High-grade power supply lightning protection system and lightning protection method based on YOLO model

Abstract

The application discloses a high-grade power supply lightning protection system based on a YOLO model, which comprises an image acquisition module, an active lightning protection module, a passive lightning protection module and a control module; wherein the image acquisition module acquires lightning movie images and transmits the images to the active protection module; the active protection module analyzes the lightning movie images based on the YOLO model, identifies and warns the thunderstorm cloud, and sends signals to the control module; the passive protection module receives instructions from the control module, limits overvoltage through a B5-grade lightning protection circuit and discharges overcurrent; and the control module is used for coordinating the work of each module. The application further discloses a lightning protection method using the system. The lightning protection system combines active warning with passive protection, meets the high-grade protection requirement, and has high warning accuracy.

Description

Technical Field

[0001] This invention belongs to the field of lightning protection and early warning technology, specifically relating to a high-level power supply lightning protection system based on the YOLO model. This invention also relates to a method for lightning protection using this system. Background Technology

[0002] Lightning protection for airborne equipment generally adopts passive protection technology, which uses dedicated lightning suppressors to prevent the indirect effects of lightning from having a significant impact on the equipment and ensure the normal operation of the equipment or system. Most lightning protection designs are designed for low to medium levels of lightning indirect effects protection, while high-level lightning indirect effects protection has a low actual application rate due to high testing requirements, high costs, and high technical difficulty.

[0003] Existing lightning warning technologies are mainly divided into three categories: First, warning technologies based on atmospheric electric field monitoring, which typically use atmospheric electric field meters for lightning warnings. These methods are mechanically complex, susceptible to electromagnetic interference, and prone to zero-point drift. Second, warning technologies based on lightning location, which use multi-station sensors to capture lightning electromagnetic pulses and utilize time-difference positioning technology to determine the lightning's location, intensity, and type. This method can only detect lightning that has already occurred and cannot warn of the initial charge accumulation during thunderstorms. Third, multi-source data fusion warning technologies, which require the use of multiple independent instruments to measure different parameters and combine data sharing with meteorological bureaus to improve forecast accuracy. Therefore, these technologies have high measurement costs, complex maintenance and operation, limited independent operation capabilities, and a single system costing tens to hundreds of thousands of yuan, with significant regular maintenance expenses.

[0004] In addition, most existing surge protection circuits are designed for RTCAO-160G standard A3 and A4 levels, adopting a "basic protection + experience design" mode, which is difficult to meet the needs of high-level protection and lacks a linkage mechanism with intelligent early warning, resulting in insufficient protection flexibility and effectiveness. Summary of the Invention

[0005] The purpose of this invention is to provide a high-level power surge protection system based on the YOLO model, which combines active early warning with passive protection to meet high-level protection requirements and has a high early warning accuracy.

[0006] Another object of the present invention is to provide a method for lightning protection using the above-described lightning protection system.

[0007] The technical solution adopted in this invention is:

[0008] A high-level power surge protection system based on the YOLO model includes an image acquisition module, an active surge protection module, a passive surge protection module, and a control module. The image acquisition module acquires lightning images and transmits them to the active protection module. The active protection module analyzes the lightning images based on the YOLO model, identifies and warns of thunderstorm clouds, and sends signals to the control module. The passive protection module receives instructions from the control module and limits overvoltage and discharges overcurrent through B5-level surge protection circuits. The control module coordinates the operation of each module.

[0009] The invention is further characterized by:

[0010] The image acquisition module uses a combination of wide-angle and telephoto dual cameras for collaborative observation, collecting cloud images and video data from near the ground to mid-altitude (400-2000 meters) and high altitude (4000-6000 meters), covering different scenes day and night. Among them, the wide-angle camera is responsible for large-scale initial screening of the vertical development structure of cloud clusters and the diffusion trend of cloud anvils, while the telephoto camera is responsible for focusing on the abrupt changes in cloud anvil spikes and the details of ice crystal diffusion.

[0011] The active lightning protection module uses the YOLO v8 model to preprocess the lightning images transmitted by the image acquisition module, extract cloud features through a convolutional neural network, detect thunderstorm cloud targets using a sliding window, calculate the target bounding box and confidence score, and then filter overlapping target boxes using a non-maximum suppression algorithm to output the final detection result. If a thunderstorm cloud is identified, a warning signal is sent to the control module.

[0012] Image preprocessing includes scaling and normalizing pixel values ​​to adapt them to the model's input requirements. In addition, the YOLO v8 model is trained using Python programming based on the Darknet deep learning network.

[0013] The passive surge protection module includes a B5-level surge protection circuit and a switching circuit. The B5-level surge protection circuit has a three-level combined structure, specifically:

[0014] First stage: Gas discharge tube and varistor are connected in series. When overvoltage occurs, the gas discharge tube conducts first, triggering the varistor to conduct in order to release lightning energy.

[0015] The second stage: inductors, used to weaken the instantaneous lightning spikes, absorb lightning energy, and work with subsequent devices to delay the lightning strike.

[0016] The third stage: diode AK6-380C or AK6-300C, where the clamping voltage of AK6-380C is 520V and the clamping voltage of AK6-300C is 470V, used to further discharge current and precisely clamp the voltage within a safe range.

[0017] The switching circuit is driven by the control module to switch between different clamping voltage diodes in the lightning protection circuit.

[0018] The specific structure of a B5 level surge protection circuit is as follows:

[0019] The positive input terminal of the power supply is connected to the gas discharge tube G1. G1 and the varistor RV1 are connected in series. The other end of RV1 is connected to the ground PE. One end of the inductor L1 is connected to G1, and the other end is connected to the switching circuit. The circuit also includes diodes D1 and D2. The switching circuit is used to switch between diodes D1 and D2. Both D1 and D2 are connected to the PE ground.

[0020] The negative input terminal of the power supply is connected to the gas discharge tube G2. G2 and the varistor RV2 are connected in series. The other end of RV2 is connected to the ground PE. One end of the inductor L2 is connected to G2, and the other end is connected to the switching circuit. The circuit also includes diodes D3 and D4. The switching circuit is used to switch between diodes D3 and D4. Both D3 and D4 are connected to the PE ground.

[0021] Among them, the gas discharge tubes G1 and G2 are model JHM091M, the varistors RV1 and RV2 are model MYG3-20K300, the inductors L1 and L2 are model JSAC1212-6R0M, the diodes D1 and D3 are model AK6-380C, and the diodes D2 and D4 are model AK6-300C.

[0022] The specific structure of the switching circuit is as follows:

[0023] The GPIO1 terminal controlling the positive switching of the power supply is connected to resistor R2. The other end of R2 and GND are connected to the two input terminals of optocoupler U1. The two output terminals of optocoupler U1 are connected to VCC5V and resistor R3 respectively. The other end of R3 is connected to GND. One end of capacitor C1 is connected to R3 and the gate of MOSFET Q1, and the other end is connected to GND. The source of MOSFET Q1 is connected to GND, and the drain is connected to the output terminal of relay KJ1. The power supply input terminal of relay KJ1 is connected to VCC5V. Terminal K1 is connected to the rear end of inductor L1. Terminals K2 and K3 are connected to diodes D1 and D2 respectively.

[0024] The GPIO2 terminal controlling the negative switching of the power supply is connected to resistor R5. The other end of R5 and GND are connected to the two input terminals of optocoupler U2. The two output terminals of optocoupler U2 are connected to VCC5V and resistor R6 respectively. The other end of R6 is connected to GND. One end of capacitor C2 is connected to R6 and the gate of MOSFET Q2, and the other end is connected to GND. The source of MOSFET Q2 is connected to GND, and the drain is connected to the output terminal of relay KJ2. The power supply input terminal of relay KJ2 is connected to VCC5V. Terminal K4 is connected to the back end of inductor L2. Terminals K5 and K6 are connected to diodes D3 and D4 respectively.

[0025] After receiving the early warning signal from the active lightning protection module, the control module triggers an alarm and sends a switching command to the passive lightning protection module. After the thunderstorm cloud passes, the control module resets the circuit of the passive lightning protection module and stores the monitoring data for model optimization.

[0026] Another technical solution adopted in this invention is:

[0027] The method for lightning protection using the above system is as follows:

[0028] After the system starts up, the image acquisition module continuously collects image data of sky clouds and transmits it to the active lightning protection module in real time. The active lightning protection module processes and analyzes the data using the YOLO v8 model to identify thunderstorm cloud characteristics. If a thunderstorm cloud is detected, it outputs a warning signal to the control module. The control module triggers an alarm and sends a command to the passive lightning protection module. After receiving the command, the passive lightning protection module switches its circuit action to discharge overcurrent and clamp the residual voltage to a safe range. After the thunderstorm cloud passes, the control module receives a release signal, controls the passive lightning protection module circuit to reset, and stores the monitoring data in the database for subsequent model optimization.

[0029] The beneficial effects of this invention are:

[0030] (1) Meets high-level protection requirements

[0031] The surge protection circuit of this invention conforms to RTCA / DO-160G standard B5 level, capable of withstanding 1600V voltage and 1600A current surges, with a residual peak voltage of 444V, ensuring that equipment hardware is not damaged, functions are not interrupted, signals are not interfered with, and the system does not fail. Furthermore, the method of this invention adapts to different operating conditions through a switching circuit; during a lightning warning, it switches to a low residual voltage diode, further enhancing the protection effect.

[0032] (2) Reduce deployment and maintenance costs

[0033] The active warning part of the lightning protection system of this invention relies on cameras and YOLO models, which does not require a large number of expensive sensors and has low deployment costs. In addition, the passive lightning protection circuit is designed for the protection of 270V DC power supply, and each protection circuit only uses 4 components, resulting in low raw material costs and more convenient maintenance in the later stage.

[0034] (3) Improve the accuracy and efficiency of early warning

[0035] Existing multi-source early warning systems require separate modeling and processing of data from different sources such as radar, electric fields, and lightning location, resulting in a cumbersome structure. YOLO, as a unified perception model, can transform multiple image features into a unified thunderstorm target detection problem, greatly simplifying the fusion process. Furthermore, combined with deep neural networks, it has good fault tolerance and can compensate for problems caused by missing data. In addition, the YOLO v8 model of this invention is trained on more than 18,000 images, with an early warning response time of 25±10 minutes before lightning occurs and an accuracy of ≥90%.

[0036] (4) Enhance system intelligence and scalability

[0037] The system of this invention supports continuous optimization methods such as transfer learning and online learning, can adapt to the meteorological characteristics of different regions and seasons, and has strong generalization ability; it can store monitoring data for model iteration, has self-evolution ability, and continuously improves early warning performance.

[0038] (5) Expanding the boundaries of technology application

[0039] The method of this invention adopts a comprehensive scheme that combines active early warning and passive protection, realizing the cross-border integration of lightning protection and artificial intelligence algorithms, providing a new approach to intelligent lightning protection. In addition, the system of this invention is highly portable and is suitable for resource-constrained, remote unattended, and extremely high-safety and reliability requirements. Attached Figure Description

[0040] Figure 1 This is a schematic diagram of the lightning protection system of the present invention;

[0041] Figure 2 This is a flowchart of the YOLO v8 model in the lightning protection system of this invention;

[0042] Figure 3 This is a flowchart of the lightning protection circuit in the lightning protection system of the present invention;

[0043] Figure 4 This is a schematic diagram of the lightning protection circuit and switching circuit in the lightning protection system of the present invention;

[0044] Figure 5 This is a schematic diagram of the residual voltage (waveform 5A) after lightning protection of the lightning protection system of the present invention. Detailed Implementation

[0045] The present invention will now be described in detail with reference to the accompanying drawings and specific embodiments.

[0046] Example 1:

[0047] This invention relates to a high-level power surge protection system based on the YOLO model, such as... Figure 1As shown, it includes an image acquisition module, an active lightning protection module, a passive lightning protection module, and a control module. The image acquisition module acquires lightning images and provides stable data source support; the active protection module uses the YOLO model to achieve accurate identification and early warning of thunderstorm clouds; the passive protection module limits overvoltage and discharge current through B5-level lightning protection circuits; the control module coordinates the work of each module; through the comprehensive mode of "active early warning + passive protection", lightning disaster prevention is achieved together.

[0048] Example 2:

[0049] Based on Example 1, the specific structure, function, and working principle of each module are as follows:

[0050] The image acquisition module employs a combination of wide-angle and telephoto lenses for collaborative observation, with complementary data from both cameras ensuring comprehensive monitoring. Specifically, the wide-angle observation unit is responsible for initial screening of the vertical development structure of cloud clusters and the diffusion trend of cloud anvils, using a wide-angle lens to scan the sky; the telephoto tracking unit focuses on the abrupt changes in cloud anvil spikes and the details of ice crystal diffusion.

[0051] The wide-angle camera is a Hikvision DS-2DY9C412IAX-D model, featuring a built-in 4mm F1.0 wide-angle lens with a 142° field of view, covering near-ground to mid-level cloud evolution. It boasts starlight-level night vision, intelligently locking onto fast-moving cumulonimbus clouds, and a horizontal rotation speed of 80° / s to respond to storm development, focusing on monitoring the accumulation of low-altitude clouds at 400-2000 meters. The telephoto camera is a Hikvision iDS-2DYHS455XS-D model, with a 6.8-550mm super telephoto lens (48x optical zoom), penetrating rain and fog to observe cloud top height. It supports day and night collaborative observation, switching to thermal imaging mode at night to monitor thermal convection in the charge separation layer at 4000-6000 meters.

[0052] Example 3:

[0053] Building upon Example 2, the active lightning protection module employs the YOLO (You Only Look Once) algorithm, specifically the YOLO v8 model. YOLO is a single-stage object detection algorithm based on deep learning. Its key feature is that it transforms the object detection task into an end-to-end regression problem. Unlike traditional object detection systems, YOLO uses a single neural network to predict the bounding boxes and class probabilities of objects in the input image. This single-stage object detection algorithm is suitable for scenarios requiring efficient processing and rapid response.

[0054] The specific workflow of the YOLO v8 model is as follows: Figure 2As shown, cloud image data is acquired through an image acquisition module. After image preprocessing, convolutional neural network feature extraction, sliding window target detection, category and location prediction, and non-maximum suppression, thunderstorm cloud characteristics are identified and a warning signal is output. The warning response time is 25 ± 10 minutes before the lightning strike. Details are as follows:

[0055] (1) Data acquisition: Receive cloud images and video data from near-ground to mid-altitude (400~2000 meters) and high-altitude (4000~6000 meters) transmitted by the image acquisition module, covering different scenes day and night.

[0056] (2) Image preprocessing: The input data is scaled and normalized to match the input requirements of the model.

[0057] (3) Feature extraction and target detection: Cloud features are extracted by convolutional neural network, and thunderstorm cloud targets are detected by sliding window. The target bounding box and confidence score are calculated.

[0058] (4) Result filtering and output: Overlapping target boxes are filtered by non-maximum suppression algorithm, and the final detection result is output. If thunderstorm clouds are detected, an early warning signal is sent to the control module.

[0059] The model training parameters are as follows:

[0060] (1) Hardware environment: Based on Windows 10 x64 operating system, configured with AMD Ryzen 9 6900HX CPU, NVIDIA GeForce RTX3060 Laptop 8G GPU and 16GB memory.

[0061] (2) Dataset: It contains more than 18,000 images related to thunderstorm clouds, including more than 5,000 nighttime images. The ratio of training set to validation set is 8:2.

[0062] (3) Training process: Based on the Darknet deep learning network, it is implemented using Python language. The model parameters are optimized through real-time data augmentation (flipping / scaling / Mosaic) and multiple rounds of iterative training. Finally, the threshold and NMS parameters are adjusted through validation set testing to ensure recognition accuracy.

[0063] Example 4:

[0064] Based on Example 3, the passive surge protection module includes a B5-level surge protection circuit and a switching circuit. The B5-level surge protection circuit is designed for a DC 270V power supply, and its operation is as follows: Figure 3 As shown, the circuit structure is as follows Figure 4 As shown, a three-level composite structure is adopted:

[0065] First stage: Gas discharge tube JHM091M and varistor MYG3-20K300 are connected in series. Under normal operating voltage, the gas discharge tube has high insulation resistance. Under overvoltage, it conducts first, and the gap between the two electrodes discharges and breaks down, turning into a conductive state, reducing the voltage to a low residual voltage level, triggering the varistor to conduct, and conducting the lightning energy to the metal casing for discharge.

[0066] The second stage is the JSAC1212-6R0M inductor. On one hand, it weakens the instantaneous lightning spike and absorbs lightning energy; on the other hand, it works with the subsequent diode to delay the lightning strike, ensuring that the varistor activates first and the diode activates later, preventing diode damage.

[0067] The third stage: diodes (AK6-380C or AK6-300C). These further discharge residual voltage and current from the previous stage, precisely clamping the voltage to a safe range. The clamping voltage of AK6-380C is 520V, and the clamping voltage of AK6-300C is 470V.

[0068] like Figure 4 As shown, the switching circuit consists of an MCU, relays, optocouplers, and switching transistors. Driven by commands from the control module, the MCU controls the relays to switch between diodes AK6-380C and AK6-300C in the surge protection circuit.

[0069] When there is no lightning warning: GPIO output is low, optocoupler is not conducting, switching transistor is not conducting, relay is engaged at positions K1-K2 and K4-K5, and diode AK6-380C with higher operating voltage is connected to the line.

[0070] When a lightning warning is received: GPIO outputs a high level, optocoupler conducts, switching transistor conducts, relays switch to K1-K3 and K4-K6, and diode AK6-300C with a lower operating voltage is connected to the line to improve the protection effect.

[0071] The specific structure of the B5 level surge protection circuit and switching circuit is as follows:

[0072] The positive input terminal of the 270V power supply is connected to the gas discharge tube G1. The gas discharge tube G1 and the varistor RV1 are connected in series. The other end of the varistor RV1 is connected to the ground PE to form a lightning current discharge path. This is the first-level lightning protection circuit. One end of the inductor L1 is connected to the gas discharge tube G1, and the other end is connected to the K1 terminal of the relay KJ1 in the switching circuit to form the second-level lightning protection circuit.

[0073] The negative input terminal of the 270V power supply is connected to the gas discharge tube G2. The gas discharge tube G2 and the varistor RV2 are connected in series. The other end of the varistor RV2 is connected to the ground PE to form a lightning current discharge path. This is the first-level lightning protection circuit. One end of the inductor L2 is connected to the gas discharge tube G2, and the other end is connected to the K4 terminal of the relay KJ2 in the switching circuit to form the second-level lightning protection circuit.

[0074] In the 270V power supply forward switching circuit, terminal K1 of relay KJ1 is connected to the rear end of inductor L1, forming the third-level protection circuit. Terminals K2 and K3 of relay KJ1 are connected to diodes D1 and D2, respectively, to switch between diodes with different turn-on and clamping voltages. The other ends of diodes D1 and D2 are connected to PE ground to discharge lightning current. Terminal GPIO1, which controls the 270V power supply forward switching, is connected to resistor R2. The other end of resistor R2 and GND are connected to the two input terminals of optocoupler U1. The two output terminals of optocoupler U1 are connected to VCC5V and resistor R3, respectively. The other end of resistor R3 is connected to GND. One end of capacitor C1 is connected to R3 and the gate of MOSFET Q1, and the other end is connected to GND. The source of MOSFET Q1 is connected to GND. The power supply input terminal of relay KJ1 is connected to VCC5V, and the output terminal is connected to the drain of MOSFET Q1.

[0075] Similarly, the 270V power supply negative switching circuit uses a relay KJ2. Terminal K4 of relay KJ2 is connected to the rear end of inductor L2, forming the third-level protection circuit. Terminals K5 and K6 of relay KJ2 are connected to diodes D3 and D4, respectively, to switch between diodes with different turn-on and clamping voltages. The other ends of diodes D3 and D4 are connected to PE ground to discharge lightning current. GPIO2, controlling the 270V power supply negative switching, is connected to resistor R5. The other end of resistor R5 and GND are connected to the two input terminals of optocoupler U2. The two output terminals of optocoupler U2 are connected to VCC5V and resistor R6, with the other end of resistor R6 connected to GND. One end of capacitor C2 is connected to R6 and the gate of MOSFET Q2, and the other end is connected to GND. The source of MOSFET Q2 is connected to GND. The power supply input terminal of relay KJ2 is connected to VCC5V, and its output terminal is connected to the drain of MOSFET Q2.

[0076] The main function of optocouplers U1 and U2 is to achieve isolated transmission of electrical signals.

[0077] Among them, the gas discharge tubes G1 and G2 are model JHM091M, the varistors RV1 and RV2 are model MYG3-20K300, the inductors L1 and L2 are model JSAC1212-6R0M, the diodes D1 and D3 are model AK6-380C with a clamping voltage of 520V, and the diodes D2 and D4 are model AK6-300C with a clamping voltage of 470V.

[0078] When encountering indirect interference from lightning, transient overvoltage pulses and transient overcurrent pulses are generated. The lightning protection circuit can quickly discharge the instantaneous overcurrent of lightning through the grounding terminal, and then limit the overvoltage through the internal lightning protection device. Combined with the lightning warning signal of the active lightning protection module, the lightning protection circuit is switched according to the signal command to control the residual energy of the indirect effect of lightning within a certain range and protect the downstream circuit to the maximum extent.

[0079] After testing, the surge protection circuit meets the requirements of Chapter 22, Level B5 of the RTCA / DO-160G standard, and can withstand a voltage surge of 1600V and a current surge of 1600A, with a residual peak voltage of 444V. Figure 5 It has a good protective effect.

[0080] The control module is the core coordination unit of the system, responsible for scheduling the work of each module, signal processing, data storage, and alarm functions. Upon receiving a warning signal from the active lightning protection module, the control module immediately triggers an alarm and sends a switching command to the passive lightning protection module. After the thunderstorm cloud passes, it receives a risk clearance signal and controls the switching circuit to reset to normal operation. Simultaneously, it stores cloud feature data at preset time intervals to support model optimization. The communication interface is equipped with a lightning protection circuit combining a gas discharge tube and ESD devices to ensure the module itself is protected from lightning strikes and operates stably.

[0081] Example 5:

[0082] Based on Example 4, the workflow of the high-level power surge protection system based on the YOLO model of the present invention is as follows:

[0083] (1) After the system starts, the image acquisition module continuously acquires image data of sky clouds and transmits it to the active lightning protection module in real time;

[0084] (2) The active lightning protection module analyzes and processes the data through the YOLO v8 model to identify the characteristics of thunderstorm clouds; if a thunderstorm cloud is detected, the active lightning protection module outputs a warning signal to the control module, and the control module triggers an alarm and sends a switching command.

[0085] (3) After receiving the instruction, the passive surge protection module switches the circuit to switch the surge protection circuit to the AK6-300C channel and enters the high protection state;

[0086] (4) When the indirect effect of lightning causes transient overvoltage and overcurrent, the B5 level lightning protection circuit starts three-level protection: the first-level gas discharge tube and varistor quickly conduct to discharge energy, the second-level inductor weakens the energy spike, and the third-level diode clamps the residual voltage to the safe range of 444V to protect the downstream equipment.

[0087] (5) After the thunderstorm cloud passes, the control module receives the release signal, the control switching circuit is reset, and the system returns to the normal working mode of AK6-380C; at the same time, the control module stores the monitoring data to the database for subsequent model optimization.

[0088] Example 6:

[0089] Based on Example 5, the installation and deployment requirements for each module in the lightning protection system of the present invention are as follows:

[0090] The passive surge protection module must be connected in series with the power supply interface to be protected, and the grounding terminal must be reliably grounded to ensure the smooth dissipation of lightning energy. The image acquisition module is installed at a high, unobstructed location to maximize the sky observation range and avoid obstruction from buildings, trees, etc., that could affect cloud data acquisition. The active surge protection module uses a Jetson chip as its AI computing core and connects directly to the control module via a communication interface to ensure real-time data transmission.

Claims

1. A high-level power lightning protection system based on a YOLO model, characterized in that, The lightning protection device comprises an image acquisition module, an active lightning protection module, a passive lightning protection module and a control module; the image acquisition module acquires lightning images and transmits the images to the active lightning protection module; the active lightning protection module analyzes the lightning images based on a YOLO model, identifies and warns the thunderstorm cloud, and sends signals to the control module; the passive lightning protection module receives instructions from the control module, limits overvoltage through a B5-grade lightning protection circuit, and discharges overcurrent; and the control module is used for coordinating the work of the modules. The passive lightning protection module comprises a B5-grade lightning protection circuit and a switching circuit, the B5-grade lightning protection circuit is a three-stage combined structure, and specifically comprises: The first stage: a gas discharge tube and a voltage-dependent resistor are connected in series, the gas discharge tube is turned on first when overvoltage occurs, and the voltage-dependent resistor is triggered to be turned on to discharge lightning energy; The second stage: an inductor is used for weakening lightning instantaneous peaks and absorbing lightning energy, and cooperates with the third-stage device to play a time delay role; The third stage: a diode AK6-380C or AK6-300C, wherein the clamping voltage of the AK6-380C is 520V, and the clamping voltage of the AK6-300C is 470V, and the diode is used for further discharging current and accurately clamping voltage to a safe range; The switching circuit is driven by the control module to realize switching of different clamping voltage diodes in the lightning protection circuit; The specific structure of the B5-grade lightning protection circuit is as follows: The positive input end of the power supply is connected with a gas discharge tube G1, the G1 and a voltage-dependent resistor RV1 are connected in series, and the other end of the RV1 is connected with the ground PE; one end of an inductor L1 is connected with the G1, and the other end is connected with the switching circuit; the circuit further comprises diodes D1 and D2, the switching of the diodes D1 and D2 is realized by the switching circuit, and the D1 and D2 are both connected with the PE ground; The negative input end of the power supply is connected with a gas discharge tube G2, the G2 and a voltage-dependent resistor RV2 are connected in series, and the other end of the RV2 is connected with the ground PE; one end of an inductor L2 is connected with the G2, and the other end is connected with the switching circuit; the circuit further comprises diodes D3 and D4, the switching of the diodes D3 and D4 is realized by the switching circuit, and the D3 and D4 are both connected with the PE ground; Wherein, the model of the gas discharge tubes G1 and G2 is JHM091M, the model of the voltage-dependent resistors RV1 and RV2 is MYG3-20K300, the model of the inductors L1 and L2 is JSAC1212-6R0M, the model of the diodes D1 and D3 is AK6-380C, and the model of the diodes D2 and D4 is AK6-300C; The specific structure of the switching circuit is as follows: The GPIO1 end of the control power supply positive switching is connected with a resistor R2, the other end of the R2 is connected with the GND and two input ends of an optical coupler U1, two output ends of the optical coupler U1 are connected with a VCC 5V and a resistor R3 respectively, the other end of the R3 is connected with the GND, one end of a capacitor C1 is connected with the R3 and a gate of a MOS tube Q1, the other end of the C1 is connected with the GND, a source of the MOS tube Q1 is connected with the GND, a drain is connected with an output end of a relay KJ1, a power supply input end of the relay KJ1 is connected with the VCC 5V, a K1 end is connected with a rear end of the inductor L1, a K2 end and a K3 end are connected with diodes D1 and D2 respectively. The GPIO2 end of the control power negative switching is connected with the resistor R5, the other end of the R5 is connected with the GND and the two input ends of the photo-coupler U2, the two output ends of the photo-coupler U2 are connected with the VCC 5V and the resistor R6 respectively, the other end of the R6 is connected with the GND, one end of the capacitor C2 is connected with the R6 and the gate of the MOS tube Q2, the other end of the capacitor C2 is connected with the GND, the source of the MOS tube Q2 is connected with the GND, the drain of the MOS tube Q2 is connected with the output end of the relay KJ2, the power supply input end of the relay KJ2 is connected with the VCC 5V, the K4 end is connected with the rear end of the inductor L2, the K5 and K6 ends are connected with the diodes D3 and D4 respectively.

2. The high-level power lightning protection system based on YOLO model according to claim 1, characterized in that, The image acquisition module adopts wide-angle + long-focus dual-camera lens combination cooperative observation to collect cloud cluster image and video data from near ground to hollow 400-2000 meters and high altitude 4000-6000 meters, covering different scenes in day and night; wherein, the wide-angle camera is responsible for large-scale preliminary screening of cloud cluster vertical development structure and cloud anvil diffusion trend, and the long-focus camera is responsible for focusing on cloud anvil burr mutation area and ice crystal diffusion details. 3.The high-level power lightning protection system based on YOLO model according to claim 1, wherein, The active lightning prevention module adopts YOLO v8 model to pre-process the lightning movie image transmitted by the image acquisition module, extracts cloud cluster features through convolutional neural network, detects lightning target through sliding window, calculates target boundary box and confidence, and then screens overlapping target box through non-maximum suppression algorithm to output the final detection result, and sends a warning signal to the control module if a thunderstorm cloud is identified.

4. The high-level power lightning protection system based on YOLO model according to claim 3, characterized in that, The pre-processing of the image includes scaling and normalizing the pixel value of the image to adapt to the model input requirements; in addition, based on the Darknet deep learning network, the YOLO v8 model is trained by programming in Python language.

5. The high-level power lightning protection system based on YOLO model according to claim 1, characterized in that, After receiving the warning signal of the active lightning prevention module, the control module triggers an alarm prompt and sends a switching instruction to the passive lightning prevention module, and controls the passive lightning prevention module circuit to reset after the thunderstorm cloud passes, and stores the monitoring data for model optimization.

6. A method of lightning protection using the system according to any one of claims 1 to 5, characterized in that, Specifically: After the system starts, the image acquisition module continuously acquires sky cloud cluster image data and transmits it to the active lightning prevention module in real time; the active lightning prevention module processes and analyzes the data through the YOLO v8 model to identify thunderstorm cloud characteristics, and outputs a warning signal to the control module if a thunderstorm cloud is detected; the control module triggers an alarm and sends an instruction to the passive lightning prevention module, the passive lightning prevention module receives the instruction, switches the circuit, discharges the overcurrent, and clamps the residual voltage to a safe range; after the thunderstorm cloud passes, the control module receives a release signal, controls the passive lightning prevention module circuit to reset, and stores the monitoring data of this time to the database for subsequent model optimization.

Images