Lightning current waveform parameter determination method and device based on reversible neural network, equipment and storage medium

Abstract

The invention discloses a lightning current waveform parameter determination method and device based on a reversible neural network, equipment and a storage medium, and belongs to the technical field of power grid disaster prevention and reduction. Carrying out Gaussian distribution sampling processing on the target lightning current amplitude by using a reversible neural network model to obtain a plurality of hidden variables, and carrying out reverse mapping in combination with the target lightning current amplitude to obtain a plurality of lightning current waveform parameter combinations; performing constraint correction on each lightning current waveform parameter combination to obtain a plurality of corrected lightning current waveform parameter combinations; determining a plurality of parameter combination intervals based on a preset interval division rule; counting the corrected lightning current waveform parameter combination to obtain a probability value of each parameter combination interval; and determining lightning current waveform parameters of the target power transmission line based on the parameter combination interval with the highest probability value. According to the invention, the problem of low lightning current waveform parameter deduction accuracy in the prior art can be solved.

Description

Technical Field

[0001] This invention relates to the field of power grid disaster prevention and mitigation technology, and in particular to a method, apparatus, equipment and storage medium for determining lightning current waveform parameters based on a reversible neural network. Background Technology

[0002] Currently, in the field of power system lightning protection, traditional technologies have always focused on the positive correlation between lightning current waveform parameters and tower lightning withstand levels, establishing a mapping relationship between parameters and lightning withstand performance through electromagnetic simulation or empirical models recommended by international standards. However, in engineering practice, there is an urgent need to solve the problem of inverse deduction of potential lightning current parameter combinations and their probability distributions from known lightning withstand levels, and this problem has long lacked an effective solution.

[0003] Existing methods for inverse derivation of lightning current waveform parameters have fundamental flaws: for example, inverse derivation using empirical formulas can only output a single solution, failing to address complex scenarios involving multiple waveform parameter combinations corresponding to the same lightning withstand level; inverse derivation based on machine learning methods such as generative adversarial networks lacks physical constraints, resulting in over 30% of the generated results violating basic discharge laws such as the wavefront time must be shorter than the wave tail time, requiring significant manual screening costs; while the Monte Carlo simulation exhaustive method is theoretically sound, it requires thousands of hours of computing power when dealing with hundreds of thousands of parameter combinations and cannot construct continuous probability density functions. Therefore, developing a novel inverse derivation system integrating data-driven technologies has become a key breakthrough in improving the intelligence level of power grid lightning protection. Summary of the Invention

[0004] This invention provides a method, apparatus, device, and storage medium for determining lightning current waveform parameters based on a reversible neural network, which can solve the problem of low accuracy in the extrapolation of lightning current waveform parameters in the prior art.

[0005] To address the aforementioned technical problems, this invention provides a method for determining lightning current waveform parameters based on a reversible neural network, comprising:

[0006] The target lightning withstand current amplitude of the target transmission line was obtained by conducting a lightning strike simulation test on the target transmission line.

[0007] The target lightning current amplitude is input into a reversible neural network model so that the reversible neural network model generates the lightning current waveform parameters of the target transmission line.

[0008] The reversible neural network model generates the lightning current waveform parameters of the target transmission line by including the following steps:

[0009] The target lightning withstand current amplitude is sampled using a Gaussian distribution to obtain several latent variables;

[0010] The target lightning withstand current amplitude and several latent variables are inversely mapped to obtain several combinations of lightning current waveform parameters; wherein, the lightning current waveform parameter combinations include wavefront time, wave tail time and waveform type encoding;

[0011] Constraints are applied to each of the lightning current waveform parameter combinations to obtain several corrected lightning current waveform parameter combinations.

[0012] Based on a preset interval division rule, several parameter combination intervals are determined; wherein, the parameter combination intervals include wavefront intervals, wave tail intervals, and waveform types; wherein, the waveform types include double exponential waves, Heidler waves, and double oblique waves;

[0013] Statistical analysis was performed on several of the modified lightning current waveform parameter combinations to obtain the probability value of each parameter combination interval.

[0014] Based on the parameter combination range with the highest probability value, the lightning current waveform parameters of the target transmission line are determined.

[0015] As a preferred embodiment, the step of obtaining the target lightning withstand current amplitude of the target transmission line by conducting a lightning strike simulation test on the target transmission line includes:

[0016] Obtain the structural and electrical parameters of the target transmission line; the structural parameters include conductor type, insulator string type, and tower type; the electrical parameters include voltage level and grounding resistance value.

[0017] Based on the structural and electrical parameters of the target transmission line, a simulation model of the transmission line is constructed;

[0018] According to several preset current amplitudes, lightning currents are applied to the simulation model of the transmission line respectively, and the lightning current waveforms are recorded.

[0019] By judging the flashover situation of the lightning current waveform, the target lightning withstand current amplitude of the target transmission line is determined from several preset current amplitudes.

[0020] As a preferred embodiment, the training process of the reversible neural network model is as follows:

[0021] Training data is obtained from a preset database; wherein the training data includes several training samples; the training samples include historical lightning current waveform parameter combinations and historical lightning withstand current amplitudes;

[0022] The training data is normalized to obtain normalized training data;

[0023] The normalized training data is input into the original reversible neural network model, which is then trained to map the normalized training data and calculate the total loss value. The model parameters of the original reversible neural network model are adjusted according to the total loss value until the total loss value converges, thus forming a reversible neural network model.

[0024] As a preferred embodiment, the process of training the original reversible neural network model to map the normalized training data and calculate the total loss value includes:

[0025] Based on the combination of historical lightning current waveform parameters, a forward mapping is performed to obtain the predicted lightning withstand current amplitude and calculate the forward loss value.

[0026] The historical lightning withstand current amplitude was sampled using a Gaussian distribution to obtain several training latent variables;

[0027] Based on the historical lightning current amplitude and several training latent variables, an inverse mapping is performed to obtain the predicted lightning current waveform parameter combination, and the inverse loss value is calculated.

[0028] Based on the positive loss value and the negative loss value, a total loss value is generated.

[0029] As a preferred embodiment, the constraint correction of each lightning current waveform parameter combination yields several corrected lightning current waveform parameter combinations, including:

[0030] For each combination of lightning current waveform parameters, probability values ​​for several waveform types are determined based on the waveform type encoding;

[0031] The waveform type with the highest probability value is determined as the corrected waveform type;

[0032] Determine the time constraints based on the modified waveform type;

[0033] Based on the aforementioned time constraints, the wavefront time and wave tail time are constrained and corrected to obtain the corrected wavefront time and corrected wave tail time.

[0034] Based on the corrected wavefront time, the corrected wave tail time, and the corrected waveform type, a combination of corrected lightning current waveform parameters is generated.

[0035] As a preferred embodiment, the step of determining several parameter combination intervals based on preset interval division rules includes:

[0036] Based on preset interval division rules, several wavehead intervals, several wave tail intervals, and several waveform types are determined; wherein, the waveform types include double exponential waves, Heidler waves, and double oblique waves;

[0037] By performing a full permutation and combination of several wavefront intervals, several wave tail intervals, and several waveform types, several parameter combination intervals are obtained.

[0038] As a preferred embodiment, the step of statistically analyzing several combinations of corrected lightning current waveform parameters to obtain probability values ​​for each parameter combination interval includes:

[0039] For each parameter combination interval, select the lightning current waveform parameter combination from several corrected lightning current waveform parameter combinations where the corrected wavefront time is in the wavefront interval of the parameter combination interval, the corrected wavetail time is in the wavetail interval of the parameter combination interval, and the corrected waveform type is the waveform type of the parameter combination interval. The determination condition meets the parameter combination.

[0040] Divide the number of parameter combinations that meet the conditions by the number of parameter combinations of the corrected lightning current waveform to obtain the probability value of the parameter combination interval.

[0041] Accordingly, the present invention provides a lightning current waveform parameter determination device based on a reversible neural network, comprising: a lightning current amplitude acquisition module and a waveform parameter generation module;

[0042] The lightning withstand current amplitude acquisition module is used to obtain the target lightning withstand current amplitude of the target transmission line by conducting a lightning strike simulation test on the target transmission line;

[0043] The waveform parameter generation module is used to input the target lightning withstand current amplitude into the reversible neural network model so that the reversible neural network model generates the lightning current waveform parameters of the target transmission line.

[0044] The waveform parameter generation module includes a sampling unit, a mapping unit, a correction unit, a division unit, a statistics unit, and a parameter generation unit.

[0045] The sampling unit is used to perform Gaussian distribution sampling processing on the target lightning withstand current amplitude to obtain several hidden variables;

[0046] The mapping unit is used to perform inverse mapping on the target lightning withstand current amplitude and several latent variables to obtain several combinations of lightning current waveform parameters; wherein, the combinations of lightning current waveform parameters include wavefront time, wavetail time and waveform type encoding;

[0047] The correction unit is used to perform constraint correction on each of the lightning current waveform parameter combinations to obtain several corrected lightning current waveform parameter combinations.

[0048] The division unit is used to determine several parameter combination intervals based on preset interval division rules; wherein, the parameter combination intervals include wavefront intervals, wavetail intervals, and waveform types; wherein, the waveform types include double exponential waves, Heidler waves, and double oblique waves;

[0049] The statistical unit is used to perform statistics on several combinations of the corrected lightning current waveform parameters to obtain the probability value of each parameter combination interval.

[0050] The parameter generation unit is used to determine the lightning current waveform parameters of the target transmission line based on the parameter combination range with the highest probability value.

[0051] The present invention also provides a terminal device, comprising: a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, it implements the steps of the lightning current waveform parameter determination method based on a reversible neural network of the present invention.

[0052] The present invention also provides a computer-readable storage medium item, comprising: a stored computer program, wherein when the computer program is running, the device on which the computer-readable storage medium is located executes the steps of the lightning current waveform parameter determination method based on the reversible neural network of the present invention.

[0053] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0054] This invention provides a method for determining lightning current waveform parameters based on a reversible neural network. The method involves conducting a lightning strike simulation test on a target transmission line to obtain the target lightning current amplitude. This target lightning current amplitude is then input into a reversible neural network model, which performs the following steps: Gaussian distribution sampling is applied to the target lightning current amplitude to obtain multiple latent variables; inverse mapping is performed between the target lightning current amplitude and the latent variables to obtain multiple combinations of lightning current waveform parameters; constraint corrections are applied to each combination of lightning current waveform parameters to obtain multiple corrected combinations; multiple parameter combination intervals are determined based on a preset interval division rule; statistical analysis is performed on the corrected lightning current waveform parameter combinations to obtain the probability value of each parameter combination interval; and the lightning current waveform parameters of the target transmission line are determined based on the parameter combination interval with the highest probability value. This invention constructs a reversible neural network model that can generate multiple combinations of lightning current waveform parameters based on the lightning withstand level. By constraining and correcting the combinations of lightning current waveform parameters, the legality of the parameters is ensured. The model intelligently calculates the probability of each combination of lightning current waveform parameters and determines the combination of lightning current waveform parameters with the highest probability, effectively improving the accuracy of the finally determined lightning current waveform parameters and providing accurate data for lightning protection measures of the target transmission line. Attached Figure Description

[0055] To more clearly illustrate the technical solution of this application, the drawings used in the embodiments 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 from these drawings without creative effort.

[0056] Figure 1 A flowchart illustrating an embodiment of the lightning current waveform parameter determination method based on a reversible neural network provided by the present invention;

[0057] Figure 2 A flowchart illustrating an embodiment of the reversible neural network model training method provided by the present invention;

[0058] Figure 3 This is a schematic diagram of an embodiment of the lightning current waveform parameter determination device based on a reversible neural network provided by the present invention. Detailed Implementation

[0059] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings of the embodiments. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0060] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains; the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the application; the terms “comprising” and “having”, and any variations thereof, in the specification, claims, and foregoing description of the drawings are intended to cover non-exclusive inclusion.

[0061] In the description of the embodiments of this application, technical terms such as "first" and "second" are used only to distinguish different objects and should not be construed as indicating or implying relative importance or implicitly specifying the number, specific order, or primary and secondary relationship of the indicated technical features. In the description of the embodiments of this application, "multiple" means two or more, unless otherwise explicitly defined.

[0062] 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.

[0063] In the description of the embodiments in this application, the term "and / or" is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this document generally indicates that the preceding and following related objects have an "or" relationship.

[0064] In the description of the embodiments of this application, the term "multiple" refers to two or more (including two), similarly, "multiple sets" refers to two or more (including two sets), and "multiple pieces" refers to two or more (including two pieces).

[0065] In the description of the embodiments of this application, unless otherwise expressly specified and limited, technical terms such as "installation," "connection," "joining," and "fixing" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral part; they can refer to a mechanical connection or an electrical connection; 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. For those skilled in the art, the specific meaning of the above terms in the embodiments of this application can be understood according to the specific circumstances.

[0066] See Figure 1 To address the problem of low accuracy in the extrapolation of lightning current waveform parameters in existing technologies, an embodiment of the present invention provides a method for determining lightning current waveform parameters based on a reversible neural network. This method includes steps 101 to 102, each of which is detailed below:

[0067] Step 101: By conducting a lightning strike simulation test on the target transmission line, the target lightning withstand current amplitude of the target transmission line is obtained.

[0068] In this embodiment of the invention, lightning withstand level is a key concept in power system lightning protection, used to measure the ability of electrical equipment or lines to withstand lightning overvoltages without insulation damage. Lightning withstand level is typically expressed as the maximum lightning current amplitude that a transmission line can withstand (i.e., the lightning withstand current amplitude); the higher the value, the stronger the lightning protection performance of the equipment or line. Therefore, by conducting lightning strike simulation tests on the target transmission line, the target lightning withstand current amplitude can represent the lightning withstand level of the target transmission line, and the lightning current waveform parameters of the target transmission line can be subsequently deduced based on the target lightning withstand current amplitude.

[0069] As a preferred embodiment, the target lightning withstand current amplitude of the target transmission line is obtained by conducting a lightning strike simulation test on the target transmission line, including:

[0070] Obtain the structural and electrical parameters of the target transmission line; the structural parameters include conductor type, insulator string type, and tower type; the electrical parameters include voltage level and grounding resistance value.

[0071] Based on the structural and electrical parameters of the target transmission line, a simulation model of the transmission line is constructed;

[0072] According to several preset current amplitudes, lightning currents are applied to the simulation model of the transmission line respectively, and the lightning current waveforms are recorded.

[0073] By judging the flashover situation of the lightning current waveform, the target lightning withstand current amplitude of the target transmission line is determined from several preset current amplitudes.

[0074] In this embodiment of the invention, a lightning strike simulation test is conducted on the target transmission line. First, a simulation model of the target transmission line is constructed. Then, a lightning current is applied to the simulation model, and the lightning current waveform is observed to determine the target lightning current withstand amplitude of the target transmission line. Specifically, constructing the transmission line simulation model requires obtaining the structural and electrical parameters of the target transmission line. Structural parameters include conductor type (steel-cored aluminum stranded wire, carbon fiber composite conductor, etc.), insulator string type (porcelain insulator, glass insulator, composite insulator), and tower type (straight-line tower / tension tower), reflecting the characteristics of the line's physical structure. Electrical parameters include voltage level (110kV / 220kV / 500kV, etc.) and grounding resistance value, reflecting the line's electrical performance. The transmission line simulation model is built according to the actual line structure of the target transmission line, ensuring that the geometric dimensions and electrical connections are consistent with the actual structure.

[0075] In this embodiment of the invention, after constructing a transmission line simulation model, a lightning current is applied to the simulation model at a preset current amplitude (e.g., 20kA) for multiple tests, and the lightning current waveform is recorded simultaneously to determine whether flashover occurs (e.g., insulator string breakdown, conductor-to-tower discharge). If no flashover occurs in multiple tests, the current amplitude is increased (e.g., 25kA) and the test is repeated; if flashover occurs in multiple tests, the current level is decreased (e.g., 15kA) and the test is repeated; if some flashovers and some non-flashovers occur in multiple tests, the applied current amplitude is determined to be a critical value and is identified as the target lightning current amplitude for the target transmission line.

[0076] Step 102: Input the target lightning current amplitude into the reversible neural network model so that the reversible neural network model generates the lightning current waveform parameters of the target transmission line.

[0077] As a preferred embodiment, the training process of the reversible neural network model is as follows:

[0078] Training data is obtained from a preset database; wherein the training data includes several training samples; the training samples include historical lightning current waveform parameter combinations and historical lightning withstand current amplitudes;

[0079] The training data is normalized to obtain normalized training data;

[0080] The normalized training data is input into the original reversible neural network model, which is then trained to map the normalized training data and calculate the total loss value. The model parameters of the original reversible neural network model are adjusted according to the total loss value until the total loss value converges, thus forming a reversible neural network model.

[0081] In this embodiment of the invention, the reversible neural network model is composed of reversible block structures, which can perform forward and reverse computations. The training process of the reversible neural network model is as follows: first, multiple training samples are obtained from a preset database, and each training sample is normalized to obtain normalized training samples. The normalized training samples are then input into the original reversible neural network model for model training. When the model converges, a reversible neural network model is formed.

[0082] In this embodiment of the invention, the training data includes multiple training samples. Each training sample includes a combination of historical lightning current waveform parameters and a historical lightning current amplitude. The combination of historical lightning current waveform parameters includes historical wavefront time, historical wavetail time, and historical waveform type. A training sample can be represented as:

[0083]

[0084] In the formula, For the training dataset; The wavehead time (in μs) of the i-th training sample; w represents the tail time (in μs) of the i-th training sample. (i) Encode the waveform type of the i-th training sample (dimension n×1), where n is the number of waveform types; Let N be the lightning current amplitude of the i-th training sample (unit: kA); N is the number of training samples.

[0085] Assuming the waveform types include double exponential waves, Heidler waves, and double-slant waves, then when the waveform type is a double exponential wave, the waveform type code is [1,0,0]. T When the waveform type is a Heidler wave, the waveform type is encoded as [0,1,0]. T When the waveform type is a double-slant wave, the waveform type code is [0,0,1]. T .

[0086] In this embodiment of the invention, the training data can be normalized using the following formula to obtain normalized training data:

[0087]

[0088] In the formula, Let be the normalized wavefront time of the i-th training sample; Let be the normalized wavetail time of the i-th training sample; Let be the normalized withstand current amplitude of the i-th training sample; the values ​​in the formula can be set according to the actual situation.

[0089] In a preferred embodiment, training the original reversible neural network model to map the normalized training data and calculating the total loss value includes:

[0090] Based on the combination of historical lightning current waveform parameters, a forward mapping is performed to obtain the predicted lightning withstand current amplitude and calculate the forward loss value.

[0091] The historical lightning withstand current amplitude was sampled using a Gaussian distribution to obtain several training latent variables;

[0092] Based on the historical lightning current amplitude and several training latent variables, an inverse mapping is performed to obtain the predicted lightning current waveform parameter combination, and the inverse loss value is calculated.

[0093] Based on the positive loss value and the negative loss value, a total loss value is generated.

[0094] In this embodiment of the invention, after normalized training data is input into the original reversible neural network model, the original reversible neural network model is trained to perform forward mapping and inverse mapping based on the normalized training data. Specifically, the forward mapping is to obtain the predicted lightning withstand current amplitude based on the combination of historical lightning current waveform parameters.

[0095]

[0096] In the formula, f is the predicted lightning withstand current amplitude for the i-th training sample; θ It is a positive function.

[0097] By constructing a positive loss function, the positive loss value can be calculated after each positive mapping to assist model training. The mean squared error is used as the positive loss, and the calculation is performed by combining the predicted lightning current amplitude and the actual lightning current amplitude of each training sample.

[0098]

[0099] In the formula, Let be the actual withstand current amplitude of the i-th training sample.

[0100] Inverse mapping derives the predicted lightning current waveform parameter combination based on historical lightning current amplitude values. First, a Gaussian distribution is used to sample multiple training latent variables from the historical lightning current amplitude values:

[0101]

[0102] In the formula, z is the latent vector; I is the identity matrix; this formula means that the training latent variable z is obtained by random sampling from a normal distribution with a mean of 0 and a covariance of the identity matrix.

[0103] After determining multiple training latent variables z, an inverse mapping is performed based on these training latent variables z and historical lightning current amplitudes to obtain the predicted lightning current waveform parameter combination:

[0104]

[0105] In the formula, Let be the predicted wavefront time of the i-th training sample; Let be the predicted wave tail time for the i-th training sample; The predicted waveform type for the i-th training sample; It is the inverse function.

[0106] By constructing an inverse loss, the inverse loss value can be calculated after each inverse mapping to assist model training, using KL divergence as the inverse loss:

[0107]

[0108] In the formula, The parameter distribution generated for the model; This represents the conditional distribution of parameters in the database.

[0109] In this embodiment of the invention, the total loss of the original reversible neural network model includes not only the forward loss and the inverse loss, but also a physical constraint penalty term:

[0110]

[0111] In the formula, λ is the penalty coefficient, which can be taken as 0.5; M is the number of generated combinations of predicted lightning current waveform parameters. The physical constraint penalty term can constrain the predicted wavefront time in the generated combinations of predicted lightning current waveform parameters to be less than the predicted wavetail time, thereby improving the rationality of the combinations of predicted lightning current waveform parameters.

[0112] Based on the forward loss, the reverse loss, and the physical constraint penalty term, the total loss can be constructed:

[0113]

[0114] In the formula, Total loss; This is a positive loss; This is a reverse loss; This is a physical constraint penalty term.

[0115] See Figure 2 This is a flowchart illustrating an embodiment of the reversible neural network model training method provided by the present invention. The original reversible neural network model is trained using training data. The forward loss, inverse loss, and physical constraint loss are calculated separately, and the three are added together to calculate the total loss. The KL divergence is then used to determine whether it has converged. If it has not converged, parameter optimization is performed, and training is iteratively continued. If it has converged, training ends, and a reversible neural network model is generated.

[0116] Step 102 includes steps 201 to 206, each of which is detailed below:

[0117] Step 201: Perform Gaussian distribution sampling on the target lightning withstand current amplitude to obtain several latent variables.

[0118] In this embodiment of the invention, after the lightning withstand current amplitude is input into the reversible neural network model, the target lightning withstand current amplitude is first sampled using a Gaussian distribution to obtain multiple latent variables. For example, 10 samples are taken from the Gaussian distribution. 5 There are several latent variables, and these latent variables all conform to... That is, the mean and covariance of the Gaussian distribution are the same as those during model training, both of which are obtained by randomly sampling multiple latent variables from a normal distribution with a mean of 0 and a covariance of the identity matrix.

[0119] Step 202: Perform inverse mapping on the target lightning current amplitude and several latent variables to obtain several combinations of lightning current waveform parameters; wherein, the combinations of lightning current waveform parameters include wavefront time, wave tail time and waveform type encoding.

[0120] In this embodiment of the invention, based on the trained reversible neural network model, multiple combinations of lightning current waveform parameters are generated according to the generated latent variables and the target lightning withstand current amplitude. Each latent variable corresponds to one combination of lightning current waveform parameters.

[0121] Step 203: Perform constraint correction on each of the lightning current waveform parameter combinations to obtain several corrected lightning current waveform parameter combinations.

[0122] As a preferred embodiment, the lightning current waveform parameter combinations are constrained and corrected to obtain several corrected lightning current waveform parameter combinations, including:

[0123] For each combination of lightning current waveform parameters, probability values ​​for several waveform types are determined based on the waveform type encoding;

[0124] The waveform type with the highest probability value is determined as the corrected waveform type;

[0125] Determine the time constraints based on the modified waveform type;

[0126] Based on the aforementioned time constraints, the wavefront time and wave tail time are constrained and corrected to obtain the corrected wavefront time and corrected wave tail time.

[0127] Based on the corrected wavefront time, the corrected wave tail time, and the corrected waveform type, a combination of corrected lightning current waveform parameters is generated.

[0128] In this embodiment of the invention, after obtaining multiple combinations of lightning current waveform parameters, constraints are applied to each parameter in the combinations to improve the rationality of the lightning current waveform parameter combinations. The constraint correction includes time constraint correction for the wavefront and wave tail times, and determination of the waveform type based on waveform type encoding. Specifically, assuming there are three waveform types—double exponential wave, Heidler wave, and double-slant wave—the generated waveform type encoding is represented as [a, b, c]. TLet a, b, and c be the probability values ​​corresponding to the three waveform types. Therefore, by comparing the probability values ​​of each waveform type, the waveform type of the current lightning current waveform parameter combination can be determined, and this waveform type can be designated as the correction type. For example, if the waveform type is coded as [20%, 10%, 70%]... T If the waveform type of this lightning current waveform parameter combination is a double-slant wave, then the time constraints for the wavefront and wave tail times include basic time constraints such as the wavefront time must be less than the wave tail time, the wavefront time must be within a preset wavefront interval, and the wave tail time must be within a preset wave tail interval. In addition, corresponding correlation time constraints between the wavefront and wave tail times are set for different waveform types. Therefore, after determining the modified waveform type, the correlation time constraints are derived based on the modified waveform type, and combined with the basic time constraints, the time constraints for this lightning current waveform parameter combination are obtained. As an example, the correlation time constraints can be expressed as:

[0129]

[0130] In the formula, min(i) is the minimum value function, and max(·) is the maximum value function. This formula means that the wavefront time is at most 95% of the wavetail time, and the wavetail time is at least 105% of the wavefront time. In other words, this associated time constraint forces the wavetail time to be at least 5% greater than the wavefront time, while the wavefront time is at most 95% of the wavetail time, to prevent the difference between the two from being too large or too small, and to ensure that the parameters are within a reasonable relative range.

[0131] As an example, the basic time constraints that the wavefront time must be within a preset wavefront interval and the wavetail time must be within a preset wavetail interval can be expressed as:

[0132] T1=clip(T1, 1μs, 10μs), T2=clip(T2, 20μs, 200μs)

[0133] In the formula, clip(·) is the truncation function, 1μs is the minimum wavefront time, 10μs is the maximum wavefront time, 20μs is the minimum wavetail time, and 200μs is the maximum wavetail time. This formula means that when the wavefront time or wavetail time is less than its corresponding minimum value, it is truncated to the minimum value; when the wavefront time or wavetail time is greater than its corresponding maximum value, it is truncated to the maximum value; when the wavefront time or wavetail time is between its corresponding minimum and maximum value, the original value remains unchanged. By using the truncation function, the value of the variable can be kept within a reasonable range, making the result more controllable and more in line with practical significance.

[0134] Step 204: Based on the preset interval division rules, determine several parameter combination intervals; wherein, the parameter combination intervals include wavefront intervals, wavetail intervals, and waveform types; wherein, the waveform types include double exponential waves, Heidler waves, and double oblique waves.

[0135] As a preferred embodiment, based on a preset interval division rule, several parameter combination intervals are determined, including:

[0136] Based on preset interval division rules, several wavehead intervals, several wave tail intervals, and several waveform types are determined; wherein, the waveform types include double exponential waves, Heidler waves, and double oblique waves;

[0137] By performing a full permutation and combination of several wavefront intervals, several wave tail intervals, and several waveform types, several parameter combination intervals are obtained.

[0138] In this embodiment of the invention, the wavefront time range can be divided into multiple wavefront intervals based on the wavefront time value range and the partition step size; similarly, the wavetail time range can be divided into multiple wavetail intervals based on the wavetail time value range and the partition step size. By performing a full permutation and combination of these wavefront intervals, wavetail intervals, and waveform types, multiple parameter combination intervals can be obtained. For example, assuming the wavefront time value range is [1,10] and the partition step size is 1, the wavefront intervals are: [1,2), [2,3),...,[9,10]. Assuming the wavetail time value range is [20,200] and the partition step size is 10, the wavetail intervals are: [20,30), [30,40),...,[190,200]. The waveform types include double exponential waves, Heidler waves, and double-angled waves. After performing all permutations and combinations, the generated parameter combination interval can be represented as ([a,b), [c,d), k), where [a,b) is the wavefront interval, [c,d) is the wave tail interval, and k is the waveform type.

[0139] Step 205: Statistically analyze several combinations of the modified lightning current waveform parameters to obtain the probability value of each parameter combination interval.

[0140] As a preferred embodiment, the probability values ​​for each parameter combination interval are obtained by statistically analyzing several of the modified lightning current waveform parameter combinations, including:

[0141] For each parameter combination interval, select the lightning current waveform parameter combination from several corrected lightning current waveform parameter combinations where the corrected wavefront time is in the wavefront interval of the parameter combination interval, the corrected wavetail time is in the wavetail interval of the parameter combination interval, and the corrected waveform type is the waveform type of the parameter combination interval. The determination condition meets the parameter combination.

[0142] Divide the number of parameter combinations that meet the conditions by the number of parameter combinations of the corrected lightning current waveform to obtain the probability value of the parameter combination interval.

[0143] In this embodiment of the invention, after determining multiple parameter combination intervals, for each parameter combination interval, the number of corrected lightning current waveform parameter combinations falling within that parameter combination interval is counted. By dividing this number by the total number of corrected lightning current waveform parameter combinations, the probability value of that parameter combination interval can be calculated.

[0144] Step 206: Determine the lightning current waveform parameters of the target transmission line based on the parameter combination range with the highest probability value.

[0145] In this embodiment of the invention, after calculating the probability values ​​of each parameter combination interval, the parameter combination interval with the highest probability value is determined, and the lightning current waveform parameters of the target transmission line are determined to fall within this parameter combination interval. Then, probability calculations are performed on each modified lightning current waveform parameter combination within the parameter combination interval, and the modified lightning current waveform parameter combination with the highest probability value is determined as the lightning current waveform parameters of the target transmission line.

[0146] In this embodiment of the invention, after determining the lightning current waveform parameters of the target transmission line, the lightning protection measures for the target transmission line can be determined based on the wavefront time, wave tail time, and waveform type of the lightning current waveform parameters. For example, assuming a small wavefront time (e.g., 1–3 μs) and a medium wave tail time (e.g., 20–100 μs), it indicates that the lightning current rises rapidly, and the main threat is insulation breakdown caused by the steepness of the lightning wave. In this case, the lightning protection measures for the target transmission line can be determined as follows: increasing the line-to-ground gap, optimizing insulation coordination, selecting surge arresters with faster response speeds (e.g., ZnO surge arresters), and reducing the suspension height of the insulator string (reducing coupling). Assuming a medium wavefront time (e.g., 4–8 μs) and a large wave tail time (e.g., 100–500 μs), it indicates a long discharge duration, significant thermal effect, and a large impact on the grounding system. In this case, the lightning protection measures for the target transmission line can be determined as follows: thickening the grounding down conductor, increasing the length or number of grounding electrodes, using large-capacity energy dissipation type SPDs, and considering multi-point grounding and ring grounding networks.

[0147] Implementing the above embodiments has the following effects:

[0148] This invention provides a method for determining lightning current waveform parameters based on a reversible neural network. The method involves conducting a lightning strike simulation test on a target transmission line to obtain the target lightning current amplitude. This target lightning current amplitude is then input into a reversible neural network model, which performs the following steps: Gaussian distribution sampling is applied to the target lightning current amplitude to obtain multiple latent variables; inverse mapping is performed between the target lightning current amplitude and the latent variables to obtain multiple combinations of lightning current waveform parameters; constraint corrections are applied to each combination of lightning current waveform parameters to obtain multiple corrected combinations; multiple parameter combination intervals are determined based on a preset interval division rule; statistical analysis is performed on the corrected lightning current waveform parameter combinations to obtain the probability value of each parameter combination interval; and the lightning current waveform parameters of the target transmission line are determined based on the parameter combination interval with the highest probability value. This invention constructs a reversible neural network model that can generate multiple combinations of lightning current waveform parameters based on the lightning withstand level. By constraining and correcting the combinations of lightning current waveform parameters, the legality of the parameters is ensured. The model intelligently calculates the probability of each combination of lightning current waveform parameters and determines the combination of lightning current waveform parameters with the highest probability, effectively improving the accuracy of the finally determined lightning current waveform parameters and providing accurate data for lightning protection measures of the target transmission line.

[0149] like Figure 3 As shown, based on the above method embodiments, corresponding apparatus embodiments are provided;

[0150] One embodiment of the present invention provides a lightning current waveform parameter determination device based on a reversible neural network, comprising: a lightning current amplitude acquisition module and a waveform parameter generation module;

[0151] The lightning withstand current amplitude acquisition module is used to obtain the target lightning withstand current amplitude of the target transmission line by conducting a lightning strike simulation test on the target transmission line;

[0152] The waveform parameter generation module is used to input the target lightning withstand current amplitude into the reversible neural network model so that the reversible neural network model generates the lightning current waveform parameters of the target transmission line.

[0153] The waveform parameter generation module includes a sampling unit, a mapping unit, a correction unit, a division unit, a statistics unit, and a parameter generation unit.

[0154] The sampling unit is used to perform Gaussian distribution sampling processing on the target lightning withstand current amplitude to obtain several hidden variables;

[0155] The mapping unit is used to perform inverse mapping on the target lightning withstand current amplitude and several latent variables to obtain several combinations of lightning current waveform parameters; wherein, the combinations of lightning current waveform parameters include wavefront time, wavetail time and waveform type encoding;

[0156] The correction unit is used to perform constraint correction on each of the lightning current waveform parameter combinations to obtain several corrected lightning current waveform parameter combinations.

[0157] The division unit is used to determine several parameter combination intervals based on preset interval division rules; wherein, the parameter combination intervals include wavefront intervals, wavetail intervals, and waveform types; wherein, the waveform types include double exponential waves, Heidler waves, and double oblique waves;

[0158] The statistical unit is used to perform statistics on several combinations of the corrected lightning current waveform parameters to obtain the probability value of each parameter combination interval.

[0159] The parameter generation unit is used to determine the lightning current waveform parameters of the target transmission line based on the parameter combination range with the highest probability value.

[0160] It is understood that the above-described device embodiments correspond to the method embodiments of the present invention, and can implement the lightning current waveform parameter determination method based on a reversible neural network provided by any of the above-described method embodiments of the present invention.

[0161] It should be noted that the device embodiments described above are merely illustrative, and some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Furthermore, in the accompanying drawings of the device embodiments provided by this invention, the connection relationships between modules indicate that they have communication connections, which can specifically be implemented as one or more communication buses or signal lines. Those skilled in the art can understand and implement this without any creative effort.

[0162] Based on the above embodiments of the lightning current waveform parameter determination method based on reversible neural networks, another embodiment of the present invention provides a terminal device, which includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the lightning current waveform parameter determination method based on reversible neural networks of any embodiment of the present invention.

[0163] For example, in this embodiment, the computer program can be divided into one or more modules, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules may be a series of computer program instruction segments capable of performing a specific function, which describe the execution process of the computer program in the terminal device.

[0164] The terminal device may be a desktop computer, laptop, handheld computer, or cloud server, etc. The terminal device may include, but is not limited to, a processor and a memory.

[0165] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor can be a microprocessor or any conventional processor. The processor is the control center of the terminal device, connecting all parts of the terminal device via various interfaces and lines.

[0166] Based on the above-described method embodiments, another embodiment of the present invention provides a computer-readable storage medium including a stored computer program, wherein, when the computer program is executed, it controls the device where the computer-readable storage medium is located to execute the lightning current waveform parameter determination method based on a reversible neural network as described in any of the above-described method embodiments of the present invention.

[0167] The modules / units integrated in the device / terminal equipment, if implemented as software functional units and sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc.

[0168] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A method for determining lightning current waveform parameters based on a reversible neural network, characterized in that, include: The target lightning withstand current amplitude of the target transmission line was obtained by conducting a lightning strike simulation test on the target transmission line. The target lightning current amplitude is input into a reversible neural network model so that the reversible neural network model generates the lightning current waveform parameters of the target transmission line. The reversible neural network model generates the lightning current waveform parameters of the target transmission line by including the following steps: The target lightning withstand current amplitude is sampled using a Gaussian distribution to obtain several latent variables; The target lightning withstand current amplitude and several latent variables are inversely mapped to obtain several combinations of lightning current waveform parameters; wherein, the lightning current waveform parameter combinations include wavefront time, wavetail time and waveform type encoding; Constraints are applied to each of the lightning current waveform parameter combinations to obtain several corrected lightning current waveform parameter combinations. Based on preset interval division rules, several parameter combination intervals are determined; wherein, the parameter combination intervals include wavefront intervals, wavetail intervals, and waveform types; wherein, the waveform types include double exponential waves, Heidler waves, and double oblique waves; Statistical analysis was performed on several of the modified lightning current waveform parameter combinations to obtain the probability value of each parameter combination interval. Based on the parameter combination range with the highest probability value, the lightning current waveform parameters of the target transmission line are determined.

2. The method for determining lightning current waveform parameters based on a reversible neural network according to claim 1, characterized in that, The process of obtaining the target lightning withstand current amplitude of the target transmission line by conducting a lightning strike simulation test includes: Obtain the structural and electrical parameters of the target transmission line; the structural parameters include conductor type, insulator string type, and tower type; the electrical parameters include voltage level and grounding resistance value. Based on the structural and electrical parameters of the target transmission line, a simulation model of the transmission line is constructed; According to several preset current amplitudes, lightning currents are applied to the simulation model of the transmission line respectively, and the lightning current waveforms are recorded. By judging the flashover situation of the lightning current waveform, the target lightning withstand current amplitude of the target transmission line is determined from several preset current amplitudes.

3. The method for determining lightning current waveform parameters based on a reversible neural network according to claim 1, characterized in that, The training process of the reversible neural network model is as follows: Training data is obtained from a preset database; wherein the training data includes several training samples; the training samples include historical lightning current waveform parameter combinations and historical lightning withstand current amplitudes; The training data is normalized to obtain normalized training data; The normalized training data is input into the original reversible neural network model, which is then trained to map the normalized training data and calculate the total loss value. The model parameters of the original reversible neural network model are adjusted according to the total loss value until the total loss value converges, thus forming a reversible neural network model.

4. The method for determining lightning current waveform parameters based on a reversible neural network according to claim 3, characterized in that, The process of training the original reversible neural network model by mapping the normalized training data and calculating the total loss value includes: Based on the combination of historical lightning current waveform parameters, a forward mapping is performed to obtain the predicted lightning withstand current amplitude and calculate the forward loss value. The historical lightning withstand current amplitude was sampled using a Gaussian distribution to obtain several training latent variables; Based on the historical lightning current amplitude and several training latent variables, an inverse mapping is performed to obtain the predicted lightning current waveform parameter combination, and the inverse loss value is calculated. Based on the positive loss value and the negative loss value, a total loss value is generated.

5. The method for determining lightning current waveform parameters based on a reversible neural network according to claim 1, characterized in that, The constraint correction of each lightning current waveform parameter combination yields several corrected lightning current waveform parameter combinations, including: For each combination of lightning current waveform parameters, probability values ​​for several waveform types are determined based on the waveform type encoding; The waveform type with the highest probability value is determined as the corrected waveform type; Determine the time constraints based on the modified waveform type; Based on the aforementioned time constraints, the wavefront time and wave tail time are constrained and corrected to obtain the corrected wavefront time and corrected wave tail time. Based on the corrected wavefront time, the corrected wave tail time, and the corrected waveform type, a combination of corrected lightning current waveform parameters is generated.

6. The method for determining lightning current waveform parameters based on a reversible neural network according to claim 1, characterized in that, The determination of several parameter combination intervals based on preset interval division rules includes: Based on preset interval division rules, several wavehead intervals, several wave tail intervals, and several waveform types are determined; wherein, the waveform types include double exponential waves, Heidler waves, and double oblique waves; By performing a full permutation and combination of several wavefront intervals, several wave tail intervals, and several waveform types, several parameter combination intervals are obtained.

7. The method for determining lightning current waveform parameters based on a reversible neural network according to claim 1, characterized in that, The step of statistically analyzing several combinations of modified lightning current waveform parameters to obtain probability values ​​for each parameter combination interval includes: For each parameter combination interval, select the lightning current waveform parameter combination from several corrected lightning current waveform parameter combinations where the corrected wavefront time is in the wavefront interval of the parameter combination interval, the corrected wavetail time is in the wavetail interval of the parameter combination interval, and the corrected waveform type is the waveform type of the parameter combination interval. The determination condition meets the parameter combination. Divide the number of parameter combinations that meet the conditions by the number of parameter combinations of the corrected lightning current waveform to obtain the probability value of the parameter combination interval.

8. A device for determining lightning current waveform parameters based on a reversible neural network, characterized in that, include: Lightning current amplitude acquisition module and waveform parameter generation module; The lightning withstand current amplitude acquisition module is used to obtain the target lightning withstand current amplitude of the target transmission line by conducting a lightning strike simulation test on the target transmission line; The waveform parameter generation module is used to input the target lightning withstand current amplitude into the reversible neural network model so that the reversible neural network model generates the lightning current waveform parameters of the target transmission line. The waveform parameter generation module includes a sampling unit, a mapping unit, a correction unit, a division unit, a statistics unit, and a parameter generation unit. The sampling unit is used to perform Gaussian distribution sampling processing on the target lightning withstand current amplitude to obtain several hidden variables; The mapping unit is used to perform inverse mapping on the target lightning withstand current amplitude and several latent variables to obtain several combinations of lightning current waveform parameters; wherein, the combinations of lightning current waveform parameters include wavefront time, wavetail time and waveform type encoding; The correction unit is used to perform constraint correction on each of the lightning current waveform parameter combinations to obtain several corrected lightning current waveform parameter combinations. The division unit is used to determine several parameter combination intervals based on preset interval division rules; wherein, the parameter combination intervals include wavefront intervals, wavetail intervals, and waveform types; wherein, the waveform types include double exponential waves, Heidler waves, and double oblique waves; The statistical unit is used to perform statistics on several combinations of the corrected lightning current waveform parameters to obtain the probability value of each parameter combination interval. The parameter generation unit is used to determine the lightning current waveform parameters of the target transmission line based on the parameter combination range with the highest probability value.

9. A terminal device, characterized in that, The method includes a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the lightning current waveform parameter determination method based on any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, include: A stored computer program, wherein, when the computer program is executed, it controls the device containing the computer-readable storage medium to perform the lightning current waveform parameter determination method based on a reversible neural network as described in any one of claims 1-7.

Images