A substation flood multi-source fusion early warning method and device

Abstract

The application provides a transformer substation flood multi-source fusion early warning method and device, and relates to the field of transformer substation environment monitoring. The method comprises the following steps: obtaining a water level height value and a physical contact height value of a target transformer substation, and capturing a video frame containing a physical capacitance electrode water gauge in real time; determining a visual water level value by using the video frame containing the physical capacitance electrode water gauge; calculating a comprehensive fusion water level of the target transformer substation based on the water level height value, the physical contact height value and the visual water level value, and determining the risk level of the target transformer substation according to the comprehensive fusion water level and a preset equipment height mapping table. The application can improve the accuracy of transformer substation water level monitoring, and can more accurately evaluate the actual risk degree of the transformer substation facing flood disasters.

Description

Technical Field

[0001] This application relates to the field of substation environmental monitoring, and in particular to a method and device for multi-source fusion early warning of floods in substations. Background Technology

[0002] Substations, as crucial nodes in power systems for energy transformation and distribution, typically house key electrical facilities such as transformer foundations, high-voltage switchgear, and cable tunnels. These facilities are highly sensitive to environmental humidity and water depth. Water accumulation or abnormally high water levels within the substation can lead to decreased insulation performance of electrical equipment, corrosion of metal components, and in severe cases, even electrical short circuits, thereby affecting the safe operation of the power grid. Therefore, real-time monitoring of water levels within substations and timely early warning systems are of paramount importance for ensuring the safe operation of power facilities.

[0003] Currently, common technical solutions for flood control monitoring in substations mainly include fixed monitoring devices and manual inspection methods. Fixed monitoring devices are typically constructed through civil engineering projects, requiring on-site excavation, sensor installation, and the laying of mains power and communication lines. They collect water level or related hydrological parameters through permanently installed sensors. Manual inspection methods rely on maintenance personnel periodically entering the site to assess water accumulation using visual inspection or simple measuring tools. In addition, some technical solutions incorporate pressure sensors, ultrasonic sensors, and other detection methods to collect water level information and upload it to a backend system for analysis and processing.

[0004] However, the above traditional solutions have the following inherent defects: First, the accuracy of water level sensing is insufficient. Monitoring by a single type of sensor is easily affected by environmental interference. Pressure level gauges are affected by silt, air bubbles, and dynamic pressure, resulting in measurement deviations. Video recognition lacks a stable physical calibration benchmark and is easily affected by light, water surface reflection, and lens moisture, leading to distorted monitoring data. Second, the risk assessment logic is imperfect. Risk is judged only by combining equipment elevation and is not effectively coupled with real-time sensor data, which easily leads to false alarms and missed alarms. Third, the environmental adaptability is poor. The strong electromagnetic environment of substations can easily interfere with weak electrical communication equipment, leading to communication abnormalities or equipment damage. Fourth, the emergency response capability is weak. Most systems rely on cloud data analysis, and communication terminals cannot independently complete risk assessment, thus losing emergency early warning capabilities.

[0005] In summary, the aforementioned deficiencies directly result in the inability of existing substation flood early warning technologies to meet the requirements for high-precision, high-reliability, and high-adaptability flood control monitoring. Summary of the Invention

[0006] This application provides a multi-source fusion early warning method and device for floods in substations to solve the problem of low accuracy in existing flood early warning and monitoring of substations.

[0007] In a first aspect, this application provides a multi-source fusion early warning method for flooding in substations. The method is applied to a portable flood early warning device for substations. The device includes an integrated mobile enclosure, a three-source redundant sensing unit, an autonomous power supply and temperature control scheduling unit, a hardware-isolated 4G communication unit, and a containerized edge intelligence platform. The method includes: Acquire the water level and physical contact height of the target substation, and capture video frames containing the physical capacitance electrode water gauge in real time; The visual water level value is determined using the video frame containing the physical capacitance electrode water gauge. Based on the water level height value, the physical contact height value, and the visual water level value, the comprehensive integrated water level of the target substation is calculated, and the risk level of the target substation is determined according to the comprehensive integrated water level and the preset equipment height mapping table.

[0008] Secondly, this application provides a substation flood multi-source fusion early warning device, which is applied to a portable flood early warning device for substations. The device includes an integrated mobile housing, a three-source redundant sensing unit, an autonomous power supply and temperature control scheduling unit, a hardware-isolated 4G communication unit, and a containerized edge intelligent platform. The device includes: The acquisition module is used to acquire the water level height and physical contact height of the target substation, and to capture video frames containing the physical capacitance electrode water gauge in real time. The determination module is used to determine the visual water level value using the video frame containing the physical capacitance electrode water gauge; The risk level determination module is used to calculate the comprehensive integrated water level of the target substation based on the water level height value, the physical contact height value, and the visual water level value, and to determine the risk level of the target substation according to the comprehensive integrated water level and the preset equipment height mapping table.

[0009] This application provides a method and device for multi-source fusion early warning of flooding in substations. This application acquires data from three different sources: water level height, physical contact height, and visual water level, and fuses them to obtain a comprehensive fusion water level. The multi-source data complements and verifies each other, effectively reducing the errors and uncertainties that may exist with a single data source, thereby significantly improving the accuracy of substation water level monitoring and providing a reliable basis for accurate flood disaster early warning. Secondly, the risk level of the target substation is determined based on the comprehensive fusion water level and a preset equipment height mapping table. This determination method considers the impact of multiple factors on substation safety, compared to relying solely on a single water level data to determine risk. Compared to other risk levels, this approach is more scientific and comprehensive. It can not only more accurately assess the actual risk level of a substation facing floods, but also buy valuable response time for substation maintenance personnel, effectively reducing the impact of floods on substation equipment and power system operation, and improving the reliability and stability of the power system. At the same time, this application helps to improve the flood disaster emergency early warning system of the power system, enhance the emergency response capability and handling efficiency of the power system in the face of floods, and enable power system maintenance departments to respond to floods more proactively, rationally allocate emergency resources and manpower, and improve the disaster resistance capability and operational stability of the entire power system through real-time monitoring and accurate early warning. Attached Figure Description

[0010] To more clearly illustrate the technical solutions in the embodiments of this application, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0011] Figure 1 This is a flowchart illustrating the implementation of the substation flood multi-source fusion early warning method provided in this application embodiment; Figure 2 This is a schematic diagram of the structure of the portable flood early warning device for substations provided in the embodiments of this application; Figure 3 This is a schematic diagram of the bottom of the integrated mobile housing provided in this application embodiment; Figure 4 This is a schematic diagram of the internal layout of the integrated mobile housing provided in the embodiments of this application; Figure 5 This is a schematic diagram of the substation flood multi-source fusion early warning device provided in the embodiments of this application. Detailed Implementation

[0012] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of this application. However, those skilled in the art will understand that this application may also be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods have been omitted so as not to obscure the description of this application with unnecessary detail.

[0013] To make the objectives, technical solutions, and advantages of this application clearer, the following description will be provided in conjunction with the accompanying drawings and specific embodiments.

[0014] Figure 1 The implementation flowchart of the substation flood multi-source fusion early warning method provided in the embodiments of this application is described in detail below: In addressing the flood control monitoring needs of substations, existing technical solutions have certain limitations in practical applications. First, fixed monitoring devices typically rely on prior infrastructure construction, and their deployment involves civil engineering, power supply access, and communication line laying, resulting in a long installation cycle. In scenarios requiring rapid deployment of monitoring equipment, such as temporary construction areas, renovation areas of old substations, or emergency rescue operations, it is difficult to achieve timely monitoring coverage.

[0015] Secondly, existing monitoring equipment mostly relies on mains power or conventional energy storage batteries for power. When continuous rainfall leads to insufficient photovoltaic power, or under low temperature conditions, the effective capacity of the batteries may decrease significantly, thus affecting the continuous working capability of the equipment. Under extreme weather conditions, the monitoring equipment may experience insufficient power supply.

[0016] To address the shortcomings of existing monitoring equipment in terms of deployment flexibility, power supply stability, and adaptability to complex environments, this application also designs a portable flood early warning device for substations, referring to... Figure 2 , Figure 3 and Figure 4 As shown, the device includes an integrated mobile housing 1, a three-source redundant sensing unit (pressure-type water level acquisition subunit 10, physical capacitor electrode water gauge 11, and video acquisition subunit 12), an autonomous power supply and temperature control scheduling unit 20, a hardware-isolated 4G communication unit 25, and a containerized edge intelligent platform 30.

[0017] Specifically, for the integrated mobile housing 1: the integrated mobile housing 1 serves as the physical carrier of the entire machine, and its external dimensions are precisely defined as follows: This specification is designed to meet the mobile deployment requirements of typical cable trenches and narrow inspection passages in substations. The enclosure is made of aerospace-grade aluminum alloy. The outer surface of the enclosure is coated with a layer of [thickness missing] using thermal spraying technology. to The UV-resistant fluorocarbon coating not only improves the chemical stability of the equipment in long-term outdoor high-light environments, but also reduces the adhesion rate of water mist and dust through its superhydrophobic properties.

[0018] Secondly, regarding the physical stability design of the integrated mobile enclosure 1, a 15 kg steel counterweight 2 is rigidly fixed to the bottom of the enclosure via a cross-shaped reinforcing rib structure. The center of gravity of this steel counterweight 2 has been precisely calibrated to ensure that when the equipment is in a waterlogged environment, the physical center of gravity of the entire unit is always at least 5 cm below the center of buoyancy. For example, when the water level rises to 25 cm, although the enclosure itself will generate upward buoyancy, the equipment as a whole can still maintain a vertical and stable state due to the balance of the high-density steel counterweight at the bottom, thereby avoiding monitoring interruption caused by floating or overturning.

[0019] In addition, as a key protection indicator in this embodiment, the overall protection level of the integrated mobile enclosure 1 is set to IP54. To achieve this level, a D-shaped waterproof sealing rubber ring 3 is embedded at the joint surface between the enclosure cover and the main body. This rubber ring is made of modified EPDM rubber and provides continuous compressive stress in the locked state. Meanwhile, considering the heat dissipation requirements of the internal electronic modules, two sets of louvered heat dissipation vents 4 are symmetrically arranged on the side wall of the enclosure. Each vent is fitted with a layer of stainless steel insect-proof mesh 5. The mesh count of this stainless steel insect-proof mesh 5 is limited to 40 meshes, corresponding to an aperture of 0.5 mm, which ensures air convection efficiency while physically blocking small insects and raindrops splashing from within the substation. At the bottom of the housing, four shock-absorbing and noise-reducing rubber wheels are symmetrically installed. Each rubber wheel integrates a mechanical locking mechanism based on the principle of eccentric cam. The operator only needs to step on it with one foot to complete the dual locking of the radial rotation axis and the horizontal steering axis of the wheel, ensuring that the equipment position will not shift by centimeters under the tilting shear force generated by heavy rain.

[0020] At the top of the integrated mobile housing 1, a three-section pneumatic lifting rod 7 is connected via a three-axis adjustable gimbal 8 with three degrees of freedom. This pneumatic lifting rod 7 utilizes pre-filled high-pressure nitrogen for lifting power, with a maximum lifting height set at 2.5 meters. By adjusting the height of the pneumatic lifting rod 7, the video acquisition subunit 12 at the top can obtain the optimal overhead view, thus fully covering the physical landmarks of the area to be measured.

[0021] For the three-source redundant sensing unit: The three-source redundant sensing unit in this application embodiment is the core hardware architecture for realizing high-precision water level sensing, specifically composed of a pressure-type water level acquisition subunit 10, a physical capacitive electrode water gauge 11, and a video acquisition subunit 12.

[0022] The core component of the pressure-type water level acquisition subunit 10 is the diffused silicon pressure sensor 13 (located inside the pressure-type water level acquisition subunit 10). Figures 2-4 (Not specified in the text) The diffused silicon pressure sensor has an internal temperature compensation circuit, and its range is set from 0 to 5 meters, with a comprehensive measurement accuracy of 0.1% of full scale. The original electrical signal (i.e., voltage signal) generated by the diffused silicon pressure sensor is electrically isolated by an opto-isolator 14, converted into a signed digital signal, and then connected to the back-end processing system. The isolation voltage level of the opto-isolator 14 is not less than 2500V RMS, thereby effectively cutting off the impact of low potential rise in the substation on the core logic board.

[0023] The physical capacitive electrode water level gauge 11, serving as the physical calibration reference in this embodiment, is vertically fixed beside the cable trench or main transformer foundation in the area to be measured. Its substrate is made of 316L stainless steel, possessing extremely strong electrochemical corrosion resistance. High-contrast physical graduation lines are fabricated on the front side of the substrate using laser etching. Inside the substrate, multiple capacitive sensing electrodes 15, spaced 5 mm apart, are arrayed vertically. Each capacitive sensing electrode 15 is covered with a 0.1 mm thick polytetrafluoroethylene (PTFE) film 16. In this embodiment, the PTFE film 16 utilizes its extremely low surface energy to prevent false triggering of capacitive sensing due to droplet adhesion after water level recedes, ensuring real-time measurement during the liquid level descent process.

[0024] For the video acquisition subunit 12: The video acquisition subunit 12 includes an 8-megapixel, 4K resolution dual-light full-color camera 17. This camera supports Power over Ethernet (PoE) protocol, enabling single-cable transmission of video signals and power. The camera integrates an infrared supplementary light array 18 with a center wavelength of 850 nanometers, and a white light illuminator 19 with a color rendering index of not less than 90 and a color temperature of 5500 Kelvin. In low-light conditions or when visibility is reduced due to heavy rainfall, the central control system will automatically retrieve parameters from the illuminance sensor to trigger the white light illuminator 19 to enhance the saliency of the physical capacitive electrode water level gauge 11 in the video frame.

[0025] For the autonomous power supply and temperature control scheduling unit 20: Energy assurance in this embodiment is achieved by the autonomous power supply and temperature control scheduling unit 20. This unit includes two parallel 12V, 20Ah nanoscale colloidal energy storage batteries 21. These batteries use fumed silica as the electrolyte and possess excellent low-temperature charge and discharge efficiency. The energy replenishment terminal is a flexible thin-film solar module 22 installed on the outside of the casing cover. The electrical energy generated by this module is converted by a maximum power point tracking photovoltaic controller 23. This photovoltaic controller 23 employs a high-efficiency interleaved parallel Buck circuit topology, with a measured conversion efficiency of 97.6%.

[0026] Considering the frigid operating conditions of substations in northern China and high-altitude areas, this embodiment of the application attaches a layer of polyimide electric heating film 24 to the outer wall of the colloidal energy storage battery 21. When the containerized edge intelligent platform 30 detects that the internal ambient temperature is below 0 degrees Celsius, it will prioritize allocating redundant power generated by photovoltaics to drive the heating film, and utilize the battery's own constant temperature logic to ensure the activity of the electrolyte.

[0027] For the hardware-isolated 4G communication unit 25: The hardware-isolated 4G communication unit 25 reconstructs communication reliability at the physical layer. The RS485 isolation circuit 26 integrates three independent transient voltage suppression diode arrays 29, with a response time controlled to within 1 nanosecond, effectively clamping nanometer-level transient common-mode voltages generated by the operation of high-voltage disconnect switches in substations. More importantly, it is configured with physically isolated dual gigabit Ethernet ports. The first port 27 is dedicated to receiving large-volume data streams from the video acquisition subunit 12 via an independent PoE switch chip, while the second port 28 is independently connected to the industrial-grade wireless communication terminal. This physical-layer bandwidth and link isolation design prevents interference from high-definition video streams to critical service warning signaling during sudden congestion.

[0028] For the containerized edge intelligence platform 30: The algorithm scheduling center in this embodiment is the containerized edge intelligence platform 30. This platform has a built-in industrial-grade processor with a core frequency of 1.6GHz, 16GB of memory, and a 256GB solid-state drive, and is externally equipped with a dedicated neural network hardware acceleration engine (Neural Processing Unit, NPU). The platform runs an embedded Linux system based on a streamlined kernel, and all business logic runs in isolation as container images. This containerized architecture supports remote hot updates based on digital signature verification, enabling maintenance personnel to issue customized water level risk mapping models in real time according to the equipment height characteristics of substations at different voltage levels.

[0029] This application proposes a multi-source fusion early warning method for substation flooding, applied to the aforementioned designed portable substation flooding early warning device. The method specifically includes: In step 101, the water level height and physical contact height of the target substation are obtained, and video frames containing the physical capacitor electrode water gauge are captured in real time.

[0030] In this embodiment of the application, the water level height and physical contact height values ​​of the target substation are simultaneously collected by the three-source redundant sensing unit in the device, and video frames containing the physical capacitor electrode water gauge 11 are captured at the same time.

[0031] In one possible implementation, obtaining the water level height and physical contact height of the target substation may include: The pressure-type water level acquisition subunit is used to acquire the purified water pressure at a preset unit frequency and convert the purified water pressure into a water level height value. The dielectric constant is collected using a physical capacitance electrode water gauge, and the physical contact height is determined based on the dielectric constant.

[0032] Optionally, firstly, the pressure of the target substation is collected by the pressure-type water level acquisition subunit 10 at a preset unit frequency (e.g., 1 Hz), and then the water pressure is converted into a water level height value.

[0033] In one possible implementation, a pressure-type water level acquisition subunit is used to acquire the purified water pressure at a preset unit frequency and convert the purified water pressure into a water level height value, which may include: The purified water pressure is converted into a voltage signal using a diffused silicon pressure sensor. After the voltage signal is transmitted to the containerized edge intelligence platform through the opto-isolator, the water level height value is calculated.

[0034] Optionally, the pressure-type water level acquisition subunit 10 uses a diffused silicon pressure sensor 13 to sense the clean water pressure at the bottom of the water body, and then uses a Wheatstone bridge inside the diffused silicon pressure sensor 13 to convert the sensed clean water pressure into a weak voltage signal. Due to the large temperature difference in the substation, the temperature compensation circuit integrated inside the diffused silicon pressure sensor 13 corrects zero-point drift in real time, outputting a standardized 4-20mA current signal or an RS485 digital signal. The diffused silicon pressure sensor 13 transmits the converted voltage signal to the containerized edge intelligent platform 30 through an opto-isolator 14 with an isolation voltage of not less than 2500V. The containerized edge intelligent platform 30 triggers a sampling command at a frequency of 1Hz. Upon receiving the voltage signal... Then, the voltage signal is input into the static pressure formula to calculate the water level height value. The static pressure formula is:

[0035] in, This represents the water level height. It is a voltage signal. This is an ambient atmospheric correction value. For water density, This is the acceleration due to gravity.

[0036] Then, the dielectric constant of the capacitance sensing electrode is detected by the physical capacitance electrode water gauge 11, and the physical contact height value is output according to the step-like jump of the dielectric constant.

[0037] In one possible implementation, an array of capacitive sensing electrodes is embedded within the physical capacitive electrode water gauge; the dielectric constant is acquired through the physical capacitive electrode water gauge, and the physical contact height value is determined based on the dielectric constant, which may include: The dielectric constant between multiple sets of adjacent capacitive sensing electrodes is obtained by scanning the array of embedded capacitive sensing electrodes inside the physical capacitive electrode water gauge. If there is a dielectric constant jump among multiple sets of adjacent capacitive sensing electrodes, the number of the last capacitive sensing electrode that experiences a dielectric constant jump is taken as the current water level boundary. The physical contact height value is determined by combining the current water level boundary with the physical elevation scale of each capacitive sensing electrode.

[0038] Optionally, the physical capacitive electrode water gauge 11 determines the precise physical location of water level contact by detecting changes in dielectric constant. Specifically, the arrayed capacitive sensing electrodes (spaced 5mm apart) embedded within the water gauge are periodically scanned by the containerized edge intelligent platform 30. When the water level is not in contact with the electrodes, the medium between the electrodes is air, and the dielectric constant is approximately 1. When water covers the electrodes, the medium becomes water, and the dielectric constant is approximately 80. The containerized edge intelligent platform 30 determines whether there is a dielectric constant jump by measuring the slope or frequency offset of the charge-discharge curves of each electrode. That is, it searches the electrode state from bottom to top and uses the last electrode that experiences a state jump (i.e., from no load to coverage) as the current water level boundary. Since each electrode corresponds to a unique physical elevation scale, the physical contact height value is directly output. .

[0039] It should be noted that this embodiment also utilizes the superhydrophobicity of the 0.1mm polytetrafluoroethylene film, allowing water to quickly slide off during the water level drop. This ensures that the electrode capacitance value can rapidly return from a high level to an unloaded state, avoiding false high water level readings.

[0040] In one possible implementation, real-time capture of video frames containing the physical capacitive electrode water gauge may include: The video acquisition subunit captures video frames containing the physical capacitance electrode water gauge in real time.

[0041] Optionally, in this embodiment, the video acquisition subunit 12 is responsible for acquiring raw visual data including the physical calibration reference (i.e., the physical capacitance electrode water level gauge 11). Specifically, the containerized edge intelligent platform 30 monitors the photoresistor parameters built into the camera 17 in real time. When visibility is extremely low due to darkness or heavy rain, the system instructs the infrared fill light array 18 (for infrared mode) or the white light illumination lamp 19 (for full-color mode) to ensure that the laser scale on the surface of the physical capacitance electrode water level gauge 11 is clearly visible. Then, the camera continuously captures a 4K full-color video stream, i.e., video frames including the physical capacitance electrode water level gauge 11, at a frame rate of 25fps (i.e., one frame every 40ms).

[0042] The containerized edge intelligence platform 30 does not process all redundant frames, but instead captures a high-definition still image of the current moment as the raw material for feature extraction based on the 1Hz system heartbeat cycle.

[0043] In one possible implementation, after acquiring the water level and physical contact height values ​​of the target substation and capturing video frames containing the physical capacitive electrode water gauge in real time, the method may further include: To eliminate the impact of response delays and transmission losses from different sensors on the fusion results, this application also employs a "timestamp marking and alignment" technique based on a high-precision system clock (RTC), namely: The containerized edge intelligence platform 30 is equipped with a high-precision hardware timer. At the beginning of each acquisition cycle, the system issues a global broadcast interrupt signal.

[0044] Data mounting and caching: The water level height value, with a timestamp Tk, is stored in a memory buffer. The physical contact height value with a timestamp Tk completes the logical locking; The video frames extracted by the video acquisition subunit 12, with timestamps Tk, are stored in the neural network input queue.

[0045] The system sets an alignment window threshold (e.g.) Only the three data streams whose time stamp errors are within the threshold range will be selected.

[0046] If a certain path (such as video stream delay) prevents data registration from being completed within the predetermined window, the system will automatically mark that period as a "single point of failure" and rely on the other two data streams to perform emergency judgments to ensure the continuity of the monitoring logic.

[0047] In step 102, the visual water level value is determined using video frames containing the physical capacitance electrode water gauge.

[0048] In this embodiment, the containerized edge intelligence platform 30 is used to process video frames containing physical capacitor electrodes 11 captured in real time by the three-source redundant sensing unit to obtain visual water level values.

[0049] In one possible implementation, the containerized edge intelligence platform includes an object detection model built on a YOLOv8 network and a semantic segmentation model built on a U-Net architecture; it uses video frames containing physical capacitive electrode water gauges to determine visual water level values, including: Input the video frame into the target detection model and output the scale position of the physical capacitor electrode water gauge in the video frame; The video frame with the determined scale position of the physical capacitor electrode water gauge in the video frame is input into the semantic segmentation model to identify the water surface pixel line in the video frame. Calculate the physical resolution per unit pixel by using the scale position of the physical capacitor electrode water gauge in the video frame; The visual water level value is determined based on the scale position of the physical capacitive electrode water gauge in the video frame, the water surface pixel line, and the physical resolution per unit pixel.

[0050] Optionally, in this embodiment, the containerized edge intelligence platform 30 calls a target detection model built on a YOLOv8 network to identify the scale position of the physical capacitive electrode water gauge 11 in the video frame, and calls a semantic segmentation model built on a U-Net architecture to perform semantic segmentation on the water area in the video frame. The visual water level value is calculated based on the pixel mapping relationship between the segmentation edge and the scale position. .

[0051] This application embodiment utilizes the neural network hardware acceleration engine built into the containerized edge intelligence platform 30 to perform parallel semantic parsing on the high-definition video frames output by the video acquisition subunit 12, namely: The containerized edge intelligence platform 30 first preprocesses the captured key video frame images (such as histogram equalization to enhance contrast in rainy and foggy weather), and then inputs them into the target detection model pre-loaded in the container.

[0052] The target detection model extracts features of the physical capacitor electrode water gauge 11 at different scales through its feature pyramid network, and can accurately identify the overall bounding box of the physical capacitor electrode water gauge 11.

[0053] The target detection model not only identifies the main body of the water gauge but also extracts key "feature anchor points" on it, namely the laser-etched physical scale lines. The system automatically locks onto the set of scale marks at known heights closest to the water surface (e.g., 30cm, 50cm, and 100cm scale lines) and outputs their centroid coordinates in the pixel coordinate system. .

[0054] By setting a confidence threshold of 0.8 or higher, suspected interference caused by rain splashes and floating debris is filtered out, ensuring that the reference coordinates for subsequent calculations are absolutely reliable.

[0055] While locking the water gauge's scale position, the system calls the semantic segmentation model in parallel to process the same frame of image, aiming to isolate the water area from the complex substation background.

[0056] The semantic segmentation model employs a symmetric encoder-decoder structure. The encoder extracts high-dimensional features such as water ripples, reflections, and colors; the decoder restores spatial resolution through skip connections, enabling binary classification of each pixel (water vs. non-water).

[0057] The semantic segmentation model outputs a mask image, where pixels in the water area have a value of 1, and the rest have a value of 0. The system uses edge detection operators (such as Canny) to extract the upper edges of the connected regions where the water body contacts the air / water level.

[0058] To address the potential for momentary image jitter caused by substation switch operations, the system compares the segmentation results of three consecutive frames and uses median filtering to remove transient edge spikes caused by wave fluctuations, resulting in smooth and stable water surface pixel lines. .

[0059] Then, a "non-linear mapping" calibration from pixel to physical space is performed. This is the core step in the embodiments of this application to solve visual errors, aiming to eliminate measurement deviations caused by camera pitch and lens distortion.

[0060] The distance in the pixel coordinate system between two known scale lines (such as the 50cm and 100cm scales) identified by the object detection model. Calculate the physical resolution R per pixel (unit: mm / pixel) under the current operating conditions:

[0061] in, The fixed physical distance between known scale graduations (e.g., 500mm). The scale is 50cm. The scale is 100cm.

[0062] By combining the pitch angle parameters of the enclosure measured by the built-in six-axis inertial measurement unit, the pixel coordinates are geometrically corrected using a projection transformation matrix to ensure that even if the device is deployed on uneven ground, the unit resolution R can still accurately reflect the vertical height.

[0063] Finally, using the scale of a known anchor point that has been identified. Using a 100cm scale line as a reference, calculate the water surface pixel lines. With the pixel coordinates of the anchor point vanchor difference .

[0064] The visual water level value was calculated using the height conversion formula. ,Right now:

[0065] The final output of this embodiment is the visual water level value. It also includes an initial weighting coefficient based on the model prediction confidence level, which will be used for subsequent dynamic confidence level allocation.

[0066] In this embodiment, the calculation of water level height does not rely on a preset fixed scale, but instead uses a physical capacitive electrode water gauge 11 as a "dynamic scale". Even if the height of the pneumatic lifting rod 7 changes, or the device is moved to a different distance, the system can achieve self-calibration in each frame by recalculating the pixel resolution R, thereby ensuring that the water level monitoring of the escort substation achieves an extremely high accuracy of ±1.8cm.

[0067] In step 103, the comprehensive integrated water level of the target substation is calculated based on the water level height, physical contact height, and visual water level. The risk level of the target substation is then determined according to the comprehensive integrated water level and the preset equipment height mapping table.

[0068] In this embodiment of the application, the comprehensive fusion water level of the target substation is calculated using the water level height value, physical contact height value and visual water level value obtained in steps 101 and 102. Then, the containerized edge intelligent platform 30 matches the comprehensive fusion water level with the preset equipment height mapping table to generate the corresponding risk level.

[0069] In one possible implementation, the comprehensive integrated water level of the target substation is calculated based on the water level height, physical contact height, and visual water level, and may include: Input the water level height, physical contact height, and visual water level value into the first formula to calculate the comprehensive integrated water level. The first formula is:

[0070] in, To comprehensively integrate water levels, the estimated value of the water level is the one that most closely approximates the real-time physical state of the monitoring point after heterogeneous redundancy verification. This represents the water level height. This is the physical contact height value. Visual water level value. The confidence level of the pressure-type water level acquisition subunit. The confidence level of the physical capacitance electrode water gauge. The confidence level of the video acquisition subunit.

[0071] in, , , The confidence level is dynamically assigned and adjusted, specifically as follows: In this embodiment, the confidence coefficient of each sensing source (i.e., the three-source redundant sensing unit) is calculated in real time. For the video acquisition subunit 12, the Laplace variance of the video frame is calculated. If the variance value is less than 100, it is determined that the current ambient illumination is insufficient or there is rain or fog obstruction, and the confidence coefficient of the visual acquisition subunit is adjusted accordingly. The confidence level is set to 0.2; if the proportion of the obstructed area on the surface of the physical capacitance electrode water gauge 11 exceeds 50%, the confidence level of the physical capacitance electrode water gauge 11 will be adjusted. Set to 0.1; under normal operating conditions, set the confidence levels as follows: It is 0.9. It is 0.8. It is 0.7.

[0072] in, Setting it to 0.9 is based on the principle of Newtonian hydrostatic pressure. As long as it doesn't break, its continuity and linearity are the best, so it is used as the first benchmark. The value of 0.8 is based on the dielectric constant jump, which is less affected by water pollution, but it is limited by the 5mm discrete spacing and has an inherent rounding error, so it is ranked second. The value of 0.7 is based on visual feature extraction. It is most affected by lighting, rain, lens moisture, and algorithm jitter, and has the highest volatility. Therefore, it is used as an auxiliary reference.

[0073] Among them, the confidence level of the pressure-type water level acquisition subunit 10 The judgment logic is as follows: While pressure sensors are generally stable, they are most vulnerable to zero-point drift caused by siltation and reading oscillations caused by dynamic water flow in flood-prone environments. Therefore, the confidence level in this embodiment is... It cannot be static; its judgment logic is usually based on the following two dimensions of "signal health detection": (1) Stability determination based on signal fluctuations (oscillation detection): The system calculates the water level height in real time. Rolling standard deviation over the past 10 seconds (sampling frequency 1Hz) .

[0074] like (e.g., 5cm) indicates that there are violent fluctuations or air bubble interference in the water. At this time, the pressure signal no longer reflects pure static water pressure. The value was linearly reduced from the initial 0.9 to 0.4.

[0075] (2) Self-check judgment based on range boundary (failure detection): The system monitors the output current of the diffused silicon pressure sensor 13.

[0076] If the current signal is below 3.8mA (open circuit) or above 20.2mA (over-range / short circuit), the sensor hardware is considered faulty, and the circuit should be directly switched off. Force it to be set to 0 to completely remove the source from the fusion formula.

[0077] (3) Judgment based on multi-source trend bias (siltation detection): System Comparison and , The long-term mean deviation.

[0078] If within a period of up to 1 hour... The readings continued to be longer than If the elevation of the (capacitive contact type) sensor is more than 5cm and there is no trend of rainfall, it is determined that the sensor has accumulated silt. Adjust to 0.3.

[0079] To further enhance robustness, this application also introduces a residual verification mechanism, which performs real-time comparisons. , , If, relative to the deviation from the overall mean, the residual of a sensor reading exceeds three times the global standard deviation, the sensor is deemed to have hardware damage or sporadic physical interference (such as a spider web obscuring the lens or the sensor being entangled in a plastic bag). The system dynamically sets the weight of the sensor to zero and generates a sensor sub-health report. The specific process is as follows: Suppose that at the current moment, the system has collected three data points: (Pressure type) (Capacitive type) (Video recognition).

[0080] Calculate the arithmetic mean :

[0081] Calculate the "residuals" of each sensing source. The residual refers to the deviation between the value measured by each sensor and the collective average, i.e.: The residual: .

[0082] The residual: .

[0083] The residual: .

[0084] Calculate the standard deviation of this set of data. :

[0085] Suppose the business scenario involves a sudden downpour at a substation, and the video camera is blocked by a plastic bag.

[0086] at this time, (Pressure is normal) (Capacitor is normal) (The video detected the edge of the plastic bag, error).

[0087] but: Mean: .

[0088] Residual: ; ; .

[0089] Standard deviation: .but .

[0090] at this time Although it's large, it doesn't exceed... .

[0091] Once the judgment is triggered, the system considers... Due to physical interference, .

[0092] The corresponding dynamic confidence adjustment logic is as follows:

[0093] in, This is the weight attenuation coefficient, which can be adjusted based on the residual, oscillation ratio, or image sharpness.

[0094] In one possible implementation, the risk levels include Level 1 Attention, Level 2 Early Warning, Level 3 Alarm, and Emergency. Determining the risk level of the target substation based on a comprehensive mapping table of water level and pre-set equipment height may include: If the overall water level is lower than the first water level, the risk level of the target substation will be determined as Level 1 concern. If the overall water level is greater than or equal to the first water level and less than the second water level, the risk level of the target substation will be determined as a level two early warning state. If the overall water level is greater than or equal to the second water level and less than the third water level, the risk level of the target substation will be determined as a level three alarm state. If the overall water level is greater than or equal to the third water level, the risk level of the target substation will be determined as an emergency state.

[0095] Optionally, assuming the first water level is 0.3m, the second water level is 0.5m, and the third water level is 1.2m, this embodiment of the application uses the containerized edge intelligence platform 30 to comprehensively integrate the water levels. The risk level is generated by matching the data with a preset equipment height mapping table. when When this is detected, it is determined to be a Level 1 concern state, and the system executes local time-series data storage on a minute-by-minute basis; when When the system detects a level 2 warning, the wireless communication module pushes a warning to the mobile app and increases the sensor sampling frequency to 5 Hz. At this time, because this height typically corresponds to the edge of the opening in a typical cable trench at a power station, the system determines it to be in a level three alarm state. At this point, the local audible and visual alarm is immediately activated on the edge side, and a real-time video pop-up is pushed to the central monitoring platform via a wireless link; when... At that time, the water level had approached the main transformer's heat sink and core foundation. The system determined that it was in an emergency and immediately invoked the encrypted security protocol to directly issue a shutdown recommendation instruction containing characteristic evidence to the dispatch center.

[0096] In one possible implementation, the method may further include: The system reads the remaining percentage of charge in the gel energy storage battery 21 in real time. When the charge drops to 30%, a first-level power limit is implemented, disabling the real-time preview of the video stream and retaining only the intermittently captured static frames, compressing the frame rate to 5 frames per second. When the charge further drops below 15%, the system executes underlying hard-cutoff logic via an electromagnetic relay, shutting down the high-power video acquisition module 12 and the edge inference core, retaining only the microampere-level pressure gauge acquisition circuit and the low-power communication link based on narrowband Internet of Things (NB-IoT). Simultaneously, in conjunction with the constant temperature strategy of the polyimide electric heating film 24, the internal resistance of the battery remains stable under extremely cold conditions.

[0097] In one possible implementation, the method may further include: During the period when wireless communication links are interrupted due to a disaster, the edge computing side independently performs the entire process of risk assessment and alarms using its local computing power. All alarm logs, water level trend curves, and video clips of critical moments are written to a 256-gigabyte solid-state drive in real time. Once the restoration of 4G / 5G signal is detected, the system automatically initiates a timestamp-based incremental retransmission mechanism to ensure the data integrity of the management center database.

[0098] In one possible implementation, before calculating the integrated water level of the target substation based on the water level height, physical contact height, and visual water level, the method may further include: The system utilizes a built-in six-axis inertial measurement unit in real time. This unit monitors the pitch and roll angles of the integrated mobile housing 1. If tilting of the housing is detected due to uneven ground or localized foundation settlement caused by heavy rain, the system automatically establishes a three-dimensional projection transformation matrix to perform geometric compensation correction on the water level readings identified in the video.

[0099] This application embodiment achieves extremely high rapid deployment capability and scenario adaptability. Through the integrated mobile housing 1, combined with steel counterweights and pneumatic lifting rods, a single operator can complete the equipment deployment within 5 minutes without any civil construction or mains power connection.

[0100] Furthermore, by constructing a highly reliable, zero-false-alarm sensing system, this application's embodiments abandon the traditional single-source water level detection method and adopt a three-source redundancy architecture of pressure, capacitive electrodes, and visual recognition. A dynamic confidence fusion algorithm based on Laplace variance and occlusion rate detection is also introduced. Real-world test data shows that even in extreme environments such as siltation and dense fog interference, this algorithm can still control the water level identification error within ±1.8 cm, with a measured false-alarm rate of zero, significantly improving the authority of the early warning system.

[0101] Secondly, the embodiments of this application possess exceptionally long survivability in extreme environments. Utilizing the low-temperature charging and discharging characteristics of gel energy storage batteries, combined with a polyimide electric heating temperature control strategy and a graded power scheduling mechanism, the embodiments of this application can maintain uninterrupted operation of critical water level monitoring functions for over 51 hours in extremely cold environments of -20 degrees Celsius and in situations where prolonged overcast and rainy weather results in no solar power replenishment, ensuring that the substation survives the most dangerous disaster window period.

[0102] Then, a robust design was implemented specifically for the electromagnetic environment unique to substations. Through the combined effects of RS485 opto-isolation, dual gigabit Ethernet physical isolation, and a TVS transient suppression array, this device successfully passed the Level 4 electrical fast transient / burst immunity test under the GB / T17626.4 standard, effectively preventing electromagnetic interference generated by switching operations from damaging the communication link.

[0103] Simultaneously, it achieves intelligent closed-loop at the edge. This device does not rely on logic delivery from cloud servers. Even in the event of network outages caused by disasters, it can still perform independent risk level assessments and trigger local audible and visual alarms through locally running YOLOv8 and U-Net container images, ensuring the absolute integrity of the emergency response chain under extreme conditions.

[0104] Taking the flood control application in the renovation area of ​​a certain 110kV old substation as an example, the feasibility of the embodiments of this application is verified.

[0105] The equipment in this embodiment was deployed in the low-lying cable trench construction area of ​​the site. This area lacked mains power access and was located within the high-frequency electromagnetic radiation zone of the expanding substation. During the experiment, it encountered historically rare short-duration heavy rainfall, with the maximum water depth reaching 60 centimeters.

[0106] The deployment process was completed by a single maintenance team member. This person pushed the integrated mobile unit 1 to the designated position, locked the wheels using the mechanical locking mechanism 6, and deployed the pneumatic lifting rod 7. The entire deployment, power-on, and heartbeat docking process with the remote platform took only 4 minutes and 20 seconds, fully demonstrating the device's portability and rapid deployment capabilities.

[0107] During the monitoring process, due to the large amount of sediment carried by the heavy rain, the probe of the pressure-type water level acquisition subunit 10 experienced partial siltation. At this time, The reading showed a zero-point drift of approximately 3.5 cm. Simultaneously, due to the dense rain, the image quality of the video stream degraded, and the Laplace variance dropped to 82. At this point, the system adjusted the confidence level of the visual acquisition subunit. The value was lowered to 0.2, and the weighting coefficient of the physical capacitance electrode water gauge 11 was increased due to the pressure gauge residual test triggering the calculation. The comprehensive fused water level was obtained through multi-source fusion calculation. Compared with the benchmark water level measured manually using a total station, the error is only 1.2 centimeters, which fully meets the engineering requirements for power flood control.

[0108] Regarding communication stability, the main transformer underwent two switching operations during the experiment. Due to the clamping effect of the RS485 isolation circuit 26 of the hardware isolation communication module 25 and the TVS array 29, the system's water level message sampling curve did not show any abnormal jumps or disconnections / reconnections. In contrast, the conventional RS485 water level gauge (without physical isolation design), deployed as a comparison, experienced a severe communication timeout at the moment of switching, with a disconnection duration of up to 3 minutes.

[0109] In addition, during the 72nd hour of continuous rain and no solar power replenishment, the device successfully implemented the secondary power consumption limit strategy, shutting down video streaming. Relying on the NB-IoT link, even with only 8% battery power remaining, it still accurately captured the critical alarm information that the water level had exceeded the height of the cable trench.

[0110] To further demonstrate the non-obviousness of this embodiment, Table 1 lists a comparison of key indicators between this embodiment and the prior art: Table 1. Performance Comparison between the Equipment in this Embodiment and Traditional Monitoring Equipment

[0111] As can be seen from Table 1, this embodiment, through the systematic integration of mechanics, sensing, power supply and edge algorithms, not only improves deployment efficiency by nearly an order of magnitude, but also demonstrates superior performance in perception accuracy and robustness under extreme conditions.

[0112] This application provides a multi-source fusion early warning method for flooding in substations. It acquires data from three different sources: water level height, physical contact height, and visual water level, and fuses these data to obtain a comprehensive fusion water level. The multi-source data complements and verifies each other, effectively reducing the errors and uncertainties that may exist with a single data source, thereby significantly improving the accuracy of substation water level monitoring and providing a reliable basis for accurate flood disaster early warning. Secondly, it determines the risk level of the target substation based on a mapping table between the comprehensive fusion water level and a preset equipment height. This determination method considers the impact of multiple factors on substation safety, which is superior to relying solely on a single water level data to determine risk. Compared to other methods, this approach is more scientific and comprehensive. It can not only more accurately assess the actual risk level of substations facing floods, but also buy valuable response time for substation maintenance personnel, effectively reducing the impact of floods on substation equipment and power system operation, and improving the reliability and stability of the power system. At the same time, this application helps to improve the flood disaster emergency early warning system of the power system, enhance the emergency response capability and handling efficiency of the power system in the face of floods, and enable power system maintenance departments to respond to floods more proactively through real-time monitoring and accurate early warning, rationally allocate emergency resources and manpower, and improve the disaster resistance capability and operational stability of the entire power system.

[0113] It should be understood that the sequence number of each step in the above embodiments does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0114] The following are device embodiments of this application. For details not described in detail, please refer to the corresponding method embodiments described above.

[0115] Figure 5A schematic diagram of the substation flood multi-source fusion early warning device provided in an embodiment of this application is shown. For ease of explanation, only the parts related to the embodiment of this application are shown, and are described in detail below: like Figure 5 As shown, the substation flood multi-source integrated early warning device 5 includes: The acquisition module 51 is used to acquire the water level height value and physical contact height value of the target substation, and to capture video frames containing the physical capacitance electrode water gauge in real time. Module 52 is used to determine the visual water level value using video frames containing physical capacitive electrode water gauges. The risk level determination module 53 is used to calculate the comprehensive integrated water level of the target substation based on the water level height value, physical contact height value and visual water level value, and to determine the risk level of the target substation according to the comprehensive integrated water level and the preset equipment height mapping table.

[0116] This application provides a multi-source fusion early warning device for substation flooding. It acquires data from three different sources: water level height, physical contact height, and visual water level, and fuses these data to obtain a comprehensive fusion water level. The multi-source data complements and verifies each other, effectively reducing the errors and uncertainties that may exist with a single data source, thereby significantly improving the accuracy of substation water level monitoring and providing a reliable basis for accurate flood disaster early warning. Secondly, it determines the risk level of the target substation based on a mapping table between the comprehensive fusion water level and a preset equipment height. This determination method considers the impact of multiple factors on substation safety, compared to relying solely on a single water level data to determine risk. Compared to other methods, this approach is more scientific and comprehensive. It can not only more accurately assess the actual risk level of substations facing floods, but also buy valuable response time for substation maintenance personnel, effectively reducing the impact of floods on substation equipment and power system operation, and improving the reliability and stability of the power system. At the same time, this application helps to improve the flood disaster emergency early warning system of the power system, enhance the emergency response capability and handling efficiency of the power system in the face of floods, and enable power system maintenance departments to respond to floods more proactively through real-time monitoring and accurate early warning, rationally allocate emergency resources and manpower, and improve the disaster resistance capability and operational stability of the entire power system.

[0117] In one possible implementation, the three-source redundant sensing unit includes a pressure-type water level acquisition subunit and a physical capacitive electrode water gauge, the physical capacitive electrode water gauge being vertically fixed to the water body area to be measured in the target substation; the acquisition module can be used for: The pressure-type water level acquisition subunit is used to acquire the purified water pressure at a preset unit frequency and convert the purified water pressure into a water level height value. The dielectric constant is collected using a physical capacitance electrode water gauge, and the physical contact height is determined based on the dielectric constant.

[0118] In one possible implementation, the pressure-type water level acquisition subunit includes a diffused silicon pressure sensor and an opto-isolator; the acquisition module can also be used for: The purified water pressure is converted into a voltage signal using a diffused silicon pressure sensor. After the voltage signal is transmitted to the containerized edge intelligence platform through the opto-isolator, the water level height value is calculated.

[0119] In one possible implementation, the acquisition module can also be used for: The voltage signal is input into the static pressure formula to calculate the water level height. The static pressure formula is as follows:

[0120] in, This represents the water level height. It is a voltage signal. This is an ambient atmospheric correction value. For water density, This is the acceleration due to gravity.

[0121] In one possible implementation, the physical capacitive electrode water gauge contains an array of embedded capacitive sensing electrodes; the acquisition module can also be used for: The dielectric constant between multiple sets of adjacent capacitive sensing electrodes is obtained by scanning the array of embedded capacitive sensing electrodes inside the physical capacitive electrode water gauge. If there is a dielectric constant jump among multiple sets of adjacent capacitive sensing electrodes, the number of the last capacitive sensing electrode that experiences a dielectric constant jump is taken as the current water level boundary. The physical contact height value is determined by combining the current water level boundary with the physical elevation scale of each capacitive sensing electrode.

[0122] In one possible implementation, the three-source redundant sensing unit further includes a video acquisition subunit; real-time capture of video frames containing physical capacitance electrodes and water level gauges, including: The video acquisition subunit captures video frames containing the physical capacitance electrode water gauge in real time.

[0123] In one possible implementation, the containerized edge intelligence platform includes an object detection model built on a YOLOv8 network and a semantic segmentation model built on a U-Net architecture; the determination module can be used for: Input the video frame into the target detection model and output the scale position of the physical capacitor electrode water gauge in the video frame; The video frame with the determined scale position of the physical capacitor electrode water gauge in the video frame is input into the semantic segmentation model to identify the water surface pixel line in the video frame. Calculate the physical resolution per unit pixel by using the scale position of the physical capacitor electrode water gauge in the video frame; The visual water level value is determined based on the scale position of the physical capacitive electrode water gauge in the video frame, the water surface pixel line, and the physical resolution per unit pixel.

[0124] In one possible implementation, the level determination module can be used for: Input the water level height, physical contact height, and visual water level value into the first formula to calculate the comprehensive integrated water level. The first formula is:

[0125] in, In order to comprehensively integrate water levels, This represents the water level height. This is the physical contact height value. Visual water level value. The confidence level of the pressure-type water level acquisition subunit. The confidence level of the physical capacitance electrode water gauge. The confidence level of the video acquisition subunit.

[0126] In one possible implementation, the risk levels include Level 1 Attention, Level 2 Warning, Level 3 Alarm, and Emergency; the level determination module can also be used for: If the overall water level is lower than the first water level, the risk level of the target substation will be determined as Level 1 concern. If the overall water level is greater than or equal to the first water level and less than the second water level, the risk level of the target substation will be determined as a level two early warning state. If the overall water level is greater than or equal to the second water level and less than the third water level, the risk level of the target substation will be determined as a level three alarm state. If the overall water level is greater than or equal to the third water level, the risk level of the target substation will be determined as an emergency state.

[0127] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0128] Those skilled in the art will recognize that the templates, units, and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0129] If the module / unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the above embodiments of the multi-source flood early warning method for substations. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, a read-only memory, a random access memory, an electrical carrier signal, a telecommunication signal, and a software distribution medium, etc.

[0130] The above-described embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be included within the protection scope of the present invention.

Claims

1. A multi-source fusion early warning method for flooding in substations, characterized in that, The method is applied to portable flood early warning equipment in substations. The equipment includes an integrated mobile enclosure, a three-source redundant sensing unit, an autonomous power supply and temperature control scheduling unit, a hardware-isolated 4G communication unit, and a containerized edge intelligence platform. The method includes: Acquire the water level and physical contact height of the target substation, and capture video frames containing the physical capacitance electrode water gauge in real time; The visual water level value is determined using the video frame containing the physical capacitance electrode water gauge. Based on the water level height value, the physical contact height value, and the visual water level value, the comprehensive integrated water level of the target substation is calculated, and the risk level of the target substation is determined according to the comprehensive integrated water level and the preset equipment height mapping table.

2. The substation flood multi-source fusion early warning method according to claim 1, characterized in that, The three-source redundant sensing unit includes a pressure-type water level acquisition subunit and a physical capacitive electrode water gauge, wherein the physical capacitive electrode water gauge is vertically fixed to the water body area to be measured in the target substation. The acquisition of the water level height and physical contact height of the target substation includes: The pressure-type water level acquisition subunit is used to acquire the purified water pressure at a preset unit frequency, and the purified water pressure is converted into the water level height value; The dielectric constant is collected using the physical capacitance electrode water gauge, and the physical contact height value is determined based on the dielectric constant.

3. The substation flood multi-source fusion early warning method according to claim 2, characterized in that, The pressure-type water level acquisition subunit includes a diffused silicon pressure sensor and an opto-isolator; the step of using the pressure-type water level acquisition subunit to acquire the purified water pressure at a preset unit frequency and converting the purified water pressure into the water level height value includes: The purified water pressure is converted into a voltage signal by the diffused silicon pressure sensor. After the voltage signal is transmitted to the containerized edge intelligent platform through the opto-isolator, the water level height value is calculated.

4. The substation flood multi-source fusion early warning method according to claim 3, characterized in that, The step of calculating the water level height value after the voltage signal is transmitted to the containerized edge intelligence platform through the opto-isolator includes: The voltage signal is input into the static pressure formula to calculate the water level height. The static pressure formula is as follows: in, The water level height value, The voltage signal, This is an ambient atmospheric correction value. For water density, This is the acceleration due to gravity.

5. The substation flood multi-source fusion early warning method according to claim 2, characterized in that, The physical capacitance electrode water gauge has an array of embedded capacitive sensing electrodes inside; the process of acquiring the dielectric constant through the physical capacitance electrode water gauge and determining the physical contact height value based on the dielectric constant includes: By scanning the array of embedded capacitive sensing electrodes inside the physical capacitive electrode water gauge, the dielectric constant between multiple sets of adjacent capacitive sensing electrodes can be obtained. If there is a dielectric constant jump among multiple sets of adjacent capacitive sensing electrodes, the number of the last capacitive sensing electrode that experiences a dielectric constant jump is taken as the current water level boundary. The physical contact height value is determined by combining the current water level boundary with the physical elevation scale of each capacitive sensing electrode.

6. The substation flood multi-source fusion early warning method according to claim 2, characterized in that, The three-source redundant sensing unit further includes a video acquisition subunit; the real-time capture of video frames containing physical capacitance electrode water gauges includes: The video acquisition subunit captures video frames containing physical capacitance electrodes and water gauges in real time.

7. The substation flood multi-source fusion early warning method according to claim 1, characterized in that, The containerized edge intelligence platform includes a target detection model built on the YOLOv8 network and a semantic segmentation model built on the U-Net architecture; the step of determining the visual water level value using the video frame containing the physical capacitance electrode water gauge includes: The video frame is input into the target detection model, and the scale position of the physical capacitor electrode water gauge in the video frame is output. The video frame in which the scale position of the physical capacitor electrode water gauge in the video frame has been determined is input into the semantic segmentation model to identify the water surface pixel line in the video frame. The physical resolution per unit pixel is calculated using the scale position of the physical capacitor electrode water gauge in the video frame. The visual water level value is determined based on the scale position of the physical capacitive electrode water gauge in the video frame, the water surface pixel line, and the physical resolution per unit pixel.

8. The substation flood multi-source fusion early warning method according to claim 1, characterized in that, The calculation of the comprehensive integrated water level of the target substation based on the water level height value, the physical contact height value, and the visual water level value includes: The water level height value, the physical contact height value, and the visual water level value are input into the first formula to calculate the comprehensive integrated water level. The first formula is: in, The comprehensive water level is the aforementioned integrated water level. The water level height value, The physical contact height value. The visual water level value, The confidence level of the pressure-type water level acquisition subunit. The confidence level of the physical capacitance electrode water gauge. The confidence level of the video acquisition subunit.

9. The substation flood multi-source fusion early warning method according to claim 1, characterized in that, The risk levels include Level 1 Attention, Level 2 Early Warning, Level 3 Alarm, and Emergency. The step of determining the risk level of the target substation based on the integrated water level and the preset equipment height mapping table includes: If the comprehensive water level is lower than the first water level, the risk level of the target substation will be determined as the first level of concern. If the integrated water level is greater than or equal to the first water level and the integrated water level is less than the second water level, then the risk level of the target substation is determined to be the Level II early warning state. If the integrated water level is greater than or equal to the second water level and the integrated water level is less than the third water level, then the risk level of the target substation is determined to be the third-level alarm state. If the comprehensive water level is greater than or equal to the third water level, then the risk level of the target substation is determined to be the emergency state.

10. A substation flood multi-source integrated early warning device, characterized in that, The device is applied to a portable flood early warning system for substations. The device includes an integrated mobile housing, a three-source redundant sensing unit, an autonomous power supply and temperature control scheduling unit, a hardware-isolated 4G communication unit, and a containerized edge intelligence platform. The device comprises: The acquisition module is used to acquire the water level height and physical contact height of the target substation, and to capture video frames containing the physical capacitance electrode water gauge in real time. The determination module is used to determine the visual water level value using the video frame containing the physical capacitance electrode water gauge; The risk level determination module is used to calculate the comprehensive integrated water level of the target substation based on the water level height value, the physical contact height value, and the visual water level value, and to determine the risk level of the target substation according to the comprehensive integrated water level and the preset equipment height mapping table.

Images