Lightning stroke identification method and system based on multi-modal sensor information fused with C4.5 classifier

Abstract

The invention discloses a lightning stroke identification method and system based on multi-modal sensing information fused with a C4.5 classifier, and the method comprises the steps: S1, training a model: collecting a considerable number of samples, employing a cost complexity pruning (CCP) method, employing the samples to generate a C4.5 classifier on a PC, controlling the depth of a tree within five layers, and obtaining a training model; the trained model can be conveniently transplanted to a single chip microcomputer for use; s2, system initialization: initializing each module to enable the system to enter a low-power-consumption mode, and waiting for a trigger source to wake up the system by a single-chip microcomputer; s3, collecting data; s4, analyzing data: analyzing data in an off-chip memory in real time, discretizing sampling values, and analyzing temperature, humidity, a peak value of a lightning stroke sensor, rise time of the lightning stroke sensor, fall time of the lightning stroke sensor, a ratio of a middle rise point to a mean rise point, a voltage variation of a solar cell panel and a magnetic field variation parameter; and inputting the result into a trained C4.5 classifier model to obtain whether lightning stroke exists at the moment.

Description

Technical Field

[0001] The present invention belongs to the technical field of embedded intelligent detection terminals, and in particular relates to a lightning strike identification method and system based on a multimodal sensor information fusion C4.5 classifier. Background Art

[0002] During the transmission and distribution of electricity, natural factors can cause instability or even damage to power transmission circuits. Lightning strikes, in particular, pose a significant threat to power transmission. To effectively prevent the impact of lightning on power transmission, effective lightning strike detection is necessary. In practical applications, single lightning current detection sensors are susceptible to environmental influences, such as birds and sharp discharges. Simply identifying a lightning strike based solely on the current curve results in low sensitivity to environmental interference factors and poor anti-interference capabilities.

[0003] In summary, there is an urgent need to design a lightning strike identification method that uses environmental factors as the basis for judging lightning strike events and improves the accuracy of lightning strike identification. Summary of the Invention

[0004] In order to solve the above technical problems existing in the prior art, the present invention provides a lightning strike identification method and system based on multimodal sensor information fusion C4.5 classifier, which has the advantages of improving environmental anti-interference ability and improving the accuracy of tower lightning strike identification.

[0005] The technical solution adopted in the present invention is:

[0006] The first aspect of the present invention relates to a lightning strike identification method based on a multimodal sensor information fusion C4.5 classifier, characterized in that the identification method comprises the following steps:

[0007] S1. Training model

[0008] Collect a considerable number of samples and use the cost-complexity pruning (CCP) method to generate a C4.5 classifier on a PC. Keep the tree depth within 5 layers to facilitate porting the trained model to a single-chip microcomputer.

[0009] S2. System initialization

[0010] Initialize each module, put the system into low-power mode, and wait for the trigger source to wake up the system;

[0011] S3. Collect data

[0012] Use temperature sensors, humidity sensors, tower lightning current detection circuits, solar panels, and magnetic field strength sensors to collect ambient temperature, humidity, tower current, light intensity, and surrounding magnetic field strength near the tower, and convert the collected data into digital quantities and store them in off-chip memory.

[0013] S4. Analyze data

[0014] The data in the off-chip memory is analyzed in real time, the sampled values ​​are discretized, and the temperature, humidity, peak value of the lightning sensor, rise time of the lightning sensor, fall time of the lightning sensor, ratio of the rising midpoint to the rising mean, voltage change of the solar panel, and magnetic field change parameters are analyzed and input into the trained C4.5 classifier model to determine whether there is a lightning strike at this time.

[0015] Furthermore, in step S3, a large current detection circuit is used to connect the pole tower. The circuit uses a magnetic induction coil to convert the current value into a voltage value, performs low-pass filtering through a second-order low-pass filter, amplifies the signal through an integrated operational amplifier, and connects it to the single-chip microcomputer; the single-chip microcomputer uses its own ADC analog-to-digital conversion module to convert analog data into digital quantities, and stores the current sampling value in an off-chip memory.

[0016] Furthermore, in step S3, the voltage detection circuit is connected through a capacitor, and the detection circuit uses the resistor voltage divider function to step down and collect the voltage output by the solar panel. The follower implemented by the integrated amplifier is connected using a capacitor. The microcontroller can collect the voltage changes of the solar panel through its own ADC digital-to-analog conversion module to reflect the photoelectric changes in the current situation, and store the sampled values ​​in the off-chip memory.

[0017] Furthermore, in step S3, a magnetic field strength sensor is used to collect the surrounding magnetic field strength, and the output analog quantity of the magnetic field strength sensor is connected to the integrated differential amplifier circuit, and after passing through the RC low-pass filter circuit, it is sent to the ADC acquisition port of the microcontroller; the microcontroller uses the ADC to collect the voltage value of the sensor to reflect the change in magnetic field strength, and stores the collected information in the off-chip memory.

[0018] Furthermore, in step S3, the solar panel voltage acquisition circuit includes a voltage ADC acquisition circuit and a dynamic peak detection circuit; the ADC acquisition circuit converts light intensity into a voltage value and converts the voltage from an analog signal to a digital signal; the dynamic peak detection circuit detects the peak value of the dynamic voltage output by the solar panel.

[0019] Furthermore, in step S1, the trigger source wake-up system mode includes one or more combinations of leakage current external interrupt trigger wake-up, solar panel dynamic voltage peak trigger wake-up or magnetic field change trigger wake-up; at least one of the triggers will wake up the microcontroller to collect information on various environmental variables.

[0020] Furthermore, comparator circuits are added at the ends of the high current detection circuit, the solar panel dynamic peak detection circuit, and the magnetic field strength sensor acquisition circuit, and the comparison results are sent to different external interrupt trigger pins of the microcontroller, which are used as lightning trigger sources to wake up the low-power module of the entire system.

[0021] The second aspect of the present invention relates to a lightning strike identification system based on multimodal sensor information fusion C4.5 classifier, characterized by comprising:

[0022] A single-chip microcomputer is used to process the collected data;

[0023] The tower lightning current detection circuit is used to detect the current on the tower; the magnetic field detection circuit is used to detect the change of the magnetic field near the equipment;

[0024] Magnetic field sensor, collects the surrounding magnetic field strength;

[0025] 4G communication module, used to achieve remote communication between the device and the server;

[0026] Temperature sensor, used to collect temperature values ​​near the device;

[0027] A humidity sensor is used to collect humidity values ​​near the device;

[0028] A voltage peak detection circuit is used to detect the dynamic voltage peak of the solar panel;

[0029] Solar panels, used to charge the battery;

[0030] Batteries, used to power the entire system;

[0031] And off-chip memory is used to store environmental sampling data.

[0032] Furthermore, the single chip microcomputer uses an M4 core embedded single chip microcomputer, which has multiple interfaces, ADC analog quantity acquisition, SDIO, and I2C functions, and can realize fast processing of data.

[0033] Compared with the prior art, the beneficial effects of the present invention are embodied in:

[0034] 1. The present invention uses a variety of sensors to obtain environmental information based on various factors in the environment, and performs lightning strike identification by classifying and processing the information in combination with the C4.5 classifier algorithm.

[0035] 2. The present invention combines multiple environmental factors such as temperature, humidity, tower current, light, magnetic field, etc. to predict lightning strike events. Using environmental factors as the basis for judging lightning strike events improves the accuracy of lightning strike identification.

[0036] 3. The present invention is designed with a low-power mode, requiring no external power supply, relying solely on batteries and solar panels to provide system power, and can operate continuously. The system is normally in low-power mode, and when awakened, it will trigger an external interrupt to perform data collection, analysis, upload, and other operations. After the operation is completed, it will enter low-power mode again.

[0037] 4. When the present invention wakes up, if it determines that it is triggered by an external interrupt, it determines that there may be a lightning strike trigger at this time, and it is necessary to collect information and analyze the data in the off-chip memory in real time, discretize the sampling values, etc., and analyze the temperature, humidity, peak value of the lightning strike sensor, rise time of the lightning strike sensor, fall time of the lightning strike sensor, ratio of the rising midpoint to the rising mean, voltage change of the solar panel, magnetic field change and other parameters, and input them into the trained C4.5 classifier model to determine whether there is a lightning strike at this time. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] Figure 1 It is a flow chart of the identification method of the present invention.

[0039] Figure 2 It is a schematic diagram of the lightning strike identification system of the present invention. DETAILED DESCRIPTION

[0040] The following describes the specific implementation of the embodiment of the present invention in detail with reference to the accompanying drawings. It should be understood that the specific implementation described herein is only used to illustrate and explain the embodiment of the present invention and is not used to limit the embodiment of the present invention.

[0041] It should be noted that, in the absence of conflict, the embodiments of the present invention and the features therein may be combined with each other.

[0042] The present invention will be described in detail below with reference to the accompanying drawings and in conjunction with exemplary embodiments.

[0043] Example 1

[0044] refer to Figure 1 The present invention provides a lightning strike identification method based on multimodal sensor information fusion C4.5 classifier, the identification method comprising the following steps:

[0045] S1. Training model

[0046] Collect a considerable number of samples and use the cost-complexity pruning (CCP) method to generate a C4.5 classifier on a PC. Keep the tree depth within 5 layers to facilitate porting the trained model to a single-chip microcomputer.

[0047] S2. System initialization

[0048] Initialize each module, put the system into low-power mode, and wait for the trigger source to wake up the system;

[0049] S3. Collect data

[0050] Use temperature sensors, humidity sensors, tower lightning current detection circuits, solar panels, and magnetic field strength sensors to collect ambient temperature, humidity, tower current, light intensity, and surrounding magnetic field strength near the tower, and convert the collected data into digital quantities and store them in off-chip memory.

[0051] S4. Analyze data

[0052] The data in the off-chip memory is analyzed in real time, the sampled values ​​are discretized, and the temperature, humidity, peak value of the lightning sensor, rise time of the lightning sensor, fall time of the lightning sensor, ratio of the rising midpoint to the rising mean, voltage change of the solar panel, and magnetic field change parameters are analyzed and input into the trained C4.5 classifier model to determine whether there is a lightning strike at this time.

[0053] In this embodiment, in step S3, a large current detection circuit is used to connect the pole tower. The circuit uses a magnetic induction coil to convert the current value into a voltage value, performs low-pass filtering through a second-order low-pass filter, amplifies the signal through an integrated operational amplifier, and connects it to the single-chip microcomputer; the single-chip microcomputer uses its own ADC analog-to-digital conversion module to convert analog data into digital quantities, and stores the current sampling value in an off-chip memory.

[0054] In this embodiment, in step S3, a voltage detection circuit is connected via a capacitor, and the detection circuit uses a resistor voltage divider function to step down and collect the voltage output by the solar panel. A follower implemented by connecting a capacitor to an integrated amplifier is used, and the single-chip microcomputer can collect the voltage changes of the solar panel through its own ADC digital-to-analog conversion module to reflect the photoelectric changes in the current situation, and store the sampled values ​​in an off-chip memory.

[0055] In this embodiment, in step S3, a magnetic field strength sensor is used to collect the surrounding magnetic field strength, and the output analog quantity of the magnetic field strength sensor is connected to the integrated differential amplifier circuit, and after passing through the RC low-pass filter circuit, it is sent to the ADC acquisition port of the microcontroller; the microcontroller uses the ADC to collect the voltage value of the sensor to reflect the change in magnetic field strength, and stores the collected information in an off-chip memory.

[0056] In this embodiment, in step S3, the solar panel voltage acquisition circuit includes a voltage ADC acquisition circuit and a dynamic peak detection circuit; the ADC acquisition circuit converts light intensity into a voltage value and converts the voltage from an analog signal to a digital signal; the dynamic peak detection circuit detects the peak value of the dynamic voltage output by the solar panel.

[0057] In this embodiment, in step S1, the trigger source wake-up system mode includes one or more combinations of leakage current external interrupt trigger wake-up, solar panel dynamic voltage peak trigger wake-up or magnetic field change trigger wake-up; at least one of the triggers will wake up the microcontroller to collect information on various environmental variables.

[0058] In this embodiment, comparator circuits are added to the ends of the high current detection circuit, the solar panel dynamic peak detection circuit, and the magnetic field strength sensor acquisition circuit, respectively. The comparison results are sent to different external interrupt trigger pins of the microcontroller, which are used as lightning trigger sources to wake up the low-power module of the entire system.

[0059] This embodiment combines multiple environmental factors such as temperature, humidity, tower current, light, magnetic field, etc. to predict lightning strike events. Using environmental factors as the basis for determining lightning strike events improves the accuracy of lightning strike identification, as shown in Tables 1 and 2.

[0060] Table 1 is the sample data collection and discretization processing results of this embodiment.

[0061]

[0062] Table 2 shows the corresponding relationship between the simulated sampling values ​​and the discretized variables in this embodiment.

[0063]

[0064] Example 2

[0065] refer to Figure 2 The present invention provides a lightning strike identification system based on a multimodal sensor information fusion C4.5 classifier, comprising:

[0066] A single-chip microcomputer is used to process the collected data;

[0067] The tower lightning current detection circuit is used to detect the current on the tower; the magnetic field detection circuit is used to detect the change of the magnetic field near the equipment;

[0068] Magnetic field sensor, collects the surrounding magnetic field strength;

[0069] 4G communication module, used to achieve remote communication between the device and the server;

[0070] Temperature sensor, used to collect temperature values ​​near the device;

[0071] A humidity sensor is used to collect humidity values ​​near the device;

[0072] A voltage peak detection circuit is used to detect the dynamic voltage peak of the solar panel;

[0073] Solar panels, used to charge the battery;

[0074] Batteries, used to power the entire system;

[0075] And off-chip memory is used to store environmental sampling data.

[0076] In this embodiment, the single chip microcomputer uses an M4 core embedded single chip microcomputer, which has multiple interfaces, ADC analog quantity acquisition, SDIO, and I2C functions, and can realize fast data processing.

[0077] Although the embodiments of the present invention have been shown and described above, it will be understood that the above embodiments are illustrative and are not to be construed as limitations on the present invention. A person skilled in the art may change, modify, replace and modify the above embodiments within the scope of the present invention.

Claims

1. A lightning strike identification method based on multimodal sensor information fusion C4.5 classifier, characterized in that: The identification method includes the following steps: S1. Training model Collect a considerable number of samples and use the cost-complexity pruning (CCP) method to generate a C4.5 classifier on a PC. Keep the tree depth within 5 layers to facilitate porting the trained model to a single-chip microcomputer. S2. System initialization Initialize each module, put the system into low-power mode, and wait for the trigger source to wake up the system; S3. Collect data Use temperature sensors, humidity sensors, tower lightning current detection circuits, solar panels, and magnetic field strength sensors to collect ambient temperature, humidity, tower current, light intensity, and surrounding magnetic field strength near the tower, and convert the collected data into digital quantities and store them in off-chip memory. S4. Analyze data The data in the off-chip memory is analyzed in real time, the sampled values ​​are discretized, and the temperature, humidity, peak value of the lightning sensor, rise time of the lightning sensor, fall time of the lightning sensor, ratio of the rising midpoint to the rising mean, voltage change of the solar panel, and magnetic field change parameters are analyzed and input into the trained C4.5 classifier model to determine whether there is a lightning strike at this time.

2. The lightning strike identification method based on multimodal sensor information fusion C4.5 classifier according to claim 1 is characterized in that: In step S3, a high current detection circuit is connected to the tower. The circuit uses a magnetic induction coil to convert the current value into a voltage value, performs low-pass filtering through a second-order low-pass filter, amplifies the signal through an integrated operational amplifier, and connects it to the single-chip microcomputer. The microcontroller uses its own ADC analog-to-digital conversion module to convert analog data into digital quantities and stores the sampled values ​​of the current in an off-chip memory.

3. The lightning strike identification method based on multimodal sensor information fusion C4.5 classifier according to claim 1 is characterized in that: In step S3, the voltage detection circuit is connected through a capacitor. The detection circuit uses the resistor voltage divider function to step down and collect the voltage output by the solar panel. The follower implemented by the integrated amplifier is connected using a capacitor. The microcontroller can collect the voltage changes of the solar panel through its own ADC digital-to-analog conversion module to reflect the photoelectric changes in the current situation and store the sampled values ​​in the off-chip memory.

4. The lightning strike identification method based on multimodal sensor information fusion C4.5 classifier according to claim 1, characterized in that: In step S3, a magnetic field intensity sensor is used to collect the surrounding magnetic field intensity, and the output analog quantity of the magnetic field intensity sensor is connected to the integrated differential amplifier circuit, passed through the RC low-pass filter circuit, and then sent to the ADC acquisition port of the single-chip microcomputer; The single chip microcomputer uses ADC to collect the voltage value of the sensor to reflect the change of magnetic field strength, and stores the collected information in the off-chip memory.

5. The method for identifying lightning strikes based on multimodal sensor information fusion C4.5 classifier according to claim 2, characterized in that: In step S3, the solar panel voltage acquisition circuit includes a voltage ADC acquisition circuit and a dynamic peak detection circuit; the ADC acquisition circuit converts light intensity into a voltage value and converts the voltage from an analog signal to a digital signal; the dynamic peak detection circuit detects the peak value of the dynamic voltage output by the solar panel.

6. The lightning strike identification method based on multimodal sensor information fusion C4.5 classifier according to claim 1, characterized in that: In step S1, the trigger source wake-up system mode includes one or more combinations of leakage current external interrupt trigger wake-up, solar panel dynamic voltage peak trigger wake-up or magnetic field change trigger wake-up; at least one of the triggers will wake up the microcontroller to collect information on various environmental variables.

7. The method for identifying lightning strikes based on multimodal sensor information fusion C4.5 classifier according to claim 6, characterized in that: Comparator circuits are added at the ends of the high current detection circuit, the solar panel dynamic peak detection circuit, and the magnetic field strength sensor acquisition circuit, and the comparison results are sent to different external interrupt trigger pins of the microcontroller, which are used as lightning trigger sources to wake up the low-power module of the entire system.

8. A lightning strike identification system based on multimodal sensor information fusion C4.5 classifier, characterized by: include: A single-chip microcomputer is used to process the collected data; Tower lightning current detection current, used to detect the current on the tower; The magnetic field detection circuit is used to detect changes in the magnetic field near the device; Magnetic field sensor, collects the surrounding magnetic field strength; 4G communication module, used to achieve remote communication between the device and the server; Temperature sensor, used to collect temperature values ​​near the device; A humidity sensor is used to collect humidity values ​​near the device; A voltage peak detection circuit is used to detect the dynamic voltage peak of the solar panel; Solar panels, used to charge the battery; Batteries, used to power the entire system; And off-chip memory is used to store environmental sampling data.

9. The lightning strike identification system based on multimodal sensor information fusion C4.5 classifier according to claim 1, characterized in that: The single chip microcomputer uses an M4 core embedded single chip microcomputer, which has multiple interfaces, ADC analog quantity acquisition, SDIO, and I2C functions, and can realize fast processing of data.

Images