A distribution network asset state monitoring method based on a three-dimensional real scene model

Abstract

The present application relates to the technical field of distribution network asset state monitoring, and particularly relates to a distribution network asset state monitoring method based on a three-dimensional real scene model. The method comprises: obtaining a transient current impulse signal of a high-voltage switch cabinet through a current acquisition device, calculating a thermal diffusion delay time according to the insulating medium property, obtaining a temperature matrix of a target area for difference and threshold filtering when the delay time is reached, and obtaining an initial thermal transmission area; calculating a two-dimensional center pixel coordinate of the initial thermal transmission area, stopping feature updating when the continuous displacement is less than the tolerance, and locking it as a static space reference coordinate; obtaining a three-dimensional fault source coordinate in the three-dimensional real scene model, monitoring the temperature rise rate of the three-dimensional fault source coordinate, and generating a trip power-off instruction when the limit is exceeded. The present application triggers an electrical signal and stops updating the coordinate, eliminates the position drift error caused by thermal diffusion, reduces the computing power consumption, and improves the positioning accuracy.

Description

Technical Field

[0001] This invention relates to the field of distribution network asset status monitoring technology, specifically to a distribution network asset status monitoring method based on a three-dimensional real-scene model. Background Technology

[0002] For core electromechanical assets in power distribution networks and substations, including cable splices inside high-voltage switchgear, moving and stationary contacts of vacuum circuit breakers, busbar connection fittings, and high-voltage bushings of transformers, during the long-term service of these high-voltage electrical equipment, the internal metal contacts will experience increased micro-roughness due to continuous current load, oxidation corrosion, or mechanical stress vibration, leading to an abnormal increase in contact resistance. According to Joule's law, when a large load current flows through the aforementioned poorly contacted areas, a large amount of Joule heat will be generated. If not intervened in time, it will cause serious electrical accidents such as insulation melting, phase-to-phase short circuits, or even equipment burnout.

[0003] To achieve real-time, non-contact status monitoring of the aforementioned equipment, the industry primarily uses infrared thermal imagers to acquire a two-dimensional surface temperature matrix, which is then spatially mapped to a three-dimensional real-world model. This allows for the marking of overheating fault points in the three-dimensional view. For example, Chinese patent application CN113609716A describes an infrared temperature measurement visualization simulation method and device for substation equipment. This method creates a three-dimensional model of the equipment and combines it with infrared thermal imaging data to achieve three-dimensional temperature visualization. Another example is Chinese patent application CN111179423B, which describes a three-dimensional infrared image generation method based on two-dimensional infrared images. This method extracts the contours of two-dimensional infrared images and maps them to surface elements of a three-dimensional model to generate a three-dimensional temperature distribution map.

[0004] However, in actual electrical operating environments, existing pure vision tracking and mapping technologies have theoretical flaws and cannot effectively serve industrial control systems, specifically manifested in the following three core pain points: Firstly, the spatial coordinate drift caused by the hysteresis of thermal conduction in the insulating medium and anisotropic thermal diffusion. In real power distribution network assets, the heated metal contacts are tightly wrapped by thick insulating sleeves, such as epoxy resin castings and silicone rubber insulating sleeves. When the internal metal contacts generate a large amount of heat due to a sudden increase in load current, it takes several minutes to tens of minutes for the heat to penetrate the insulating sleeve and be conducted to the equipment surface. Furthermore, due to the uneven thickness of the insulating layer during manufacturing and the different air convection heat dissipation conditions in different directions of the equipment, the heat follows the anisotropic thermal diffusion law when it is conducted outward, that is, the heat will preferentially transmit along the path with the lower thermal resistance. Over time, the high-temperature area on the surface of the equipment will undergo severe irregular deformation in the direction of lower thermal resistance. At this time, the highest temperature point captured by the infrared detector will continuously shift in two-dimensional space, gradually moving away from the actual internal heating point, resulting in a serious coordinate drift error.

[0005] Secondly, existing monitoring algorithms suffer from a technical bias of continuously tracking the highest surface temperature, leading to severe misalignment in 3D reprojection. Currently known spatial registration algorithms in this field are all based on an inherent data-driven logic, assuming that the point with the highest temperature within the current field of view is the core of the fault. Faced with the aforementioned irreversible coordinate drift, existing technologies still extract distorted, hysteretic thermal features from each frame to match the fixed geometric topology of the 3D static device. This control logic causes distortion in the 3D mapping matrix, misreporting a real joint fault as overheating of the adjacent normal equipment casing, misleading on-site maintenance personnel's judgment, and easily triggering malfunctions or failures to operate by the relay protection unit.

[0006] Third, in order to avoid overlooking potential overheating issues, existing infrared and 3D fusion monitoring systems keep infrared cameras and edge computing gateways operating at extremely high frequencies of continuous feature extraction, image segmentation, and high-dimensional tensor projection calculations around the clock. Under the limited edge hardware computing power of power distribution rooms, this can easily lead to overload and overheating of the motherboard's central processing unit. At the same time, pure vision systems are prone to misinterpreting environmental interference such as heat emitted by personnel during inspections or sunlight refraction through windows as equipment overheating, resulting in a high false alarm rate and reducing the engineering usability of the monitoring device. Summary of the Invention

[0007] To address the high false alarm rate of existing technologies, this invention proposes a method for monitoring the status of distribution network assets based on a three-dimensional real-scene model. This method includes: S1. Acquire the instantaneous current of the power supply circuit. When the rate of change of the instantaneous current is greater than a preset current mutation threshold, activate the image acquisition device to acquire preset insulation medium property parameters inside the high-voltage switchgear and calculate the thermal diffusion delay time. S2. After the thermal diffusion delay time, control the image acquisition device to acquire the surface temperature matrix, preprocess the surface temperature matrix, extract the initial heat transmission zone, and then determine the static spatial reference coordinates. S3. Acquire a preset three-dimensional real-scene model and three-dimensional spatial geometric nodes. Iteratively calculate the static spatial reference coordinates and the three-dimensional spatial geometric nodes to determine the three-dimensional fault source coordinates. S4. Determine the monitoring field of view with the three-dimensional fault source coordinates as the center, extract the temperature rise rate within the monitoring field of view, and generate a trip power-off command when the temperature rise rate is greater than a preset heat resistance limit.

[0008] This invention utilizes the instantaneous rate of change of current as a pre-triggered condition to activate the image acquisition device, avoiding the high computational cost of blind scanning around the clock. Simultaneously, it introduces a thermal diffusion delay time, extracting the initial thermal transmission zone and solidifying the static spatial reference coordinates the instant heat is conducted out of the surface. This reduces the positional drift error of the equipment insulation layer caused by thermal conduction hysteresis and lateral diffusion. Finally, it combines a three-dimensional real-scene model to locate the heat source and cut off the fault circuit, improving the accuracy and security of power distribution network status monitoring.

[0009] Furthermore, the method for monitoring the status of distribution network assets based on a three-dimensional real-scene model further includes: acquiring the discrete current time series generated by analog-to-digital conversion of the power supply circuit, performing time-domain differentiation on the discrete current time series to obtain the rate of change of the instantaneous current, generating a transient current impulse signal with a timestamp when the rate of change of the instantaneous current is greater than the preset current mutation threshold, and sending the transient current impulse signal to the control terminal of the image acquisition device to release its dormant state.

[0010] Furthermore, the calculation of the thermal diffusion delay time includes: extracting the insulation thickness variable and the inherent thermal diffusivity from the insulation medium property parameters, and calculating the thermal diffusion delay time characterizing the internal heat generation through the insulation medium to the outer surface, based on the solid heat transfer mechanism and in combination with the insulation thickness variable and the inherent thermal diffusivity.

[0011] Further, the extraction of the initial heat transmission zone includes: obtaining a historical reference temperature matrix; subtracting the surface temperature matrix from the historical reference temperature matrix pixel by pixel to obtain a temperature increment value matrix; and overwriting the pixels in the temperature increment value matrix whose values ​​are lower than a preset noise threshold to zero to obtain the initial heat transmission zone after filtering out static background environmental heat.

[0012] This invention eliminates static background interference caused by sunlight or slow room temperature rise in substations by subtracting the current surface temperature matrix from the historical reference temperature matrix pixel by pixel and filtering with a preset noise threshold. This allows for accurate extraction of dynamic heating characteristics and effectively reduces the false alarm rate caused by environmental factors.

[0013] Further, determining the static spatial reference coordinates includes: using the temperature increment of non-zero pixels within the initial thermal transmission zone as a spatial weight, performing a weighted arithmetic mean calculation on the image plane coordinates of all non-zero pixels to obtain the two-dimensional center pixel coordinates; extracting the spatial displacement of the two-dimensional center pixel coordinates; and when the spatial displacement is less than a preset anti-shake tolerance, writing the current two-dimensional center pixel coordinates into a register to solidify them as the static spatial reference coordinates.

[0014] This invention uses temperature increment as a weight to calculate two-dimensional center coordinates, and writes them into a register for solidification when the displacement is less than the anti-shake tolerance. This ensures that the calibration coordinates will not jump or shift when dealing with severely deformed surface hot spots, thus guaranteeing the stability and uniqueness of the spatial reference coordinates.

[0015] Further, determining the coordinates of the three-dimensional fault source includes: constructing an evaluation function with the objective of minimizing the geometric error between the static spatial reference coordinates and the three-dimensional spatial geometric nodes, performing a nonlinear iterative approximation operation to obtain the rotation and translation parameters of the image acquisition device relative to the three-dimensional real scene model; after the iteration converges, the three-dimensional spatial geometric node with the smallest error is taken as the coordinates of the three-dimensional fault source.

[0016] Furthermore, the method for monitoring the status of distribution network assets based on a three-dimensional real-scene model further includes: using the coordinates of the three-dimensional fault source as the geometric anchor point, combining the rotation parameters and the translation parameters to determine the monitoring field of view, extracting the highest temperature value inside the monitoring field of view, and simultaneously calculating the temperature increase per unit time as the temperature rise rate.

[0017] Furthermore, when the temperature rise rate exceeds the preset heat resistance limit, a trip power-off command is generated. Specifically, when the highest temperature value in the surface temperature matrix exceeds the heat resistance melting threshold of the insulating material, an electrical protection trip signal is generated, and a hardware pulse is output to the microcomputer measurement and control device through the serial communication interface to drive the tripping coil of the corresponding feeder circuit breaker to perform the action to cut off the main circuit.

[0018] This invention outputs hardware pulses to a microcomputer-based monitoring and control device via a serial communication interface, driving the corresponding circuit breaker to trip and open, thus shortening the equipment's fault response time and reducing the probability of major electrical accidents.

[0019] Furthermore, the preset current mutation threshold includes: recording the current change rate caused by normal load fluctuations under fault-free steady-state operation of the high-voltage switchgear, and extracting its upper limit of confidence interval as the normal fluctuation baseline; recording the current change rate caused by large current inrush under short-circuit conditions, and extracting its lower limit of distribution as the fault mutation baseline; and calculating the current mutation threshold based on the normal fluctuation baseline and the fault mutation baseline.

[0020] Furthermore, the method for monitoring the status of distribution network assets based on a three-dimensional real-scene model further includes: recording the time delay data consumed from acquiring the instantaneous current to determining the coordinates each time the static spatial reference coordinates are determined, constructing the time delay data into an observation vector, and then using a state prediction filtering algorithm to iteratively correct the inherent thermal diffusivity in order to compensate for the thermal conduction characteristic shift caused by the aging of the equipment insulation medium.

[0021] This invention utilizes a state prediction filtering algorithm to iteratively correct the inherent thermal diffusivity, which can automatically compensate for changes in thermal conductivity caused by the aging of the insulating medium during long-term service of high-voltage switchgear, thereby enabling the equipment to maintain high-precision time delay waiting capability.

[0022] The present invention has the following technical effects: This invention implements cross-modal asynchronous wake-up through underlying electrical signals, maintaining a dormant state when there are no sudden load changes in the power supply circuit, thus reducing the computational burden on edge devices. It calculates the delay time using heat transfer laws, captures spatial features at the distortion-free moment in the early stage of internal heating, and then performs a coordinate stop update operation to prevent subsequent lateral thermal diffusion from interfering with three-dimensional calibration, achieving high-precision three-dimensional fault tracing and location. It uses an inter-frame differential mechanism to remove environmental pseudo-heat sources and transforms the spatial tracing results into tripping control commands for circuit breakers, constructing a rigorous data acquisition, monitoring and measurement, and electromechanical execution system, enhancing the daily operation and maintenance support level of distribution network assets. Attached Figure Description

[0023] Figure 1 This is a flowchart of a method for monitoring the status of distribution network assets based on a three-dimensional real-scene model, provided by an embodiment of the present invention; Figure 2 This is a comparison diagram of the effects of the prior art provided in the embodiments of the present invention and the present invention. Detailed Implementation

[0024] This invention provides a method for monitoring the status of distribution network assets based on a three-dimensional real-scene model, referring to... Figure 1 This includes steps S1-S4: S1: Data Acquisition and Preprocessing.

[0025] Specifically, the instantaneous current of the power supply circuit is acquired. When the rate of change of the instantaneous current exceeds the preset current mutation threshold, the image acquisition device is activated to acquire the preset insulation medium property parameters inside the high-voltage switchgear and calculate the thermal diffusion delay time.

[0026] First, configure the hardware environment. In this embodiment, Rogowski coil-type high-frequency current transformers are preferably installed on the incoming busbars of each phase of the high-voltage switchgear in the distribution network as the front-end current acquisition device. Under normal operation without faults, the image acquisition device and its associated graphics computing process at the back end are kept in a suspended sleep state to reduce the bus power consumption of the edge computing gateway. The high-frequency current transformer captures the alternating magnetic field flowing through the busbar of the target electrical equipment in real time and continuously converts it into a millivolt-level analog current signal.

[0027] It should be noted that this embodiment uses an uncooled focal plane array infrared thermal imager as the image acquisition device. The detector array of this image acquisition device has a resolution of 640×512, a spectral response band of 8-14μm, and is equipped with a motorized focusing lens. Since the interior of a high-voltage switchgear is usually in an extremely narrow and completely dark sealed state, traditional visible light cameras cannot work properly. However, this image acquisition device can directly receive the thermal radiation energy emitted from the surface of the target equipment and convert it into a surface temperature matrix with specific temperature values ​​within the microprocessor. Furthermore, the uncooled focal plane array has extremely low static power consumption in standby mode. Those skilled in the art will know that the above hardware model and parameters are only preferred examples. In practical applications, using an industrial-grade PTZ camera with dual-spectrum fusion function to replace a single infrared thermal imager also falls within the protection scope of this invention.

[0028] In order to convert the continuous analog signal into a digital format that the microprocessor can recognize, this embodiment uses a high-speed analog-to-digital conversion channel built into the edge computing gateway for signal processing. The high-speed analog-to-digital conversion channel continuously captures the above analog current signal at a sampling frequency of 1000Hz and adds a timestamp to each captured sampling data point, thereby generating a discrete current time series arranged in chronological order in memory.

[0029] The microprocessor inside the edge computing gateway reads the discrete current time series in real time and performs a first-order derivative operation in the time domain to monitor the current change trend. Specifically, the derivative operation process is as follows: the microprocessor extracts the discrete current values ​​corresponding to two adjacent timestamps and performs a subtraction calculation. Then, the current difference obtained by subtraction is divided by the sampling time interval between the two timestamps. Through this discrete difference division operation, the current increment per unit time is obtained, and this increment is the instantaneous current change rate.

[0030] Subsequently, the instantaneous current change rate is compared with the preset current mutation threshold in real time. When the instantaneous current change rate is greater than the current mutation threshold, according to the underlying circuit operation mechanism, it is determined that a large load step event has occurred in the current power supply circuit, and a high-power load has started or a short circuit impact has occurred in the target electrical equipment.

[0031] It should be noted that the preset current surge threshold represents the maximum allowable normal current increment of the power supply circuit per unit time. This embodiment uses a statistical calibration experiment for calibration, and the specific steps are as follows: First, prepare fault-free negative samples. Extract discrete current time series data of the target high-voltage switchgear for 30 consecutive days under fault-free steady-state operation and under the condition of violent peak and valley load alternation. Perform first-order difference division on these data to obtain the set of current change rates. Fit the set to a normal distribution law and extract its 99th percentile value as the normal fluctuation baseline.

[0032] Next, prepare positive samples of abnormal impacts, obtain waveform files of various phase-to-phase short circuits, grounding faults and direct start-up of high-power inductive loads recorded in the historical operation and maintenance log, perform the same first-order difference division calculation on these time series data containing large current inrush currents to obtain the initial current change rate set, and extract the distribution lower bound value in the set as the fault mutation baseline.

[0033] Finally, the optimal balance point is found. After obtaining the normal fluctuation baseline and the fault mutation baseline, the microprocessor performs a linear step traversal within the numerical range of the two with a step size of 0.1. By comparing the false alarm rate and missed alarm rate under different values, the obtained optimal balance point value is used as the current mutation threshold and sent to the register of the edge computing gateway for solidification.

[0034] Upon confirmation of a load step event, a transient current surge signal containing the current time is immediately generated. The microprocessor, through its internal hardware interrupt controller, sends this transient current surge signal as the highest priority interrupt instruction to the main control pin of the image acquisition device, which is in sleep mode. After receiving the high-level interrupt instruction, the image acquisition device immediately exits sleep mode and enters working mode, thus completing the asynchronous wake-up operation based on the underlying electrical signals.

[0035] After the image acquisition device is successfully started, any feature extraction operation is intercepted and suspended to prevent premature capture. First, the insulation medium property parameters corresponding to the heating contacts inside the target high-voltage switchgear are obtained from the local configuration database. In this embodiment, the insulation medium property parameters obtained include the thickness variable of the insulation sheath and the thermal diffusivity of the insulation material at the factory.

[0036] To predict when heat will be conducted from the inside to the outside, the thermal diffusion delay time needs to be calculated based on the solid heat transfer mechanism. The microprocessor obtains the value of the insulation sheath thickness variable and calculates its square value, using this square value as the dividend. At the same time, the product of the inherent thermal diffusivity and 4 is calculated, and this product is used as the divisor. The dividend is divided by the divisor, and the quotient is the thermal diffusion delay time required for internal heat to penetrate the insulation medium and be conducted to the outer surface of the device. This thermal diffusion delay time is then sent to the hardware timer module to provide a time reference for subsequent accurate image capture.

[0037] S2: Static space reference coordinates are determined.

[0038] Specifically, after the thermal diffusion delay time, the image acquisition device is controlled to acquire the surface temperature matrix, the surface temperature matrix is ​​preprocessed, the initial thermal transmission zone is extracted, and then the static spatial reference coordinates are determined.

[0039] After the thermal diffusion delay time in the hardware timer returns to zero, the microprocessor immediately sends a capture command to the control bus of the image acquisition device, controlling the infrared detector array to output the full-frame thermal radiation distribution data at the current moment, thereby obtaining the surface temperature matrix. In order to eliminate the interference of static background heat sources such as the slow rise of the substation ambient temperature or sunlight shining through the window, a background stripping preprocessing operation needs to be performed: the microprocessor retrieves the historical reference temperature matrix cached in the random access memory before starting the image acquisition device. This historical reference temperature matrix is ​​the normal noise temperature of the equipment under the condition of no large current sudden impact. The current surface temperature matrix is ​​subtracted from the historical reference temperature matrix pixel by pixel to obtain a numerical matrix reflecting the temperature increment caused by the sudden load.

[0040] Subsequently, noise truncation is performed on the temperature increment value matrix. The microprocessor iterates through all pixels in the temperature increment value matrix, overwriting pixels with values ​​below a preset noise threshold to 0. The remaining non-zero pixel connected regions constitute the initial heat transmission zone after filtering out static background environmental heat. It should be noted that this noise threshold is obtained through an environmental background noise statistical calibration experiment. The specific steps are as follows: the target high-voltage switchgear is continuously operated under rated no-load conditions for 72 hours. An image acquisition device continuously acquires a sequence of environmental infrared images at a period of 1 minute. The temperature difference between pixels at the same position between two adjacent frames is calculated, and the probability density distribution of all pixel temperature differences is statistically analyzed. The upper limit deviation value covering the 99.7% confidence interval in the probability density distribution curve is extracted and set as the preset noise threshold.

[0041] After extracting the initial heat transmission zone, the microprocessor uses the temperature increment of non-zero pixels within the initial heat transmission zone as spatial weight to calculate the weighted arithmetic mean of the image plane coordinates of all non-zero pixels. The specific calculation logic is as follows: multiply the horizontal coordinate value of each non-zero pixel by its temperature increment value and sum them globally, then divide by the sum of the temperature increment values ​​of all non-zero pixels to obtain the comprehensive horizontal coordinate. Similarly, the comprehensive vertical coordinate is calculated, and the two are combined to obtain the two-dimensional center pixel coordinates. It can be seen that the higher the temperature increment of a pixel, the greater the pull on the final coordinates, so that the obtained two-dimensional center pixel coordinates can fit the thermodynamic core of the heat source, rather than the geometric center of an irregular contour.

[0042] After obtaining the two-dimensional center pixel coordinates, the microprocessor tracks the movement trajectory of the coordinates within five consecutive image sampling frames, extracts and calculates the spatial displacement of the two-dimensional center pixel coordinates between adjacent frames. When the spatial displacement is continuously less than the preset anti-shake tolerance, the microprocessor determines that the current thermal transmission area is in the stable period of the initial thermal penetration stage. At this time, the current two-dimensional center pixel coordinates are directly written into the dedicated protection register and solidified as static spatial reference coordinates.

[0043] It should be noted that the preset image stabilization tolerance was calibrated through an interference evaluation experiment. The specific steps are as follows: with the equipment in a cold shutdown state, an alternating electromagnetic vibration with a frequency of 50Hz is applied to the metal cabinet of the high-voltage switchgear using a mechanical vibrator. A high-speed industrial camera is used to record the pixel jitter trajectory of a high-contrast reference target within the field of view of the image acquisition device. The maximum pixel offset of the target in the horizontal and vertical directions is extracted, and a 10% safety margin is added to this value for calculation. The result is calibrated as the preset image stabilization tolerance.

[0044] After the coordinate solidification operation is completed, the subsequent image feature update calculation for the current heat transmission area is stopped, and the large-area distortion and irregular spread interference of surface hot spots caused by uneven thermal resistance inside the insulating medium are cut off. No matter how the high temperature area in the subsequent video stream is deformed and drifted, the most accurate positioning point initially captured is always maintained.

[0045] S3: Three-dimensional iterative mapping and fault source determination.

[0046] Specifically, a preset 3D real-world model and 3D spatial geometric nodes are obtained, and the static spatial reference coordinates and 3D spatial geometric nodes are iteratively calculated to determine the coordinates of the 3D fault source.

[0047] First, a 3D laser scanner is used to perform an offline real-scene scan of the interior of the target high-voltage switchgear and the surrounding station environment to obtain high-precision 3D point cloud topology data of the equipment. Then, specific 3D coordinate values ​​representing high-risk cable lap joints, moving and stationary contacts of vacuum circuit breakers, and busbar connection hardware are extracted from the 3D point cloud topology data. These extracted coordinate data of specific parts are summarized into a set of 3D spatial geometric nodes and stored in the local configuration database of the edge computing gateway.

[0048] Next, based on the perspective projection law of pinhole imaging, a geometric mapping mechanism between the two-dimensional pixel plane and the three-dimensional spatial environment is established. In order to obtain the position of the two-dimensional coordinates in the three-dimensional space, it is necessary to determine an evaluation function with the goal of minimizing the error. The microprocessor assumes various posture parameters, calculates the position of the two-dimensional scattered points generated when the three-dimensional spatial geometric nodes are projected onto the two-dimensional pixel plane, calculates the sum of the squares of the straight-line distances between these theoretical two-dimensional scattered points and the static spatial reference coordinates, and takes minimizing the sum of the squares of the straight-line distances as the goal of the above evaluation function.

[0049] Subsequently, the microprocessor calls the arithmetic logic unit within the chip to perform nonlinear iterative approximation calculations. In this embodiment, the Levenberg-Marquardt algorithm is used for iterative optimization of spatial attitude. During the iterative calculation, the rotation and translation parameters representing the spatial orientation of the image acquisition device relative to the three-dimensional real-scene model are continuously corrected. In each iteration, the partial derivatives of the evaluation function with respect to each spatial attitude parameter are solved to generate the Jacobian matrix, thereby determining the optimal gradient direction that can rapidly reduce the reprojection error. The values ​​of the rotation and translation parameters are continuously updated along this optimal gradient direction.

[0050] During the nonlinear iterative approximation operation, the geometric error difference between two consecutive iterations is continuously calculated. This geometric error difference is compared with the set convergence accuracy extreme value. In this embodiment, the convergence accuracy extreme value is preferably calibrated by the inherent spatial resolution of the image acquisition device. Specifically, the calibration method is to extract the equivalent size of a single pixel of the image acquisition device at the current monitoring field distance, and take 10% of the equivalent size of the single pixel as the convergence accuracy extreme value. When the geometric error difference calculated by two consecutive iterations is less than the convergence accuracy extreme value, the operation is determined to have converged. At this time, the spatial alignment error has been limited to the sub-pixel level.

[0051] After iterative convergence, the microprocessor substitutes the final determined rotation and translation parameters, as well as the focal length and optical center of the image acquisition device, into the pinhole imaging perspective projection equation. This pinhole imaging perspective projection equation is the aforementioned geometric mapping mechanism for converting three-dimensional space to a two-dimensional plane. Using this pinhole imaging perspective projection equation, the reprojection error of all nodes in the three-dimensional spatial geometric node set is checked, and the theoretical projection position of each node in the set under this equation is calculated. Then, the planar straight-line distance between each theoretical projection position and the static spatial reference coordinates is calculated, and the specific three-dimensional spatial geometric node with the smallest planar straight-line distance is extracted. This specific three-dimensional spatial geometric node is determined as the coordinate of the three-dimensional fault source of actual heating inside the high-voltage switchgear.

[0052] S4: Monitoring view frame determination and trip power-off command execution.

[0053] Specifically, the monitoring field of view is determined with the three-dimensional fault source coordinates as the center, the temperature rise rate within the monitoring field of view is extracted, and a trip power-off command is generated when the temperature rise rate exceeds the preset heat resistance limit.

[0054] After obtaining the coordinates of the three-dimensional fault source, the monitoring logic changes. Instead of blindly searching for features in the global field of view using two-dimensional infrared images, the three-dimensional fault source coordinates are used as geometric anchor points. The rotation and translation parameters obtained in S3 are used to perform a reverse projection operation on each frame of the input infrared image stream. Through this reverse projection operation, a rectangular area with a fixed area is obtained in the two-dimensional image, which serves as the monitoring field of view. In this embodiment, the length and width of the monitoring field of view are preferably set to a range of 20 pixels including the center pixel and its surrounding extension. This local fixed-point monitoring mode reduces the computational load of the microprocessor and eliminates temperature interference from other normal devices in the field of view.

[0055] Subsequently, the highest temperature values ​​of the pixels within the monitoring field of view are extracted cyclically. The microprocessor reads the current clock, records the time difference between two consecutive extractions of the highest temperature value, calculates the difference between the two consecutive extractions of the highest temperature value, and divides the difference by the time difference to obtain the temperature increase per unit time. This temperature increase per unit time is used as the real-time temperature rise rate.

[0056] The highest temperature value and the rate of temperature rise are compared in real time with the preset heat resistance limit, which includes the heat resistance melting threshold of the insulation material and the abnormal temperature rise surge threshold. In this embodiment, the highest safe operating temperature value of the insulation sheath material inside the target high-voltage switchgear is extracted by consulting the factory technical manual and used as the heat resistance melting threshold of the insulation material. At the same time, the abnormal temperature rise surge threshold is preferably set to 15℃ / min.

[0057] When the microprocessor determines that the rate of temperature rise is greater than the threshold for abnormal temperature rise, or that the highest temperature value is greater than the heat resistance melting threshold of the insulating material, it immediately generates a high-level trip power-off command to trigger the power-off protection.

[0058] Upon receiving the control command, the edge computing gateway sends an action signal to the external opto-isolation relay board via the serial communication interface. After the opto-isolation relay board is activated, it outputs a high-level DC hardware pulse to the hard-wired terminal of the microcomputer monitoring and control device of the target high-voltage switchgear. After receiving the hardware pulse, the microcomputer monitoring and control device excites the tripping coil of the upper-level vacuum circuit breaker of the corresponding feeder circuit. The tripping coil performs a mechanical tripping action, instantly separating the moving and stationary contacts of the main circuit, thereby cutting off the fault power supply circuit that causes the equipment to overheat.

[0059] It should be noted that in the above embodiments, the inherent thermal diffusivity parameter used to calculate the thermal diffusion delay time is derived from the manufacturer's instructions for the insulating material. However, in actual industrial settings, after decades of service, high-voltage switchgear may develop microscopic air gaps or localized thinning within the insulating resin, causing a shift in its actual thermal conductivity characteristics. To compensate for this error caused by equipment aging, this embodiment adds a historical feedback adaptive update mechanism after executing S2. Specifically, each time the microprocessor successfully extracts and determines the static spatial reference coordinates, it retrieves the system log and calculates the time from the moment the transient current impact signal is received until the moment the static spatial reference coordinates are successfully written into the dedicated protection register. The microprocessor continuously accumulates the latency data consumed during the period in its local database, and constructs a one-dimensional observation vector from the latency data of the most recent 10 times. Subsequently, the microprocessor calls a state prediction filtering algorithm, preferably the Kalman filter algorithm in this embodiment, to use the locally configured inherent thermal diffusivity as a state variable. Iterative correction calculations are performed on this state variable using the aforementioned observation vector, and the corrected value is written back to the local configuration database for use in the next thermal diffusivity delay time calculation. Through this mechanism, automatic tracking and parameter compensation of the aging process of the insulating medium can be achieved, ensuring that the capture timing remains in the optimal state throughout the entire life cycle of the equipment.

[0060] Figure 2 This is a comparison diagram of the effects of the prior art and the present invention provided in the embodiments of the present invention. It can be seen that as heat diffuses laterally within the insulator, the highest temperature coordinates extracted by the prior art continuously deviate from the actual internal heat source, and the positioning deviation increases exponentially. In contrast, the present invention intercepts and solidifies the coordinates within the stable window at the initial stage of heat penetration, preventing interference from subsequent thermal drift, and converging the positioning deviation and keeping it at an extremely low level.

Claims

1. A method for monitoring the status of distribution network assets based on a three-dimensional real-scene model, characterized in that, include: S1, acquire the instantaneous current of the power supply circuit. When the rate of change of the instantaneous current is greater than the preset current change threshold, start the image acquisition device, acquire the preset insulation medium property parameters inside the high-voltage switchgear, and calculate the thermal diffusion delay time. S2, after the heat diffusion delay time, control the image acquisition device to acquire the surface temperature matrix, preprocess the surface temperature matrix, extract the initial heat transmission area, and then determine the static spatial reference coordinates. S3, obtain the preset three-dimensional real scene model and three-dimensional spatial geometric nodes, and perform iterative calculations on the static spatial reference coordinates and the three-dimensional spatial geometric nodes to determine the coordinates of the three-dimensional fault source; S4. Using the coordinates of the three-dimensional fault source as the center, determine the monitoring field of view, extract the temperature rise rate within the monitoring field of view, and generate a trip power-off command when the temperature rise rate is greater than the preset heat resistance limit.

2. The method for monitoring the status of distribution network assets based on a three-dimensional real-scene model according to claim 1, characterized in that, The method further includes: acquiring a discrete current time series generated by analog-to-digital conversion of the power supply circuit; performing time-domain differentiation on the discrete current time series to obtain the rate of change of the instantaneous current; generating a transient current impulse signal with a timestamp when the rate of change of the instantaneous current is greater than the preset current mutation threshold; and sending the transient current impulse signal to the control terminal of the image acquisition device to release its sleep state.

3. The method for monitoring the status of distribution network assets based on a three-dimensional real-scene model according to claim 1, characterized in that, The calculation of the thermal diffusion delay time includes: extracting the insulation thickness variable and the inherent thermal diffusivity from the insulation medium property parameters, and calculating the thermal diffusion delay time characterizing the internal heat generation through the insulation medium to the outer surface based on the solid heat transfer mechanism and the insulation thickness variable and the inherent thermal diffusivity.

4. The method for monitoring the status of distribution network assets based on a three-dimensional real-scene model according to claim 1, characterized in that, The extraction of the initial thermal transmission zone includes: obtaining a historical reference temperature matrix; subtracting the surface temperature matrix from the historical reference temperature matrix pixel by pixel to obtain a temperature increment value matrix; and overwriting the pixels in the temperature increment value matrix whose values ​​are lower than a preset noise threshold to zero to obtain the initial thermal transmission zone after filtering out static background environmental heat.

5. The method for monitoring the status of distribution network assets based on a three-dimensional real-scene model according to claim 1, characterized in that, The determination of the static spatial reference coordinates includes: using the temperature increment of the non-zero pixels inside the initial thermal transmission zone as the spatial weight, performing a weighted arithmetic mean calculation on the image plane coordinates of all non-zero pixels to obtain the two-dimensional center pixel coordinates; extracting the spatial displacement of the two-dimensional center pixel coordinates; and when the spatial displacement is less than the preset anti-shake tolerance, writing the current two-dimensional center pixel coordinates into a register to solidify them as the static spatial reference coordinates.

6. The method for monitoring the status of distribution network assets based on a three-dimensional real-scene model according to claim 1, characterized in that, The determination of the three-dimensional fault source coordinates includes: constructing an evaluation function with the objective of minimizing the geometric error between the static spatial reference coordinates and the three-dimensional spatial geometric nodes; performing a nonlinear iterative approximation operation to obtain the rotation and translation parameters of the image acquisition device relative to the three-dimensional real scene model; and after the iteration converges, taking the three-dimensional spatial geometric node with the smallest error as the three-dimensional fault source coordinates.

7. The method for monitoring the status of distribution network assets based on a three-dimensional real-scene model according to claim 6, characterized in that, The method further includes: using the three-dimensional fault source coordinates as geometric anchor points, combining the rotation parameters and the translation parameters to determine the monitoring field of view, extracting the highest temperature value inside the monitoring field of view, and simultaneously calculating the temperature increase per unit time as the temperature rise rate.

8. The method for monitoring the status of distribution network assets based on a three-dimensional real-scene model according to claim 1, characterized in that, When the temperature rise rate exceeds the preset heat resistance limit, a trip power-off command is generated. Specifically, when the highest temperature value in the surface temperature matrix exceeds the heat resistance melting threshold of the insulating material, an electrical protection trip signal is generated. A hardware pulse is output to the microcomputer measurement and control device through the serial communication interface to drive the tripping coil of the corresponding feeder circuit breaker to perform the action to cut off the main circuit.

9. The method for monitoring the status of distribution network assets based on a three-dimensional real-scene model according to claim 1, characterized in that, The preset current mutation threshold includes: recording the current change rate caused by normal load fluctuations under fault-free steady-state operation of the high-voltage switchgear, and extracting its upper limit of confidence interval as the normal fluctuation baseline; recording the current change rate caused by large current inrush under short-circuit conditions, and extracting its lower limit of distribution as the fault mutation baseline; and calculating the current mutation threshold based on the normal fluctuation baseline and the fault mutation baseline.

10. The method for monitoring the status of distribution network assets based on a three-dimensional real-scene model according to claim 3, characterized in that, The method further includes: recording the time delay data consumed from acquiring the instantaneous current to determining the coordinates each time the static spatial reference coordinates are determined, constructing the time delay data into an observation vector, and then using a state prediction filtering algorithm to iteratively correct the inherent thermal diffusivity to compensate for the thermal conduction characteristic shift caused by the aging of the equipment insulation medium.

Images