should exhibit during healthy operation.

[0035] The physical node location mapping library, the device thermal coupling network diagram, and the subsystem thermal flow topology diagram are integrated to generate a spatial topology coupling model.

[0036] It should be noted that, finally, the three data structures—the physical node location mapping library, the device thermal coupling network diagram, and the subsystem thermal flow topology diagram—are logically integrated and packaged into a unified dataset, which generates the spatial topology coupling model, providing comprehensive static prior knowledge for subsequent analysis.

[0037] The fixed infrared camera's installation pose information is a set of data describing the camera's position and attitude in the valve hall's global coordinate system. Specifically, it includes three-dimensional coordinate values ​​and three rotation angles: pitch, yaw, and roll. This data is based on precise calibration through on-site surveys, such as measurements using laser rangefinders and gyroscopes. Key monitoring nodes are a pre-selected set of equipment components or specific locations on components, whose temperature changes are crucial for assessing equipment health. In this method, key monitoring nodes are limited to two categories: first, electrical connection points prone to overheating due to poor contact or overload, such as valve tower terminals; and second, fluid path inflection points that reflect the operating status of fluids such as cooling systems, such as cooling water pipe bends. The estimated imaging area is a rectangular frame drawn on the two-dimensional infrared image plane for each key monitoring node. Its coordinates and dimensions are derived from three-dimensional projection, with a certain redundancy margin reserved to accommodate slight changes in viewing angle. The physical node location mapping library is a data lookup table with a key-value pair structure. The key is a unique identifier for a key monitoring node, and the value is the image coordinates of its corresponding estimated imaging area. This library establishes a preliminary association between the physical identity of equipment and its image location. Physical connections and spatial adjacencies between equipment are topological relationships between components derived from 3D design drawings. The former refers to physical contact between components, while the latter refers to the mutual radiative heat exchange between component surfaces in space. Heat conduction paths and heat radiation paths are logical connections established between nodes with physical or spatial adjacency relationships to simulate two basic heat transfer methods. The equipment thermal coupling network diagram is a graph data structure where nodes represent key monitoring nodes and other related components, and edges represent heat conduction or heat radiation paths between nodes. A cooling loop is a closed or open fluid circulation system in a valve hall used to remove heat from the operating equipment, such as the pure water cooling system of a converter valve. Design parameters refer to a series of performance indicators and physical parameters provided by the manufacturer when the equipment leaves the factory, ensuring the safe and efficient operation of the equipment under rated conditions. These include rated voltage, current, cooling water flow rate, and standard operating temperature. These settings are based on thousands of laboratory tests and simulation optimizations. A subsystem thermal flux topology diagram is another graph data structure that specifically describes the heat flow patterns within an associated subsystem. Nodes in the diagram are ordered according to the fluid flow direction, and the ideal temperature gradient values ​​are labeled on the edges.

[0038] For example, to construct a spatial topology coupling model of a thyristor converter valve, its three-dimensional design drawing file is first loaded. The installation pose information of the fixed infrared camera is obtained as coordinates X=5.2m, Y=3.1m, Z=4.5m, pitch angle=-15°, yaw angle=35°, roll angle=0.1°. In the three-dimensional model, a thyristor terminal located on the upper arm of phase A is selected as a key monitoring node and marked as "Node_T_A1_Upper_Terminal". Through projection calculation, its estimated imaging area in the infrared image is obtained as a rectangular box [(85,120),(95,130)]. The identifier "Node_T_A1_Upper_Terminal" and the coordinates "(85,120,95,130)" are stored in the physical node position mapping library. Next, analysis revealed that "Node_T_A1_Upper_Terminal" and its adjacent busbar "Busbar_A1" are physically connected by bolts, creating a heat conduction path; simultaneously, it has a spatial adjacency with the valve tower surface of the opposite phase B, creating a heat radiation path. Based on material properties and distance, heat transfer parameters were calculated and recorded, and these relationships were stored in the equipment's thermal coupling network diagram. Then, for its pure water cooling loop, design parameters showed that the standard temperature of the cooling water entering the loop was 40℃, and the temperature at the outlet was 45℃. The ideal temperatures of the three series-connected radiator nodes ("Node_Rad_1", "Node_Rad_2", "Node_Rad_3") should be 41.5℃, 43℃, and 44.5℃ respectively. Therefore, an ideal temperature gradient sequence of +1.5℃, +1.5℃, +1.5℃ was defined. This information was used to construct the subsystem's heat flow topology diagram. Finally, the physical node location mapping library, device thermal coupling network diagram, and subsystem thermal flow topology diagram generated above are packaged to generate a spatial topology coupling model.

[0039] S2. Real-time spatial reference anchor point identification: Acquire real-time infrared thermal images and identify multiple spatial reference anchor points in the real-time infrared thermal images to generate a real-time anchor point coordinate set.

[0040] In a specific embodiment of the present invention, the specific steps for generating a real-time anchor point coordinate set include: calling a preset lightweight recognition operator to analyze the real-time infrared thermal image.

[0041] By using a lightweight recognition operator, equipment components with constant positions and prominent shapes within real-time infrared thermal images are identified as spatial reference anchor points.

[0042] Extract and output the image coordinates of the spatial reference anchor points to generate a real-time anchor point coordinate set.

[0043] It should be noted that the implementation process begins with acquiring a single frame from the video stream transmitted from a fixed infrared camera on-site; this image is the real-time infrared thermal image. This real-time infrared thermal image, in the form of a pixel matrix, is input into a pre-trained lightweight recognition operator. Once invoked, the lightweight recognition operator immediately performs a rapid scan and analysis of the entire real-time infrared thermal image. Its core task is to find and match the visual features of one or more predefined sets of spatial reference anchors within the real-time infrared thermal image. These anchors are selected during system deployment based on principles such as their constant position in the valve hall environment, clear outlines, high temperature contrast, and resistance to obstruction by routine inspection personnel or mobile devices. The recognition process is completed through a series of efficient image feature comparison algorithms. The lightweight recognition operator disregards the temperature of the anchor itself, focusing only on its shape, edges, and relative texture within the real-time infrared thermal image. Once the lightweight recognition operator successfully identifies these spatial reference anchors, it immediately calculates the precise position of each anchor in the current image coordinate system, typically outputting the pixel coordinates of the center point of the anchor's bounding rectangle. Finally, the image coordinates of all successfully identified spatial reference anchor points are compiled into a list or array-structured data set, which serves as the final output of this step: the real-time anchor point coordinate set.

[0044] In this context, a real-time spatial reference anchor point refers to a specific instance of a spatial reference anchor point that is successfully identified and located at a given moment in the current real-time infrared thermal image. The lightweight recognition operator is an optimized computer vision algorithm model characterized by low computational resource consumption, high recognition speed, and the ability to be deployed on edge computing devices, specifically designed for rapidly identifying specific targets from input images. In this method, the lightweight recognition operator is pre-trained to specifically recognize the features of selected spatial reference anchor points. Its training dataset contains several infrared images of the valve hall acquired under different lighting conditions, seasons, and operating conditions. Spatial reference anchor points are a set of physical objects pre-selected during the system configuration phase. These objects must possess characteristics such as constant position, prominent shape, and resistance to occlusion, such as the metal flange on top of a supporting insulator in the valve hall, a specific door handle of an equipment cabinet, or a fixed sign on a wall. Image coordinates are numerical pairs used to uniquely locate a pixel position on a two-dimensional image plane, typically represented as (x, y), where x represents the horizontal pixel position and y represents the vertical pixel position.

[0045] Following the example from the previous step, the system acquires a real-time infrared thermal image and calls a pre-trained lightweight recognition operator. This lightweight recognition operator was designed to identify the top cover of the support insulator (model BZL-550) and the upper right corner of the equipment cabinet "Cabinet_C2". After scanning the real-time infrared thermal image, the lightweight recognition operator successfully matched these two spatial reference anchor points. Subsequently, it extracted the center image coordinates of the support insulator top cover as (255, 410) and the image coordinates of the upper right corner of the equipment cabinet as (850, 150). These two coordinate pairs are combined into a real-time anchor point coordinate set, specifically [(255, 410), (850, 150)], and output it for use in the next step.

[0046] S3. Coordinate derivation of key monitoring nodes: Input the real-time anchor point coordinate set and the spatial topology coupling model, and calculate the precise mapping position of each key monitoring node in the real-time infrared thermal image through geometric transformation to generate a node positioning coordinate map.

[0047] In a specific embodiment of the present invention, the specific steps of calculating the precise mapping position of each key monitoring node in the real-time infrared thermal image through geometric transformation to generate a node positioning coordinate map include: obtaining the preset precise relative position relationship between each key monitoring node and the spatial reference anchor point from the spatial topology coupling model.

[0048] Based on the real-time anchor point coordinate set and the precise relative position relationship, a geometric transformation model that can describe the changes in the current view is calculated.

[0049] A geometric transformation model is applied to the reference positions of all key monitoring nodes to calculate and generate a node positioning coordinate map containing the precise mapping positions of each key monitoring node.

[0050] It should be noted that the implementation process first calls the generated spatial topology coupling model and extracts a subset of data—the physical node position mapping library—from it. Simultaneously, it receives and inputs the real-time anchor point coordinate set generated in the second step. Next, the algorithm enters the core geometric solution stage. The physical node position mapping library not only stores the estimated imaging areas of each key monitoring node, but more importantly, it contains the precise and invariant relative spatial positional relationships between all key monitoring nodes and all spatial reference anchor points during model construction. This relationship is pre-stored in the form of coordinate differences or relative vectors. The system pairs the real-time coordinates of each anchor point in the real-time anchor point coordinate set with the corresponding reference coordinates in the physical node position mapping library, forming multiple pairs of current and reference position coordinate points. Based on these paired coordinate points, a geometric transformation model, such as affine transformation, is used to inversely calculate a mathematical transformation matrix that describes the change from the reference view to the current real-time infrared thermal image view. This transformation matrix captures the overall image translation, rotation, scaling, or distortion caused by slight camera vibrations or thermal expansion and contraction. After calculating the transformation matrix, the system iterates through each key monitoring node recorded in the physical node location mapping library. It extracts the reference coordinates of each key monitoring node in the reference view and then applies the calculated transformation matrix to perform a coordinate transformation operation. The result of this operation is the precise mapped position of the key monitoring node in the current real-time infrared thermal image. This process is repeated for all key monitoring nodes in the library, ultimately generating a detailed list of coordinates, where each key monitoring node's unique identifier is matched with precise image coordinates. This list constitutes the node location coordinate map.

[0051] It should also be noted that the core of this step is to perform geometric transformation calculations, which can be explained using a simplified affine transformation formula, as follows: This is a matrix operation expression. and A vector representing image coordinates is typically a 3x1 homogeneous coordinate column vector containing x, y, and 1, with the dimension being pixel position. It is a 3x3 affine transformation matrix, and it is a dimensionless transformation operator. After the operation... The dimension remains pixel position, so the dimensions are consistent. This represents the precise mapping coordinate vector of the key monitoring node in the current real-time infrared thermal image. This represents the affine transformation matrix calculated based on the reference position and real-time position of the spatial reference anchor point. This affine transformation matrix is ​​obtained by solving for at least three pairs of non-collinear anchor point coordinates. It is set on the assumption that the change in the camera's field of view can be approximated by a linear two-dimensional transformation. This represents the coordinate vector in the reference image used when constructing the physical node location mapping library for this key monitoring node.

[0052] The precise relative position relationship is pre-calculated and stored data describing the geometric relationship between any key monitoring node and a spatial reference anchor point in three-dimensional physical space. This relationship is constant under the premise that the device structure remains unchanged and is projected onto the reference image coordinate system as a calculation benchmark. Geometric transformation is a mathematical operation that calculates a function or matrix that describes this point-position mapping relationship based on a set of known input and output point pairs, and uses this function or matrix to deduce the mapped positions of other unknown points. The precise mapped position is the specific pixel coordinates of a key monitoring node in a specific frame of the current real-time infrared thermal image, calculated through geometric transformation, with a higher accuracy than the estimated imaging area in the first step. The node positioning coordinate map is a data mapping table with a key-value pair data structure. The key is the unique identifier of each key monitoring node, and the value is the precise mapped position coordinates of that node calculated in this step, used to provide accurate target positioning for subsequent temperature extraction.

[0053] Following the example from the previous step, the system invokes the physical node location mapping library in the spatial topology coupling model and receives the real-time anchor point coordinate set generated in the previous step as [(255,410),(850,150)]. The physical node location mapping library records the reference coordinates of these two spatial reference anchor points in the reference image as (250,400) and (860,145), respectively. The system uses these two pairs of coordinates ((250,400) to (255,410) and (860,145) to (850,150)) to calculate a geometric transformation matrix. Next, the system retrieves the reference coordinates (90,125) of the key monitoring node "Node_T_A1_Upper_Terminal" from the physical node location mapping library. Applying the transformation matrix calculated earlier to these reference coordinates, the system calculates its precise mapping position in the current real-time infrared thermal image as (96,134). After repeating this process for all other key monitoring nodes, a node location coordinate map is generated, which contains a record named {“Node_T_A1_Upper_Terminal”:(96,134)}.

[0054] S4. Estimation of cross-thermal pollution components: Based on the node location coordinate map, extract the measured temperature values ​​of each key monitoring node from the real-time infrared thermal image, and record the nodes with abnormal measured temperature values ​​as abnormal nodes. Call the spatial topology coupling model to trace its neighboring heat source nodes in order to calculate and generate cross-thermal pollution components.

[0055] In a specific embodiment of the present invention, the specific steps for calculating and generating cross-thermal pollution components include: calling the device thermal coupling network diagram in the spatial topology coupling model and tracing all neighboring heat source nodes that have thermal interaction paths with the abnormal nodes.

[0056] Obtain the current measured temperature value of each nearby heat source node.

[0057] By combining the heat transfer coefficients and path parameters recorded in the equipment thermal coupling network diagram, the total heat contribution transferred from all neighboring heat source nodes to the abnormal nodes is calculated to generate cross-heat contamination components.

[0058] It should be noted that the implementation process begins with receiving the node location coordinate map. Based on the precise location information provided by this map, the system locates each key monitoring node on the real-time infrared thermal image and reads the average temperature value of that pixel or its neighborhood as the measured temperature value of that key monitoring node. After reading the temperature of all key monitoring nodes, the system checks whether the measured temperature value of each node exceeds the preset upper limit of normal operating temperature. For any node whose measured temperature value is determined to be abnormal, the system will initiate a traceability analysis. It will call the generated spatial topology coupling model and extract the device thermal coupling network diagram from it. In this thermal coupling network diagram, with the abnormal node as the center, all nodes directly connected to it through heat conduction or heat radiation paths are found; these found nodes are the neighboring heat source nodes. Subsequently, the system obtains the current measured temperature value of each of these neighboring heat source nodes. Combining this with the heat transfer coefficients and path parameters pre-stored in the thermal coupling network diagram, the system performs a heat transfer calculation. Specifically, it estimates the temperature rise caused by each neighboring heat source node through its corresponding heat transfer path to the anomalous node due to its own temperature. The sum of the temperature rise effects caused by all neighboring heat source nodes is defined as the comprehensive background thermal impact on the anomalous node. This value is the final output of this step, namely the cross-thermal contamination component.

[0059] It should also be noted that the core of this step lies in quantifying the heat contribution from nearby heat sources, which can be estimated using a linear superposition model, as shown in the following formula: ,in, Represents the cross-contamination component; For the first Measured temperatures of nearby heat source nodes; The ambient reference temperature is typically the temperature of a temperature measurement point farthest from any heat source within the equipment area, or a set standard ambient temperature such as 25 degrees Celsius, used to establish a temperature difference benchmark; Σ represents the summation of the effects of all identified nearby heat source nodes; Is with the first The heat transfer coefficient of a path connecting adjacent heat source nodes is a dimensionless weighted value. It is determined based on the combined effects of conduction, radiation, and convection recorded in the equipment's thermal coupling network diagram, and is obtained through heat transfer simulation experiments. For example, the heat transfer coefficient of a tightly bolted adjacent node... The value might be set to 0.15, while a distant radiating node's... The value may be as low as 0.02.

[0060] The measured temperature value is the temperature reading of the key monitoring node directly read from the current real-time infrared thermal image after accurate positioning using the node location coordinate map. The cross-contamination component is a quantified value in temperature, specifically used to indicate how much of a node's measured temperature is contributed by surrounding heat-generating devices through heat conduction and radiation, rather than by its own malfunction. Proximity heat source nodes are all other nodes in the device thermal coupling network diagram that have a direct heat conduction or radiation path connected to a specified abnormal node. The heat transfer coefficient and path parameters are pre-existing data in the device thermal coupling network diagram, used to describe the efficiency and mode of heat transfer between two nodes. The heat transfer coefficient is as described in the formula section above, while path parameters may include physical parameters such as distance, contact area, and surface emissivity. These parameters collectively determine the value of the heat transfer coefficient.

[0061] Following the example from the previous step, the system, based on the node location coordinate map, reads the measured temperature of the critical monitoring node "Node_T_A1_Upper_Terminal" as 75℃ from the real-time infrared thermal image. This temperature exceeds the alarm threshold of 60℃ and is therefore identified as an abnormal node. The system then calls the device thermal coupling network diagram and finds two adjacent heat source nodes directly connected to "Node_T_A1_Upper_Terminal": "Busbar_A1" and the opposite valve tower surface "Surface_B1". The system reads the current measured temperature of "Busbar_A1" as 70℃ and the current measured temperature of "Surface_B1" as 55℃. From the device thermal coupling network diagram, the corresponding heat transfer coefficients k are found to be 0.1 and 0.05, respectively. Assuming an ambient reference temperature of 25℃, the temperature contribution from "Busbar_A1" is calculated to be 4.5℃, and the temperature contribution from "Surface_B1" is 1.5℃. Superimposing the two values ​​yields a total heat contribution of 6.0℃ for this anomalous node. Finally, the system generates a cross-thermal contamination component for "Node_T_A1_Upper_Terminal" with a value of 6.0℃.

[0062] S5. Calculation of Net Heating Value of Nodes: Subtract the corresponding cross-heat contamination component from the measured temperature value of the abnormal node, and use the result as the net heating value that characterizes the degree of heating caused by the fault of the abnormal node.

[0063] In a specific embodiment of the present invention, after calculating and generating the net calorific value that characterizes the degree of heat generation of the abnormal node itself, the method further includes: repeatedly performing the estimation of cross-thermal contamination components and the calculation of net calorific values ​​for all key monitoring nodes to generate a list of net calorific values ​​containing all key monitoring nodes.

[0064] It should be noted that, firstly, the system receives the cross-thermal contamination component generated for a specific anomalous node in step S4. Simultaneously, the system retrieves the measured temperature value of that anomalous node again. Next, a core subtraction operation is performed. Specifically, the measured temperature value of the anomalous node is used as the minuend, and its corresponding cross-thermal contamination component is used as the subtrahend; the two are subtracted. The physical meaning of this subtraction operation is that it removes all background heat transferred from nearby heat sources from the total observed temperature; the remaining portion is the heat independently generated by the anomalous node's abnormal operating condition or material degradation. The calculated difference is given a new technical name: net calorific value. This net calorific value is a key indicator for assessing whether a single node truly has a fault. Finally, the system associates this calculated net calorific value with the unique identifier of the anomalous node and repeats this process for all analyzed key monitoring nodes, ultimately compiling the net calorific values ​​of all key monitoring nodes into a list. This list containing the net calorific values ​​of all key monitoring nodes is the output of this step.

[0065] Following the example from the previous step, the system receives the cross-heat contamination component calculated for the critical monitoring node "Node_T_A1_Upper_Terminal," with a value of 6.0℃. Simultaneously, the system obtains the measured temperature of this node as 75℃. Next, the system performs a subtraction operation, subtracting 6.0℃ from 75℃. The result is 69℃. This result of 69℃ is confirmed as the net calorific value of "Node_T_A1_Upper_Terminal." This value is then recorded and added to a net calorific value list along with the net calorific values ​​of other critical monitoring nodes, forming a record such as {"Node_T_A1_Upper_Terminal":69.0℃}.

[0066] S6. System-level thermal imbalance index assessment: The net calorific value of all nodes belonging to the same related subsystem is selected, and the model is compared with the ideal temperature gradient preset in the spatial topology coupling model to quantify and generate a thermal imbalance index to assess the overall health status of the same related subsystem.

[0067] In a specific embodiment of the present invention, the specific steps of quantifying and generating a thermal imbalance index to assess the overall health status of the same related subsystem include: selecting the net calorific value of all nodes belonging to the same related subsystem from the net calorific value list.

[0068] It should be noted that the implementation process first calls the created spatial topology coupling model and filters out the subsystem heat flow topology diagram from it. At the same time, the system will filter out all key monitoring nodes that belong to the same specific related subsystem, such as the same cooling loop, and extract the net heat value corresponding to these nodes from the net heat value list generated in step S5.

[0069] Based on the preset node order in the spatial topology coupling model, the selected net calorific values ​​are arranged to form the actual temperature gradient curve.

[0070] It should be noted that, subsequently, based on the predefined upstream and downstream order of nodes in the subsystem heat flow topology diagram, the real-time net calorific value of these selected nodes is arranged. This forms an ordered temperature data sequence, which, when plotted in a coordinate system, constitutes the actual temperature gradient curve, intuitively reflecting the actual distribution and changes of heat from upstream to downstream within the subsystem at the current moment.

[0071] The actual temperature gradient curve is compared with the ideal temperature gradient to quantify the degree of deviation between the two, thereby generating a thermal imbalance index.

[0072] It's important to note that the system then performs a pattern comparison between the newly generated actual temperature gradient curve and the ideal temperature gradient pre-existing in the subsystem's heat flow topology diagram. This comparison is not a simple numerical subtraction, but rather calculates the root mean square error (RMSE) between the actual and ideal temperature gradient curves. This calculation quantifies the degree of deviation between the actual temperature distribution and the temperature distribution under ideal healthy conditions. This quantified deviation value is defined as a comprehensive index, namely the thermal imbalance index. A higher thermal imbalance index indicates a more severe deviation of the subsystem's overall operating state from design standards, potentially indicating a systemic risk of failure rather than a localized problem at a single node.

[0073] It should also be noted that when calculating the root mean square error between the actual temperature gradient curve and the ideal temperature gradient curve, the temperature values ​​of corresponding nodes in the two curves are first subtracted to obtain the temperature difference at each node. Then, these temperature differences are squared to eliminate the negative sign and emphasize larger deviations. Next, the squared differences of all nodes are summed and divided by the number of nodes to obtain the mean squared difference. Finally, the square root of the mean squared difference is taken to quantify the overall deviation between the actual temperature distribution and the temperature distribution under ideal healthy conditions. This quantified deviation value is the root mean square error, which is used as a thermal imbalance index to assess the overall health of the subsystem.

[0074] S7. Composite Early Warning Decision Generation: Based on the preset alarm threshold and early warning threshold, the net calorific value and thermal imbalance index are judged respectively, and the judgment results are integrated to generate a composite early warning decision.

[0075] In a specific embodiment of the present invention, the specific steps of integrating the judgment results to generate a composite early warning decision include: comparing the net heat value of any key monitoring node with a preset alarm threshold to determine whether there is a real heat fault in a single node.

[0076] The thermal imbalance index of any related subsystem is compared with a preset warning threshold to determine whether there are potential system-level operational risks.

[0077] The results of assessing actual overheating failures at a single node and potential operational risks at the system level are integrated to generate composite early warning decisions.

[0078] It's important to note that the first step in the implementation process is threshold setting. Based on the operating specifications provided by the equipment manufacturer, industry safety standards, and long-term operational experience, two key numerical boundaries are set for the net calorific value and the thermal imbalance index, namely the alarm threshold and the warning threshold. Next, the system enters the decision-making phase. The net calorific value of each node in the generated net calorific value list is compared with the preset alarm threshold. If the net calorific value of any node exceeds this alarm threshold, the system determines that there are one or more definite real calorific failures caused by the node itself. This is the node-level fault determination. Next, the system compares the generated thermal imbalance index for a specific subsystem with the preset warning threshold. If the thermal imbalance index exceeds the warning threshold, the system determines that the overall operating status of the subsystem has deviated from a healthy track, posing a potential system-level operational risk that could lead to larger-scale failures in the future. This is the system-level risk determination. Finally, the system integrates the above two independent determination results. The integration method is to summarize all determined node-level real faults and system-level status deviation information into a unified report or data structure to form a composite warning decision. This decision clearly identifies which specific component has failed and which system is experiencing an overall abnormality, thus providing comprehensive and accurate guidance for operations and maintenance personnel.

[0079] The alarm threshold is a specific value set for the net calorific value, which can be set to 70℃. Once the net calorific value of a node exceeds this value, a high-level fault alarm is triggered. This threshold is set based on statistical analysis of historical temperature rise data before the burnout of similar equipment, and is usually taken as 80% of the critical temperature that leads to irreversible degradation of equipment performance. The warning threshold is a value set for the thermal imbalance index, which can be set to 0.7. When the thermal imbalance index exceeds this value, a suggestive risk warning is triggered, indicating that although the system has not experienced a definitive failure, its operating status is not ideal and requires attention. This threshold is set based on long-term monitoring of the changes in the thermal imbalance index during the evolution of the system from health to failure, selecting an inflection point value that can significantly distinguish between normal fluctuations and abnormal trends. A single node's actual calorific failure refers to a temperature rise event exceeding the safety limit, confirmed to be caused by the node itself after cross-contamination correction. System-level potential operational risks refer to an unstable state in which the cooperative state or thermodynamic balance of the entire functional subsystem has been disrupted, even if no individual component is currently clearly damaged, indicating a possible future failure.

[0080] Reference Figure 2The second aspect of the present invention provides an infrared temperature measurement and early warning system for valve hall equipment based on a lightweight detection network, comprising: a spatial topology coupling model construction module, a real-time spatial reference anchor point identification module, a key monitoring node coordinate deduction module, a cross-thermal pollution component estimation module, a node net calorific value separation calculation module, a system-level thermal imbalance index evaluation module, and a composite early warning decision generation module.

[0081] Both the spatial topology coupling model construction module and the real-time spatial reference anchor point identification module are connected to the key monitoring node coordinate deduction module. The spatial topology coupling model construction module, the real-time spatial reference anchor point identification module, and the key monitoring node coordinate deduction module are all connected to the cross-thermal pollution component estimation module. The cross-thermal pollution component estimation module is connected to the node net calorific value separation calculation module. The node net calorific value separation calculation module is connected to the system-level thermal imbalance index assessment module. Both the node net calorific value separation calculation module and the system-level thermal imbalance index assessment module are connected to the composite early warning decision generation module.

[0082] The spatial topology coupling model construction module generates a spatial topology coupling model, which establishes a digital twin of the physical nodes of the device and the thermal interaction path for the monitoring perspective of the fixed infrared camera.

[0083] The real-time spatial reference anchor point identification module acquires a real-time infrared thermal image and identifies multiple spatial reference anchor points in the real-time infrared thermal image to generate a real-time anchor point coordinate set.

[0084] The key monitoring node coordinate derivation module takes into account the real-time anchor point coordinate set and the spatial topology coupling model, and calculates the precise mapping position of each key monitoring node in the real-time infrared thermal image through geometric transformation to generate a node positioning coordinate map.

[0085] The cross-thermal pollution component estimation module extracts the measured temperature values ​​of each key monitoring node from the real-time infrared thermal image based on the node location coordinate map, and records nodes with abnormal measured temperature values ​​as abnormal nodes. It then calls the spatial topology coupling model to trace its neighboring heat source nodes in order to calculate and generate the cross-thermal pollution component.

[0086] The node net calorific value separation calculation module subtracts the corresponding cross-heat contamination component from the measured temperature value of the abnormal node, and the calculation result is used as the net calorific value characterizing the degree of fault heating of the abnormal node itself.

[0087] The system-level thermal imbalance index assessment module filters out the net calorific value of all nodes belonging to the same related subsystem, and performs pattern comparison based on the preset ideal temperature gradient in the spatial topology coupling model to quantify and generate a thermal imbalance index to assess the overall health status of the same related subsystem.

[0088] The composite early warning decision generation module determines the net calorific value and thermal imbalance index based on preset alarm thresholds and early warning thresholds, and integrates the determination results to generate a composite early warning decision.

[0089] The above content is merely an example and illustration of the concept of the present invention. Those skilled in the art can make various modifications or additions to the specific embodiments described, or use similar methods to replace them, as long as they do not deviate from the concept of the invention or exceed the scope defined by the present invention, and all such modifications and additions should fall within the protection scope of the present invention.

Claims

1. A method for infrared temperature measurement and early warning of valve hall equipment based on a lightweight detection network, characterized in that, include: S1. Spatial topology coupling model construction: Generate a spatial topology coupling model, which establishes a digital twin of the physical nodes of the device and the thermal interaction path for the monitoring perspective of the fixed infrared camera; S2. Real-time spatial reference anchor point identification: acquire real-time infrared thermal images and identify multiple spatial reference anchor points in the real-time infrared thermal images to generate a real-time anchor point coordinate set; S3. Coordinate deduction of key monitoring nodes: Input the real-time anchor point coordinate set and the spatial topology coupling model, and calculate the precise mapping position of each key monitoring node in the real-time infrared thermal image through geometric transformation to generate a node positioning coordinate map. S4. Estimation of cross-thermal pollution components: Based on the node location coordinate map, extract the measured temperature values ​​of each key monitoring node from the real-time infrared thermal image, and record the nodes with abnormal measured temperature values ​​as abnormal nodes. Call the spatial topology coupling model to trace its neighboring heat source nodes in order to calculate and generate cross-thermal pollution components. S5. Calculation of net calorific value of nodes: Subtract the corresponding cross-heat contamination component from the measured temperature value of the abnormal node, and use the result as the net calorific value that characterizes the degree of calorific value of the abnormal node's own fault. S6. System-level thermal imbalance index assessment: The net calorific value of all nodes belonging to the same related subsystem is screened out, and the pattern is compared according to the ideal temperature gradient preset in the spatial topology coupling model to quantify and generate a thermal imbalance index to assess the overall health status of the same related subsystem. S7. Composite Early Warning Decision Generation: Based on the preset alarm threshold and early warning threshold, the net calorific value and thermal imbalance index are judged respectively, and the judgment results are integrated to generate a composite early warning decision.

2. The infrared temperature measurement and early warning method for valve hall equipment based on a lightweight detection network according to claim 1, characterized in that, The specific steps for generating the spatial topological coupling model include: By integrating the 3D design drawings of the valve hall equipment with the installation pose information of the fixed infrared camera, the estimated imaging area of ​​each key monitoring node in the infrared image is calibrated to generate a physical node position mapping library. Based on the physical connections and spatial adjacency between devices, the heat conduction and heat radiation paths between key monitoring nodes are analyzed and quantified to construct a device thermal coupling network diagram. Based on the design parameters of the associated subsystems, the ideal temperature gradient between key monitoring nodes is defined to construct the heat flow topology of the subsystems; The physical node location mapping library, the device thermal coupling network diagram, and the subsystem thermal flow topology diagram are integrated to generate a spatial topology coupling model.

3. The infrared temperature measurement and early warning method for valve hall equipment based on a lightweight detection network according to claim 2, characterized in that, The specific steps for generating the real-time anchor point coordinate set include: The preset lightweight recognition operator is invoked to analyze the real-time infrared thermal image; Using a lightweight recognition operator, equipment components with constant positions and prominent shapes within real-time infrared thermal images are identified as spatial reference anchor points. Extract and output the image coordinates of the spatial reference anchor points to generate a real-time anchor point coordinate set.

4. The infrared temperature measurement and early warning method for valve hall equipment based on a lightweight detection network according to claim 3, characterized in that, The specific steps for calculating the precise mapping position of each key monitoring node in the real-time infrared thermal image through geometric transformation to generate a node positioning coordinate map include: The precise relative positional relationship between each key monitoring node and the spatial reference anchor point is obtained from the spatial topology coupling model. Based on the real-time anchor point coordinate set and the precise relative position relationship, a geometric transformation model that can describe the changes in the current view is calculated. A geometric transformation model is applied to the reference positions of all key monitoring nodes to calculate and generate a node positioning coordinate map containing the precise mapping positions of each key monitoring node.

5. The infrared temperature measurement and early warning method for valve hall equipment based on a lightweight detection network according to claim 2, characterized in that, The specific steps for calculating and generating cross-thermal contamination components include: Call the device thermal coupling network diagram in the spatial topology coupling model to trace all neighboring heat source nodes that have thermal interaction paths with the abnormal node; Obtain the current measured temperature values ​​of each nearby heat source node; By combining the heat transfer coefficients and path parameters recorded in the equipment thermal coupling network diagram, the total heat contribution transferred from all neighboring heat source nodes to the abnormal nodes is calculated to generate cross-heat contamination components.

6. The infrared temperature measurement and early warning method for valve hall equipment based on a lightweight detection network according to claim 5, characterized in that, After calculating and generating the net calorific value that characterizes the degree of heat generation of the abnormal node's own fault, the process also includes: repeatedly performing the estimation of cross-heat contamination components and the calculation of net calorific values ​​for all key monitoring nodes to generate a list of net calorific values ​​for all key monitoring nodes.

7. The infrared temperature measurement and early warning method for valve hall equipment based on a lightweight detection network according to claim 6, characterized in that, The specific steps for quantifying and generating a thermal imbalance index to assess the overall health status of the same related subsystem include: From the list of net calorific values, filter out the net calorific values ​​of all nodes belonging to the same related subsystem; Based on the preset node order in the spatial topological coupling model, the selected net calorific values ​​are arranged to form the actual temperature gradient curve. The actual temperature gradient curve is compared with the ideal temperature gradient to quantify the degree of deviation between the two, thereby generating a thermal imbalance index.

8. The infrared temperature measurement and early warning method for valve hall equipment based on a lightweight detection network according to claim 1, characterized in that, The specific steps for integrating the judgment results to generate a composite early warning decision include: The net heat value of any key monitoring node is compared with a preset alarm threshold to determine whether there is a real overheating fault in a single node. The thermal imbalance index of any related subsystem is compared with a preset warning threshold to determine whether there are potential system-level operational risks. The results of assessing actual overheating failures at a single node and potential operational risks at the system level are integrated to generate composite early warning decisions.

9. An infrared temperature measurement and early warning system for valve hall equipment based on a lightweight detection network, characterized in that, include: The spatial topology coupling model construction module generates a spatial topology coupling model, which establishes a digital twin of the physical nodes of the device and the thermal interaction path for the monitoring perspective of the fixed infrared camera. The real-time spatial reference anchor point identification module acquires real-time infrared thermal images and identifies multiple spatial reference anchor points in the real-time infrared thermal images to generate a real-time anchor point coordinate set. The key monitoring node coordinate derivation module takes as input a real-time anchor point coordinate set and a spatial topology coupling model, and calculates the precise mapping position of each key monitoring node in the real-time infrared thermal image through geometric transformation to generate a node positioning coordinate map. The cross-thermal pollution component estimation module extracts the measured temperature values ​​of each key monitoring node from the real-time infrared thermal image based on the node location coordinate map, and records nodes with abnormal measured temperature values ​​as abnormal nodes. It then calls the spatial topology coupling model to trace its neighboring heat source nodes in order to calculate and generate the cross-thermal pollution component. The node net calorific value separation calculation module subtracts the corresponding cross-heat contamination component from the measured temperature value of the abnormal node, and the calculation result is used as the net calorific value characterizing the degree of fault heating of the abnormal node itself. The system-level thermal imbalance index assessment module filters out the net calorific value of all nodes belonging to the same related subsystem, and performs pattern comparison based on the preset ideal temperature gradient in the spatial topology coupling model to quantitatively generate a thermal imbalance index to assess the overall health status of the same related subsystem. The composite early warning decision generation module determines the net calorific value and thermal imbalance index based on preset alarm thresholds and early warning thresholds, and integrates the determination results to generate a composite early warning decision.

Images