Power distribution network fault waveform generation method and system, storage medium and equipment

Abstract

The embodiment of the invention discloses a power distribution network fault waveform generation method, and the method comprises the steps: collecting key parameters of a power distribution network fault scene, and carrying out the preprocessing of the key parameters, and generating a structured fault condition vector; calculating a zero-mode equivalent capacitance ratio as a target amplitude ratio according to the key parameters of the power distribution network fault scene, and calculating a resonance angular frequency difference as a target frequency difference according to the key parameters of the power distribution network fault scene; configuring a time sequence convolution layer of a depth generator based on the target amplitude ratio and the target frequency difference, and determining an optimal depth generator according to the configured time sequence convolution layer and the structured fault condition vector; performing constraint calculation on the target amplitude ratio and the target frequency difference through the optimal depth generator to generate a three-phase current waveform; and performing COMTRADE standard format packaging on the three-phase current waveform to generate a fault waveform file.

Description

Technical Field

[0001] This invention relates to the field of power system technology, and in particular to a method, system, storage medium and device for generating fault waveforms in power distribution networks. Background Technology

[0002] Currently, with the widespread integration of distributed power sources with high penetration rates, the transient processes of faults in active distribution networks exhibit high complexity and uncertainty, posing a severe challenge to the accuracy of fault analysis technologies. To address this challenge, the industry commonly employs deep learning models such as generative adversarial networks (GANs) to generate distribution network fault waveforms through data-driven methods, aiming to provide data support for fault location algorithms and protection device testing. However, these mainstream methods have fundamental limitations: their generation models are often treated as black boxes, overly relying on the statistical characteristics of historical data while lacking explicit modeling of the transient physical laws of the power grid. This results in generated waveforms that, while approximating each other in the time domain, fail to accurately reflect the key physical differences in zero-mode currents upstream and downstream of the fault point.

[0003] The core flaw of traditional deep generation methods lies in their lack of physical consistency. Specifically, existing technologies fail to coordinate the design of key scenario conditions such as the relative location of the fault and the distributed power supply access method with the internal structure of the model, and also fail to effectively embed core physical constraints such as the amplitude ratio determined by the zero-mode equivalent capacitance ratio and the frequency difference determined by the circuit resonant angular frequency. As a result, the generated waveforms cannot meet the stringent requirements of fault segment location algorithms for the physical realism of transient features, greatly limiting their applicability and reliability in engineering practice.

[0004] Therefore, in the simulation and testing of active distribution network faults with incomplete information, there is an urgent need for a new method that can deeply integrate transient physical mechanisms and data-driven capabilities. This method needs to break through the traditional black-box generation framework and ensure that the generated fault waveforms not only have visual similarity, but also maintain consistency with the transient response laws of the real power grid in terms of physical essence, thereby providing a high-quality and highly reliable data foundation for advanced fault analysis functions. Summary of the Invention

[0005] Therefore, it is necessary to propose a method for generating fault waveforms in power distribution networks to address the above problems.

[0006] A method for generating fault waveforms in a power distribution network, the method comprising the following steps: Key parameters of power distribution network fault scenarios are collected, and the key parameters are preprocessed to generate a structured fault condition vector. The zero-mode equivalent capacitance ratio is calculated based on the key parameters of the power distribution network fault scenario, and is used as the target amplitude ratio. The resonant angular frequency difference is also calculated based on the key parameters of the power distribution network fault scenario, and is used as the target frequency difference. The temporal convolutional layer of the depth generator is configured based on the target amplitude ratio and the target frequency difference, and the optimal depth generator is determined based on the configured temporal convolutional layer and the structured fault condition vector. The optimal depth generator performs constraint calculations on the target amplitude ratio and the target frequency difference to generate a three-phase current waveform. The three-phase current waveforms are encapsulated in the COMTRADE standard format to generate a fault waveform file.

[0007] In the above scheme, the key parameters of the distribution network fault scenario include: Fault type, relative location of fault, total length of distribution network line, zero-mode capacitance parameters per unit length of line, zero-mode inductance parameters per unit length of line, transition resistance, type of distributed power source, location of distributed power source connection, and penetration rate of distributed power source.

[0008] In the above scheme, the preprocessing of the key parameters to generate a structured fault condition vector specifically includes: The relative location of the fault, the total length of the distribution network line, the zero-mode capacitance parameter per unit length of the line, the zero-mode inductance parameter per unit length of the line, the transition resistance, and the penetration rate of distributed power sources are normalized to generate normalized values. The fault type, distributed power source type, and distributed power source connection location are encoded to generate binary vectors. The normalized value is concatenated with the binary vector to generate a structured fault condition vector.

[0009] In the above scheme, the step of calculating the zero-mode equivalent capacitance ratio based on the distribution network fault scenario as the target amplitude ratio specifically includes: Based on the relative location of the fault and the total length of the distribution network lines in the aforementioned distribution network fault scenario, calculate the upstream and downstream line lengths of the fault point:

[0010]

[0011] in, The length of the line upstream of the fault point, For the relative location of the fault, For the total length of the distribution network lines, The length of the line downstream of the fault point; Obtain the zero-mode capacitance parameters per unit length of the line, and calculate the upstream and downstream zero-mode equivalent capacitances of the fault point based on these parameters:

[0012]

[0013] in, For the zero-mode equivalent capacitance upstream of the fault point, The zero-mode capacitance per unit length of the line. For the zero-mode equivalent capacitance downstream of the fault point, The length of the line downstream of the fault point, The length of the line upstream of the fault point; The ratio of the upstream zero-mode equivalent capacitance to the downstream zero-mode equivalent capacitance is set as the target amplitude ratio:

[0014] in, For the target amplitude ratio, The zero-mode equivalent capacitance upstream of the fault point. The zero-mode equivalent capacitance downstream of the fault point.

[0015] In the above scheme, the step of calculating the resonant angular frequency difference based on the key parameters of the distribution network fault scenario as the target frequency difference specifically includes: Calculate the zero-mode equivalent inductance upstream and downstream of the fault point based on the zero-mode inductance parameters per unit length of the line:

[0016]

[0017] in, For the zero-mode equivalent inductance upstream of the fault point, The length of the line upstream of the fault point, The zero-mode inductance parameter per unit length of the line, For the zero-mode equivalent inductance downstream of the fault point, The length of the line downstream of the fault point; Calculate the resonant angular frequency of the upstream circuit and the downstream circuit of the fault point using the following formulas:

[0018]

[0019] in, The resonant angular frequency of the upstream circuit at the fault point, The resonant angular frequency of the downstream circuit of the fault point, For the zero-mode equivalent inductance upstream of the fault point, For the zero-mode equivalent inductance downstream of the fault point, For the zero-mode equivalent capacitance upstream of the fault point, The zero-mode equivalent capacitance downstream of the fault point; The difference between the resonant angular frequency of the upstream circuit and the resonant angular frequency of the downstream circuit at the fault point is taken as the target frequency difference:

[0020] in, For the target frequency difference, The resonant angular frequency of the upstream circuit at the fault point, This is the resonant angular frequency of the circuit downstream of the fault point.

[0021] In the above scheme, the temporal convolutional layer that configures the depth generator based on the target amplitude ratio and target frequency difference, and determines the optimal depth generator according to the configured temporal convolutional layer and the structured fault condition vector, specifically includes: Based on the relative location of the fault, the feature channels within the depth generator used for processing temporal features are divided into upstream channel groups of the fault point. and downstream channel group of the fault point ; Using the target amplitude ratio For the upstream channel group of the fault point and downstream channel group of the fault point Configure the corresponding upstream attention weights and downstream attention weights for the fault points:

[0022]

[0023] in, For the target amplitude ratio, For upstream attention weights of the fault point, The downstream attention weight of the fault point; According to the target frequency difference The numerical range is used to dynamically configure the kernel dilation rate of the temporal convolutional layer and generate the configured temporal convolutional layer. The structured fault condition vector is input into the depth generator and FiLM modulated to generate residual blocks; The upstream channel group, the downstream channel group, the configured temporal convolutional layer, and the residual block are integrated to generate an optimal depth generator.

[0024] In the above scheme, the step of generating a three-phase current waveform by constraining the target amplitude ratio and the target frequency difference through the optimal depth generator specifically includes: The structured fault condition vector, random noise, preset sampling frequency, preset number of sampling points, and fault start time are input into the optimal depth generator to generate initial values ​​of the three-phase current waveform. The initial values ​​of the three-phase current waveforms are subjected to phase mode transformation to obtain the zero-mode current waveform; Based on the relative location of the fault, the zero-mode current waveform is divided into the zero-mode current waveform upstream of the fault point and the zero-mode current waveform downstream of the fault point. Calculate the actual amplitude ratio and actual frequency difference based on the zero-mode current waveform upstream of the fault point and the zero-mode current waveform downstream of the fault point. Based on the squared error between the target amplitude ratio and the actual amplitude ratio, and the squared error between the target frequency difference and the actual frequency difference, a modal physical consistency constraint loss function is constructed.

[0025] in, As the first preset weight, As the second preset weight, Let the modal physical consistency constraint loss function be . For the target amplitude ratio, For actual amplitude ratio, For target frequency difference, This represents the actual frequency difference. By minimizing the modal physical consistency constraint loss function The parameters of the optimal depth generator are adjusted, and the initial values ​​of the three-phase current waveform are iterated until the modal physical consistency constraint loss function is reached. Below the preset threshold The output current waveform meets the physical consistency requirements.

[0026] This application also proposes a distribution network fault waveform generation system, the system comprising: a structured fault condition vector determination unit, a physical feature waveform output unit, a generator construction unit, and a waveform output unit; The structured fault condition vector determination unit is used to collect key parameters of the power distribution network fault scenario, and to normalize and encode the key parameters to generate a structured fault condition vector. The physical characteristic waveform output unit is used to calculate the ratio of zero-mode equivalent capacitance, generate the target amplitude ratio, and calculate the resonant angular frequency to generate the target frequency difference based on the power distribution network fault scenario. The generator construction unit is used to configure temporal convolutional layers based on the target amplitude ratio and the target frequency difference, and to construct a deep generator by combining the structured fault condition vector. The waveform output unit is used to perform modal physical consistency constraint calculation on the target amplitude ratio and target frequency difference through the depth generator to generate a three-phase current waveform; and to encapsulate the three-phase current waveform in the COMTRADE standard format to generate a fault waveform file.

[0027] This application also proposes a readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the following steps: Key parameters of power distribution network fault scenarios are collected, and the key parameters are preprocessed to generate a structured fault condition vector. The zero-mode equivalent capacitance ratio is calculated based on the key parameters of the power distribution network fault scenario, and is used as the target amplitude ratio. The resonant angular frequency difference is also calculated based on the key parameters of the power distribution network fault scenario, and is used as the target frequency difference. The temporal convolutional layer of the depth generator is configured based on the target amplitude ratio and the target frequency difference, and the optimal depth generator is determined based on the configured temporal convolutional layer and the structured fault condition vector. The optimal depth generator performs constraint calculations on the target amplitude ratio and the target frequency difference to generate a three-phase current waveform. The three-phase current waveforms are encapsulated in the COMTRADE standard format to generate a fault waveform file.

[0028] This application also proposes a computer device, including a memory and a processor, wherein the memory stores a computer program, and the computer program is executed by the processor in the following steps: Key parameters of power distribution network fault scenarios are collected, and the key parameters are preprocessed to generate a structured fault condition vector. The zero-mode equivalent capacitance ratio is calculated based on the key parameters of the power distribution network fault scenario, and is used as the target amplitude ratio. The resonant angular frequency difference is also calculated based on the key parameters of the power distribution network fault scenario, and is used as the target frequency difference. The temporal convolutional layer of the depth generator is configured based on the target amplitude ratio and the target frequency difference, and the optimal depth generator is determined based on the configured temporal convolutional layer and the structured fault condition vector. The optimal depth generator performs constraint calculations on the target amplitude ratio and the target frequency difference to generate a three-phase current waveform. The three-phase current waveforms are encapsulated in the COMTRADE standard format to generate a fault waveform file.

[0029] The embodiments of this invention offer the following advantages: By pre-calculating two key physical quantities—the target amplitude ratio and the target frequency difference—under fault scenarios, and embedding them as core constraints into the configuration of the temporal convolutional layer of the deep generator and the entire waveform optimization process, this invention ensures that the generated three-phase current waveform not only closely approximates the actual waveform recording in appearance but also possesses a high degree of realism in the zero-mode transient characteristics upstream and downstream of the fault point, fundamentally solving the problem of insufficient physical consistency of the generated waveform. Attached Figure Description

[0030] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0031] in: Figure 1 This is a schematic diagram of a method for generating fault waveforms in a power distribution network in one embodiment. Detailed Implementation

[0032] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0033] In the following description, numerous specific details are set forth in order to provide a more thorough understanding of the invention; however, it will be apparent to those skilled in the art that the invention may be practiced without one or more of these details; in other instances, certain technical features well-known in the art have not been described in order to avoid confusion with the invention. It should be understood that the invention can be practiced in different forms and should not be construed as limited to the embodiments set forth herein; rather, these embodiments are provided to make the disclosure thorough and complete and to fully convey the scope of the invention to those skilled in the art.

[0034] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the invention. When used herein, the singular forms “a,” “an,” and “the” are also intended to include the plural forms, unless the context clearly indicates otherwise. The terms “comprising” and / or “including,” when used in this specification, identify the presence of said features, integers, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups. When used herein, the term “and / or” includes any and all combinations of the associated listed items.

[0035] To fully understand the present invention, a detailed structure will be presented in the following description in order to illustrate the technical solution proposed by the present invention; optional embodiments of the present invention are described in detail below, however, in addition to these detailed descriptions, the present invention may have other embodiments.

[0036] like Figure 1 As shown, in one embodiment, a method for generating distribution network fault waveforms is provided. This method includes steps S101 to S105, detailed below: S101. Collect key parameters of the power distribution network fault scenario and preprocess the key parameters to generate a structured fault condition vector. Specifically, preprocessing includes data normalization and one-hot encoding. By collecting key parameters such as fault type, location, and transition resistance, and performing preprocessing such as normalization and one-hot encoding, heterogeneous information from different sources and with different dimensions can be transformed into numerical vectors that can be directly processed by deep learning models, ensuring the accuracy and consistency of all subsequent calculations and model construction.

[0037] Preferably, before preprocessing, the key parameters of the power distribution network fault scenario are cleaned and outlier removed.

[0038] In some embodiments, key parameters for distribution network fault scenarios include: Fault type, relative location of fault, total length of distribution network line, zero-mode capacitance parameters per unit length of line, zero-mode inductance parameters per unit length of line, transition resistance, type of distributed power source, location of distributed power source connection, and penetration rate of distributed power source.

[0039] In some embodiments, preprocessing of key parameters to generate a structured fault condition vector specifically includes: The relative location of the fault, the total length of the distribution network line, the zero-mode capacitance parameters per unit length of the line, the zero-mode inductance parameters per unit length of the line, the transition resistance, and the penetration rate of distributed power sources are normalized to generate normalized values. The fault type, distributed power source type, and distributed power source connection location are encoded to generate binary vectors. The normalized value and the binary vector are concatenated to generate a structured fault condition vector.

[0040] Preferably, the relative fault location is the ratio of the distance from the fault point to the busbar to the total length of the line, with a value between 0.0 and 1.0, which is directly used as the normalized value of the relative fault location. The transition resistance is transformed by natural logarithm to obtain the logarithmic transition resistance, and then the logarithmic transition resistance is transformed to the range of 0.0 to 1.0 through linear mapping to obtain the normalized value of the transition resistance. The distributed generation penetration rate is processed. The distributed generation penetration rate is a percentage of the installed capacity of distributed generation to the maximum load of the distribution network. The percentage form is converted to decimal form to obtain the normalized value of the distributed generation penetration rate.

[0041] Preferably, key parameters of the power distribution network fault scenario are converted into binary vectors through one-hot encoding, and a structured fault condition vector is generated by combining the normalized values ​​of the key parameters.

[0042] Specifically, fault types are encoded using one-hot encoding. Fault types include single-phase grounding and phase-to-phase short circuit. Single-phase grounding is encoded as [1, 0], and phase-to-phase short circuit is encoded as [0, 1], resulting in a fault type encoding vector. Distributed power source types are also encoded using one-hot encoding. Distributed power source types include inverter type and rotating type. Inverter type is encoded as [1, 0], and rotating type is encoded as [0, 1], resulting in a distributed power source type encoding vector. Distributed power source connection locations are also encoded using one-hot encoding. Distributed power source connection locations include intact lines, upstream of the fault point, and downstream of the fault point. Insulated lines are encoded as [1, 0, 0], upstream of the fault point is encoded as [0, 1, 0], and downstream of the fault point is encoded as [0, 0, 1], resulting in a distributed power source connection location encoding vector. The fault type encoding vector, the distributed power source type encoding vector, and the distributed power source connection location encoding vector are concatenated sequentially to generate a structured fault condition vector.

[0043] S102. Calculate the zero-mode equivalent capacitance ratio based on the key parameters of the distribution network fault scenario, and use it as the target amplitude ratio. Calculate the resonant angular frequency difference based on the key parameters of the distribution network fault scenario, and use it as the target frequency difference. In fact, the amplitude ratio and frequency difference directly reflect the physical differences in zero-mode current upstream and downstream of the fault point. Abstract fault scenario parameters are transformed into target amplitude ratios and target frequency differences with clear physical meaning through circuit principles. Through this step, the method is no longer a black-box data fitting approach, but rather has clear physical laws as a benchmark, solving the problem of lack of physical consistency in the background technology.

[0044] In some embodiments, the zero-mode equivalent capacitance ratio is calculated based on the distribution network fault scenario as the target amplitude ratio, specifically including: Based on the relative location of the fault and the total length of the distribution network lines in the fault scenario, calculate the lengths of the upstream and downstream lines of the fault point:

[0045]

[0046] in, The length of the line upstream of the fault point, For the relative location of the fault, For the total length of the distribution network lines, The length of the line downstream of the fault point; Obtain the zero-mode capacitance parameters per unit length of the line, and calculate the zero-mode equivalent capacitance upstream and downstream of the fault point based on them:

[0047]

[0048] in, For the zero-mode equivalent capacitance upstream of the fault point, The zero-mode capacitance per unit length of the line. For the zero-mode equivalent capacitance downstream of the fault point, The length of the line downstream of the fault point, The length of the line upstream of the fault point; The ratio of the upstream zero-mode equivalent capacitance to the downstream zero-mode equivalent capacitance is set as the target amplitude ratio:

[0049] in, For the target amplitude ratio, The zero-mode equivalent capacitance upstream of the fault point. The zero-mode equivalent capacitance downstream of the fault point.

[0050] Preferably, the lengths of the upstream and downstream lines of the fault point are obtained based on the relative location of the fault in the distribution network fault scenario and in combination with the topology of the distribution network. Specifically, the topology of the distribution network is obtained, which includes the total line length from the bus to the end of the line; the relative location of the fault is obtained, which is the ratio of the distance from the fault point to the bus to the total line length; the product of the relative location of the fault and the total line length is taken as the distance from the fault point to the bus, which is the upstream line length of the fault point; finally, the difference between the total line length and the upstream line length of the fault point is taken as the downstream line length of the fault point.

[0051] Based on the length of the upstream and downstream lines of the fault point, the zero-mode equivalent capacitance upstream and downstream of the fault point is obtained through zero-mode equivalent capacitance calculation. Specifically, the unit zero-mode capacitance is obtained, which is the zero-mode capacitance value corresponding to each kilometer of the distribution network line; the product of the line length upstream of the fault point and the unit zero-mode capacitance is taken as the upstream zero-mode equivalent capacitance of the fault point; the product of the line length downstream of the fault point and the unit zero-mode capacitance is taken as the downstream zero-mode equivalent capacitance of the fault point.

[0052] The ratio of the upstream zero-mode equivalent capacitance to the downstream zero-mode equivalent capacitance at the fault point is used as the target amplitude ratio.

[0053] In some embodiments, the resonant angular frequency difference is calculated based on key parameters of the distribution network fault scenario as the target frequency difference, specifically including: Calculate the zero-mode equivalent inductance upstream and downstream of the fault point based on the zero-mode inductance parameters per unit length of the line:

[0054]

[0055] in, For the zero-mode equivalent inductance upstream of the fault point, The length of the line upstream of the fault point, The zero-mode inductance parameter per unit length of the line, For the zero-mode equivalent inductance downstream of the fault point, The length of the line downstream of the fault point; Calculate the resonant angular frequency of the upstream circuit and the downstream circuit of the fault point using the following formulas:

[0056]

[0057] in, The resonant angular frequency of the upstream circuit at the fault point, The resonant angular frequency of the downstream circuit of the fault point, For the zero-mode equivalent inductance upstream of the fault point, For the zero-mode equivalent inductance downstream of the fault point, For the zero-mode equivalent capacitance upstream of the fault point, The zero-mode equivalent capacitance downstream of the fault point; The difference between the resonant angular frequency of the upstream circuit and the resonant angular frequency of the downstream circuit at the fault point is taken as the target frequency difference:

[0058] in, For the target frequency difference, The resonant angular frequency of the upstream circuit at the fault point, This is the resonant angular frequency of the circuit downstream of the fault point.

[0059] S103. Configure the temporal convolutional layer of the depth generator based on the target amplitude ratio and target frequency difference, and determine the optimal depth generator based on the configured temporal convolutional layer and the structured fault condition vector. This step deeply embeds prior physical knowledge into the network structure of the deep generator, giving it the ability to perceive physical phenomena. Specifically, it configures temporal convolutional layers by initializing attention weights using the target amplitude ratio and setting the convolutional kernel dilation rate using the target frequency difference. This essentially tells the generator which features to focus on and how to capture high-frequency or low-frequency information. Through this physically guided configuration, the final generator is a model tailored to the specific fault scenario and best conforms to the laws of physics.

[0060] Specifically, the internal feature channels used to represent temporal features in the deep generator are obtained; then the relative position of the fault is obtained, and the internal feature channels are divided into two parts according to the time dimension. The first part corresponds to the upstream region of the fault point, and the second part corresponds to the downstream region of the fault point. The division position is determined by the product of the relative position of the fault and the total length of the internal feature channels; the first part is taken as the upstream channel of the fault point, and the second part is taken as the downstream channel of the fault point.

[0061] It should be noted that internal feature channels refer to the channel dimensions used to transmit and process temporal features within the convolutional layers or residual blocks during the forward propagation of the depth generator.

[0062] Based on the target amplitude ratio, attention weights are initialized for the upstream and downstream channels of the fault point to obtain the initialized upstream and downstream channels of the fault point. Specifically, the attention weight of the upstream channel of the fault point is calculated based on the target amplitude ratio. The attention weight of the upstream channel of the fault point is equal to the ratio of the target amplitude ratio to the sum of the target amplitude ratio and 1. The attention weight of the downstream channel of the fault point is calculated. The attention weight of the downstream channel of the fault point is equal to the ratio of 1 to the sum of the target amplitude ratio and 1. The calculated attention weight of the upstream channel of the fault point is assigned to the attention mechanism parameter of the upstream channel of the fault point. The calculated attention weight of the downstream channel of the fault point is assigned to the attention mechanism parameter of the downstream channel of the fault point. After the assignment is completed, the initialized upstream and downstream channels of the fault point are obtained.

[0063] In some embodiments, the temporal convolutional layers of the depth generator are configured based on the target amplitude ratio and the target frequency difference. The optimal depth generator is determined based on the configured temporal convolutional layers and the structured fault condition vector. Specifically, this includes: Based on the relative location of the fault, the feature channels within the depth generator used for processing temporal features are divided into upstream channel groups of the fault point. and downstream channel group of the fault point ; Using the target amplitude ratio upstream channel group of the fault point and downstream channel group of the fault point Configure the corresponding upstream attention weights and downstream attention weights for the fault points:

[0064]

[0065] in, For the target amplitude ratio, For upstream attention weights of the fault point, The downstream attention weight of the fault point; Based on the target frequency difference The numerical range is used to dynamically configure the kernel dilation rate of the temporal convolutional layer and generate the configured temporal convolutional layer. The structured fault condition vector is input into the depth generator, FiLM modulation is performed, and residual blocks are generated. The upstream channel group, the downstream channel group, the configured temporal convolutional layer, and the residual block are integrated to generate the optimal depth generator.

[0066] Preferably, the high-frequency response capability required for the temporal convolutional layer is determined based on the target frequency difference; the larger the target frequency difference, the stronger the required high-frequency response capability. The convolutional kernel dilation rate is set based on the high-frequency response capability; the stronger the high-frequency response capability, the larger the convolutional kernel dilation rate. The channels of the temporal convolutional layer are divided into upstream channel group and downstream channel group based on the relative position of the fault. The downstream channel group is configured to use a larger convolutional kernel dilation rate to enhance the high-frequency feature extraction capability, while the upstream channel group is configured to use a smaller convolutional kernel dilation rate to retain low-frequency amplitude features. After completing the setting of the convolutional kernel dilation rate and the channel grouping configuration, the configured temporal convolutional layer is generated.

[0067] It should be noted that the convolution kernel expansion rate is set based on the high-frequency transient response requirements of the downstream region of the fault point reflected by the target frequency difference; the value is usually an integer power of 2 from 1 to 128, such as (1, 2, 4, 8, 16, 32, 64, 128).

[0068] The structured fault condition vector is input into the depth generator, FiLM modulation is performed, and residual blocks are generated. Specifically, the structured fault condition vector is input into the fully connected subnetwork corresponding to each residual block in the depth generator. In each fully connected subnetwork, a scaling factor vector and an offset factor vector specific to the residual block are generated based on the structured fault condition vector. Inside the residual block, the scaling factor vector is multiplied element-wise with the intermediate feature map of the residual block along the channel dimension, and the offset factor vector is added element-wise with the intermediate feature map of the residual block along the channel dimension to complete FiLM modulation. The modulated intermediate feature map is used as the output of the residual block. The above operations are performed sequentially on all residual blocks in the depth generator to finally generate the residual block.

[0069] It should be noted that the FiLM mechanism is a general method for conditionalizing deep neural networks. The core operation of the FiLM mechanism is as follows: ; in, Represents a residual block; Represents the scaling factor vector; Represents the offset factor vector; This represents the intermediate feature map.

[0070] The initialized upstream and downstream channels of the fault point, the configured temporal convolutional layer and residual block are integrated, assembled and parameter-bound to generate a depth generator.

[0071] Specifically, the upstream and downstream channels of the initial fault point are used as feature inputs and connected to the configured temporal convolutional layer; the output of the configured temporal convolutional layer is connected to the first residual block; then the output of the first residual block is connected to each subsequent residual block in sequence to form a cascaded structure; during the connection process, the structured fault condition vector is applied to the intermediate feature map of each residual block through the FiLM mechanism; after all connections and parameter bindings are completed, the optimal depth generator is formed.

[0072] S104. The target amplitude ratio and target frequency difference are constrained and calculated using the optimal depth generator to generate a three-phase current waveform. Iterative optimization of the generated results through physical consistency constraints ensures the physical accuracy of the final waveform. Even if the generator possesses physical sensing capabilities, its initial output may still contain deviations. This step calculates the actual amplitude ratio and actual frequency difference of the generated waveform and constructs a loss function with the target value to repeatedly correct the waveform. This forms a second layer of protection, ensuring that the final output three-phase current waveform is not only morphologically realistic but also highly accurate in its inherent physical characteristics, meeting the stringent requirements of downstream algorithms.

[0073] In some embodiments, a three-phase current waveform is generated by constraining the target amplitude ratio and the target frequency difference using an optimal depth generator, specifically including: The structured fault condition vector, random noise, preset sampling frequency, preset number of sampling points and fault start time are input into the optimal depth generator to generate the initial values ​​of the three-phase current waveform. The initial values ​​of the three-phase current waveforms are transformed by phase mode to obtain the zero-mode current waveform; Based on the relative location of the fault, the zero-mode current waveform is divided into the zero-mode current waveform upstream of the fault point and the zero-mode current waveform downstream of the fault point. Calculate the actual amplitude ratio and actual frequency difference based on the zero-mode current waveform upstream and downstream of the fault point; Based on the squared error between the target amplitude ratio and the actual amplitude ratio, and the squared error between the target frequency difference and the actual frequency difference, a modal physical consistency constraint loss function is constructed.

[0074] in, As the first preset weight, As the second preset weight, Let the modal physical consistency constraint loss function be . For the target amplitude ratio, For actual amplitude ratio, For target frequency difference, This represents the actual frequency difference. By minimizing the modal physical consistency constraint loss function Adjust the parameters of the optimal depth generator and iterate the initial values ​​of the three-phase current waveforms until the modal physical consistency constraint loss function is reached. Below the preset threshold The output current waveform meets the physical consistency requirements.

[0075] Preferably, if the above modal physical consistency constraint loss function Not lower than the preset threshold Then, random noise is regenerated, input into the optimal depth generator, and combined with the input parameters to generate initial values ​​for the three-phase current waveform. The process of generating the three-phase current waveform continues until the modal physical consistency constraint loss function is reached. Below the preset threshold The output current waveform meets the physical consistency requirements.

[0076] In some embodiments, a structured fault condition vector is input into the FiLM modulation part of the depth generator; a random noise vector with a dimension matching the input layer of the depth generator is generated; a preset sampling frequency and a preset number of sampling points are used as time dimension control parameters for the output layer of the depth generator; the fault start time is then used as a time reference and embedded into the output alignment part of the depth generator; the optimal depth generator scales and offsets the feature maps of each residual block based on the structured fault condition vector, and simultaneously upsamples the random noise layer by layer to generate time-domain signals; finally, the initial values ​​of the three-phase current waveforms containing phase A current, phase B current, and phase C current are output.

[0077] The preset sampling frequency is based on the frequency band characteristics of the transient signal of the power distribution network fault and the recording requirements of the COMTRADE standard, and is usually set in the range of 10kHz-12.8kHz; the preset number of sampling points is based on a complete power frequency cycle and the duration of a typical transient process, and is usually set in the range of 2048-4096 points.

[0078] Based on the initial values ​​of the three-phase current waveforms, the zero-mode current waveform is obtained through Karrenbauer phase-mode transformation. Specifically, based on the initial values ​​of the three-phase current waveforms, including the A-phase current sampling sequence, the B-phase current sampling sequence, and the C-phase current sampling sequence; the A-phase current sampling sequence, the B-phase current sampling sequence, and the C-phase current sampling sequence are aligned in time to form a three-phase current vector; then the three-phase current vector is multiplied by the Karrenbauer phase mode transformation matrix, which is: ; The first row of the calculated transformation result is the zero-mode current waveform; the final output is the zero-mode current waveform.

[0079] The zero-mode current waveform is divided into the zero-mode current waveform upstream of the fault point and the zero-mode current waveform downstream of the fault point. The maximum amplitude is calculated, and the maximum amplitude of the zero-mode current waveform upstream of the fault point and the maximum amplitude of the zero-mode current waveform downstream of the fault point are generated. The actual amplitude ratio is obtained. Specifically, the relative position of the zero-mode current waveform and the fault in the structured fault condition vector is obtained. The zero-mode current waveform is divided into two segments in chronological order: the first segment corresponds to the upstream region of the fault point, and the second segment corresponds to the downstream region of the fault point. The division point is determined by the product of the relative position of the fault and the total number of sampling points of the zero-mode current waveform. The first segment is taken as the zero-mode current waveform upstream of the fault point, and the second segment is taken as the zero-mode current waveform downstream of the fault point. The absolute value of the zero-mode current waveform upstream of the fault point is taken to find the maximum value, which is the maximum amplitude of the zero-mode current waveform upstream of the fault point. The absolute value of the zero-mode current waveform downstream of the fault point is taken to find the maximum value, which is the maximum amplitude of the zero-mode current waveform downstream of the fault point. The ratio of the maximum amplitude of the zero-mode current waveform upstream of the fault point to the maximum amplitude of the zero-mode current waveform downstream of the fault point is taken as the actual amplitude ratio.

[0080] The zero-mode current waveform is divided into the zero-mode current waveform upstream of the fault point and the zero-mode current waveform downstream of the fault point. The dominant frequency is calculated, the dominant frequency of the zero-mode current waveform upstream of the fault point and the dominant frequency of the zero-mode current waveform downstream of the fault point are generated, and the actual frequency difference is obtained. Specifically, the relative position of the zero-mode current waveform and the fault in the structured fault condition vector is obtained; the zero-mode current waveform is divided into two segments in chronological order, the first segment corresponding to the upstream region of the fault point and the second segment corresponding to the downstream region of the fault point, the division point being determined by the product of the relative position of the fault and the total number of sampling points of the zero-mode current waveform; the first segment is taken as the upstream zero-mode current waveform of the fault point, and the second segment is taken as the downstream zero-mode current waveform of the fault point; Hilbert-Huang transform is performed on the upstream zero-mode current waveform of the fault point to extract the instantaneous frequency sequence, and the center frequency of the frequency band with the largest amplitude energy in the instantaneous frequency sequence is counted as the dominant frequency of the upstream zero-mode current waveform of the fault point; Hilbert-Huang transform is performed on the downstream zero-mode current waveform of the fault point to extract the instantaneous frequency sequence, and the center frequency of the frequency band with the largest amplitude energy in the instantaneous frequency sequence is counted as the dominant frequency of the downstream zero-mode current waveform of the fault point; the difference between the dominant frequency of the downstream zero-mode current waveform and the dominant frequency of the upstream zero-mode current waveform of the fault point is taken as the actual frequency difference.

[0081] Based on the squared error between the target amplitude ratio and the actual amplitude ratio, and the squared error between the target frequency difference and the actual frequency difference, a modal physical consistency constraint loss is constructed. Specifically, based on the target amplitude ratio and the actual amplitude ratio, the difference between the target amplitude ratio and the actual amplitude ratio is calculated, and the difference is squared to obtain the amplitude ratio squared error; the target frequency difference and the actual frequency difference are obtained; the difference between the target frequency difference and the actual frequency difference is calculated, and the difference is squared to obtain the frequency difference squared error; the sum of the amplitude ratio squared error and the frequency difference squared error is used as the modal physical consistency constraint loss.

[0082] Based on the modal physical consistency constraint loss, the initial value of the three-phase current waveform is iteratively optimized until the error threshold is met, and the three-phase current waveform is generated. Specifically, it is determined whether the modal physical consistency constraint loss meets the preset error threshold. If the modal physical consistency constraint loss does not meet the preset error threshold, random noise is regenerated. The structured fault condition vector, the newly generated random noise, the preset sampling frequency, the preset number of sampling points, and the fault start time are input into the depth generator again to generate new initial values ​​for the three-phase current waveform. Based on the new initial values ​​for the three-phase current waveform, the complete process of Karrenbauer phase mode transformation, zero-mode current waveform division, maximum amplitude and main frequency calculation, actual amplitude ratio and actual frequency difference acquisition, and modal physical consistency constraint loss construction is repeated. The above judgment and regeneration operations are executed cyclically until the modal physical consistency constraint loss meets the preset error threshold. When the modal physical consistency constraint loss meets the preset error threshold, the current initial values ​​for the three-phase current waveform are output as the three-phase current waveform.

[0083] It should be noted that the preset error threshold is defined based on the tolerance range of the zero-mode current characteristic accuracy of the active distribution network fault location algorithm, and is typically set to a relative error of less than 5%. The active distribution network fault location algorithm is a calculation method used in distribution networks containing distributed power sources to automatically identify and determine the fault location based on the amplitude ratio and frequency difference of the transient zero-mode current.

[0084] S105. Encapsulate the three-phase current waveforms in the COMTRADE standard format to generate a fault waveform file.

[0085] COMTRADE is a common transient data exchange standard in the power system field. Encapsulating the generated waveforms in this format means that the file can be directly read and used by various power system analysis software, relay protection testing devices, and fault location algorithms. This bridges the final link from algorithm generation to industrial application, enabling this invention to truly serve the practical engineering needs of distribution network fault simulation and testing.

[0086] In some embodiments, the three-phase current waveforms are discretized and sampled to generate current sampling sequences for each phase; Specifically, based on the three-phase current waveform, the sampling time interval and the total number of sampling moments are determined according to the preset sampling frequency and the preset number of sampling points. Starting from the fault initiation moment, the A-phase current is sampled sequentially with the sampling time interval as the step size to obtain the A-phase current sampling sequence. The B-phase current is sampled at the same sampling moment to obtain the B-phase current sampling sequence. Finally, the C-phase current is sampled at the same sampling moment to obtain the C-phase current sampling sequence.

[0087] Based on the sampling frequency, number of sampling points, number of channels, and fault start time of the discretized sampling, write the configuration file format to generate the COMTRADE configuration file; Specifically, based on the sampling frequency, number of sampling points, number of channels, and fault start time, the following steps are taken: Following the COMTRADE configuration file format, the following steps are performed: First, write the file identifier line, which includes the fault waveform file name and fault start time. Next, write the data configuration line, which includes the number of channels, number of sampling points, and sampling frequency. Then, write the channel definition lines, defining channel 1 as IA (phase A current), channel 2 as IB (phase B current), and channel 3 as IC (phase C current). Each line includes the channel number, signal name, phase, unit, multiplier factor, and offset. After writing all lines, the COMTRADE configuration file is generated.

[0088] Arrange the current sampling sequences of each phase in time order to generate a COMTRADE data file; Specifically, the A-phase current sampling sequence, B-phase current sampling sequence, and C-phase current sampling sequence are obtained; then the total number of sampling points is determined; next, starting from the first sampling time, the values ​​of the A-phase current sampling sequence, B-phase current sampling sequence, and C-phase current sampling sequence at the first sampling time are extracted sequentially to form a row of data; the above extraction operation is repeated for each sampling time in chronological order until the Nth sampling time, where N is the total number of sampling points; all rows of data are arranged in chronological order to form a numerical matrix containing N rows and 3 columns per row; the numerical matrix is ​​saved as a text file to generate a COMTRADE data file.

[0089] Combine the COMTRADE configuration file and the COMTRADE data file into a fault waveform file.

[0090] Specifically, based on the COMTRADE configuration file and COMTRADE data file, the COMTRADE configuration file and COMTRADE data file are placed in the same directory. This ensures that the filenames of the COMTRADE configuration file and the COMTRADE data file have the same prefix, only differing in extension. The COMTRADE configuration file has the extension .cfg, and the COMTRADE data file has the extension .dat. After completing the file naming and directory organization, a fault waveform file is generated.

[0091] In summary, this invention achieves quantitative modeling of the transient physical characteristics of active distribution network faults by generating target amplitude ratios and target frequency differences. By constructing an optimal depth generator, it enables the generator to have prior perception capabilities of the zero-mode current differences between upstream and downstream of the fault point, thereby generating three-phase current waveforms that conform to actual transient characteristics. Ultimately, this supports the high-precision operation of the fault section location algorithm and effectively improves the engineering applicability of fault simulation and testing in incomplete information active distribution networks.

[0092] This application also proposes a distribution network fault waveform generation system, which includes: a structured fault condition vector determination unit, a physical feature waveform output unit, a generator construction unit, and a waveform output unit; The structured fault condition vector determination unit is used to collect key parameters of the distribution network fault scenario, and to normalize and encode the key parameters to generate a structured fault condition vector. The physical characteristic waveform output unit is used to calculate the ratio of zero-mode equivalent capacitance based on the fault scenario of the distribution network, generate the target amplitude ratio, calculate the resonant angular frequency, and generate the target frequency difference. The generator building unit is used to configure temporal convolutional layers based on the target amplitude ratio and the target frequency difference, and to build a deep generator by combining the structured fault condition vector. The waveform output unit is used to perform modal physical consistency constraint calculations on the target amplitude ratio and target frequency difference through the depth generator to generate three-phase current waveforms; and to encapsulate the three-phase current waveforms in the COMTRADE standard format to generate fault waveform files.

[0093] This application also proposes a readable storage medium storing a computer program, which, when executed by a processor, causes the processor to perform the following steps: Collect key parameters of power distribution network fault scenarios and preprocess the key parameters to generate structured fault condition vectors; The zero-mode equivalent capacitance ratio is calculated based on the key parameters of the distribution network fault scenario, and is used as the target amplitude ratio. The resonant angular frequency difference is also calculated based on the key parameters of the distribution network fault scenario, and is used as the target frequency difference. The temporal convolutional layer of the depth generator is configured based on the target amplitude ratio and the target frequency difference. The optimal depth generator is determined based on the configured temporal convolutional layer and the structured fault condition vector. The three-phase current waveform is generated by constraining the target amplitude ratio and target frequency difference using the optimal depth generator. The three-phase current waveforms are encapsulated in the COMTRADE standard format to generate fault waveform files.

[0094] This application also proposes a computer device, including a memory and a processor, wherein the memory stores a computer program, and the computer program is executed by the processor in the following steps: Collect key parameters of power distribution network fault scenarios and preprocess the key parameters to generate structured fault condition vectors; The zero-mode equivalent capacitance ratio is calculated based on the key parameters of the distribution network fault scenario, and is used as the target amplitude ratio. The resonant angular frequency difference is also calculated based on the key parameters of the distribution network fault scenario, and is used as the target frequency difference. The temporal convolutional layer of the depth generator is configured based on the target amplitude ratio and the target frequency difference. The optimal depth generator is determined based on the configured temporal convolutional layer and the structured fault condition vector. The three-phase current waveform is generated by constraining the target amplitude ratio and target frequency difference using the optimal depth generator. The three-phase current waveforms are encapsulated in the COMTRADE standard format to generate fault waveform files.

[0095] Those skilled in the art will understand that implementing all or part of the processes in the above embodiments can be accomplished by instructing related hardware through a computer program. The program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments described above. Any references to memory, storage, databases, or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), dual data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), Rambus direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0096] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0097] The embodiments described above are merely illustrative of 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 application's patent. 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. The embodiments disclosed above are merely preferred embodiments of the present invention and should not be construed as limiting the scope of the present invention. Therefore, equivalent variations made according to the claims of this invention are still within the scope of this invention.

Claims

1. A method of generating a fault waveform for a power distribution network, characterized by, The method comprises: Collecting key parameters of a power distribution network fault scene, and preprocessing the key parameters to generate a structured fault condition vector; Calculating a zero-mode equivalent capacitance ratio as a target amplitude ratio according to the key parameters of the power distribution network fault scene, and calculating a resonant angle frequency difference as a target frequency difference according to the key parameters of the power distribution network fault scene; Configuring a time convolution layer of a depth generator based on the target amplitude ratio and the target frequency difference, and determining an optimal depth generator according to the configured time convolution layer and the structured fault condition vector; Conducting constraint calculation on the target amplitude ratio and the target frequency difference through the optimal depth generator to generate a three-phase current waveform; Packaging the three-phase current waveform in a COMTRADE standard format to generate a fault waveform file.

2. The power distribution network fault waveform generation method of claim 1, wherein, The key parameters of the power distribution network fault scene comprise: Fault type, fault relative position, total length of a power distribution network line, zero-mode capacitance parameter of a line unit length, zero-mode inductance parameter of a line unit length, transition resistance, distributed power type, distributed power access position, and distributed power penetration rate.

3. The power distribution network fault waveform generation method of claim 2, wherein, The preprocessing of the key parameters to generate a structured fault condition vector specifically comprises: Normalizing the fault relative position, total length of a power distribution network line, zero-mode capacitance parameter of a line unit length, zero-mode inductance parameter of a line unit length, transition resistance, and distributed power penetration rate to generate normalized values; Encoding the fault type, distributed power type, and distributed power access position to generate a binary vector; Splicing the normalized values and the binary vector to generate a structured fault condition vector.

4. The power distribution network fault waveform generation method of claim 3, wherein, The calculation of a zero-mode equivalent capacitance ratio as a target amplitude ratio according to the power distribution network fault scene specifically comprises: Based on the fault relative position and the total length of a power distribution network line in the power distribution network fault scene, calculating the length of an upstream line and the length of a downstream line at a fault point: wherein, is the length of the line upstream of the fault point, is the relative position of the fault, is the total length of the lines of the distribution network, is the length of the line downstream of the fault point; Obtaining the zero-mode capacitance parameter of a line unit length, and calculating the upstream zero-mode equivalent capacitance at the fault point and the downstream zero-mode equivalent capacitance at the fault point according to the zero-mode capacitance parameter of a line unit length: wherein, C0is the zero-mode equivalent capacitance upstream of the fault point, C1is the zero-mode capacitance parameter per unit length of the line, C2is the zero-mode equivalent capacitance downstream of the fault point, L2is the length of the line downstream of the fault point, L1is the length of the line upstream of the fault point; The ratio of the upstream zero-mode equivalent capacitance to the downstream zero-mode equivalent capacitance is set as a target amplitude ratio: wherein, is the target amplitude ratio, is the zero-mode equivalent capacitance upstream of the fault point, is the zero-mode equivalent capacitance downstream of the fault point.

5. The power distribution network fault waveform generation method of claim 4, wherein, The calculation of a resonant angle frequency difference as a target frequency difference according to the key parameters of the power distribution network fault scene specifically comprises: Calculating the upstream zero-mode equivalent inductance at the fault point and the downstream zero-mode equivalent inductance at the fault point according to the zero-mode inductance parameter of a line unit length: wherein, L0is the zero-mode equivalent inductance upstream of the fault point, L is the line length upstream of the fault point, L0is the zero-mode inductance per unit length of the line, L0is the zero-mode equivalent inductance downstream of the fault point, L is the line length downstream of the fault point; Calculating the resonant angle frequency of an upstream loop at the fault point and the resonant angle frequency of a downstream loop at the fault point according to the following formula: wherein, is the resonant angular frequency of the circuit upstream of the fault point, is the resonant angular frequency of the circuit downstream of the fault point, is the zero-mode equivalent inductance upstream of the fault point, is the zero-mode equivalent inductance downstream of the fault point, is the zero-mode equivalent capacitance upstream of the fault point, is the zero-mode equivalent capacitance downstream of the fault point; Differencing the resonant angle frequency of the upstream loop at the fault point and the resonant angle frequency of the downstream loop at the fault point as the target frequency difference: wherein is the target frequency difference, is the resonant angular frequency of the circuit upstream of the fault point, is the resonant angular frequency of the circuit downstream of the fault point.

6. The power distribution network fault waveform generation method of claim 5, wherein, The configuration of a time convolution layer of a depth generator based on the target amplitude ratio and the target frequency difference, and the determination of an optimal depth generator according to the configured time convolution layer and the structured fault condition vector specifically comprise: the feature channels inside the depth generator for processing the time-series features are divided into an upstream channel group and a downstream channel group according to the relative position of the fault point and the fault point ; using the target amplitude ratio for the upstream channel group of the failure point and the downstream channel group of the failure point configure the corresponding failure point upstream attention weight, failure point downstream attention weight: wherein, is a target amplitude ratio, is an attention weight upstream of the fault point, is an attention weight downstream of the fault point; According to the target frequency difference The numerical range of the value, the dynamic configuration of the convolution kernel expansion rate of the timing convolution layer, and the generation of the configured timing convolution layer; Inputting the structured fault condition vector into the depth generator to generate a residual block through FiLM modulation; Integrating the upstream channel group at the fault point, the downstream channel group at the fault point, the configured time convolution layer, and the residual block to generate the optimal depth generator.

7. The power distribution network fault waveform generation method of claim 6, wherein, The target amplitude ratio and the target frequency difference are calculated by the optimal depth generator to generate a three-phase current waveform, and the method specifically comprises the following steps: The structured fault condition vector, random noise, preset sampling frequency, preset sampling point number and fault starting time are input into the optimal depth generator to generate a three-phase current waveform initial value; The three-phase current waveform initial value is subjected to phase-mode transformation to obtain a zero-mode current waveform; The zero-mode current waveform is divided into an upstream zero-mode current waveform and a downstream zero-mode current waveform according to the fault relative position; Actual amplitude ratio and actual frequency difference are calculated according to the upstream zero-mode current waveform and the downstream zero-mode current waveform; A modal physical consistency constraint loss function is constructed according to the square error of the target amplitude ratio and the actual amplitude ratio and the square error of the target frequency difference and the actual frequency difference; wherein, is a first preset weight, is a second preset weight, is a modal physical consistency constraint loss function, is a target amplitude ratio, is an actual amplitude ratio, is a target frequency difference, is an actual frequency difference; by minimizing the modal physical consistency constraint loss function , adjusting parameters of the optimal depth generator, iterating the three-phase current waveform initial value until the modal physical consistency constraint loss function is lower than a preset threshold , outputting the three-phase current waveform satisfying the physical consistency requirement.

8. A power distribution network fault waveform generation system, characterized by, The system comprises a structured fault condition vector determination unit, a physical characteristic waveform output unit, a generator construction unit and a waveform output unit. The structured fault condition vector determination unit is configured to collect key parameters of a power distribution network fault scene, normalize and encode the key parameters, and generate a structured fault condition vector. The physical characteristic waveform output unit is configured to calculate the ratio of zero-mode equivalent capacitance according to the power distribution network fault scene, generate a target amplitude ratio, and calculate a resonant angular frequency to generate a target frequency difference. The generator construction unit is configured to configure a time series convolution layer based on the target amplitude ratio and the target frequency difference, and construct a deep generator in combination with the structured fault condition vector. The waveform output unit is configured to calculate a modal physical consistency constraint of the target amplitude ratio and the target frequency difference by the deep generator to generate a three-phase current waveform, and package the three-phase current waveform in a COMTRADE standard format to generate a fault waveform file.

9. A readable storage medium, storing a computer program, characterized in that, The computer program is executed by the processor to make the processor execute the steps of the method of any one of claims 1 to 7. 10.A computer device, comprising a memory and a processor, and characterized in that, The memory stores a computer program, and the computer program is executed by the processor to make the processor execute the steps of the method of any one of claims 1 to 7. The memory stores a computer program, and the computer program is executed by the processor to make the processor execute the steps of the method of any one of claims 1 to 7.

Images