Waveform identification method of lightning electromagnetic wave and related equipment

Abstract

The embodiment of the invention discloses a lightning electromagnetic wave waveform identification method and related equipment, and the method comprises the steps: calling a plurality of typical reference waveform sequences from a preset reference type library, calculating a similarity measurement value of a sampling sequence and each reference waveform sequence, and taking a difference value between a similarity maximum value and a similarity secondary maximum value as a judgment basis. Parameters such as waveform amplitude, pulse width and rising edge do not need to be used for judgment, misjudgment easily generated when different types of waveform feature intervals are overlapped is avoided, the waveform identification accuracy is improved, meanwhile, only a preset reference waveform library and similarity calculation logic are needed, complex model training and a large number of calculation resources are not needed, and the method is easy to implement. And rapid deployment in the low-computing-power online monitoring terminal is realized.

Description

Technical Field

[0001] This invention relates to the field of electromagnetic wave analysis technology, and in particular to a waveform identification method and related equipment for lightning electromagnetic waves. Background Technology

[0002] With the continuous expansion of the power grid, lightning disasters pose an increasingly serious threat to the power system, especially in areas prone to lightning strikes. Lines and equipment are highly susceptible to overvoltage or insulation flashover accidents caused by lightning strikes, leading to power outages and equipment damage, thus causing economic losses. Lightning electromagnetic wave signals, as important characterizing information of lightning activity, can reflect the physical characteristics of different types of lightning processes. Their accurate identification is of great significance for the lightning protection design, operation and maintenance, and fault diagnosis of power systems.

[0003] Currently, there are two main types of classification methods for lightning electromagnetic waves: one is based on manual parameter criteria, which makes judgments by extracting parameters such as waveform amplitude, pulse width, and rising edge, but it is prone to misjudgment when the feature intervals of different types of waveforms overlap; the other is based on deep learning-based intelligent classification methods, which can improve accuracy, but they are highly dependent on large-scale labeled data and high computing power platforms, making them difficult to deploy in low-computing-power online monitoring terminals. Summary of the Invention

[0004] In view of this, the present invention provides a waveform identification method and related equipment for lightning electromagnetic waves.

[0005] The specific technical solution of the first embodiment of the present invention is as follows: a waveform identification method for lightning electromagnetic waves, the method comprising: acquiring a sampling sequence of the lightning electromagnetic wave to be measured; calling multiple reference waveform sequences from a preset reference type library, wherein the waveform types of the reference waveform sequences include ground flash return sequence, cloud flash return sequence, initial breakdown pulse sequence, and narrow bipolar pulse sequence; calculating a similarity metric between the sampling sequence and each reference waveform sequence; acquiring the maximum and second largest similarity metric values ​​among the similarity metric values; determining whether the maximum similarity metric value is less than a preset threshold; if less than the preset threshold, outputting a preset rejection prompt message; if not less than the preset threshold, determining whether the difference between the maximum and second largest similarity metric values ​​is less than a preset interval; if less than the preset interval, determining the waveform type corresponding to the target reference waveform sequence corresponding to the maximum similarity metric value as the waveform type of the sampling sequence; if not less than the preset interval, outputting a preset rejection prompt message.

[0006] Preferably, before calculating the similarity metric between the sampled sequence and each reference waveform sequence, the method further includes: performing nonlinear time alignment and normalization processing on the sampled sequence and the reference waveform sequence to obtain an optimized sampled sequence and an optimized reference waveform sequence; then, calculating the similarity metric between the sampled sequence and each reference waveform sequence includes: calculating the similarity metric between the optimized sampled sequence and each of the optimized reference waveform sequences.

[0007] Preferably, the similarity metric is obtained using the following formula:

[0008] in, For optimized sampling sequence With the j-th optimized reference waveform sequence Similarity metric between them For the optimized sampling sequence and the j-th optimized reference waveform sequence The number of data points in the middle, For optimized sampling sequence The value of the i-th data point. For the j-th optimized reference waveform sequence The value of the i-th data point.

[0009] Preferably, after acquiring the sampling sequence of the measured lightning electromagnetic wave, the method further includes: determining whether the sampling length of the sampling sequence exceeds a preset length; if the sampling length does not exceed the preset length, then performing the step of calling multiple reference waveform sequences from a preset reference type library; if the sampling length exceeds the preset length, then resampling the sampling sequence with the preset length as the target to obtain a resampled sampling sequence, using the resampled sampling sequence as the sampling sequence, and performing the step of calling multiple reference waveform sequences from the preset reference type library.

[0010] Preferably, the step of resampling the sampled sequence with the preset length as the target to obtain the resampled sampled sequence includes: obtaining a first spectral sequence of the sampled sequence using a preset anti-aliasing low-pass window; performing an inverse discrete Fourier transform on the first spectral sequence with a length of the preset length to obtain a time-domain sequence; and extracting the real part of the time-domain sequence to obtain the resampled sampled sequence.

[0011] Preferably, obtaining the first spectral sequence of the sampling sequence using a preset anti-aliasing low-pass window includes: multiplying the sampling sequence by a preset anti-aliasing low-pass window to obtain a windowed sampling sequence, and symmetrically truncating the windowed sampling sequence while maintaining conjugate symmetry to obtain the first spectral sequence.

[0012] Preferably, if the sampling length does not exceed the preset length, the method further includes: determining whether the sampling length is equal to the preset length; if the sampling length is equal to the preset length, then performing the step of calling multiple reference waveform sequences from the preset reference type library; if the sampling length is less than the preset length, then performing a discrete Fourier transform on the sampling sequence to obtain a second spectral sequence, and zero-padding the second spectral sequence with the preset length as the target while maintaining conjugate symmetry to obtain a third spectral sequence with the preset length, using the third spectral sequence as the sampling sequence, and performing the step of calling multiple reference waveform sequences from the preset reference type library.

[0013] The specific technical solution of the second embodiment of the present invention is as follows: a waveform identification system for lightning electromagnetic waves, the system comprising: a first sequence acquisition module, a second sequence acquisition module, a similarity calculation module, a similarity extraction module, a first judgment module, a first output module, a second judgment module, a second output module, and a third output module; the first sequence acquisition module is used to acquire a sampling sequence of the lightning electromagnetic wave under test; the second sequence acquisition module is used to call various reference waveform sequences from a preset reference type library, the waveform types of the reference waveform sequences including ground flash return stroke sequence, cloud flash return stroke sequence, initial breakdown pulse sequence, and narrow bipolar pulse sequence; the similarity calculation module is used to calculate the similarity metric between the sampling sequence and each reference waveform sequence; the first sequence acquisition module is used to acquire a sampling sequence of the lightning electromagnetic wave under test; the second ... The similarity extraction module is used to obtain the maximum and second largest similarity values ​​in the similarity metric; the first judgment module is used to determine whether the maximum similarity value is less than a preset threshold; the first output module is used to output a preset rejection prompt if it is less than the preset threshold; the second judgment module is used to determine whether the difference between the maximum and second largest similarity values ​​is less than a preset interval if it is not less than the preset threshold; the second output module is used to determine the waveform type corresponding to the target reference waveform sequence corresponding to the maximum similarity value as the waveform type of the sampling sequence if it is less than the preset interval; the third output module is used to output a preset rejection prompt if it is not less than the preset interval.

[0014] The specific technical solution of the third embodiment of the present invention is as follows: a waveform identification device for lightning electromagnetic waves, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor performs the steps of the method as described in any one of the first embodiments of this application.

[0015] The specific technical solution of the fourth embodiment of the present invention is as follows: a computer-readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the processor performs the steps of the method as described in any one of the first embodiments of this application.

[0016] Implementing the embodiments of the present invention will have the following beneficial effects: This invention calls various typical reference waveform sequences from a preset reference type library and calculates the similarity metric between the sampled sequence and each reference waveform sequence. The difference between the maximum and second-largest similarity values ​​is used as the judgment criterion, eliminating the need to use parameters such as waveform amplitude, pulse width, and rising edge for judgment. This avoids misjudgment when the feature intervals of different types of waveforms overlap, thus improving the accuracy of waveform identification. At the same time, this invention only requires a preset reference waveform library and similarity calculation logic, without the need for complex model training and a large amount of computing resources, enabling rapid deployment in low-computing-power online monitoring terminals. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application 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 some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A flowchart illustrating the steps of a method for identifying the waveform of lightning electromagnetic waves; Figure 2 A waveform comparison of the similarity between the tested sequence and the initial breakdown pulse sequence; Figure 3 The waveform diagram shows the difference between the tested sequence and the initial breakdown pulse sequence. Figure 4 The waveform diagram shows the similarity between the tested sequence and the ground flash return stroke sequence. Figure 5 The waveform of the difference between the measured sequence and the ground flash return stroke sequence is shown. Figure 6 This is a schematic diagram of the structure of a waveform identification system for lightning electromagnetic waves; Among them, 201 is the first sequence acquisition module; 202 is the second sequence acquisition module; 203 is the similarity calculation module; 204 is the similarity extraction module; 205 is the first judgment module; 206 is the first output module; 207 is the second judgment module; 208 is the second output module; and 209 is the third output module. Detailed Implementation

[0019] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application.

[0020] The terms "first," "second," etc., used in the specification, claims, and drawings of this application are used to distinguish different objects, not to describe a specific order. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion. For example, a process, method, system, product, or apparatus that includes a series of steps or modules is not limited to the listed steps or modules, but may optionally include steps or modules not listed, or may optionally include other steps or modules inherent to such processes, methods, products, or apparatus.

[0021] In this document, the term "embodiment" means that a particular feature, structure, or characteristic described in connection with an embodiment may be included in at least one embodiment of this application. The appearance of this phrase in various places throughout the specification does not necessarily refer to the same embodiment, nor is it a separate or alternative embodiment mutually exclusive with other embodiments. It will be explicitly and implicitly understood by those skilled in the art that the embodiments described herein can be combined with other embodiments.

[0022] Please see Figure 1 This is a flowchart illustrating the steps of a waveform identification method for lightning electromagnetic waves in the first embodiment of this application, thereby improving the accuracy of waveform identification. The method includes: Step 101: Obtain the sampling sequence of the measured lightning electromagnetic wave; Step 102: Call up various reference waveform sequences from the preset reference type library. The waveform types of the reference waveform sequences include ground flash return stroke sequence, cloud flash return stroke sequence, initial breakdown pulse sequence, and narrow bipolar pulse sequence. Step 103: Calculate the similarity metric between the sampled sequence and each reference waveform sequence; Step 104: Obtain the maximum and second-largest similarity metrics from the similarity metrics; Step 105: Determine whether the maximum value of the similarity metric is less than a preset threshold; Step 106: If the value is less than the preset threshold, output the preset rejection prompt message; Step 107: If it is not less than the preset threshold, then determine whether the difference between the maximum value of the similarity metric and the second largest value of the similarity metric is less than the preset interval. Step 108: If it is less than a preset interval, then the waveform type corresponding to the target reference waveform sequence corresponding to the maximum similarity metric value is determined as the waveform type of the sampling sequence; Step 109: If the interval is not less than the preset interval, output the preset rejection prompt message.

[0023] Specifically, in lightning monitoring research, sampling sequences of the measured lightning electromagnetic waves are acquired using electromagnetic wave sampling equipment. These sequences accurately record information such as amplitude changes of the lightning electromagnetic waves over a period of time. Multiple reference waveform sequences are retrieved from a pre-set reference type library. This library pre-stores different types of waveform sequences, including ground flash return stroke sequences, cloud flash return stroke sequences, initial breakdown pulse sequences, and narrow bipolar pulse sequences. These sequences are typical waveforms obtained through extensive experiments and data analysis. A similarity calculation algorithm is used to calculate the similarity metric between the sampled sequence and each reference waveform sequence. The maximum and second-largest similarity metrics are obtained, and then it is determined whether the maximum similarity metric is less than a preset threshold, which is set based on actual needs and experience. If it is less than the preset threshold, it indicates that the similarity between the sampled sequence and all reference waveform sequences is very low, and a preset rejection message is output. If it is not less than the preset threshold, it is further determined whether the difference between the maximum and second-largest similarity metrics is less than a preset interval. If the value is less than the preset interval, it indicates that the reference waveform sequence corresponding to the maximum value has a significant advantage in similarity with the sampled sequence. The waveform type corresponding to the target reference waveform sequence with the maximum similarity metric value is determined as the waveform type of the sampled sequence. If the value is not less than the preset interval, a preset rejection prompt message is output.

[0024] The method in this embodiment calls various typical reference waveform sequences from a preset reference type library and calculates the similarity metric between the sampled sequence and each reference waveform sequence. The difference between the maximum and the second largest similarity values ​​is used as the judgment criterion. It does not require the use of parameters such as waveform amplitude, pulse width, and rising edge for judgment, thus avoiding misjudgment when the feature intervals of different types of waveforms overlap, and improving the accuracy of waveform identification. At the same time, this invention only requires a preset reference waveform library and similarity calculation logic, without the need for complex model training and a large amount of computing resources, enabling rapid deployment in low-computing-power online monitoring terminals.

[0025] In a specific embodiment, before calculating the similarity metric between the sampled sequence and each reference waveform sequence, the method further includes: performing nonlinear time alignment and normalization processing on the sampled sequence and the reference waveform sequence to obtain optimized sampled sequences and optimized reference waveform sequences; then, calculating the similarity metric between the sampled sequence and each reference waveform sequence includes: calculating the similarity metric between the optimized sampled sequence and each optimized reference waveform sequence. Specifically, nonlinear time alignment processing can effectively solve the sequence misalignment problem caused by time axis distortion, making the sequences comparable in time. Normalization processing eliminates the interference of amplitude differences, allowing the similarity calculation to focus more on the shape features of the waveform. Calculating the similarity metric after these two preprocessing operations can more accurately reflect the true similarity between the sampled sequence and the reference waveform sequence, greatly improving the accuracy and reliability of lightning electromagnetic wave waveform identification.

[0026] In a specific embodiment, the similarity metric is obtained using the following formula:

[0027] in, For optimized sampling sequence With the j-th optimized reference waveform sequence Similarity metric between them For the optimized sampling sequence and the j-th optimized reference waveform sequence The number of data points in the middle, For optimized sampling sequence The value of the i-th data point. For the j-th optimized reference waveform sequence The value of the i-th data point.

[0028] In a specific embodiment, after acquiring the sampling sequence of the measured lightning electromagnetic wave, the method further includes: determining whether the sampling length of the sampling sequence exceeds a preset length; if the sampling length does not exceed the preset length, then executing the step of calling multiple reference waveform sequences from a preset reference type library; if the sampling length exceeds the preset length, then resampling the sampling sequence with the preset length as the target, obtaining a resampled sampling sequence, using the resampled sampling sequence as the sampling sequence, and executing the step of calling multiple reference waveform sequences from the preset reference type library. Specifically, the preset length is determined comprehensively based on factors such as system processing capability, the conventional length of reference waveform sequences, and analysis accuracy. If it is determined that the sampling length does not exceed the preset length, it indicates that the length of the current sampling sequence is within the range that the system can handle well. If the sampling length exceeds the preset length, in order to ensure the stability of system operation and the accuracy of analysis, the sampling sequence is resampled with the preset length as the target. During the resampling process, an appropriate interpolation or sampling algorithm, such as a linear interpolation algorithm, is used to adjust the length of the sampling sequence to the preset length while retaining the key information of the sampling sequence, thereby obtaining a resampled sampling sequence. Then, the resampled sampling sequence is used as the new sampling sequence, and the steps of calling various reference waveform sequences from the preset reference type library are executed.

[0029] In a specific embodiment, resampling the sampled sequence with the preset length as the target to obtain a resampled sampled sequence includes: obtaining a first spectral sequence of the sampled sequence using a preset anti-aliasing low-pass window; performing an inverse discrete Fourier transform of the first spectral sequence with a length of the preset length to obtain a time-domain sequence; and extracting the real part of the time-domain sequence to obtain the resampled sampled sequence. Specifically, when it is determined that the sampling length of the sampled sequence exceeds the preset length and resampling is required, the sampled sequence is processed using a preset anti-aliasing low-pass window. The preset anti-aliasing low-pass window can effectively suppress high-frequency noise and prevent spectral aliasing during subsequent transformations. By multiplying the anti-aliasing low-pass window with the sampled sequence, a first spectral sequence of the sampled sequence is obtained. An inverse discrete Fourier transform of the first spectral sequence with a length of the preset length is performed. The inverse discrete Fourier transform can convert frequency domain information back to the time domain. In this process, performing the transform according to the preset length can precisely control the length of the final time-domain sequence. Since the time-domain sequence obtained by the inverse discrete Fourier transform is usually in complex form, the real part of the time-domain sequence is extracted to obtain the resampled sequence, whose length meets the preset length requirement.

[0030] In a specific embodiment, obtaining the first spectral sequence of the sampled sequence using a preset anti-aliasing low-pass window includes: multiplying the sampled sequence by the preset anti-aliasing low-pass window to obtain a windowed sampled sequence; and symmetrically truncating the windowed sampled sequence while maintaining conjugate symmetry to obtain the first spectral sequence. Windowing the sampled sequence using a preset anti-aliasing low-pass window effectively suppresses high-frequency noise, avoids spectral aliasing in subsequent spectral analysis, and ensures the accuracy and integrity of the spectral information. Symmetrically truncating the windowed sampled sequence while maintaining conjugate symmetry not only meets the input requirements of the Discrete Fourier Transform but also reduces the amount of data and improves computational efficiency.

[0031] In a specific embodiment, if the sampling length does not exceed the preset length, the method further includes: determining whether the sampling length is equal to the preset length; if the sampling length is equal to the preset length, then performing the step of calling multiple reference waveform sequences from the preset reference type library; if the sampling length is less than the preset length, then performing a discrete Fourier transform on the sampling sequence to obtain a second spectral sequence, and zero-padding the second spectral sequence with the preset length as the target while maintaining conjugate symmetry to obtain a third spectral sequence with the preset length, using the third spectral sequence as the sampling sequence, and performing the step of calling multiple reference waveform sequences from the preset reference type library.

[0032] Specifically, if the sampling length is less than the preset length, to standardize the data length for subsequent accurate analysis, a Discrete Fourier Transform (DFT) is first performed on the sampled sequence to transform it from the time domain to the frequency domain, obtaining a second spectral sequence. Next, zero-padding is applied to the second spectral sequence while maintaining conjugate symmetry, with the preset length as the target. Conjugate symmetry is a crucial characteristic of the DFT; maintaining this characteristic ensures that the time-domain sequence obtained after the inverse transform is a real number sequence. After zero-padding, a third spectral sequence with the preset length is obtained. This third spectral sequence is then used as a new sampling sequence, and the steps of calling various reference waveform sequences from a preset reference type library are executed.

[0033] The resampling step is implemented using Fourier interpolation, and specifically includes: For a test sequence of length N Perform a Discrete Fourier Transform on (n=0,…,N-1) to obtain the spectrum. (k=0,…,N 1), its calculation formula is:

[0034] in .

[0035] When the target unified length L > N, zero-padding is performed on while maintaining conjugate symmetry to obtain a spectrum of length L ; when the target unified length L < N, is first multiplied by an anti-aliasing low-pass window , and then symmetric truncation is performed while maintaining conjugate symmetry to obtain a spectrum of length L .

[0036] Perform an inverse discrete Fourier transform of length L on the to obtain the resampled time-domain sequence (n = 0, …, L 1), and its calculation formula is:

[0037]

[0037] Take the real part of the as the resampling result of length L, where maintaining conjugate symmetry means satisfying:

[0038] Before calculating the similarity of the measured lightning electromagnetic wave sequence, it also includes performing non-linear time alignment processing on the measured sequence and the reference waveform sequence to reduce the local offset on the time axis; The non-linear alignment is based on the principle of time warping matching, and its objective function is:

[0039]

[0039] where, is the measured sequence, is the reference waveform sequence, and f (i) is the time mapping function, which is used to achieve local time alignment while ensuring the time order.

[0040]

[0040] After the alignment is completed, perform normalization processing on the measured sequence to eliminate the amplitude difference and improve the accuracy of the subsequent similarity calculation.

[0041] After the resampling step, it also includes: Perform z-score normalization processing on the measured sequence; The normalization processing includes obtaining the mean and the standard deviation of the measured sequence, and normalizing each sampling point in the measured sequence according to the following formula:

[0042]

[0042] where, is the i-th sampling point of the measured sequence, is the mean of all sampling points of the measured sequence, and ​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​The standard deviation of all sampling points in the tested sequence is denoted as . The sampling points are the z-score normalized samples of the sequence being tested.

[0043] Similarly, for the reference waveform sequence, its normalization result is:

[0044] in, For the i-th sampling point of the reference waveform sequence, The mean of all sampling points in the reference waveform sequence. The standard deviation of all sampling points in the reference waveform sequence is used. The sampling points are z-score normalized for the reference waveform sequence.

[0045] After the z-score normalization process is performed on the tested sequence, the Pearson correlation coefficient sampling sequence should also be pre-classified. The Pearson correlation coefficient sampling sequence is pre-classified according to the following formula:

[0046] Where X is the vector sequence of the sampled sequence being measured. This represents all sampled points after z-score normalization of the sequence under test.

[0047] Similarly, for the reference waveform sequence, its pre-classification process is as follows:

[0048] Where Y is the vector sequence of the sequence being measured. These are all sampled points after z-score normalization of the reference waveform sequence.

[0049] Find the maximum correlation coefficient using a linear scan algorithm. ρ max and its corresponding reference sequence index j max Finally, a threshold determination is performed when... ρ max Greater than or equal to the preset threshold θ At that time, and ρ max The closer the value is to 1, the stronger the correlation between the measured sequence and the reference waveform sequence; conversely, the weaker the correlation, the weaker the correlation.

[0050] The linear scan algorithm initializes the state variable (( ρ max , j max )←( Establish a baseline (∞, 0) to ensure that the first correlation coefficient will necessarily trigger an update; During the traversal, the algorithm employs an incremental extreme value tracking mechanism: for each index j∈{1,2,…,k}, the state update rule is executed ( ρ max , j max )←( ρ j , j )),when ρ j > ρ max If the condition is met, then the current state remains unchanged; otherwise, the current state is maintained. This process requires only a single traversal to complete the global extremum detection, ultimately yielding a mathematically rigorous optimal solution: ( ρ max , j max )=(max 1≤j≤kρj argmax 1≤j≤kρj ),satisfy ρ max = ρ jmax ≥ ρ j For all j ≠ jmax Established. Specific Implementation Example 1: In an electromagnetic wave monitoring system deployed along a high-voltage transmission line corridor, electromagnetic wave signals from a lightning event were collected on-site at a sampling rate of 10MHz. The measured waveform sequence obtained through the monitoring terminal device was used as input, and a corresponding reference waveform sequence was retrieved from a pre-built sample database. Subsequently, the measured data was processed and analyzed according to the method proposed in this invention to complete a typical lightning electromagnetic wave identification process. The specific steps are as follows: (1) Perform windowing operation on the collected lightning electromagnetic wave sequence and resample to a uniform length L=5000 points, and perform normalization processing to obtain a standardized test sequence.

[0052] (2) Retrieve similar reference waveform sequences from the reference prototype library. The reference prototype library contains four types of typical lightning waveforms: ground flashback (RS), cloud flashback (IC), initial breakdown pulse (PB), and narrow bipolar pulse (NB). Each type consists of multiple prototype sequences generated by aggregation within the class.

[0053] (3) Calculate the similarity measure between the tested sequence and each reference waveform sequence, obtain the category label corresponding to the maximum similarity, and extract the translation amount under the optimal match. In all similarity calculations, the Fast Fourier Transform is used for acceleration.

[0054] (4) Set the threshold θ=0.85 and the interval Δ=0.05. If the value corresponding to the maximum correlation coefficient is greater than the threshold and the difference between it and the second largest value is greater than the interval, then output the matching result; otherwise, output rejection.

[0055] (5) In this embodiment, the tested sequence has the highest similarity to the PB class reference prototype, with a maximum correlation coefficient of 0.91, which is higher than the threshold θ=0.85, and the difference between it and the second largest value is 0.07, which is greater than the interval Δ=0.05. Therefore, the output category is PB.

[0056] Specifically, please refer to the similarity waveform comparison between the tested sequence and the initial breakdown pulse sequence, and the difference waveform between the tested sequence and the initial breakdown pulse sequence. Figure 2 and Figure 3 Furthermore, regarding the complexity of similarity calculation, if the Pearson correlation coefficient is calculated directly in the time domain, it requires approximately [time complexity missing] when the number of sampling points L = 5000. After several multiplication and addition operations, using a fast cross-correlation implementation based on FFT, only about [number] operations are required. The computational complexity of multiple complex multiplication and addition operations is reduced by […]. Down to This achieved an acceleration of approximately 1 time.

[0057] Conclusion: The simulation and calculation process of this embodiment demonstrates that the method of the present invention can accurately identify typical lightning electromagnetic wave waveforms under limited sample conditions and low computing power. Compared with traditional methods that rely on complex neural networks, this method has the advantages of simple implementation, low computational overhead, and strong robustness to waveform translation, making it suitable for online real-time monitoring and edge computing scenarios. Specific Implementation Example 2: In an electromagnetic wave monitoring system deployed along a high-voltage transmission line corridor, electromagnetic wave signals from a lightning event were collected on-site at a sampling rate of 10MHz. The measured waveform sequence obtained through the monitoring terminal device was used as input, and a corresponding reference waveform sequence was retrieved from a pre-built sample database. Subsequently, the measured data was processed and analyzed according to the method proposed in this invention to complete a typical lightning electromagnetic wave identification process. The specific steps are as follows: (1) Perform windowing operation on the collected lightning electromagnetic wave sequence and resample to a uniform length L=5000 points, and perform normalization processing to obtain a standardized test sequence.

[0059] (2) Retrieve similar reference waveform sequences from the reference prototype library. The reference prototype library contains four typical lightning waveforms: RS, IC, PB, and NB. Each type consists of multiple prototype sequences generated by aggregation within the type.

[0060] (3) Calculate the similarity measure between the tested sequence and each reference waveform sequence, obtain the category label corresponding to the maximum similarity, and extract the translation amount under the optimal match.

[0061] (4) Set the threshold θ=0.9 and the interval Δ=0.05. If the value corresponding to the maximum correlation coefficient is greater than the threshold and the difference between it and the second largest value is greater than the interval, then output the matching result; otherwise, output rejection.

[0062] (5) In this embodiment, the tested sequence has the highest similarity to the PB class reference prototype, with a maximum correlation coefficient of 0.95, which is higher than the threshold θ=0.9, and the difference between it and the second largest value is 0.06, which is greater than the interval Δ=0.05. Therefore, the output category is PB.

[0063] Specifically, please refer to the waveform comparison of similarity between the tested sequence and the ground flash return stroke sequence, and the waveform comparison of difference between the tested sequence and the ground flash return stroke sequence. Figure 4 and Figure 5 Furthermore, regarding the complexity of similarity calculation, if the Pearson correlation coefficient is calculated directly in the time domain, it would require approximately [time period missing] when the number of sampling points L=5000. After several multiplication and addition operations, using a fast cross-correlation implementation based on FFT, only about [number] operations are required. The computational complexity of multiple complex multiplication and addition operations is reduced by […]. Down to This achieved an acceleration of approximately 14 times.

[0064] Conclusion: The simulation and calculation process of this embodiment demonstrates that the method of the present invention can accurately identify typical lightning electromagnetic wave waveforms under limited sample conditions and low computing power. Compared with traditional methods that rely on complex neural networks, this method has the advantages of simple implementation, low computational overhead, and strong robustness to waveform translation, making it suitable for online real-time monitoring and edge computing scenarios.

[0065] In a specific embodiment, please refer to Figure 6This is a schematic diagram of a waveform identification system for lightning electromagnetic waves according to a second embodiment of this application. The system includes: a first sequence acquisition module 201, a second sequence acquisition module 202, a similarity calculation module 203, a similarity extraction module 204, a first judgment module 205, a first output module 206, a second judgment module 207, a second output module 208, and a third output module 209. The first sequence acquisition module 201 is used to acquire a sampling sequence of the lightning electromagnetic wave under test. The second sequence acquisition module 202 is used to call various reference waveform sequences from a preset reference type library. The waveform types of the reference waveform sequences include ground flash return stroke sequences, cloud flash return stroke sequences, initial breakdown pulse sequences, and narrow bipolar pulse sequences. The similarity calculation module 203 is used to calculate the similarity between the sampling sequence and each reference waveform sequence. Similarity metric; the similarity extraction module 204 is used to obtain the maximum and second largest similarity metric values ​​in the similarity metric values; the first judgment module 205 is used to determine whether the maximum similarity metric value is less than a preset threshold; the first output module 206 is used to output a preset rejection prompt if it is less than the preset threshold; the second judgment module 207 is used to determine whether the difference between the maximum and second largest similarity metric values ​​is less than a preset interval if it is not less than the preset threshold; the second output module 208 is used to determine the waveform type corresponding to the target reference waveform sequence corresponding to the maximum similarity metric value as the waveform type of the sampling sequence if it is less than the preset interval; the third output module 209 is used to output a preset rejection prompt if it is not less than the preset interval.

[0066] The system in this embodiment calls various typical reference waveform sequences from a preset reference type library and calculates the similarity metric between the sampled sequence and each reference waveform sequence. The difference between the maximum and the second largest similarity values ​​is used as the judgment criterion. It does not require the use of parameters such as waveform amplitude, pulse width, and rising edge for judgment, thus avoiding misjudgment when the feature intervals of different types of waveforms overlap, and improving the accuracy of waveform identification. At the same time, the present invention only requires a preset reference waveform library and similarity calculation logic, without the need for complex model training and a large amount of computing resources, enabling rapid deployment in low-computing-power online monitoring terminals.

[0067] In a specific embodiment, the third embodiment of this application provides a waveform identification device for lightning electromagnetic waves, including a memory and a processor. The memory stores a computer program, and when the computer program is executed by the processor, the processor performs the steps of the method as described in any one of the first embodiments of this application.

[0068] In a specific embodiment, the fourth embodiment of this application provides a computer-readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the steps of the method as described in any one of the first embodiments of this application.

[0069] The above embodiments merely illustrate several implementation methods of this application, and their descriptions are relatively specific and detailed. However, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.

[0070] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments for application in other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for waveform identification of lightning electromagnetic waves, characterized in that, The method includes: Obtain the sampling sequence of the electromagnetic wave of the lightning being measured; Multiple reference waveform sequences can be retrieved from a preset reference type library. The waveform types of the reference waveform sequences include ground flash return stroke sequence, cloud flash return stroke sequence, initial breakdown pulse sequence, and narrow bipolar pulse sequence. Calculate the similarity metric between the sampled sequence and each reference waveform sequence; Obtain the maximum and second-largest similarity metrics from the similarity metrics; Determine whether the maximum value of the similarity metric is less than a preset threshold; If the value is less than the preset threshold, a preset rejection message will be output. If it is not less than the preset threshold, then determine whether the difference between the maximum value of the similarity metric and the second largest value of the similarity metric is less than the preset interval; If the interval is less than the preset interval, the waveform type corresponding to the target reference waveform sequence corresponding to the maximum similarity metric value is determined as the waveform type of the sampling sequence. If the interval is not less than the preset interval, a preset rejection message will be output.

2. The waveform identification method for lightning electromagnetic waves as described in claim 1, characterized in that, Before calculating the similarity metric between the sampled sequence and each reference waveform sequence, the method further includes: The sampling sequence and the reference waveform sequence are subjected to nonlinear time alignment and normalization processing to obtain an optimized sampling sequence and an optimized reference waveform sequence. The calculation of the similarity metric between the sampled sequence and each reference waveform sequence includes: Calculate the similarity metric between the optimized sampling sequence and each optimized reference waveform sequence.

3. The waveform identification method for lightning electromagnetic waves as described in claim 2, characterized in that, The similarity metric is obtained using the following formula: in, For optimized sampling sequence With the j-th optimized reference waveform sequence Similarity metric between them For the optimized sampling sequence and the j-th optimized reference waveform sequence The number of data points in the middle, For optimized sampling sequence The value of the i-th data point. For the j-th optimized reference waveform sequence The value of the i-th data point.

4. The waveform identification method for lightning electromagnetic waves as described in claim 1, characterized in that, After acquiring the sampling sequence of the measured lightning electromagnetic wave, the method further includes: Determine whether the sampling length of the sampling sequence exceeds a preset length; If the sampling length does not exceed the preset length, then the step of calling multiple reference waveform sequences from the preset reference type library is executed; If the sampling length exceeds the preset length, the sampling sequence is resampled with the preset length as the target to obtain the resampled sampling sequence. The resampled sampling sequence is then used as the sampling sequence, and the step of calling multiple reference waveform sequences from the preset reference type library is executed.

5. The waveform identification method for lightning electromagnetic waves as described in claim 4, characterized in that, The step of resampling the sampled sequence with the preset length as the target to obtain the resampled sampled sequence includes: The first spectral sequence of the sampling sequence is obtained using a preset anti-aliasing low-pass window; Perform an inverse discrete Fourier transform of the first spectral sequence with a length of the preset length to obtain a time-domain sequence; Extract the real part of the time-domain sequence to obtain the resampled sequence.

6. The waveform identification method for lightning electromagnetic waves as described in claim 5, characterized in that, The step of obtaining the first spectral sequence of the sampling sequence using a preset anti-aliasing low-pass window includes: The sampling sequence is multiplied by a preset anti-aliasing low-pass window to obtain a windowed sampling sequence. The windowed sampling sequence is then symmetrically truncated while maintaining conjugate symmetry to obtain the first spectral sequence.

7. The waveform identification method for lightning electromagnetic waves as described in claim 4, characterized in that, If the sampling length does not exceed the preset length, the method further includes: Determine whether the sampling length is equal to the preset length; If the sampling length is equal to the preset length, then the step of calling multiple reference waveform sequences from the preset reference type library is executed; If the sampling length is less than the preset length, a discrete Fourier transform is performed on the sampling sequence to obtain a second spectral sequence. With the preset length as the target, the second spectral sequence is zero-padded while maintaining conjugate symmetry to obtain a third spectral sequence with the preset length. The third spectral sequence is used as the sampling sequence, and the step of calling multiple reference waveform sequences from the preset reference type library is executed.

8. A waveform identification system for lightning electromagnetic waves, characterized in that, The system includes: a first sequence acquisition module, a second sequence acquisition module, a similarity calculation module, a similarity extraction module, a first judgment module, a first output module, a second judgment module, a second output module, and a third output module; The first sequence acquisition module is used to acquire the sampling sequence of the measured lightning electromagnetic wave; The second sequence acquisition module is used to call various reference waveform sequences from a preset reference type library. The waveform types of the reference waveform sequences include ground flash return sequence, cloud flash return sequence, initial breakdown pulse sequence, and narrow bipolar pulse sequence. The similarity calculation module is used to calculate the similarity metric between the sampled sequence and each reference waveform sequence; The similarity extraction module is used to obtain the maximum and second-largest similarity values ​​in the similarity metric. The first judgment module is used to determine whether the maximum value of the similarity metric is less than a preset threshold; The first output module is used to output a preset rejection prompt message if the value is less than a preset threshold. The second judgment module is used to determine whether the difference between the maximum value of the similarity metric and the second largest value of the similarity metric is less than a preset interval if the difference is not less than the preset threshold. The second output module is used to determine the waveform type of the target reference waveform sequence corresponding to the maximum similarity metric value as the waveform type of the sampling sequence if the similarity interval is less than a preset interval. The third output module is used to output a preset rejection prompt message if the interval is not less than a preset interval.

9. A waveform identification device for lightning electromagnetic waves, comprising a memory and a processor, characterized in that, The memory stores a computer program that, when executed by the processor, causes the processor to perform the steps of the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it causes the processor to perform the steps of the method as described in any one of claims 1 to 7.

Images