Light emitting diode (LED) light source

Abstract

This invention relates to the field of lightning strike location and detection technology, and particularly to a method and system for lightning strike location and detection based on current and electric field strength. The method includes S1, hardware setup; S2, data preprocessing; S3, input feature encoding; S4, cross-attention fusion; and S5, lightning strike location application. This invention employs synchronous acquisition by an electric field sensor and an electromagnetic induction current sensor, sets up a deep fusion neural network, and uses a cross-attention module to enable real-time waveform features to reflect environmental changes and historical data, achieving accurate lightning strike location based on current and electric field. Compared to simple neural network splicing, this significantly improves the accuracy and robustness of lightning strike detection, reliably analyzes and derives the peak lightning current; furthermore, the model can predict the probability of thunderstorm arrival from changes in the electric field and environment, enabling timely prevention.

Description

Technical Field

[0001] This invention relates to the field of lightning strike location detection technology, and in particular to a lightning strike location detection method and system based on current and electric field strength. Background Technology

[0002] In power systems, lightning strikes are one of the main causes of high-voltage transmission line faults. To improve the repair efficiency of lightning strikes on transmission lines and reduce the losses caused by lightning faults, the power grid needs to determine the location of the lightning strike point while taking into account both timeliness and accuracy. In existing technologies, attempts to use multiple physical quantities for lightning location mostly adopt simple feature splicing or fixed threshold judgment, failing to perform in-depth feature fusion of real-time transient waveforms, environmental field evolution, and historical experience, resulting in insufficient detection accuracy and predictive ability. Summary of the Invention

[0003] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the description and other accompanying drawings.

[0004] The purpose of this invention is to overcome the above-mentioned shortcomings and provide a lightning strike location detection method and system based on current and electric field strength. It adopts a method and system based on synchronous acquisition by electric field sensor and electromagnetic induction current sensor, sets up a deep fusion neural network, and uses a cross-attention module to enable real-time waveform features to reflect environmental changes and historical data, thereby realizing accurate lightning strike location based on current and electric field. Compared with simple neural network splicing, it significantly improves the accuracy and robustness of lightning strike detection, can reliably analyze the peak value of lightning current, and the model can predict the probability of thunderstorm arrival from changes in electric field and environment, so as to achieve timely prevention. This invention provides a lightning strike location detection method based on current and electric field strength, comprising: S1. Hardware setup: Install an electromagnetic induction lightning current sensor, an electric field strength sensor, and an environmental sensor on the high-voltage power line tower. The electromagnetic induction lightning current sensor is used to detect the current lightning current, the electric field strength sensor is used to detect the electric field strength of the current environment, and the environmental sensor is used to detect the temperature, humidity, and air pressure of the current environment. S2. Data Preprocessing: The current lightning current signal is input to the instrumentation amplifier to amplify the original weak signal. Then, it is denoised using wavelet filtering and adaptive filtering algorithms to obtain the denoised lightning current data. The current electric field signal is directly denoised using wavelet filtering and adaptive filtering algorithms to obtain the denoised electric field data. The average value of the current ambient temperature, humidity and air pressure data within a time period of t seconds is taken. S3. Input Feature Encoding: Combine lightning current waveform, electric field waveform, and environmental mean to generate a transient change sequence. The input is fed into a 1D-CNN model, undergoes two convolutions, and then is fed into an LSTM model to generate a query vector Q. The mean electric field and the rate of change of the electric field within a time interval of t seconds are then concatenated to form a short-term variation sequence. A 4-layer Transformer encoder is used to encode and generate a short-time key matrix K_s and a short-time value matrix V_s; historical lightning strike experience data is input into a 3-layer fully connected layer, and after expansion, a historical key matrix K_h and a historical value matrix V_h are generated. S4, Cross-attention fusion: In the cross-attention fusion module, the query vector Q is cross-fused with the short-time key matrix K_s, the short-time value matrix V_s, the historical key matrix K_h, and the historical value matrix V_h respectively. After concatenation, the fused features are output through a fully connected layer. The fused features are then connected to a multi-task output head to output the lightning strike probability, lightning peak current, and lightning risk prediction respectively. S5. Lightning Strike Location Application: Thresholds are set for lightning strike probability, peak lightning current, and lightning strike risk prediction. If any value exceeds its corresponding threshold, the monitoring terminal will transmit the current location information of the high-voltage power tower and the lightning strike data to the monitoring center to achieve lightning strike location.

[0005] In some embodiments, in step S3, a transient change sequence is generated by combining the lightning current waveform, the electric field waveform, and the environmental mean. Step-by-step instructions: A time period of t seconds before and after the lightning strike is extracted, and the current waveform, electric field waveform, and environmental mean values ​​are extracted simultaneously. The current waveform forms a one-dimensional sequence, the electric field waveform forms a one-dimensional sequence, and the environmental parameters, temperature, humidity, and air pressure, are averaged within the time window and expanded into three one-dimensional sequences of the same length as the current waveform sequence. The five obtained one-dimensional sequences are then sequentially spliced ​​together to form an instantaneous change sequence of size (t, 5). .

[0006] In some embodiments, a sequence of instantaneous changes of size (t, 5) is used. Input a 1D-CNN model, which contains two convolutional layers, each with 64 filters. The extracted features are then input into an LSTM model, which contains a single LSTM layer with 64 hidden units. The hidden state at the last time step is taken, and the query vector Q is output.

[0007] In some embodiments, in step S3, the average electric field value and the rate of change of the electric field within a time interval of t seconds are combined to form a short-time variation sequence. Step-by-step instructions: The electric field waveform was captured 30 minutes before the lightning strike. After denoising, the waveform sequence was extracted. Samples were taken at 1-second intervals to obtain 1800 time steps. The mean electric field and the rate of change of the electric field were extracted simultaneously to form one-dimensional sequences. After splicing, a short-time variation sequence with a shape of (1800, 2) was formed. .

[0008] In some embodiments, the cross-attention fusion module includes an environment fusion submodule and a history fusion submodule. In the environment fusion submodule, the query vector Q is fused with the short-term key matrix K_s and the short-term value matrix V_s to generate a set of one-dimensional sequences. In the history fusion submodule, the query vector Q is fused with the history key matrix K_h and the history value matrix V_h to generate another set of one-dimensional sequences. The two sets of one-dimensional sequences are concatenated and then passed through a fully connected layer to output the fusion feature F.

[0009] In some embodiments, in step S5, the lightning strike probability threshold is 0.8, the lightning strike current peak threshold is 2kA, and the lightning strike risk prediction threshold is 0.7.

[0010] A lightning strike location detection system based on current and electric field strength, the system comprising: Electromagnetic induction lightning current sensor, which is used to detect the current lightning current; An electric field strength sensor is used to detect the electric field strength of the current environment; Environmental sensors are used to detect the current temperature, humidity, and air pressure of the environment; An instrumentation amplifier, used to amplify lightning strike current signals; The signal noise reduction module is used to reduce noise and average signals such as lightning current, electric field strength, temperature, humidity and air pressure. The monitoring terminal is equipped with an AI big data model for input feature encoding and cross-attention fusion. It also has a positioning and communication module for remotely sending current location information. The monitoring center receives lightning strike data and location data from monitoring terminals and displays them through a front-end page.

[0011] By adopting the above technical solution, the beneficial effects of the present invention are: This invention employs synchronous acquisition based on electric field sensors and electromagnetic induction current sensors, sets up a deep fusion neural network, and uses a cross-attention module to enable real-time waveform features to reflect environmental changes and historical data, achieving precise lightning strike location based on current and electric field. Compared with simple neural network splicing, it significantly improves the accuracy and robustness of lightning strike detection, and can reliably analyze and obtain the peak value of lightning current; moreover, the model can predict the probability of thunderstorm arrival from changes in electric field and environment, enabling timely prevention.

[0012] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure.

[0013] Undoubtedly, such and other objects of the present invention will become more apparent after the following detailed description of the preferred embodiments, which are illustrated in various accompanying drawings and figures.

[0014] To make the above and other objects, features and advantages of the present invention more apparent and understandable, one or more preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0015] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with the embodiments of the invention to explain the invention and do not constitute a limitation thereof.

[0016] In the accompanying drawings, the same parts use the same reference numerals, and the drawings are schematic and not necessarily drawn to actual scale.

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, 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 one or more embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on such drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of the overall process of the positioning and detection method in some embodiments of the present invention; Figure 2 This is a schematic diagram of the splicing structure of instantaneous change sequences in some embodiments of the present invention; Figure 3 This is a schematic diagram of the specific structure of the LSTM model in some embodiments of the present invention; Figure 4 This is a schematic diagram of the splicing structure of short-time change sequences in some embodiments of the present invention; Figure 5 This is a schematic diagram of the specific structure of cross-attention fusion and multi-head output in some embodiments of the present invention. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0020] Furthermore, in the description of this invention, it should be understood that the terms "center," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," "outer," "axial," "radial," and "circumferential" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0021] In this invention, unless otherwise explicitly specified and limited, the terms "installation," "connection," "linking," and "fixing," etc., should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral unit; they can refer to a direct connection or an indirect connection through an intermediate medium; they can refer to the internal communication of two components or the interaction between two components. However, specifying a direct connection indicates that the two main bodies are not connected through a transitional structure, but rather formed as a whole through a connecting structure. Those skilled in the art can understand the specific meaning of the above terms in this invention according to the specific circumstances.

[0022] In this invention, unless otherwise expressly specified and limited, the first feature "on" or "below" the second feature may be in direct contact with the first and second features, or indirect contact through an intermediate medium. In the description of this specification, references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0023] Reference Figure 1 , Figure 1 This is a schematic diagram of the overall process of the positioning and detection method in some embodiments of the present invention.

[0024] According to some embodiments of the present invention, the present invention provides a lightning strike location detection method based on current and electric field strength, comprising: S1. Hardware setup: Install an electromagnetic induction lightning current sensor, an electric field strength sensor, and an environmental sensor on the high-voltage power line tower. The electromagnetic induction lightning current sensor is used to detect the current lightning current, the electric field strength sensor is used to detect the electric field strength of the current environment, and the environmental sensor is used to detect the temperature, humidity, and air pressure of the current environment. Electromagnetic induction lightning current sensors are common sensors for detecting lightning current. In this application, the current acquisition range of the electromagnetic induction lightning current sensor is 500A-20000A. The electric field sensor operates on the principle that, under the influence of an electric field, the capacitance between the electrodes changes with the electric field strength. By accurately detecting and calculating the capacitance change through a precision circuit, the electric field strength can be quantified in real time. For example, the electric field strength of a 10kV high-voltage transmission line, according to the formula... ,in, vacuum permittivity The electric field sensor is 10m away from the high-voltage power line, and the theoretical electric field is E≈1804V / m. The environmental sensor can detect various variables in the surrounding environment such as temperature, humidity, air pressure, wind direction, light, and radiation. However, for the method of this application, it is only necessary to access the temperature, humidity, and air pressure data sensed by the environmental sensor. S2. Data Preprocessing: The current lightning current signal is input to the instrumentation amplifier to amplify the original weak signal. Then, it is denoised using wavelet filtering and adaptive filtering algorithms to obtain denoised lightning current data. The current electric field signal is directly denoised using wavelet filtering and adaptive filtering algorithms to obtain denoised electric field data. The average value of the obtained current ambient temperature, humidity, and air pressure data over a time interval of t seconds is taken. The method of using wavelet filtering and adaptive filtering algorithms to denoise and filter circuit waveforms is existing technology, and the specific process will not be described in detail here. Please refer to existing technology. S3. Input Feature Encoding: Combine lightning current waveform, electric field waveform, and environmental mean to generate a transient change sequence. The input is fed into a 1D-CNN model, convolved twice, and then fed into an LSTM model to generate a query vector Q. The mean electric field and the rate of change of the electric field within a time interval of t seconds are then concatenated to form a short-time variation sequence. A 4-layer Transformer encoder is used to encode and generate a short-time key matrix K_s and a short-time value matrix V_s; historical lightning strike experience data is input into a 3-layer fully connected layer, and after expansion, a historical key matrix K_h and a historical value matrix V_h are generated. Reference Figures 2-3 , Figure 2 This is a schematic diagram of the splicing structure of instantaneous change sequences in some embodiments of the present invention; Figure 3 This is a schematic diagram of the specific structure of the LSTM model in some embodiments of the present invention.

[0025] In step S3, a sequence of instantaneous changes is generated by combining the lightning current waveform, electric field waveform, and environmental mean values. Step-by-step instructions: A time period of t seconds before and after the lightning strike is extracted, and the current waveform, electric field waveform, and environmental mean values ​​are extracted simultaneously. The current waveform forms a one-dimensional sequence, the electric field waveform forms a one-dimensional sequence, and the environmental parameters, temperature, humidity, and air pressure, are averaged within the time window and expanded into three one-dimensional sequences of the same length as the current waveform sequence. The five obtained one-dimensional sequences are then sequentially spliced ​​together to form an instantaneous change sequence of size (t, 5). It is understandable that the time t of a transient change sequence generally does not exceed 1 second. The instantaneous change sequence of size (t, 5) Input a 1D-CNN model, which contains two convolutional layers, each with 64 filters. The extracted features are then input into an LSTM model, which contains a single LSTM layer with 64 hidden units. The hidden state at the last time step is taken, and the query vector Q is output. For instantaneous change sequences To illustrate further, consider a time window of t=0.6s with 10kHz synchronous sampling. All signals have undergone denoising and filtering as described above. The current waveform is 6000x1, the electric field waveform is 6000x1, and the temperature, humidity, and air pressure are averaged within this time window and extended into a sequence of length 6000, i.e., 6000x3. These are then concatenated to obtain... =6000x5, in the 1D-CNN model, the kernel size is set to 7, the stride is 2, the activation function is ReLU, the dropout rate is 0.1, the shape of the extracted sequence is (750, 32), the hidden state ht of the last time step is output by a single-layer LSTM, and the hidden state ht is used as the query vector Q, the shape of which is (128, 1).

[0026] Reference Figure 4 , Figure 4 This is a schematic diagram of the splicing structure of short-time change sequences in some embodiments of the present invention.

[0027] In step S3, the average electric field value and the rate of change of the electric field within a time interval of t seconds are combined to form a short-time variation sequence. Step-by-step instructions: The electric field waveform was captured 30 minutes before the lightning strike. After denoising, the waveform sequence was extracted. Samples were taken at 1-second intervals to obtain 1800 time steps. The mean electric field and the rate of change of the electric field were extracted simultaneously to form one-dimensional sequences. After splicing, a short-time variation sequence with a shape of (1800, 2) was formed. The mean electric field is the average value over a 1-second window, and the rate of change of the electric field is the first-order difference over time. Two-dimensional short-time variation sequence Linear mapping to 256 dimensions yields (1800, 256). Positions are encoded using standard sine waves to inject sequence position information into the Transformer model. In a 4-layer Transformer encoder, the number of multi-head attention heads is 8, the feedforward dimension is 512, the dropout rate is 0.1, and the output shape is (1800, 256). The hidden states of the last 4 time steps are extracted to form M_s = 4x256. Short-term key matrix K_s and short-term value matrix V_s are generated through linear layers, both with shapes of (4, 256). The following is a detailed explanation of step S3, which involves inputting historical lightning strike experience data into three fully connected layers and expanding it to generate the historical key matrix K_h and the historical value matrix V_h: The input dimension of historical lightning strike data is 32-dimensional. After passing through 3 fully connected layers, the shape is (1, 256). It is then expanded to (4, 256) through an extension layer, generating the historical key matrix K_h and the historical value matrix V_h. The historical lightning strike data specifically includes multi-dimensional data such as timestamps, peak lightning current, electric field waveform, electric field change rate, temperature, humidity, and air pressure. The specific data collection situation is modified according to the actual installation requirements, and the dimensions of the corresponding historical lightning strike data will also change. In practice, the historical key matrix K_h and historical value matrix V_h can be made to correspond to the sequence size of the query vector Q, short-term key matrix K_s, and short-term value matrix V_s by adding or removing fully connected layers and by debugging. For newly installed models without actual historical lightning strike data input, training samples need to be manually imported as cold start samples and set as invalid labels. With the change of time and the addition of lightning strike events, the network learns to determine the output of lightning strikes.

[0028] Reference Figure 5 , Figure 5 This is a schematic diagram illustrating the specific structure of cross-attention fusion and multi-head output in some embodiments of the present invention.

[0029] S4, Cross-attention fusion: In the cross-attention fusion module, the query vector Q is cross-fused with the short-time key matrix K_s, the short-time value matrix V_s, the historical key matrix K_h, and the historical value matrix V_h respectively. After concatenation, the fused features are output through a fully connected layer. The fused features are then connected to a multi-task output head to output the lightning strike probability, lightning peak current, and lightning risk prediction respectively. The cross-attention fusion module includes an environment fusion submodule and a history fusion submodule. In the environment fusion submodule, the query vector Q is fused with the short-term key matrix K_s and the short-term value matrix V_s to generate a set of one-dimensional sequences. In the history fusion submodule, the query vector Q is fused with the history key matrix K_h and the history value matrix V_h to generate another set of one-dimensional sequences. The two sets of one-dimensional sequences are concatenated and then passed through a fully connected layer to output the fusion feature F. S5. Lightning Strike Location Application: Thresholds are set for lightning strike probability, lightning peak current and lightning strike risk prediction respectively. If any value exceeds its corresponding threshold, the monitoring terminal will transmit the current location information of the high-voltage power tower and the lightning strike data to the monitoring center to achieve lightning strike location. The lightning strike probability threshold is 0.8, the peak lightning current threshold is 2kA, and the lightning strike risk prediction threshold is 0.7. The neural network outputs a lightning strike probability value between 0 and 1. A threshold needs to be set to make a binary decision on "whether a lightning strike has occurred." If the threshold is too low, a large number of non-lightning transient disturbances will be misclassified as lightning strikes. If it is too high, it may miss real but waveform-distorted lightning strikes. During model training, the threshold is determined by optimization using F1-score. A threshold of 0.8 usually achieves a good balance between precision and recall. A threshold greater than 0.8 indicates that the tower where the terminal is located has been struck by lightning. The peak current of most natural lightning strikes exceeds 2kA, while induced or interference signals are usually below this value. Therefore, only when the peak current output by the model is greater than 2kA is it considered a valid lightning strike event and its amplitude is recorded. Otherwise, only the waveform is recorded but the positioning process is not triggered. The lightning strike risk prediction head outputs the risk probability for the next 6 10-minute (60-minute) windows. Thresholds need to be set according to the risk level to guide power grid dispatching and inspection on the front-end page. Different warning threshold tables can also be set for lines with different power supply criticality levels to achieve differentiated warnings.

[0030] This invention also provides a lightning strike location detection system based on current and electric field strength, the system comprising: Electromagnetic induction lightning current sensor, which is used to detect the current lightning current; An electric field strength sensor is used to detect the electric field strength of the current environment; Environmental sensors are used to detect the current temperature, humidity, and air pressure of the environment; An instrumentation amplifier, used to amplify lightning strike current signals; The signal noise reduction module is used to reduce noise and average signals such as lightning current, electric field strength, temperature, humidity and air pressure. The monitoring terminal is equipped with an AI big data model for input feature encoding and cross-attention fusion. It also has a positioning and communication module for remotely sending current location information. The monitoring center receives lightning strike data and location data from monitoring terminals and displays them through a front-end page.

[0031] It should be understood that the embodiments disclosed herein are not limited to the specific processing steps or materials disclosed herein, but should be extended to equivalent substitutions of such features as understood by those skilled in the art. It should also be understood that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting.

[0032] The term "embodiment" in this specification refers to a specific feature or characteristic described in connection with an embodiment that is included in at least one embodiment of the invention. Therefore, phrases or "embodiments" appearing in various places throughout the specification do not necessarily refer to the same embodiment.

[0033] Furthermore, the described features or characteristics can be incorporated into one or more embodiments in any other suitable manner. In the above description, specific details, such as thickness, quantity, etc., are provided to provide a comprehensive understanding of embodiments of the invention. However, those skilled in the art will understand that the invention can be implemented without the aforementioned specific details or may be implemented using other methods, components, materials, etc.

Claims

1. A lightning strike location detection method based on current and electric field strength, characterized in that, include S1. Hardware setup: Install an electromagnetic induction lightning current sensor, an electric field strength sensor, and an environmental sensor on the high-voltage power line tower. The electromagnetic induction lightning current sensor is used to detect the current lightning current, the electric field strength sensor is used to detect the electric field strength of the current environment, and the environmental sensor is used to detect the temperature, humidity, and air pressure of the current environment. S2. Data Preprocessing: The current lightning current signal is input to the instrumentation amplifier to amplify the original weak signal. Then, it is denoised using wavelet filtering and adaptive filtering algorithms to obtain the denoised lightning current data. The current electric field signal is directly denoised using wavelet filtering and adaptive filtering algorithms to obtain the denoised electric field data. The average value of the current ambient temperature, humidity and air pressure data within a time period of t seconds is taken. S3. Input Feature Encoding: Combine lightning current waveform, electric field waveform, and environmental mean to generate a transient change sequence. The input is fed into a 1D-CNN model, undergoes two convolutions, and then is fed into an LSTM model to generate a query vector Q. The mean electric field and the rate of change of the electric field within a time interval of t seconds are then concatenated to form a short-term variation sequence. A 4-layer Transformer encoder is used to encode and generate a short-time key matrix K_s and a short-time value matrix V_s; historical lightning strike experience data is input into a 3-layer fully connected layer, and after expansion, a historical key matrix K_h and a historical value matrix V_h are generated. S4, Cross-attention fusion: In the cross-attention fusion module, the query vector Q is cross-fused with the short-time key matrix K_s, the short-time value matrix V_s, the historical key matrix K_h, and the historical value matrix V_h respectively. After concatenation, the fused features are output through a fully connected layer. The fused features are then connected to a multi-task output head to output the lightning strike probability, lightning peak current, and lightning risk prediction respectively. S5. Lightning Strike Location Application: Thresholds are set for lightning strike probability, peak lightning current, and lightning strike risk prediction. If any value exceeds its corresponding threshold, the monitoring terminal will transmit the current location information of the high-voltage power tower and the lightning strike data to the monitoring center to achieve lightning strike location.

2. The lightning strike location detection method based on current and electric field strength according to claim 1, characterized in that, In step S3, a sequence of instantaneous changes is generated by combining the lightning current waveform, electric field waveform, and environmental mean values. Step-by-step instructions: A time period of t seconds before and after the lightning strike is extracted, and the current waveform, electric field waveform, and environmental mean values ​​are extracted simultaneously. The current waveform forms a one-dimensional sequence, the electric field waveform forms a one-dimensional sequence, and the environmental parameters, temperature, humidity, and air pressure, are averaged within the time window and expanded into three one-dimensional sequences of the same length as the current waveform sequence. The five obtained one-dimensional sequences are then sequentially spliced ​​together to form an instantaneous change sequence of size (t, 5). .

3. The lightning strike location detection method based on current and electric field strength according to claim 2, characterized in that, The instantaneous change sequence of size (t, 5) Input a 1D-CNN model, which contains two convolutional layers, each with 64 filters. The extracted features are then input into an LSTM model, which contains a single LSTM layer with 64 hidden units. The hidden state at the last time step is taken, and the query vector Q is output.

4. The lightning strike location detection method based on current and electric field strength according to claim 1, characterized in that, In step S3, the average electric field value and the rate of change of the electric field within a time interval of t seconds are combined to form a short-time variation sequence. Step-by-step instructions: The electric field waveform was captured 30 minutes before the lightning strike. After denoising, the waveform sequence was extracted. Samples were taken at 1-second intervals to obtain 1800 time steps. The mean electric field and the rate of change of the electric field were extracted simultaneously to form one-dimensional sequences. After splicing, a short-time variation sequence with a shape of (1800, 2) was formed. .

5. The lightning strike location detection method based on current and electric field strength according to claim 1, characterized in that, The cross-attention fusion module includes an environment fusion submodule and a history fusion submodule. In the environment fusion submodule, the query vector Q is fused with the short-term key matrix K_s and the short-term value matrix V_s to generate a set of one-dimensional sequences. In the history fusion submodule, the query vector Q is fused with the history key matrix K_h and the history value matrix V_h to generate another set of one-dimensional sequences. The two sets of one-dimensional sequences are concatenated and then passed through a fully connected layer to output the fusion feature F.

6. The lightning strike location detection method based on current and electric field strength according to claim 1, characterized in that, In step S5, the lightning strike probability threshold is 0.8, the lightning strike current peak threshold is 2kA, and the lightning strike risk prediction threshold is 0.

7.

7. A lightning strike location detection system based on current and electric field strength, characterized in that, The lightning strike location detection method based on current and electric field strength, as described in any one of claims 1-6, comprises: Electromagnetic induction lightning current sensor, which is used to detect the current lightning current; An electric field strength sensor is used to detect the electric field strength of the current environment; Environmental sensors are used to detect the current temperature, humidity, and air pressure of the environment; An instrumentation amplifier, used to amplify lightning strike current signals; The signal noise reduction module is used to reduce noise and average signals such as lightning current, electric field strength, temperature, humidity and air pressure. The monitoring terminal is equipped with an AI big data model for input feature encoding and cross-attention fusion. It also has a positioning and communication module for remotely sending current location information. The monitoring center receives lightning strike data and location data from monitoring terminals and displays them through a front-end page.

Images