A graded early warning system for thermal runaway of a converter valve and its identification method

Abstract

This application relates to the field of converter valve monitoring technology, and in particular to a graded early warning system and identification method for converter valve thermal runaway. The method acquires temperature monitoring data and operating parameters of key components of the converter valve, calculates the temperature change rate, and compensates for a baseline temperature rise threshold to obtain a dynamic early warning threshold, thereby identifying potential thermal runaway points and triggering a first-level early warning. Subsequently, based on spatial coordinates, a dual-spectrum gimbal camera is controlled to acquire infrared and visible light images, extracts infrared thermal features and visible light fault features, and performs fusion judgment to trigger a second-level verification early warning. After confirming the thermal runaway state, early warning level information is generated and linkage control commands are output, realizing graded identification and early warning of converter valve thermal runaway risk. This invention achieves early identification, accurate location, and graded early warning of converter valve thermal runaway risk, thereby improving the accuracy of thermal runaway detection and operational safety.

Description

Technical Field

[0001] The present application relates to the technical field of converter valve monitoring, and in particular, to a hierarchical early warning system for converter valve thermal runaway and an identification method thereof. Background Art

[0002] In the related art, an ultra-high voltage direct current (UHVDC) transmission system is an important technical means for realizing large-scale long-distance electric energy transmission. As a core device of the direct current transmission system, a converter valve is usually installed in a valve hall of a converter station in the form of a valve tower, and key components such as thyristors, reactors and direct current capacitors are integrated inside the converter valve. Under long-term operation conditions of high voltage and large current, the above components may generate abnormal temperature rise due to increased contact resistance, insulation aging, reduced heat dissipation performance and other reasons. If the abnormal temperature rise is not detected and handled in time, it may further develop into thermal runaway and even cause a fire accident in the valve hall. Therefore, real-time monitoring of temperature changes of key components of converter valves and effective identification and early warning in the early stage of thermal runaway have become important technical means to ensure the safe operation of UHV converter stations.

[0003] However, the existing converter valve temperature monitoring technologies still have certain limitations. In the actual operation process, factors such as the load level, ambient temperature and operating conditions of the converter valve have obvious dynamic change characteristics. The single fixed threshold method is difficult to adapt to complex working conditions, and is prone to false alarms or missed alarms, which makes it difficult to timely and accurately identify early abnormalities of thermal runaway. Meanwhile, traditional monitoring methods mostly rely on single temperature monitoring data for judgment, and lack a visual confirmation mechanism for abnormal parts and a cross-validation mechanism for multi-source information. When local components have problems such as poor contact, structural damage or local overheating, it is difficult to accurately identify the fault type only by relying on temperature information, which affects the reliability of fault judgment. In addition, after detecting temperature abnormality, the prior art usually only can issue a simple alarm, and lacks the capability of accurate positioning of the abnormal position and a hierarchical early warning and linkage control mechanism, so it is difficult to realize early identification and hierarchical response to potential thermal runaway risks, thereby affecting the operation safety and operation and maintenance efficiency of the converter valve.

[0004] In conclusion, the technical problems existing in the related art need to be improved. Summary of the Invention

[0005] A main purpose of the embodiments of the present application is to provide a hierarchical early warning system for converter valve thermal runaway and an identification method thereof, so as to realize early identification, accurate positioning and hierarchical early warning of converter valve thermal runaway risks, thereby improving the accuracy of thermal runaway detection and operation safety.

[0006] To achieve the above purpose, in one aspect of the embodiments of the present application, a hierarchical identification method for converter valve thermal runaway is provided, and the method includes the following steps:[[]END]] Acquire temperature monitoring data of key components of the converter valve and collect operating parameters of the converter valve; The temperature change rate of each monitoring point is calculated based on the temperature monitoring data, and the benchmark temperature rise threshold is compensated based on the operating condition parameters to obtain the dynamic early warning threshold; the monitoring points are respectively deployed on the key components of the converter valve. The temperature change rate is compared with the dynamic early warning threshold. When the temperature change rate exceeds the dynamic early warning threshold, the corresponding monitoring point is identified as a potential thermal runaway point and a first-level early warning is triggered. At the same time, the spatial coordinate information of the potential thermal runaway point is obtained. Based on the spatial coordinate information, the camera control command of the dual-spectrum gimbal camera is calculated using a spatial coordinate mapping algorithm, and the dual-spectrum gimbal camera is controlled to acquire infrared and visible light images of the corresponding area according to the camera control command. The infrared and visible light images are preprocessed to extract the corresponding infrared thermal features and visible light fault features; The infrared thermal features and the visible light fault features are fused together for determination, and a secondary review and warning are triggered based on the determination result; When thermal runaway is confirmed, the system generates a corresponding warning level based on the determination result and outputs the corresponding linkage control command.

[0007] In some embodiments, the operating condition parameters include load current, ambient temperature, equipment service life, and cooling system status.

[0008] In some embodiments, the dynamic early warning threshold is calculated using the following formula: ; in, Indicates the real-time dynamic warning threshold; Indicates the threshold for the rate of change of the baseline temperature rise; Indicates the total compensation coefficient; Indicates the load current compensation factor; Indicates the ambient temperature compensation factor; Indicates the equipment aging compensation factor; This represents the cooling system compensation factor.

[0009] In some embodiments, the upper and lower limits of the dynamic warning threshold include: the dynamic warning threshold is no more than 3 times the reference temperature rise rate threshold, and the dynamic warning threshold is lowered by no less than 50% of the reference temperature rise rate threshold.

[0010] In some embodiments, calculating the camera control commands for the dual-spectrum gimbal camera using a spatial coordinate mapping algorithm based on the spatial coordinate information includes: A global three-dimensional world coordinate system for the valve hall is established with the ground reference point of the valve hall as the origin, and the spatial coordinate information and the installation coordinates of each dual-spectrum gimbal camera are pre-calibrated. Based on the global three-dimensional world coordinate system, calculate the spatial vector and straight-line distance of the potential thermal runaway point relative to the dual-spectrum gimbal camera; The horizontal azimuth and pitch angles of the dual-spectrum gimbal camera pointing towards the potential thermal runaway point are calculated based on the spatial vector to obtain the steering control parameters of the dual-spectrum gimbal camera. Based on the preset mapping relationship between the straight-line distance and the step value, the focus control parameters of the dual-spectrum gimbal camera are determined. Based on the steering control parameters and the focus control parameters, control commands for the dual-spectrum gimbal camera are generated.

[0011] In some embodiments, the formulas for calculating the horizontal azimuth and elevation angles are as follows: ; ; in, This represents the horizontal azimuth angle of the dual-spectral gimbal camera corresponding to the i-th potential thermal runaway measurement point; This represents the pitch angle of the dual-spectral gimbal camera corresponding to the i-th potential thermal runaway measurement point; This represents the coordinate difference in the X direction between the potential thermal runaway measurement point and the dual-spectrum gimbal camera; This represents the coordinate difference in the Y direction between the potential thermal runaway measurement point and the dual-spectrum gimbal camera; This represents the coordinate difference in the Z direction between the potential thermal runaway measurement point and the dual-spectral gimbal camera.

[0012] In some embodiments, the step of preprocessing the infrared image and the visible light image and extracting the corresponding infrared thermal features and visible light fault features includes: The infrared image and the visible light image are denoised and electromagnetic interference filtered to obtain a denoised dual-spectral image. Lens distortion correction is performed on the denoised dual-spectral image, and pixel-level registration is performed between the infrared image and the visible light image to obtain the registered dual-spectral image. Based on the spatial coordinate information of the potential thermal runaway point, the target region image is extracted from the bispectral image, and the target region image is subjected to contrast enhancement and normalization processing. Infrared thermal features are extracted from the target region image, including hot spot temperature features, temperature change rate features, and hot spot diffusion features. Visible light fault features are extracted from the target area image, including arc flashing features, carbonization features of insulating materials, and deformation features of components.

[0013] In some embodiments, the step of fusing and determining the infrared thermal features and the visible light fault features, and triggering a secondary review and warning based on the determination result, includes: The infrared thermal features and the visible light fault features are processed into feature vectors and then normalized to obtain multidimensional feature data. The multidimensional feature data is weighted and fused based on preset feature weights to obtain the thermal runaway probability index. Based on the thermal runaway probability index, a set of evidence for judgment is constructed, and the DS evidence theory is used for comprehensive reasoning and calculation to obtain the thermal runaway judgment result. The thermal runaway determination result is compared with a preset determination threshold. When the thermal runaway determination result is greater than the determination threshold, a secondary review warning is triggered.

[0014] In some embodiments, the formula for calculating the thermal runaway probability index is as follows: ; in, Indicates the probability index of thermal runaway; Indicates the overall confidence level of infrared thermal anomalies; Indicates the confidence level of the arc flicker feature; Indicates the confidence level of the carbonization characteristics of insulating materials; Indicates the confidence level of deformation characteristics; Indicates the weighting coefficient of infrared temperature anomaly characteristics; Indicates the weighting coefficient of the arc flicker feature; Indicates the weighting coefficient of carbonization characteristics; This represents the weighting coefficient of structural deformation characteristics.

[0015] To achieve the above objectives, another aspect of this application proposes a graded early warning system for thermal runaway of a converter valve, the system comprising: The contact-type fiber optic monitoring layer is used to collect temperature monitoring data of key components of the converter valve and transmit it to the monitoring backend. The non-contact multispectral verification layer is used to acquire infrared and visible light images of the area where potential thermal runaway points are located after receiving control commands sent from the monitoring backend. The monitoring backend receives the temperature monitoring data, as well as the infrared and visible light images, and performs the following processing: The temperature change rate at each monitoring point is calculated based on the temperature monitoring data, and the reference temperature rise threshold is compensated based on the operating parameters of the converter valve to obtain the dynamic early warning threshold. The temperature change rate is compared with the dynamic early warning threshold. When the temperature change rate exceeds the dynamic early warning threshold, the corresponding monitoring point is determined as a potential thermal runaway point. Based on the spatial coordinate information of the potential thermal runaway point, the non-contact multispectral core layer is controlled to collect infrared and visible light images of the corresponding area. Image processing is performed on the infrared image and the visible light image to extract infrared thermal features and visible light fault features. The infrared thermal features and the visible light fault features are then fused and judged to obtain the thermal runaway judgment result. When thermal runaway is confirmed, a corresponding early warning level is generated and a linkage control command is output.

[0016] To achieve the above objectives, another aspect of this application provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the above-described method.

[0017] To achieve the above objectives, another aspect of the embodiments of this application proposes a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method.

[0018] To achieve the above objectives, another aspect of this application provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0019] The embodiments of this application include at least the following beneficial effects: This application provides a graded early warning system for thermal runaway of a converter valve and its identification method. This scheme obtains temperature monitoring data of key components of the converter valve and combines it with operating condition parameters to perform dynamic threshold compensation calculation on the temperature change rate. This achieves adaptive identification of temperature rise anomalies under different operating conditions, thereby avoiding the misjudgment or missed judgment problem caused by fixed threshold judgment methods under load or environmental changes, and improving the accuracy and reliability of early thermal runaway identification. After detecting a potential thermal runaway point, a spatial coordinate mapping algorithm is further used to determine the target area, and a dual-spectrum gimbal camera is controlled to acquire infrared and visible light images of the corresponding area, realizing accurate positioning of temperature anomaly points and acquisition of multi-source information. By fusing and analyzing infrared thermal features and visible light fault features, abnormal states caused by structural damage, poor contact, or local overheating can be effectively distinguished, reducing the probability of false alarms caused by a single temperature monitoring method. Meanwhile, by constructing a graded early warning mechanism, image verification is performed based on the initial abnormal temperature rise identification, and the corresponding early warning level and linkage control command are output after confirming the risk of thermal runaway. This enables early identification, accurate positioning and graded response of the thermal runaway risk of the converter valve, thereby improving the safety of converter valve operation and the level of intelligent operation and maintenance. Attached Figure Description

[0020] Figure 1 This is a schematic flowchart of a graded identification method for thermal runaway of a converter valve provided in an embodiment of this application; Figure 2 This is a schematic diagram of spatial coordinates provided in an embodiment of this application; Figure 3 This is a top view of the spatial coordinate position provided in the embodiments of this application; Figure 4 This is a side view of the spatial coordinate position provided in the embodiments of this application; Figure 5 This is a schematic diagram of a graded early warning system for thermal runaway of a converter valve provided in an embodiment of this application; Figure descriptions: 100, Converter valve tower; 101, Thyristor clamping contact surface; 102, Reactor surface; 103, DC capacitor housing; 200, Contact fiber optic monitoring layer; 201, Fluorescent fiber optic sensor; 202, Fiber optic signal aggregation end; 300, Non-contact multispectral verification layer; 301, Dual-spectrum PTZ camera; 302, PTZ drive motor; 400, Monitoring backend; 500, Station control system; 501, Station control system execution module. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of this application and are not intended to limit it. In the following description, when referring to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with those of this application; they are merely examples of apparatuses and methods consistent with some aspects of the embodiments of this application as detailed in the appended claims.

[0022] It is understood that the terms “first,” “second,” etc., used in this application may be used herein to describe various concepts, but unless otherwise stated, these concepts are not limited by these terms. These terms are only used to distinguish one concept from another. For example, without departing from the scope of the embodiments of this application, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the words “if,” “when,” or “in response to a determination” as used herein may be interpreted as “when…” or “when…” or “in response to a determination.”

[0023] As used in this application, the terms "at least one", "multiple", "each", "any", etc., "at least one" includes one, two or more, "multiple" includes two or more, "each" refers to each of the corresponding multiples, and "any" refers to any one of the multiples.

[0024] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0025] Before providing a detailed description of the embodiments of this application, some of the nouns and terms used in the embodiments of this application will be explained first. The nouns and terms used in the embodiments of this application shall be interpreted as follows: Dempster-Shafer Evidence Theory (DS) is a mathematical theory and method for fusion and reasoning of uncertain information. It achieves a comprehensive judgment on the probability of a target event by assigning and combining confidence levels to evidence from different information sources. In this application, DS evidence theory is used to fuse and reason about the evidence set formed by infrared thermal features and visible light fault features to obtain a comprehensive judgment result on whether the converter valve has experienced thermal runaway. Dempster's Rule of Combination is a computational rule in Dempster's evidence theory used to fuse information from multiple independent evidence sources. It normalizes and synthesizes the confidence levels of different pieces of evidence to obtain a comprehensive confidence distribution of the target event. In this application, Dempster's Rule of Combination is used to fuse and reason about evidence corresponding to infrared thermal features and visible light fault features, thereby obtaining a comprehensive judgment result on whether the converter valve has experienced thermal runaway.

[0026] This application provides a graded early warning system and identification method for thermal runaway of a converter valve. This system acquires temperature monitoring data of key components of the converter valve and combines it with operating parameters to perform dynamic threshold compensation calculations on the temperature change rate. This enables adaptive identification of abnormal temperature rise under different operating conditions, avoiding misjudgments or missed judgments caused by fixed threshold judgment methods under load or environmental changes, thus improving the accuracy and reliability of early thermal runaway identification. After detecting a potential thermal runaway point, a spatial coordinate mapping algorithm is further used to determine the target area, and a dual-spectrum gimbal camera is controlled to acquire infrared and visible light images of the corresponding area, achieving precise location of temperature anomalies and acquisition of multi-source information. By fusing and analyzing infrared thermal features and visible light fault features, abnormal states caused by structural damage, poor contact, or local overheating can be effectively distinguished, reducing the probability of false alarms caused by a single temperature monitoring method. Meanwhile, by constructing a graded early warning mechanism, image verification is performed based on the initial abnormal temperature rise identification, and the corresponding early warning level and linkage control command are output after confirming the risk of thermal runaway. This enables early identification, accurate positioning and graded response of the thermal runaway risk of the converter valve, thereby improving the safety of converter valve operation and the level of intelligent operation and maintenance.

[0027] This application provides a method for graded identification of thermal runaway in a converter valve, relating to the field of converter valve monitoring technology. This method can be applied to a terminal, a server, or software running on either a terminal or a server. In some embodiments, the terminal can be a smartphone, tablet, laptop, desktop computer, smart speaker, smartwatch, or vehicle terminal, but is not limited thereto. The server can be configured as an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms. The server can also be a node server in a blockchain network. The software can be an application implementing the method for graded identification of thermal runaway in a converter valve, but is not limited to the above forms.

[0028] This application can be used in a wide variety of general-purpose or special-purpose computer system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics devices, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices. This application can be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform specific tasks or implement specific abstract data types. This application can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0029] Figure 1 This is an optional flowchart of a graded identification method for thermal runaway of a converter valve provided in an embodiment of this application. Figure 1 The method may include, but is not limited to, steps S1 to S7: S1: Obtain temperature monitoring data of key components of the converter valve and collect operating parameters of the converter valve; these operating parameters include load current, ambient temperature, equipment service life, and cooling system status.

[0030] In this embodiment, contact-type fluorescent fiber optic temperature sensors are first deployed at key heat-generating locations within the converter valve tower. These key heat-generating locations include the thyristor clamping contact surfaces of each layer, the reactor surfaces, and the DC capacitor shell surfaces. The thyristor clamping contact surfaces are the most prone to abnormal localized temperature rises due to increased contact resistance during converter valve operation. The reactor surfaces are prone to continuous heating due to prolonged load operation, while the DC capacitor shell surfaces reflect internal capacitor losses and heat accumulation. Each fluorescent fiber optic temperature sensor is fixed to the surface of its corresponding key component using a surface-mount method and connected to a fiber optic acquisition card in the monitoring backend via fiber optic lines. The system continuously acquires temperature values ​​at each measurement point at a sampling frequency of 1 time per second, automatically adding the measurement point number, acquisition timestamp, and corresponding equipment level information to form a structured temperature monitoring data sequence.

[0031] Simultaneously, the monitoring backend collects the operating parameters of the converter valve. Specifically, the load current parameter is provided by the converter valve body current measuring device or the station control system, reflecting the current load level of the converter valve in real time; the ambient temperature parameter is collected by the ambient temperature and humidity monitoring unit in the valve hall, used to characterize the impact of the external environment on the heat dissipation conditions of the components; the equipment service life parameter is read from the equipment ledger database, used to reflect the degree of equipment aging; and the cooling system status parameter is determined based on the operating information of the water-cooled or air-cooled system, such as rated operation, single redundant loop activation, decreased cooling capacity, or minor fault status. After receiving the above data, the monitoring backend associates and stores the temperature monitoring data with the operating parameters according to a unified time base, providing complete input for subsequent dynamic threshold compensation calculations.

[0032] S2: Calculate the temperature change rate of each monitoring point based on the temperature monitoring data, and perform compensation calculation on the reference temperature rise threshold based on the operating condition parameters to obtain the dynamic early warning threshold; the monitoring points are respectively deployed on the key components of the converter valve; the dynamic early warning threshold shall not exceed 3 times the reference temperature rise change rate threshold, and the dynamic early warning threshold shall be adjusted down to no less than 50% of the reference temperature rise change rate threshold.

[0033] In this embodiment, the monitoring backend calculates the temperature change rate using a sliding time window method for the continuously collected temperature data at each monitoring point. After obtaining the temperature change rate, dynamic compensation calculation is performed based on a preset baseline temperature rise rate threshold. This threshold corresponds to the empirical critical value of the converter valve under rated operating conditions, standard ambient temperature, relatively new equipment conditions, and normal operating conditions of the cooling system. Subsequently, four types of factors are introduced: load current compensation, ambient temperature compensation, equipment aging compensation, and cooling system condition compensation, to correct the baseline threshold in real time. When the load current is high, larger temperature rise fluctuations are allowed during normal operation, and the threshold is adjusted accordingly; when the ambient temperature rises, the heat dissipation conditions of the components deteriorate, and the threshold is moderately increased; as the equipment's service life increases, the baseline temperature rise increases, and the threshold is also compensated accordingly; when the cooling system is derated, the threshold is further increased to avoid false alarms caused by normal cooling fluctuations. After multi-factor comprehensive compensation, the dynamic early warning threshold for the corresponding measuring point at the current moment is obtained.

[0034] To prevent excessive threshold deviations under extreme operating conditions from causing missed or false alarms, this embodiment also sets upper and lower limits for the dynamic early warning threshold. Specifically, its maximum value does not exceed three times the reference temperature rise rate threshold, and its minimum value is not less than 50% of the reference temperature rise rate threshold. For example, under rated operating conditions, if the theoretical threshold after comprehensive compensation is 5.2℃ / min, the system still limits it to within 6℃ / min; under low-load and low-temperature operating conditions, if the calculated result is too low, it will not be lower than 1℃ / min. Furthermore, during transient operating conditions such as converter valve startup, shutdown, or line switching, the current threshold is locked for 10 seconds to avoid drastic threshold changes caused by transient disturbances. Finally, the monitoring backend synchronously updates the temperature change rate and dynamic early warning threshold of each measuring point at a frequency of once per second, preparing for the first-level early warning determination.

[0035] Specifically, this embodiment uses a reference temperature rise rate threshold as a basis and introduces a multi-dimensional operating condition compensation factor to achieve real-time dynamic correction of the threshold. While ensuring the sensitivity of early detection of thermal runaway, it reduces the false alarm rate caused by normal operating condition fluctuations to 0, thus solving the adaptability defects of fixed thresholds.

[0036] The formula for calculating the dynamic early warning threshold is as follows: ; in, Indicates the real-time dynamic warning threshold; Indicates the threshold for the rate of change of the baseline temperature rise; Indicates the total compensation coefficient; Indicates the load current compensation factor; Indicates the ambient temperature compensation factor; Indicates the equipment aging compensation factor; This represents the cooling system compensation factor.

[0037] Specifically, It is a load current compensation factor used to offset the temperature rise caused by normal load fluctuations in the converter valve, and is the core compensation factor.

[0038] ; in, This refers to the real-time load current of the converter valve. This is the rated load current of the converter valve; This is the load compensation factor, with a value ranging from 0.5 to 0.8. A value of 0.5 is recommended for ±800kV converter valves, and 0.65 is recommended for ±1100kV and above converter valves. When the converter valve is operating at full load, =1.5 ( When the value is 0.5, the dynamic threshold is increased by 50% to avoid false alarms triggered by normal temperature rise under full load; when the converter valve is running under low load / no load, Approaching 1, the threshold maintains high sensitivity, capturing minute abnormal temperature rises and avoiding missed detections.

[0039] This is an ambient temperature compensation factor used to correct the impact of changes in the ambient temperature of the valve hall on the temperature rise of components.

[0040] ; in, The real-time ambient temperature of the valve hall is collected by the existing environmental sensors in the valve hall, and the unit is °C. The standard ambient temperature is 25℃. The temperature compensation coefficient is set to 0.012 / ℃, meaning that for every 10℃ deviation of the ambient temperature from the standard value, the threshold is corrected by 12%. In the high-temperature environment of the valve hall in summer, the basic temperature of the components rises, and the normal temperature rise fluctuates more, so the threshold is appropriately raised; in the low-temperature environment of winter, the threshold is appropriately lowered to ensure the ability to detect abnormal temperature rises under low-temperature conditions.

[0041] Equipment aging compensation factor It is used to correct the deviation in basic temperature rise caused by insulation aging and increased contact resistance of components after long-term operation of the converter valve.

[0042] ; in, The term represents the service life of the converter valve, in years. The aging compensation coefficient is set at 0.03 / year, meaning that for each additional year of operation, the threshold increases by 3%, with a maximum increase of 30% (for units in operation for 10 years or more). For aging valve towers in existing converter stations, this compensation factor raises the baseline of normal temperature rise caused by equipment aging, preventing false alarms triggered by normal temperature rises in aging equipment, while maintaining the detection sensitivity for abnormal temperature rises. Cooling system compensation factor It is used to correct the impact of changes in the operating status of the converter valve water-cooled / air-cooled system on the temperature rise.

[0043] When the cooling system is in rated operating condition When a single redundant loop in the cooling system is activated and the cooling power decreases, When the cooling system experiences a minor malfunction and the cooling power decreases by ≥30%, . It can offset the component temperature rise caused by normal redundancy switching of the cooling system and power fluctuations, and avoid false alarms triggered by non-faulty temperature rises.

[0044] Monitoring backend for dynamic thresholds The calculation update frequency is synchronized with the fiber optic temperature measurement data acquisition frequency, i.e., 1 time / second, ensuring that the threshold is perfectly matched with the real-time operating conditions without any delay or deviation. To avoid detection failure caused by excessive threshold offset under extreme operating conditions, hard constraints are set for the upper and lower limits of the threshold: the upper limit of the threshold is... =3× This means the dynamic threshold can be raised no more than three times the baseline threshold to prevent false alarms due to excessively high thresholds under high loads; the lower limit of the threshold is... =0.5× This means that the dynamic threshold is lowered to no less than 50% of the baseline threshold to prevent frequent false alarms caused by excessively low thresholds under low load. When the converter valve experiences transient conditions such as startup, shutdown, or line switching, the system automatically locks the current threshold for 10 seconds. After the transient process ends, it resumes real-time updates to avoid false triggering caused by sudden changes in current and temperature during condition switching.

[0045] S3: Compare the temperature change rate with the dynamic early warning threshold. When the temperature change rate exceeds the dynamic early warning threshold, determine the corresponding monitoring point as a potential thermal runaway point and trigger a first-level early warning. At the same time, obtain the spatial coordinate information of the potential thermal runaway point. In this embodiment, the monitoring backend performs a comparison calculation between the temperature change rate and the dynamic early warning threshold for each monitoring point. When the real-time temperature change rate of a certain measuring point is greater than the corresponding dynamic early warning threshold, it is considered that the measuring point has deviated from the normal temperature rise fluctuation range and has the characteristics of a precursor to thermal runaway. Therefore, the measuring point is identified as a potential thermal runaway point, and a first-level early warning is immediately triggered. The purpose of the first-level early warning is not to directly determine that the equipment has suffered a serious failure, but to identify potentially risky local temperature rise anomalies as early as possible, thus gaining a time window for subsequent spatial positioning and visual verification.

[0046] While triggering a Level 1 warning, it is also necessary to obtain the spatial coordinates of the potential thermal runaway point. To this end, a global 3D world coordinate system for the valve hall was pre-established during the deployment phase, and a unique 3D coordinate system was assigned to each fluorescent fiber optic measuring point. Therefore, when a measuring point triggers a Level 1 warning, the monitoring backend can directly retrieve its corresponding 3D spatial coordinates based on the measuring point number. These coordinates not only represent the actual physical location of the anomaly in the valve hall but also serve as the basic input for subsequent calculations of the dual-spectrum gimbal camera's turning, pitch, and focus. In this way, "temperature rise anomalies" can be quickly transformed into "locatable spatial targets," thus achieving a smooth transition from contact temperature measurement to non-contact visual verification.

[0047] S4: Calculate the camera control commands for the dual-spectrum gimbal camera based on the spatial coordinate information using a spatial coordinate mapping algorithm, and control the dual-spectrum gimbal camera to acquire infrared and visible light images of the corresponding area according to the camera control commands. The camera control commands for the dual-spectrum gimbal camera are calculated using a spatial coordinate mapping algorithm based on spatial coordinate information, including: A global three-dimensional world coordinate system for the valve hall is established with the ground reference point of the valve hall as the origin, and the spatial coordinate information and the installation coordinates of each dual-spectrum gimbal camera are pre-calibrated. Based on the global three-dimensional world coordinate system, calculate the spatial vector and straight-line distance of the potential thermal runaway point relative to the dual-spectrum gimbal camera; The horizontal azimuth and pitch angles of the dual-spectrum gimbal camera pointing towards the potential thermal runaway point are calculated based on the spatial vectors to obtain the steering control parameters of the dual-spectrum gimbal camera. Based on the mapping relationship between the preset focus distance and the step value, the focus control parameters of the dual-spectrum gimbal camera are determined. Dual-spectrum gimbal camera control commands are generated based on steering control parameters and focus control parameters.

[0048] In this embodiment, a global three-dimensional world coordinate system for the valve hall is established using a preset reference point on the valve hall floor as the origin. During the deployment phase, the installation positions of each fluorescent fiber temperature measuring point and each dual-spectrum gimbal camera are uniformly calibrated. The spatial coordinate information of each measuring point and camera installation position is pre-stored in the monitoring backend database through on-site measurement or 3D modeling. When a measuring point is identified as a potential thermal runaway point in a previous step, the monitoring backend can directly retrieve the spatial coordinates corresponding to that measuring point and combine them with the installation coordinates of each dual-spectrum gimbal camera to form a spatial positional relationship under a unified coordinate system, thereby providing an accurate spatial basis for subsequent camera turning and focusing control.

[0049] After obtaining the spatial coordinates of the potential thermal runaway point, the monitoring backend calculates the spatial vector and straight-line distance of the potential thermal runaway point relative to each dual-spectrum gimbal camera based on the global three-dimensional world coordinate system, and filters multiple candidate cameras based on the calculation results. Preferably, the target dual-spectrum gimbal camera for performing the verification task is determined according to the principles of unobstructed view priority, closer distance priority, and better viewing angle priority. Subsequently, based on the spatial vector relationship between the potential thermal runaway point and the selected camera, the horizontal azimuth and pitch angles required for the camera to point at the potential thermal runaway point are calculated and used as the turning control parameters of the dual-spectrum gimbal camera, so that the camera lens can accurately point to the target abnormal area.

[0050] After determining the camera turning control parameters, the monitoring backend further matches the pre-established mapping relationship between focusing distance and lens step value based on the straight-line distance between the potential thermal runaway point and the target dual-spectrum gimbal camera, thereby determining the corresponding focusing control parameters. Subsequently, the turning control parameters and focusing control parameters are combined to generate dual-spectrum gimbal camera control commands, which are sent to the target dual-spectrum gimbal camera via the communication link. Upon receiving the control commands, the camera completes gimbal turning and lens focusing, centered the area where the potential thermal runaway point is located in the image, and simultaneously acquires infrared and visible light images of that area, providing a data foundation for subsequent image feature extraction and thermal runaway verification.

[0051] refer to Figure 2 As shown, Figure 2 This is a schematic diagram showing the spatial relationship between the potential thermal runaway measurement point and the dual-spectral gimbal camera in the global three-dimensional world coordinate system of the valve hall. Figure 2 A global three-dimensional world coordinate system is established using the valve hall ground reference point O as the origin, where the x and y axes represent the plane direction of the valve hall, and the z axis represents the height direction. Potential thermal runaway measurement points. Spatial coordinates are represented as This measuring point corresponds to the temperature monitoring point deployed on the key component of the converter valve; the location of the dual-spectrum gimbal camera is denoted as... Its spatial coordinates are By describing the positions of the measuring points and the camera in the same three-dimensional coordinate system, the spatial geometric relationship between the two can be clarified, providing a foundation for subsequent camera attitude control and image acquisition.

[0052] like Figure 2 As shown, from the camera position Pointing to the measuring point The line connecting the camera and the potential thermal runaway point represents the spatial orientation relationship between the camera and the potential thermal runaway point. This orientation vector is used to determine the pointing angle of the dual-spectrum gimbal camera. Figure 2 The marked ( The vertical height difference () represents the difference in position between the camera and the measuring point, reflecting their positional differences in the height direction. By analyzing the horizontal displacement and vertical height difference between the camera position and the measuring point position, the required pitch adjustment trend when the camera points to the measuring point can be determined, thus enabling the camera lens to be aimed at the target area.

[0053] In the implementation of this invention, when a certain measuring point is identified as a potential thermal runaway point, the monitoring backend uses the coordinates of that measuring point... With camera coordinates The spatial relationship between the two is calculated, along with their spatial vector and straight-line distance. Based on this, the turning and focusing control parameters of the dual-spectrum gimbal camera are determined, enabling the camera to automatically turn and align with the area where the potential thermal runaway point is located. Through this spatial coordinate mapping relationship, rapid positioning and linkage between the temperature monitoring point and the visual inspection equipment can be achieved, thus providing an accurate target positioning basis for the acquisition of infrared and visible light images.

[0054] like Figure 3 As shown, Figure 3 This is a top-down view of the valve hall's global 3D world coordinate system on a horizontal plane, used to illustrate the horizontal azimuth angle when the dual-spectrum gimbal camera is pointed at a potential thermal runaway measurement point. The method for determining the position is as follows: The X-axis represents the length of the valve hall, and the Y-axis represents the width of the valve hall. In the figure, the blue point is the projected position P′c of the dual-spectrum gimbal camera on the horizontal plane, and the red point is the projected position P′i of the potential thermal runaway measurement point on the horizontal plane. The line connecting the two points represents the horizontal distance L between the camera and the measurement point.

[0055] Figure 3 Using the X-axis direction at the camera's position as the horizontal reference direction, when the dual-spectrum gimbal camera needs to point towards a potential thermal runaway measurement point, it needs to be rotated by a certain angle from the X-axis reference direction so that the camera's line of sight points to the measurement point position. This rotation angle is the azimuth angle shown in the figure. .

[0056] In the implementation of this invention, the monitoring backend determines the positional relationship between the camera projection point P′c and the measurement point projection point P′i on the horizontal plane based on the relationship between the camera installation coordinates and the spatial coordinates of the measurement point, and calculates the required azimuth angle accordingly. This parameter is used as the horizontal turning control parameter for the gimbal camera, thereby controlling the dual-spectrum gimbal camera to automatically rotate to the target area for image acquisition.

[0057] like Figure 4 As shown, Figure 4 This is a side view diagram used to illustrate how the pitch angle φ is determined when a dual-spectrum gimbal camera is pointed at a potential thermal runaway measurement point. Figure 4 The vertical axis is the height of the valve hall space, and the horizontal projection distance L is the horizontal axis. Blue dots represent the positions of the dual-spectrum gimbal camera. Red dots indicate potential thermal runaway measurement points. The line connecting the camera position and the measurement point position indicates the line of sight from the camera lens towards the target measurement point.

[0058] Figure 4 There is a height difference between the installation height of the camera and the height of the measuring point. This height difference reflects the vertical positional relationship between the camera and the target measurement point. The dashed line in the figure represents the horizontal baseline at the camera's height, and the angle formed between the line of sight from the camera to the measurement point and this horizontal baseline is the pitch angle. When the measuring point is lower than the camera height, the gimbal needs to look downwards at a certain angle; when the measuring point is higher than the camera, it needs to be raised upwards at a certain angle.

[0059] In the implementation of this invention, the monitoring backend determines the horizontal distance L and the height difference between the dual-spectrum gimbal camera and the potential thermal runaway measurement point based on their installation coordinates. Based on this, the required pitch angle for the camera to point at the target measurement point is determined. This parameter is used as the pitch control parameter for the dual-spectrum gimbal camera, thereby controlling the camera to adjust its attitude in the vertical direction so that the lens can be accurately aimed at the area where the potential thermal runaway measurement point is located, so as to achieve accurate acquisition of infrared and visible light images.

[0060] Specifically, in this embodiment, a global three-dimensional world coordinate system for the valve hall is established with the reference corner point on the floor as the origin, and the unique three-dimensional coordinates of all fluorescent fiber optic measurement points are pre-entered. and the installation coordinates of each dual-spectrum gimbal camera. Before commissioning, camera calibration is completed to obtain the camera's intrinsic parameters (focal length, principal point coordinates) and extrinsic parameters (rotation matrix R, translation vector T), establish the conversion relationship between world coordinates and camera image, and eliminate positioning errors caused by lens distortion and installation deviation.

[0061] After triggering the Level 1 warning, the monitoring backend calculates the three main control parameters of the target camera in real time using a spatial analytical geometry algorithm based on the coordinates of the suspected thermal runaway measurement point: Calculate the spatial vector of the potential thermal runaway point relative to the camera and the linear focusing distance: , , ; ; Specifically, the formulas for calculating the horizontal azimuth and elevation angles are as follows: ; ; in, It represents the horizontal azimuth angle of the dual-spectrum gimbal camera corresponding to the i-th potential thermal runaway measurement point (the angle between the camera's horizontal zero position and the measurement point direction, with a value of 0°~360°). The pitch angle of the dual-spectrum gimbal camera corresponding to the i-th potential thermal runaway measurement point is (the angle between the camera's horizontal zero position and the direction of the measurement point, with upward being positive, and a value of -90° to 90°). This represents the coordinate difference in the X direction between the potential thermal runaway measurement point and the dual-spectrum gimbal camera; This represents the coordinate difference in the Y direction between the potential thermal runaway measurement point and the dual-spectrum gimbal camera; This represents the coordinate difference in the Z direction between the potential thermal runaway measurement point and the dual-spectral gimbal camera.

[0062] Focusing parameters: based on straight-line distance Calculate the autofocus parameters to ensure clear focus on the measurement point.

[0063] Multiple cameras in the valve hall were selected for optimal verification based on three priority criteria: 1. Prioritize cameras that are not structurally obstructed by the measurement point; 2. Prioritize selecting the camera closest to the measurement point; 3. Prioritize cameras whose viewing angle is perpendicular to the surface of the measuring component.

[0064] Meanwhile, it incorporates two core error compensation mechanisms: installation attitude and gimbal mechanical backlash, ensuring stable positioning accuracy during long-term operation.

[0065] S5: Perform image preprocessing on infrared and visible light images and extract the corresponding infrared thermal features and visible light fault features; This includes preprocessing infrared and visible light images and extracting corresponding infrared thermal features and visible light fault features, including: Denoising and electromagnetic interference filtering are performed on infrared and visible light images to obtain denoised dual-spectral images; Lens distortion correction is performed on the denoised dual-spectral image, and pixel-level registration is performed between the infrared image and the visible light image to obtain the registered dual-spectral image. Based on the spatial coordinate information of potential thermal runaway points, the target region image is extracted from the bispectral image, and the target region image is subjected to contrast enhancement and normalization processing. Infrared thermal features are extracted from the target region image, including hot spot temperature features, temperature change rate features, and hot spot diffusion features. Visible light fault features are extracted from the target area image, including arc flashing features, carbonization features of insulating materials, and deformation features of components.

[0066] In this embodiment, after acquiring infrared and visible light images of the area where the potential thermal runaway point is located, the monitoring backend first preprocesses the dual-spectral images. Due to electromagnetic interference and environmental noise inside the valve hall, it is preferable to perform denoising processing and electromagnetic interference filtering on the infrared and visible light images to obtain denoised dual-spectral images. Subsequently, lens distortion correction is performed on the dual-spectral images, and pixel-level registration is performed on the infrared and visible light images according to the pre-calibrated dual-spectral correspondence to ensure that the two types of images are consistent in spatial position.

[0067] After image correction and registration are completed, the monitoring backend extracts the corresponding target region image from the bispectral image based on the spatial coordinate information of the potential thermal runaway point, and performs contrast enhancement and normalization processing on the target region image to improve the clarity and identifiability of abnormal areas in the image, thereby providing a stable image data foundation for subsequent feature extraction.

[0068] After obtaining the processed target area image, infrared thermal features and visible light fault features are extracted respectively. Infrared thermal features include hot spot temperature features, temperature change rate features, and hot spot diffusion features, which are used to characterize the temperature rise anomalies and their development trends within the target area. Visible light fault features include arc flash features, insulation material carbonization features, and component deformation features, which are used to identify discharge, carbonization, and structural anomalies, thereby providing a basis for subsequent thermal runaway state determination.

[0069] Specifically, to address the image quality degradation caused by strong electromagnetic interference, uneven lighting, and structural obstruction in the UHV valve hall, a targeted preprocessing workflow was designed to provide high-quality image materials for subsequent feature extraction.

[0070] First, image denoising and electromagnetic interference filtering are performed. For salt-and-pepper noise and Gaussian noise caused by the strong electromagnetic environment of the valve hall, a combined algorithm of adaptive median filtering and wavelet transform denoising is adopted. The adaptive median filtering with a 3×3 window is used to remove salt-and-pepper noise in the image while preserving the image edge features (hot spot boundaries, component contours) and avoiding the blurring of hot spot edges caused by traditional filtering. Then, a 3-layer sym4 wavelet transform is used to perform soft thresholding shrinkage on the high-frequency components of the image, filtering out Gaussian white noise and stripe artifacts caused by electromagnetic interference, and restoring the true details of infrared and visible light images.

[0071] Next, image distortion correction and coordinate registration are performed. For image distortion caused by the wide-angle lens of the gimbal camera, the camera intrinsic parameters are calibrated in advance, and the distortion of the real-time image is corrected by the intrinsic parameter matrix to eliminate barrel distortion at the edge of the image. For the field of view deviation of the infrared and visible light dual-band images, the homography matrix is ​​used to complete the pixel-level registration of the dual-spectrum images to ensure that the hot spot area in the infrared image and the component area in the visible light image correspond completely in pixel coordinates, with a registration error of ≤2 pixels.

[0072] Based on the spatial coordinate mapping results, the Region of Interest (ROI) is accurately extracted. The corrected image is then centered and cropped to identify the component area where the suspected thermal runaway measurement point is located as the ROI. The cropping size is 256×256 pixels to ensure that the target component is always in the center of the ROI and to eliminate interference from irrelevant background areas. The Adaptive Histogram Equalization (CLAHE) algorithm is used to enhance the contrast of the infrared image of the ROI region, limiting the upper limit of contrast to 4.0 to avoid local over-enhancement. The visible light image is then subjected to Gamma correction with a Gamma value of 0.8 to eliminate overexposure / underexposure caused by strong light and reflections.

[0073] Finally, image normalization is performed to normalize the pixel values ​​of the preprocessed infrared and visible light images to the [0,1] range, and input them uniformly into the subsequent feature extraction network to eliminate the influence of differences in pixel value ranges between different spectral bands on feature extraction.

[0074] Specifically, six core features were extracted from both infrared and visible light bands, covering the entire stage of thermal runaway faults from initial abnormal temperature rise to mid-stage arcing and late-stage carbonization. The specific extraction method is as follows: First, a pre-trained improved U-Net network is used to extract pixel-level features from the infrared ROI image, outputting three types of quantized features: The first type of feature is the core hotspot feature, which identifies local hotspots within the ROI using the K-means clustering algorithm, quantifying and outputting the hotspot's highest temperature at the center (Tmax), average hotspot temperature (Tavg), hotspot area (A), and temperature difference (ΔT) between the hotspot and the surrounding area. The core criterion is ΔT; the larger ΔT is, the higher the confidence level of thermal runaway. The second type of feature is the temperature rise gradient and diffusion feature, which calculates the temperature change rate of the hotspot by comparing five consecutive frames of infrared images. dT / dt hot spot area diffusion rate dA / dt The first characteristic is the dynamic development of thermal runaway, which is distinguished from the steady-state temperature rise under normal operating conditions; the second characteristic is the uniformity of temperature distribution, which is used to calculate the coefficient of variation of temperature distribution in the ROI region. Cv (Standard deviation / mean), the component temperature distribution is uniform during normal operation. Cv <0.1; In the initial stage of thermal runaway, the temperature distribution exhibits a non-uniform and abrupt change. Cv A significant increase in temperature is used to distinguish between normal temperature rise and abnormal localized heating.

[0075] Then, an improved target detection network was used to extract features from the visible light ROI image, outputting three types of quantized features: The first type is the arc flicker time-domain-frequency domain composite feature. In the time domain, the inter-frame difference method was used to calculate the gray-level jump amplitude of 10 consecutive frames to capture the instantaneous high-brightness flicker feature of the arc. In the frequency domain, the frequency feature of the flicker signal was extracted through Fast Fourier Transform (FFT). The arc flicker frequency is concentrated in 50~200Hz, which is different from the steady-state high-frequency / low-frequency feature of environmental reflection. The arc feature confidence g is output, with a value range of [0,1], where g=1 represents 100% detection of arc discharge feature. The second type is the carbonization feature of insulating materials. First, the RGB image is converted to HSV color space, and the color, brightness, and saturation channels are separated. For the black and burnt yellow features of the aging and carbonization of insulating materials, color moment features (mean, variance, skewness) are extracted. Then, the gray-level co-occurrence matrix (GLCM) is used to extract the texture features (contrast, correlation, entropy) of the carbonized area. The color and texture features are fused to output the carbonization feature confidence. h The value range is [0,1]. h =1 indicates 100% detection of insulation carbonization features. The third category is component deformation and melting features. Using a feature matching algorithm, the real-time ROI image is compared with pre-stored component standard template images to identify deformation, melting, and cracking features caused by high temperatures, and the deformation feature confidence score is output. m The value range is [0,1], which serves as an auxiliary feature for determining the mid-to-late stages of thermal runaway.

[0076] S6: Perform fusion judgment on infrared thermal features and visible light fault features, and trigger a secondary review and warning based on the judgment result; This includes fusing infrared thermal features and visible light fault features for judgment, and triggering a secondary review and early warning based on the judgment result, including: The infrared thermal features and visible light fault features are vectorized and normalized to obtain multidimensional feature data. The thermal runaway probability index is obtained by weighted fusion calculation of multidimensional feature data based on preset feature weights. A set of evidence for judgment is constructed based on the thermal runaway probability index, and the DS evidence theory is used for comprehensive reasoning and calculation to obtain the thermal runaway judgment result. The thermal runaway determination result is compared with the preset determination threshold. When the thermal runaway determination result is greater than the determination threshold, a secondary review warning is triggered.

[0077] In this embodiment, after obtaining infrared thermal features and visible light fault features, the monitoring backend first performs feature vectorization processing on various features and then normalizes them to eliminate differences in the dimensions and value ranges of different features. Specifically, the hot spot temperature features, temperature change rate features, and hot spot diffusion features on the infrared side, and the arc flash features, insulating material carbonization features, and component deformation features on the visible light side can be converted into corresponding feature representation quantities, and combined in a preset order to form unified multi-dimensional feature data, thereby providing unified data input for subsequent fusion and judgment.

[0078] After obtaining multidimensional feature data, the monitoring backend performs weighted fusion calculations on the multidimensional feature data based on preset feature weights to obtain a thermal runaway probability index characterizing the current abnormal state. Preferably, different weights can be set for different features according to the stage of thermal runaway development. For example, the weight of infrared thermal features can be increased in the early stage of thermal runaway, and the weight of arc flash or carbonization features can be increased in the fault development stage to improve the system's ability to identify abnormal states at different stages.

[0079] After obtaining the thermal runaway probability index, the monitoring backend further constructs a set of judgment evidence based on the probability index and uses DS evidence theory to comprehensively reason about various types of evidence to obtain the final thermal runaway judgment result. Subsequently, the judgment result is compared with the preset judgment threshold. When the judgment result is greater than the judgment threshold, the potential thermal runaway point is judged to have passed the second-level review and a second-level review warning is triggered, thereby prompting the monitoring system to enter a higher-level risk response stage.

[0080] Specifically, the formula for calculating the thermal runaway probability index is as follows: ; in, Indicates the probability index of thermal runaway; Indicates the overall confidence level of infrared thermal anomalies; Indicates the confidence level of the arc flicker feature; Indicates the confidence level of the carbonization characteristics of insulating materials; Indicates the confidence level of deformation characteristics; Indicates the weighting coefficient of infrared temperature anomaly characteristics; Indicates the weighting coefficient of the arc flicker feature; Indicates the weighting coefficient of carbonization characteristics; This represents the weighting coefficient of structural deformation characteristics.

[0081] S7: When thermal runaway is confirmed, generate the corresponding warning level information based on the judgment result and output the corresponding linkage control command.

[0082] In this embodiment, once an anomaly is confirmed to be in a thermal runaway state, the monitoring backend generates corresponding early warning level information based on the thermal runaway probability index, feature type combination, and fault development degree in the fusion judgment result. Preferably, at least two early warning levels can be distinguished: minor thermal runaway and severe thermal runaway. If only a significant infrared temperature rise anomaly is detected, but no obvious features such as electric arc, carbonization, or deformation have appeared, and the comprehensive probability index is in a high range but has not reached an extreme danger level, it is judged as minor thermal runaway; if, in addition to infrared thermal anomalies, one or more features such as electric arc, carbonization, or component deformation are detected simultaneously, and the comprehensive probability index is in a higher range, it is judged as severe thermal runaway. The monitoring backend will generate early warning information by combining the early warning level, anomaly measurement point number, anomaly component location, trigger time, and main criteria, and push it to the monitoring interface and event recording system.

[0083] While outputting early warning information, the system generates linkage control commands based on the warning level and issues execution commands through the hard-wired interface between the system and the converter station control system. For minor thermal runaway events, the monitoring backend preferentially outputs a "reduced power operation" command, allowing the converter valves to reduce heat load while maintaining continuous operation, thus buying time for subsequent maintenance and risk control. For severe thermal runaway events, the monitoring backend preferentially outputs an "emergency trip" command, prompting the station control system to immediately cut off the power supply to the relevant converter valve towers to prevent thermal runaway from further developing into a fire or escalating into a more serious power grid accident. The entire linkage process requires minimal command transmission delay, clear execution actions, and consistency with the event records of the monitoring backend.

[0084] This application embodiment also provides a graded early warning system for thermal runaway of a converter valve, the system comprising: The contact-type fiber optic monitoring layer is used to collect temperature monitoring data of key components of the converter valve and transmit it to the monitoring backend. The non-contact multispectral verification layer is used to acquire infrared and visible light images of the area where potential thermal runaway points are located after receiving control commands sent from the monitoring backend. The monitoring backend receives temperature monitoring data, as well as infrared and visible light images, and performs the following processing: The temperature change rate at each monitoring point is calculated based on the temperature monitoring data, and the reference temperature rise threshold is compensated based on the operating parameters of the converter valve to obtain the dynamic early warning threshold. The temperature change rate is compared with the dynamic early warning threshold. When the temperature change rate exceeds the dynamic early warning threshold, the corresponding monitoring point is identified as a potential thermal runaway point. Based on the spatial coordinate information of the potential thermal runaway point, the non-contact multispectral core layer is controlled to collect infrared and visible light images of the corresponding area. Image processing is performed on infrared and visible light images to extract infrared thermal features and visible light fault features. The infrared thermal features and visible light fault features are then fused and judged to obtain the thermal runaway judgment result. When thermal runaway is confirmed, a corresponding early warning level is generated and a linkage control command is output.

[0085] It is understood that the content of the above method embodiments is applicable to this system embodiment. The specific functions implemented in this system embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.

[0086] refer to Figure 5 As shown, Figure 5 This is a schematic diagram of the overall structure of the converter valve thermal runaway classification identification system of the present invention. The system is deployed at the UHV converter valve hall, focusing on key heat-generating components of the converter valve tower 100, including the thyristor clamping contact surface 101, the reactor surface 102, and the DC capacitor casing 103. Fluorescent fiber optic temperature sensors are physically attached to the surfaces of these key components, forming a contact-type fiber optic detection layer 200. Specifically, the fluorescent fiber optic sensor 201 is used to collect temperature information of each key component in real time, with a preferred measurement accuracy of ±0.5℃ and a sampling frequency of once per second. The temperature signals collected by each sensor are transmitted through fiber optic lines to the fiber optic signal aggregation terminal 202, and the aggregated temperature data is sent to the monitoring backend.

[0087] The monitoring backend 400, as the core processing unit of the system, is responsible for real-time analysis and processing of temperature monitoring data, and for performing thermal runaway identification and early warning judgment. Specifically, the monitoring backend uses AI fusion judgment algorithm and spatial coordinate mapping algorithm to perform trend analysis on the received temperature data, and combines it with the pre-entered three-dimensional spatial coordinate information of each measuring point to locate potential thermal runaway measuring points. When an abnormal temperature rise trend is detected, the monitoring backend sends control commands to the non-contact multispectral verification layer 300 through WiFi 6 or 5G communication link, calculates the pan-tilt rotation parameters according to the measuring point coordinates, and causes the dual-spectral pan-tilt camera 301 to quickly turn to the target area under the control of the pan-tilt drive motor 302, and acquire infrared and visible light images of the corresponding area; on the other hand, the monitoring backend can send linkage control commands to the station control system 500 through hard wiring.

[0088] The non-contact multispectral verification layer 300 is preferably deployed at the corner or top of the valve hall for visual verification of potential abnormal areas. The dual-spectral PTZ camera 301 can simultaneously acquire infrared and visible light images, with an infrared resolution preferably of 640×512 and a visible light resolution preferably of 1080P, which are transmitted back to the monitoring backend in real time via a wireless link. The PTZ drive motor 302 controls the camera's 360° rotation, with a response time preferably less than 0.5 seconds. After feature extraction and fusion analysis of the acquired dual-spectral images, the monitoring backend can send a linkage control command to the station control system 500 when a thermal runaway risk is confirmed. The execution module 501 then performs power reduction operation or emergency tripping, thereby realizing a multi-level early warning mechanism from contact trend detection and spatial positioning verification to linkage control response. Through this system architecture, a dual-layer detection system of "contact detection + non-contact verification" can be constructed to achieve early identification and rapid handling of the thermal runaway risk of the converter valve.

[0089] Specifically, this embodiment describes in detail the specific implementation of the system construction and hierarchical identification method of the present invention in the actual application scenario of the valve hall of the ±800kV UHV converter station. The scope of protection of the present invention is not limited to this specific embodiment.

[0090] The system hardware setup and debugging are as follows: (1) Layout of contact fiber optic monitoring layer: The converter valve tower of the ±800kV UHV converter station valve hall is selected as the application object. FOT-TS200 fluorescent fiber optic temperature sensors are evenly applied to the thyristor pressing contact surface, reactor surface and DC capacitor shell of each layer of the converter valve tower. The temperature measurement accuracy of the sensor is adjusted to ±0.5℃ and the temperature acquisition frequency is set to 1 time / second. The fiber optic signal transmission lines of all sensors are collected and connected to the fiber optic acquisition card of the monitoring background to complete the real-time transmission and debugging of temperature data, ensuring that the temperature measurement data is without delay or loss.

[0091] (2) At the four corners of the valve hall and the two unobstructed locations on the top, one dual-spectrum (infrared + visible light) PTZ camera (model: IPC-YT9800) is set up. The infrared band resolution of the camera is set to 640×512, the temperature measurement range is -20℃~300℃, the visible light band resolution is set to 1080P, the PTZ rotation angle is set to 360° omnidirectional, and the rotation response time is ≤0.5s. The camera is connected to the monitoring backend via WiFi 6 wireless signal. The automatic turning accuracy of the camera after receiving the backend command is adjusted to ensure that the spatial positioning accuracy of the camera for the converter valve tower measuring point is ±5cm.

[0092] (3) Interconnection between the monitoring backend and the station control system: Input the three-dimensional spatial coordinates of all fluorescent fiber temperature sensors in the monitoring backend, embed dynamic threshold compensation, spatial coordinate mapping, AI multi-feature fusion judgment algorithm and complete the training, and set the temperature change rate benchmark threshold to 2℃ / min. Hard-wire the monitoring backend and the converter station control system, debug the transmission and execution of early warning commands, and ensure that the backend can stably send "reduce power operation" and "emergency trip" commands to the station control system with a command transmission delay of ≤0.5s; Establish a system fault tolerance mechanism: Set up 1:1 redundant backup for fiber optic sensors, and automatically switch to the backup sensor when a single sensor fails; set up a master-slave switching mechanism for PTZ cameras, and when a single camera goes offline, the adjacent camera automatically takes over its monitoring range to ensure that the system has no monitoring blind spots.

[0093] For valve halls with higher voltage levels of ±1100kV, only the fluorescent fiber optic temperature sensor needs to be replaced with a model that can withstand higher electric field strength (FOT-TS300) to avoid interference from strong electric fields on wireless signals. No other hardware layout or algorithm parameters need to be adjusted, and the system can be directly adapted.

[0094] After the system is built and debugged, the thermal runaway graded early warning system for the converter valve is started. The identification and early warning are performed according to the following three-level logic. The specific implementation process is as follows.

[0095] Level 1 Warning (Trend Detection): (1) The monitoring backend receives temperature measurement data from each fluorescent fiber optic temperature sensor in real time. When the temperature of the fiber optic measuring point on the contact surface of a certain thyristor rises from 60℃ to 61.5℃ within 30 seconds, the temperature change rate reaches 3℃ / min, and at this moment the real-time load current of the converter valve is 80% of the rated current, and the threshold after dynamic threshold compensation is 2.4℃ / min, the measured temperature change rate exceeds the dynamic threshold; the system immediately determines it as "suspected thermal runaway", triggers a first-level warning, and records the three-dimensional spatial coordinates P( of the measuring point). x , y , z And send it to the non-contact multispectral core layer; (2) The preset threshold can be dynamically compensated according to the real-time load current I of the converter valve. During the high-power operation of the converter valve, the reference threshold will automatically rise to avoid false alarms caused by normal load fluctuations. Rated operating conditions: The real-time load current of the converter valve is 100% of the rated current, the ambient temperature of the valve hall is 25℃, the equipment has been in operation for 3 years, and the cooling system is operating at its rated capacity. Under these conditions: =1 + 0.5 × 1 = 1.5, =1, =1 + 0.03 × 3 = 1.09, =1; =1.5×1×1.09×1=1.635℃; Low-load operating conditions: The real-time load current of the converter valve is 20% of the rated current, the ambient temperature of the valve hall is 15℃, the equipment has been in operation for 1 year, and the cooling system is operating at its rated capacity. Under these conditions: =1 + 0.5 × 0.2 = 1.1, =1+0.012×(15 25) = 0.88, =1.03, =1; =1.1×0.88×1.03×1≈0.997℃, the threshold is close to the reference value, and high sensitivity is maintained.

[0096] Extreme high-temperature aging conditions: The real-time load current of the converter valve is 90% of the rated current, the ambient temperature of the valve hall is 45℃, the equipment has been in operation for 8 years, and the cooling system is operating in a single loop. Under these conditions: =1 + 0.5 × 0.9 = 1.45, =1+0.012×(45 25) = 1.24, =1 + 0.03 × 8 = 1.24, =1.15; =1.45×1.24×1.24×1.15≈2.56, triggering the upper limit constraint, which both compensates for the influence of the operating condition and avoids the threshold from floating too much.

[0097] Secondary review (spatial positioning): After triggering the Level 1 alert, the monitoring backend calculates the pan-tilt camera's turning parameters using a spatial coordinate mapping algorithm and sends control commands to the dual-spectrum pan-tilt camera at the top of the valve hall to complete the anomaly location and AI fusion judgment. The specific steps are as follows: (1) Establishment and calibration of the spatial coordinate system: Taking a certain reference corner point on the ground of the converter station valve hall as the origin O (0,0,0), establish a three-dimensional Cartesian coordinate system along the length, width, and height of the valve hall. W The monitoring backend pre-enters data for each fluorescent fiber optic temperature sensor measurement point. i In coordinate system W Physical coordinates in and gimbal camera C Installation coordinates .

[0098] (2) Calculation of gimbal camera turning parameters: For the dual-spectral gimbal camera positioned at the top of the valve hall C calibrate it in the coordinate system W The installation coordinates are Simultaneously, a spherical coordinate system is established with the camera's optical center as the origin, where the azimuth angle is... θ The pitch angle is φ (Vertical rotation), focal length parameter is F.

[0099] When a Level 1 warning is triggered, determine the coordinates of the suspected thermal runaway monitoring point i. Subsequently, the monitoring backend calculates the turning parameters of the PTZ camera using a spatial analytical geometry algorithm: Calculate the horizontal azimuth and elevation angles; Autofocus reference value: Automatically adjusts the camera lens focal length F based on the Euclidean distance between the measuring point and the camera.

[0100] (3) Precise positioning of the gimbal camera: After receiving the command, the camera completes turning and focusing within 0.5s, accurately aligning with the measuring point P, which is always in the center area of ​​the video frame; to compensate for on-site installation errors, the system uses intrinsic and extrinsic parameter matrices for image correction. By left-multiplying the transformation matrix, the world coordinates are... Converted to camera pixel coordinates ( u , v This ensures that, after the gimbal camera rotates at high speed, the suspected thermal runaway measurement point remains in the center of the video frame, providing high-quality image material for neural network feature recognition in the subsequent secondary review. The conversion formula is as follows: ; In the formula: This represents the depth value of the measured point in the camera coordinate system. K For the camera intrinsic parameter matrix, R , T Let be the camera's extrinsic rotation matrix and translation vector.

[0101] In this embodiment, by using preset installation posture compensation and gimbal backlash compensation algorithms, the on-site installation deviation and mechanical error are corrected, and the gimbal camera finally achieves a positioning accuracy of ±4.2cm for abnormal points, which meets the design requirements. At the same time, by using multi-camera matching rules, the No. 2 gimbal camera, which is the closest to the unobstructed top of the valve hall, is selected as the main verification camera, and the adjacent No. 1 camera is selected as the backup verification camera, to ensure the effectiveness of dual verification.

[0102] To ensure that the suspected thermal runaway measurement point remains in the center of the video frame after the gimbal camera rotates at high speed, high-quality image material is provided for neural network feature recognition in the subsequent secondary review.

[0103] (4) AI multi-feature fusion judgment: After the PTZ camera completes image focusing, the monitoring backend executes an AI image feature recognition algorithm for verification. The specific logic is as follows: 1) Image preprocessing and Region of Interest (ROI) extraction: After receiving the real-time video stream from the PTZ camera, the system first uses preset coordinate mapping parameters to center and crop the image, determining the region of interest (ROI) of the target converter valve assembly. Then, an adaptive histogram equalization algorithm is used to enhance the infrared and visible light images, eliminating interference from complex lighting and high-voltage field strength within the valve hall on image quality.

[0104] 2) Multi-dimensional feature extraction: The system uses a pre-trained deep convolutional neural network (Faster R-CNN architecture) to extract multiple features in parallel from the ROI region: Infrared band features: Extract pixel-level temperature gradient distribution features. Use clustering algorithms to identify the presence of non-uniformly distributed "core hot spots" and calculate the hot spot area. A The highest temperature at the center.

[0105] Visible light band characteristics: Arc flicker detection: This method uses the temporal difference method to capture instantaneous bright flickering features in the image. Arcs are usually accompanied by non-periodic abrupt changes in blue-violet light intensity. The algorithm identifies arc features by analyzing the grayscale histogram transitions between consecutive frames.

[0106] Carbonization and discoloration recognition: By using color space conversion (RGB to HSV), color moments are extracted from the yellowish-brown and black pixels that are unique to the aging of insulating materials, and compared with a standard color chart under normal conditions to identify the carbonization morphology features of the material surface.

[0107] 3) Fusion Judgment Logic: The system adopts a two-layer fusion model of adaptive weighted fusion and secondary judgment based on DS evidence theory to solve the problem that fixed weights cannot adapt to different fault stages and fluctuations in confidence of different features, thus greatly improving the accuracy of judgment.

[0108] First, an adaptive weighted fusion calculation of the thermal runaway probability index is performed. Based on the fuzzy hierarchical analysis method, the weight coefficients of each feature are dynamically adjusted according to the saliency of the features at different fault stages. The complete formula is as follows: ; in: This is the thermal runaway probability index, with a value range of [0,1]. The overall confidence level of infrared thermal anomalies is calculated by fusing the above three types of infrared features, and its value ranges from [0,1]. f =1 indicates a severe temperature anomaly; , , These represent the confidence levels for the characteristics of electric arc, carbonization, and deformation, respectively, with values ​​ranging from [0,1]. , 、 , Let ω1 + ω2 + ω3 + ω4 be the adaptive weights for each feature, satisfying ω1 + ω2 + ω3 + ω4 = 1. The dynamic adjustment rule for the weights is as follows: Table 1: Rules for Dynamic Weight Adjustment;

[0109] After introducing a weight adaptive adjustment mechanism, the system can automatically determine the stage of thermal runaway development based on the temperature rise rate of the infrared hot spot, switch the corresponding weight matrix, achieve optimal feature matching for different fault stages, and solve the problems of insufficient sensitivity in early faults and inaccurate fault determination in later faults with fixed weights.

[0110] DS Evidence Theory: Secondary Judgment and False Alarm Filtering To further reduce the probability of misjudgment under strong interference environments, the DS evidence theory is introduced to perform a secondary verification of the adaptive weighted fusion results. The specific steps are as follows: Establish an identification framework Θ={A,B}, where A represents “confirmed thermal runaway” and B represents “normal operating condition / false alarm”; The four types of features are treated as four independent evidence bodies, and the basic probability assignment (BPA) of each evidence body to A and B is calculated. Using Dempster's combination rule, the BPAs of the four pieces of evidence are fused to obtain the final fusion probabilities m(A) and m(B). The judgment rule is set as follows: when m(A) > 0.85 and m(A) > m(B), it is judged as thermal runaway confirmation and the secondary review is triggered; when m(A) ≤ 0.85, it is judged as a false alarm, the warning signal is filtered, and the system resumes normal monitoring.

[0111] Based on the fusion judgment results and the severity of thermal runaway, the following graded judgment rules are formulated for this implementation case: Mild thermal runaway: 0.85 < ≤0.95, only infrared temperature abnormality was detected, with no electric arc or carbonization characteristics, which was determined to be a slight thermal runaway. The monitoring backend sent a power reduction operation command to the station control system. Severe thermal runaway: >0.95, and at the same time, abnormal infrared temperature and electric arc / carbonization characteristics were detected, which was determined to be a serious thermal runaway. The monitoring background sent an emergency trip command to the station control system. False alarm filtering: If the value is ≤0.85, regardless of whether a single feature is detected, it is judged as a false alarm, the warning signal is automatically filtered out, the secondary review and confirmation are not triggered, the abnormal data is recorded and real-time monitoring is restored.

[0112] In this embodiment, a judgment threshold is set. =0.85, Temperature Anomaly Weighting Coefficient =0.5, arc flicker weighting coefficient =0.3, weight of carbonization discoloration =0.2 is the basic weight for adapting the UHV converter valve. The weight can be adjusted according to actual operating data. , , , respectively, are confidence functions for temperature anomalies, arc flashes, and color changes, all ranging from [0,1]. =1 indicates a severe abnormality. =1 indicates that an electric arc has been detected. =1 indicates that carbonization has been detected. When Exceeding the set threshold At this time, the system automatically eliminates false alarms caused by ambient light reflection or normal operating temperature rise, determines that the second-level review has passed, and confirms that thermal runaway has occurred.

[0113] Level 3 Response (Linkage Control): Based on the dual confirmation results from the dual-spectrum PTZ camera, the monitoring backend immediately sends an "emergency trip" command to the converter station control system. The converter station control system executes the command within 0.5 seconds, cutting off the power supply to the converter valve tower. The total response time from triggering the first-level warning to completing the trip is 2.8 seconds, enabling rapid handling of thermal runaway accidents and preventing further escalation of the accident.

[0114] In summary, the application case of the ±800kV UHV converter station valve hall demonstrates that the converter valve thermal runaway graded early warning system and its identification method of the present invention can accurately detect the initial abnormal temperature rise of the small components inside the converter valve, realize the early detection of thermal runaway, effectively avoid misjudgment and missed judgment through the dual-layer architecture verification and three-level graded early warning, and has a fast linkage response speed, which can quickly cut off the power supply in the early stage of thermal runaway, fundamentally curbing the occurrence of major accidents such as valve hall fires. The system operates stably, has strong anti-interference ability, and is suitable for the high-voltage operating environment of UHV converter stations.

[0115] This application also provides an electronic device, which includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the above-described method. This electronic device can be any smart terminal, including tablet computers, in-vehicle computers, etc.

[0116] It is understood that the content of the above method embodiments is applicable to this device embodiment. The specific functions implemented by this device embodiment are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0117] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method.

[0118] It is understood that the content of the above method embodiments is applicable to this storage medium embodiment. The specific functions implemented in this storage medium embodiment are the same as those in the above method embodiments, and the beneficial effects achieved are also the same as those achieved in the above method embodiments.

[0119] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method.

[0120] It is understood that the content of the above method embodiments is applicable to the embodiments of this program product. The specific functions implemented by the embodiments of this program product are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0121] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. Furthermore, memory may include high-speed random access memory, and may also include non-transitory memory, such as at least one disk storage device, flash memory device, or other non-transitory solid-state storage device. In some embodiments, memory may optionally include memory remotely located relative to the processor, and these remote memories can be connected to the processor via a network. Examples of such networks include, but are not limited to, the Internet, intranets, local area networks, mobile communication networks, and combinations thereof.

[0122] This application provides a graded early warning system and identification method for thermal runaway of a converter valve. This system acquires temperature monitoring data of key components of the converter valve and combines it with operating parameters to perform dynamic threshold compensation calculations on the temperature change rate. This enables adaptive identification of abnormal temperature rise under different operating conditions, avoiding misjudgments or missed judgments caused by fixed threshold judgment methods under load or environmental changes, thus improving the accuracy and reliability of early thermal runaway identification. After detecting a potential thermal runaway point, a spatial coordinate mapping algorithm is further used to determine the target area, and a dual-spectrum gimbal camera is controlled to acquire infrared and visible light images of the corresponding area, achieving precise location of temperature anomalies and acquisition of multi-source information. By fusing and analyzing infrared thermal features and visible light fault features, abnormal states caused by structural damage, poor contact, or local overheating can be effectively distinguished, reducing the probability of false alarms caused by a single temperature monitoring method. Meanwhile, by constructing a graded early warning mechanism, image verification is performed based on the initial abnormal temperature rise identification, and the corresponding early warning level and linkage control command are output after confirming the risk of thermal runaway. This enables early identification, accurate positioning and graded response of the thermal runaway risk of the converter valve, thereby improving the safety of converter valve operation and the level of intelligent operation and maintenance.

[0123] The embodiments described in this application are for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and do not constitute a limitation on the technical solutions provided by the embodiments of this application. As those skilled in the art will know, with the evolution of technology and the emergence of new application scenarios, the technical solutions provided by the embodiments of this application are also applicable to similar technical problems.

[0124] Those skilled in the art will understand that the technical solutions shown in the figures do not constitute a limitation on the embodiments of this application, and may include more or fewer steps than shown, or combine certain steps, or different steps.

[0125] Those skilled in the art will understand that all or some of the steps in the methods disclosed above, as well as the functional modules / units in the systems and devices, can be implemented as software, firmware, hardware, or suitable combinations thereof.

[0126] The terms “first,” “second,” “third,” “fourth,” etc. (if present) in the specification and accompanying drawings of this application 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 this application 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 apparatus that comprises 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 apparatus.

[0127] The preferred embodiments of the present application have been described above with reference to the accompanying drawings, but this does not limit the scope of the claims of the present application. Any modifications, equivalent substitutions, and improvements made by those skilled in the art without departing from the scope and substance of the embodiments of the present application shall be within the scope of the claims of the present application.

Claims

1. A method for graded identification of thermal runaway in a converter valve, characterized in that, The method includes the following steps: Acquire temperature monitoring data of key components of the converter valve and collect operating parameters of the converter valve; The temperature change rate of each monitoring point is calculated based on the temperature monitoring data, and the benchmark temperature rise threshold is compensated based on the operating condition parameters to obtain the dynamic early warning threshold; the monitoring points are respectively deployed on the key components of the converter valve. The temperature change rate is compared with the dynamic early warning threshold. When the temperature change rate exceeds the dynamic early warning threshold, the corresponding monitoring point is identified as a potential thermal runaway point and a first-level early warning is triggered. At the same time, the spatial coordinate information of the potential thermal runaway point is obtained. Based on the spatial coordinate information, the camera control command of the dual-spectrum gimbal camera is calculated using a spatial coordinate mapping algorithm, and the dual-spectrum gimbal camera is controlled to acquire infrared and visible light images of the corresponding area according to the camera control command. The infrared and visible light images are preprocessed to extract the corresponding infrared thermal features and visible light fault features; The infrared thermal features and the visible light fault features are fused together for determination, and a secondary review and warning are triggered based on the determination result; When thermal runaway is confirmed, the system generates a corresponding warning level based on the determination result and outputs the corresponding linkage control command.

2. The method according to claim 1, characterized in that, The operating parameters include load current, ambient temperature, equipment age, and cooling system status.

3. The method according to claim 1, characterized in that, The formula for calculating the dynamic early warning threshold is as follows: ； in, Indicates the real-time dynamic warning threshold; Indicates the threshold for the rate of change of the baseline temperature rise; Indicates the total compensation coefficient; Indicates the load current compensation factor; Indicates the ambient temperature compensation factor; Indicates the equipment aging compensation factor; This represents the cooling system compensation factor.

4. The method according to claim 3, characterized in that, The upper and lower limits of the dynamic warning threshold include: the maximum dynamic warning threshold shall not exceed 3 times the reference temperature rise rate threshold, and the minimum dynamic warning threshold shall not be lower than 50% of the reference temperature rise rate threshold.

5. The method according to claim 1, characterized in that, The step of calculating the camera control commands for the dual-spectrum gimbal camera based on the spatial coordinate information using a spatial coordinate mapping algorithm includes: A global three-dimensional world coordinate system for the valve hall is established with the ground reference point of the valve hall as the origin, and the spatial coordinate information and the installation coordinates of each dual-spectrum gimbal camera are pre-calibrated. Based on the global three-dimensional world coordinate system, calculate the spatial vector and straight-line distance of the potential thermal runaway point relative to the dual-spectrum gimbal camera; The horizontal azimuth and pitch angles of the dual-spectrum gimbal camera pointing towards the potential thermal runaway point are calculated based on the spatial vector to obtain the steering control parameters of the dual-spectrum gimbal camera. Based on the preset mapping relationship between the straight-line distance and the step value, the focus control parameters of the dual-spectrum gimbal camera are determined. Based on the steering control parameters and the focus control parameters, control commands for the dual-spectrum gimbal camera are generated.

6. The method according to claim 5, characterized in that, The formulas for calculating the horizontal azimuth and elevation angles are as follows: ； ； in, This represents the horizontal azimuth angle of the dual-spectral gimbal camera corresponding to the i-th potential thermal runaway measurement point; This represents the pitch angle of the dual-spectral gimbal camera corresponding to the i-th potential thermal runaway measurement point; This represents the coordinate difference in the X direction between the potential thermal runaway measurement point and the dual-spectrum gimbal camera; This represents the coordinate difference in the Y direction between the potential thermal runaway measurement point and the dual-spectrum gimbal camera; This represents the coordinate difference in the Z direction between the potential thermal runaway measurement point and the dual-spectral gimbal camera.

7. The method according to claim 1, characterized in that, The step of preprocessing the infrared and visible light images and extracting the corresponding infrared thermal features and visible light fault features includes: The infrared image and the visible light image are denoised and electromagnetic interference filtered to obtain a denoised dual-spectral image. Lens distortion correction is performed on the denoised dual-spectral image, and pixel-level registration is performed between the infrared image and the visible light image to obtain the registered dual-spectral image. Based on the spatial coordinate information of the potential thermal runaway point, the target region image is extracted from the bispectral image, and the target region image is subjected to contrast enhancement and normalization processing. Infrared thermal features are extracted from the target region image, including hot spot temperature features, temperature change rate features, and hot spot diffusion features. Visible light fault features are extracted from the target area image, including arc flashing features, carbonization features of insulating materials, and deformation features of components.

8. The method according to claim 1, characterized in that, The process of fusing and determining the infrared thermal features and the visible light fault features, and triggering a secondary review and warning based on the determination result, includes: The infrared thermal features and the visible light fault features are processed into feature vectors and then normalized to obtain multidimensional feature data. The multidimensional feature data is weighted and fused based on preset feature weights to obtain the thermal runaway probability index. Based on the thermal runaway probability index, a set of evidence for judgment is constructed, and the DS evidence theory is used for comprehensive reasoning and calculation to obtain the thermal runaway judgment result. The thermal runaway determination result is compared with a preset determination threshold. When the thermal runaway determination result is greater than the determination threshold, a secondary review warning is triggered.

9. The method according to claim 8, characterized in that, The formula for calculating the thermal runaway probability index is as follows: ； in, Indicates the probability index of thermal runaway; Indicates the overall confidence level of infrared thermal anomalies; Indicates the confidence level of the arc flicker feature; Indicates the confidence level of the carbonization characteristics of insulating materials; Indicates the confidence level of deformation characteristics; Indicates the weighting coefficient of infrared temperature anomaly characteristics; Indicates the weighting coefficient of the arc flicker feature; Indicates the weighting coefficient of carbonization characteristics; This represents the weighting coefficient of structural deformation characteristics.

10. A graded early warning system for thermal runaway of a converter valve, characterized in that, The system includes: The contact-type fiber optic monitoring layer is used to collect temperature monitoring data of key components of the converter valve and transmit it to the monitoring backend. The non-contact multispectral verification layer is used to acquire infrared and visible light images of the area where potential thermal runaway points are located after receiving control commands sent from the monitoring backend. The monitoring backend receives the temperature monitoring data, as well as the infrared and visible light images, and performs the following processing: The temperature change rate at each monitoring point is calculated based on the temperature monitoring data, and the reference temperature rise threshold is compensated based on the operating parameters of the converter valve to obtain the dynamic early warning threshold. The temperature change rate is compared with the dynamic early warning threshold. When the temperature change rate exceeds the dynamic early warning threshold, the corresponding monitoring point is determined as a potential thermal runaway point. Based on the spatial coordinate information of the potential thermal runaway point, the non-contact multispectral core layer is controlled to collect infrared and visible light images of the corresponding area. Image processing is performed on the infrared image and the visible light image to extract infrared thermal features and visible light fault features. The infrared thermal features and the visible light fault features are then fused and judged to obtain the thermal runaway judgment result. When thermal runaway is confirmed, a corresponding early warning level is generated and a linkage control command is output.

Images