Methods, devices, systems, media, and terminal equipment for locating medium-voltage grounding faults.

Abstract

This invention provides a method, apparatus, medium, and terminal equipment for locating medium-voltage grounding faults, relating to the field of power grid fault location technology. The location method includes: obtaining a time-series traveling wave interference signal based on historical raw voltage signals and forming a sample dataset; constructing an interference signal prediction model based on the sample dataset using an evolutionary algorithm; determining the real-time line-mode voltage traveling wave signal and using the interference signal prediction model to determine the traveling wave interference prediction signal; determining the real-time clean traveling wave signal based on the difference between the real-time line-mode voltage traveling wave signal and the traveling wave interference prediction signal; determining whether a medium-voltage grounding fault has occurred; and, if a medium-voltage grounding fault has occurred, determining the location of the medium-voltage grounding fault using the real-time clean traveling wave signal. The method provided by this invention can accurately reconstruct the fault traveling wave and perform fault detection and location accordingly, improving the anti-interference capability of medium-voltage grounding fault detection and location, and increasing the accuracy of detection and location.

Description

Technical Field

[0001] This invention relates to the field of power grid fault location technology, specifically to a medium-voltage grounding fault location method, a medium-voltage grounding fault location device, a medium-voltage grounding fault location system, a machine-readable storage medium, and a terminal device. Background Technology

[0002] In power systems, medium-voltage distribution networks are a crucial link connecting transmission networks and end users, and their safe and stable operation is of paramount importance. However, grounding faults, especially single-phase grounding faults, are a frequent type of fault in medium-voltage distribution networks. If these faults are not handled promptly, they can not only damage equipment and affect power supply reliability, but may also lead to more serious catastrophic accidents such as phase-to-phase short circuits after prolonged operation. Locating medium-voltage grounding faults has always been a technical challenge. On the one hand, medium-voltage lines typically have many branches, complex structures, and significant electromagnetic interference, posing a significant challenge to the accurate extraction and analysis of fault signals. On the other hand, especially for medium-voltage single-phase grounding faults, the initial fault current is often small, and the fault characteristics are not obvious, making them highly concealed and difficult to detect in a timely manner, which also increases the difficulty of subsequent location.

[0003] In existing technologies, methods based on changes in electrical quantities at the time of a fault can be broadly categorized into two main types: transient methods and steady-state methods. Transient methods primarily focus on the brief but abundant transient signals generated at the moment of a fault, such as traveling waves. This method captures, processes, and analyzes these unique traveling wave signals to extract the fault location information they carry, thereby achieving fault location. Steady-state methods, on the other hand, focus on analyzing the changes in amplitude, phase angle, harmonics, and other steady-state electrical quantities in the power grid after a fault occurs, using these changes to estimate the fault location.

[0004] However, transient methods are easily affected by noise interference, difficulties in traveling wave extraction, and limitations in wavefront identification accuracy in practical applications, leading to unstable positioning results or large errors. Steady-state methods often require long data windows, are sensitive to changes in system operation, and have limited positioning accuracy when fault characteristics are not obvious. Therefore, improving the anti-interference capability and accuracy of medium-voltage grounding fault location has become an urgent technical problem to be solved in this field. Summary of the Invention

[0005] To address the technical problems of weak anti-interference capability and low accuracy in medium-voltage grounding fault location in existing technologies, this invention provides a medium-voltage grounding fault location method, a medium-voltage grounding fault location device, a medium-voltage grounding fault location system, a machine-readable storage medium, and a terminal device. Using this method, the fault traveling wave can be accurately reconstructed and the fault detected and located accordingly, thereby improving the anti-interference capability of medium-voltage grounding fault detection and location, and increasing the accuracy of detection and location.

[0006] To achieve the above objectives, the first aspect of the present invention provides a method for locating medium-voltage grounding faults. This method includes: obtaining a time-series traveling wave interference signal based on historical raw voltage signals and forming a sample dataset; wherein the sample dataset is established based on the time-series traveling wave interference signal and its corresponding acquisition time; constructing an interference signal prediction model based on the sample dataset using an evolutionary algorithm; determining a real-time line-mode voltage traveling wave signal and using the interference signal prediction model to determine a predicted traveling wave interference signal; determining a real-time clean traveling wave signal based on the difference between the real-time line-mode voltage traveling wave signal and the predicted traveling wave interference signal; determining whether a medium-voltage grounding fault has occurred; and, if a medium-voltage grounding fault has occurred, determining the location of the medium-voltage grounding fault using the real-time clean traveling wave signal.

[0007] Furthermore, the step of obtaining the time-series traveling wave interference signal based on the historical original voltage signal and forming a sample dataset includes: performing a phase-mode transformation on the historical original voltage signal to obtain a historical line-mode voltage traveling wave signal; performing a Fourier expansion on the historical line-mode voltage traveling wave signal to separate historical traveling wave interference signals of different frequencies; merging the historical traveling wave interference signals of different frequencies according to the acquisition time of the corresponding historical original voltage signal to obtain the time-series traveling wave interference signal; and obtaining the sample dataset based on the mapping between the acquisition time of the historical original voltage signal corresponding to the time-series traveling wave interference signal and the time-series traveling wave interference signal.

[0008] Further, the construction of the interference signal prediction model based on the sample dataset using an evolutionary algorithm includes the following steps: 1) determining the evolution termination condition; 2) generating an initial model population; wherein the initial model population includes multiple model individuals; 3) determining the fitness of each model individual based on the sample dataset; wherein the fitness of a model individual is used to characterize the prediction performance of that model individual; 4) determining whether the current model population meets the evolution termination condition; 5) if it is determined that the current model population does not meet the evolution termination condition, performing an evolutionary operation on the current model population to generate the next generation model population, and returning to step 3) to continue iteration; 6) if it is determined that the current model population meets the evolution termination condition, identifying the model individual with the highest fitness in the current model population as the interference signal prediction model.

[0009] Furthermore, the evolution termination condition includes: the fitness of at least one model individual in the model population is greater than a set fitness; or the number of iterations is greater than a set number.

[0010] Further, the generation of the initial model population includes: constructing a function symbol set and a terminal symbol set; wherein the variables in the terminal symbol set are selected from the data collection time dataset; generating multiple individual models based on the function symbol set and the terminal symbol set to obtain the initial model population; wherein each individual model consists of two parts, a first sequence and a second sequence, wherein the elements in the first sequence are selected from the function symbol set and the terminal symbol set, and the elements in the second sequence are selected from the terminal symbol set.

[0011] Furthermore, each individual model satisfies the following constraints: Where n is the length of the first sequence in the model individual, and m is the length of the second sequence in the model individual. This represents the maximum number of parameters processed by the function symbols in the function symbol set.

[0012] Furthermore, the fitness of individual models is determined in the following way: ;in, Let be the fitness of the i-th model individual; This represents the number of samples in the sample dataset. The predicted value of the interference signal is calculated by the i-th model individual based on the acquisition time included in the j-th sample. Let be the value of the interference signal contained in the j-th sample.

[0013] Furthermore, the evolutionary operation includes at least one of the following operations: retention operation, selection operation, replication operation, mutation operation, insertion operation, and recombination operation.

[0014] Furthermore, the interference signal prediction model is an explicit calculation model.

[0015] Furthermore, determining whether a medium-voltage grounding fault has occurred includes: training a grounding fault detection model; and inputting the real-time pure traveling wave signal into the grounding fault detection model to determine whether a medium-voltage grounding fault has occurred.

[0016] Further, the training of the ground fault detection model includes: performing a phase-mode transformation on the historical raw voltage signal to obtain a historical line-mode voltage traveling wave signal; wherein the historical raw voltage signal is collected from both fault and non-fault periods; using the interference signal prediction model to determine a historical traveling wave interference prediction signal corresponding to the historical line-mode voltage traveling wave signal; determining a historical clean traveling wave signal based on the difference between the historical line-mode voltage traveling wave signal and its corresponding historical traveling wave interference prediction signal; assigning a fault state label to the historical clean traveling wave signal based on the fault state of the historical raw voltage signal to which the historical clean traveling wave signal is based; wherein the fault state label includes a fault label and a non-fault label; constructing a detection sample dataset based on the historical clean traveling wave signal and its corresponding fault state label, and using the detection sample dataset to perform supervised training on a deep neural network to construct the ground fault detection model.

[0017] Furthermore, the deep neural network adopts a network architecture that integrates temporal neural networks and convolutional neural networks.

[0018] Furthermore, the step of determining the location of a medium-voltage grounding fault using a real-time clean traveling wave signal includes: performing wavelet transform on the real-time clean traveling wave signal to determine the wavelet transform modulus maxima; using the wavelet transform modulus maxima to determine the arrival time of the fault traveling wave front; and using the arrival time of the fault traveling wave front to determine the location of the medium-voltage grounding fault.

[0019] A second aspect of the present invention provides a medium-voltage grounding fault location device, comprising: an interference signal extraction and sample set preparation module, used to obtain a time-series traveling wave interference signal based on historical raw voltage signals and form a sample dataset; wherein the sample dataset is established based on the time-series traveling wave interference signal and its corresponding acquisition time; an interference signal prediction model construction module, used to construct an interference signal prediction model based on the sample dataset using an evolutionary algorithm; an interference signal prediction module, used to determine the real-time line-mode voltage traveling wave signal and use the interference signal prediction model to determine the traveling wave interference prediction signal; a real-time clean traveling wave signal calculation module, used to determine the real-time clean traveling wave signal based on the difference between the real-time line-mode voltage traveling wave signal and the traveling wave interference prediction signal; a fault determination module, used to determine whether a medium-voltage grounding fault has occurred; and a fault location module, used to determine the location of the medium-voltage grounding fault using the real-time clean traveling wave signal when it is determined that a medium-voltage grounding fault has occurred.

[0020] Furthermore, the step of obtaining the time-series traveling wave interference signal based on the historical original voltage signal and forming a sample dataset includes: performing a phase-mode transformation on the historical original voltage signal to obtain a historical line-mode voltage traveling wave signal; performing a Fourier expansion on the historical line-mode voltage traveling wave signal to separate historical traveling wave interference signals of different frequencies; merging the historical traveling wave interference signals of different frequencies according to the acquisition time of the corresponding historical original voltage signal to obtain the time-series traveling wave interference signal; and obtaining the sample dataset based on the mapping between the acquisition time of the historical original voltage signal corresponding to the time-series traveling wave interference signal and the time-series traveling wave interference signal.

[0021] Further, the construction of the interference signal prediction model based on the sample dataset using an evolutionary algorithm includes the following steps: 1) determining the evolution termination condition; 2) generating an initial model population; wherein the initial model population includes multiple model individuals; 3) determining the fitness of each model individual based on the sample dataset; wherein the fitness of a model individual is used to characterize the prediction performance of that model individual; 4) determining whether the current model population meets the evolution termination condition; 5) if it is determined that the current model population does not meet the evolution termination condition, performing an evolutionary operation on the current model population to generate the next generation model population, and returning to step 3) to continue iteration; 6) if it is determined that the current model population meets the evolution termination condition, identifying the model individual with the highest fitness in the current model population as the interference signal prediction model.

[0022] Furthermore, the evolution termination condition includes: the fitness of at least one model individual in the model population is greater than a set fitness; or the number of iterations is greater than a set number.

[0023] Further, the generation of the initial model population includes: constructing a function symbol set and a terminal symbol set; wherein the variables in the terminal symbol set are selected from the data collection time dataset; generating multiple individual models based on the function symbol set and the terminal symbol set to obtain the initial model population; wherein each individual model consists of two parts, a first sequence and a second sequence, wherein the elements in the first sequence are selected from the function symbol set and the terminal symbol set, and the elements in the second sequence are selected from the terminal symbol set.

[0024] Furthermore, each individual model satisfies the following constraints: Where n is the length of the first sequence in the model individual, and m is the length of the second sequence in the model individual. This represents the maximum number of parameters processed by the function symbols in the function symbol set.

[0025] Furthermore, the fitness of individual models is determined in the following way: ;in, Let be the fitness of the i-th model individual; This represents the number of samples in the sample dataset. The predicted value of the interference signal is calculated by the i-th model individual based on the acquisition time included in the j-th sample. Let be the value of the interference signal contained in the j-th sample.

[0026] Furthermore, the evolutionary operation includes at least one of the following operations: retention operation, selection operation, replication operation, mutation operation, insertion operation, and recombination operation.

[0027] Furthermore, the interference signal prediction model is an explicit calculation model.

[0028] Furthermore, determining whether a medium-voltage grounding fault has occurred includes: training a grounding fault detection model; and inputting the real-time pure traveling wave signal into the grounding fault detection model to determine whether a medium-voltage grounding fault has occurred.

[0029] Further, the training of the ground fault detection model includes: performing a phase-mode transformation on the historical raw voltage signal to obtain a historical line-mode voltage traveling wave signal; wherein the historical raw voltage signal is collected from both fault and non-fault periods; using the interference signal prediction model to determine a historical traveling wave interference prediction signal corresponding to the historical line-mode voltage traveling wave signal; determining a historical clean traveling wave signal based on the difference between the historical line-mode voltage traveling wave signal and its corresponding historical traveling wave interference prediction signal; assigning a fault state label to the historical clean traveling wave signal based on the fault state of the historical raw voltage signal to which the historical clean traveling wave signal is based; wherein the fault state label includes a fault label and a non-fault label; constructing a detection sample dataset based on the historical clean traveling wave signal and its corresponding fault state label, and using the detection sample dataset to perform supervised training on a deep neural network to construct the ground fault detection model.

[0030] Furthermore, the deep neural network adopts a network architecture that integrates temporal neural networks and convolutional neural networks.

[0031] Furthermore, the step of determining the location of a medium-voltage grounding fault using a real-time clean traveling wave signal includes: performing wavelet transform on the real-time clean traveling wave signal to determine the wavelet transform modulus maxima; using the wavelet transform modulus maxima to determine the arrival time of the fault traveling wave front; and using the arrival time of the fault traveling wave front to determine the location of the medium-voltage grounding fault.

[0032] A third aspect of the present invention provides a medium-voltage grounding fault location system, the system comprising: the medium-voltage grounding fault location device described above, used for fault determination and / or fault location; and a processing device, used for receiving fault determination results and / or fault location results from the medium-voltage grounding fault location device, and for monitoring, analyzing or processing the power system based on the fault determination results and / or fault location results.

[0033] A fourth aspect of the present invention provides a machine-readable storage medium storing instructions that, when executed by a processor, configure the processor to perform the medium-voltage grounding fault location method described above.

[0034] The fifth aspect of the present invention provides a terminal device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the medium-voltage grounding fault location method described above.

[0035] The present invention has at least the following technical effects through the technical solution provided by the present invention:

[0036] The medium-voltage grounding fault location method of this invention first obtains a time-series traveling wave interference signal based on historical raw voltage signals and forms a sample dataset. Based on the sample dataset, an interference signal prediction model is constructed using an evolutionary algorithm. Next, the real-time line-mode voltage traveling wave signal is determined, and the traveling wave interference prediction signal is determined using the interference signal prediction model. Based on the difference between the real-time line-mode voltage traveling wave signal and the traveling wave interference prediction signal, the real-time clean traveling wave signal is determined. Then, it is determined whether a medium-voltage grounding fault has occurred. If a medium-voltage grounding fault is determined, the location of the medium-voltage grounding fault is determined using the real-time clean traveling wave signal. The method provided by this invention can accurately reconstruct the fault traveling wave and perform fault detection and location accordingly, improving the anti-interference capability of medium-voltage grounding fault detection and location, and increasing the accuracy of detection and location.

[0037] Other features and advantages of the present invention will be described in detail in the following detailed description section. Attached Figure Description

[0038] The accompanying drawings are provided to further illustrate embodiments of the present invention and form part of the specification. They are used together with the following detailed description to explain the embodiments of the present invention, but do not constitute a limitation thereof. In the drawings:

[0039] Figure 1 A flowchart illustrating a medium-voltage grounding fault location method provided in an embodiment of the present invention;

[0040] Figure 2 This is a schematic diagram of a medium-voltage grounding fault location device provided in an embodiment of the present invention. Detailed Implementation

[0041] The specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are for illustration and explanation only and are not intended to limit the scope of the present invention.

[0042] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other.

[0043] In this invention, unless otherwise stated, directional terms such as "upper," "lower," "top," and "bottom" are generally used to describe the relative positions of components in relation to the directions shown in the accompanying drawings or in relation to the vertical, perpendicular, or gravitational directions.

[0044] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0045] Please refer to Figure 1 The first aspect of this invention provides a method for locating medium-voltage grounding faults, comprising: S101: obtaining a time-series traveling wave interference signal based on historical raw voltage signals and forming a sample dataset; wherein the sample dataset is established based on the time-series traveling wave interference signal and its corresponding acquisition time; S102: constructing an interference signal prediction model based on the sample dataset using an evolutionary algorithm; S103: determining the real-time line-mode voltage traveling wave signal and using the interference signal prediction model to determine the traveling wave interference prediction signal; S104: determining the real-time clean traveling wave signal based on the difference between the real-time line-mode voltage traveling wave signal and the traveling wave interference prediction signal; S105: determining whether a medium-voltage grounding fault has occurred; S106: if a medium-voltage grounding fault has been determined to have occurred, determining the location of the medium-voltage grounding fault using the real-time clean traveling wave signal.

[0046] Specifically, step S101 is executed first: the time-series traveling wave interference signal is obtained based on the historical original voltage signal, and a sample dataset is formed; wherein, the sample dataset is established based on the time-series traveling wave interference signal and its corresponding acquisition time.

[0047] Furthermore, the step of obtaining the time-series traveling wave interference signal based on the historical original voltage signal and forming a sample dataset includes: performing a phase-mode transformation on the historical original voltage signal to obtain a historical line-mode voltage traveling wave signal; performing a Fourier expansion on the historical line-mode voltage traveling wave signal to separate historical traveling wave interference signals of different frequencies; merging the historical traveling wave interference signals of different frequencies according to the acquisition time of the corresponding historical original voltage signal to obtain the time-series traveling wave interference signal; and obtaining the sample dataset based on the mapping between the acquisition time of the historical original voltage signal corresponding to the time-series traveling wave interference signal and the time-series traveling wave interference signal.

[0048] Specifically, in this embodiment of the invention, signal acquisition points are first deployed at both ends of the medium-voltage bus or feeder of the power grid to continuously acquire historical raw voltage signals. , Including three-phase voltage signals, This is the original historical voltage signal of phase A. This is the original historical voltage signal for phase B. This is the historical original voltage signal for phase C. (Regarding the historical original voltage signal...) Phase-mode transformation is performed to obtain the line-mode voltage traveling wave signal. In this embodiment, the Karrenbauer transform matrix is ​​used to transform the historical original voltage signal. Phase-mode transformation is performed to obtain the line-mode voltage traveling wave signal. , .

[0049] For line mode voltage traveling wave signal Perform a Fourier expansion to decompose the signal into multiple high-frequency periodic signals. Then, rank all the high-frequency periodic signals according to the acquisition time of their corresponding historical original voltage signals. The data were merged to obtain a time-series traveling wave interference signal dataset. , The collected time data was compiled into a collection time dataset. Collect the time dataset. Data and time-series traveling wave interference signal dataset The data in the dataset is mapped to obtain the sample dataset. .

[0050] Next, step S102 is executed: based on the sample dataset, an interference signal prediction model is constructed using an evolutionary algorithm.

[0051] Further, the construction of the interference signal prediction model based on the sample dataset using an evolutionary algorithm includes the following steps: 1) determining the evolution termination condition; 2) generating an initial model population; wherein the initial model population includes multiple model individuals; 3) determining the fitness of each model individual based on the sample dataset; wherein the fitness of a model individual is used to characterize the prediction performance of that model individual; 4) determining whether the current model population meets the evolution termination condition; 5) if it is determined that the current model population does not meet the evolution termination condition, performing an evolutionary operation on the current model population to generate the next generation model population, and returning to step 3) to continue iteration; 6) if it is determined that the current model population meets the evolution termination condition, identifying the model individual with the highest fitness in the current model population as the interference signal prediction model.

[0052] Furthermore, the evolution termination condition includes: the fitness of at least one model individual in the model population is greater than a set fitness; or the number of iterations is greater than a set number.

[0053] Further, the generation of the initial model population includes: constructing a function symbol set and a terminal symbol set; wherein the variables in the terminal symbol set are selected from the data collection time dataset; generating multiple individual models based on the function symbol set and the terminal symbol set to obtain the initial model population; wherein each individual model consists of two parts, a first sequence and a second sequence, wherein the elements in the first sequence are selected from the function symbol set and the terminal symbol set, and the elements in the second sequence are selected from the terminal symbol set.

[0054] Furthermore, each individual model satisfies the following constraints: Where n is the length of the first sequence in the model individual, and m is the length of the second sequence in the model individual. This represents the maximum number of parameters processed by the function symbols in the function symbol set.

[0055] Furthermore, the fitness of individual models is determined in the following way: ;in, Let be the fitness of the i-th model individual; This represents the number of samples in the sample dataset. The predicted value of the interference signal is calculated by the i-th model individual based on the acquisition time included in the j-th sample. Let be the value of the interference signal contained in the j-th sample.

[0056] Furthermore, the evolutionary operation includes at least one of the following operations: retention operation, selection operation, replication operation, mutation operation, insertion operation, and recombination operation.

[0057] Furthermore, the interference signal prediction model is an explicit calculation model.

[0058] Specifically, in this embodiment of the invention, the evolution termination condition is first determined. In this embodiment, the evolution termination condition is: the fitness of at least one model individual in the model population is greater than a set fitness, or the number of iterations is greater than a set number. A function symbol set is then constructed. and end symbol set function symbol set Includes operators, elementary functions and other mathematical symbols, and the set of terminal symbols. This includes variables, constants, and parameterless operators. Here, all variables can be derived from the data acquisition time dataset. Based on function symbol set and end symbol set Multiple individual models are generated to obtain the initial model population: each individual model is a sequence. ,in, , That is, each model individual consists of two parts: a first sequence and a second sequence. The first sequence... The elements in the set are derived from the function symbol set. and end symbol set , the second sequence The elements in the set are derived from the end symbol set. When generating each individual model, constraints must be met. Where n is the length of the first sequence in the model individual, and m is the length of the second sequence in the model individual. for The maximum number of parameters processed by the function symbol in the code.

[0059] Using the time-series traveling wave interference signal data in the sample dataset as the actual detection values, calculate the fitness of each individual model in the current model population:

[0060] ;

[0061] in, Let be the fitness of the i-th model individual; This represents the number of samples in the sample dataset. The predicted value of the interference signal is calculated by the i-th model individual based on the acquisition time included in the j-th sample. Let be the numerical value of the interference signal contained in the j-th sample. Fitness can be used to evaluate the deviation between the predicted value and the actual detected value of an individual model.

[0062] After calculation, it is determined whether the fitness of at least one model in the model population is greater than a set fitness, or whether the number of iterations is greater than a set number. If neither condition is met, the current model population is evolved to obtain the next generation of models, and the iteration continues. If the fitness of at least one model in the model population is greater than a set fitness, or the number of iterations is greater than a set number, the evolution ends, and the model with the highest fitness in the current model population is determined as the interference signal prediction model. The interference signal prediction model is an explicit computational model, consisting of one or more analytical expressions. Future times are input to the analytical expressions to directly calculate and output the traveling wave interference prediction signal.

[0063] Evolutionary operations include at least one of the following: retention, selection, replication, mutation, insertion, and recombination. Retention is used to directly retain the model individual with the highest current fitness to the next generation of the model population. Selection is used to select superior model individuals from the model population based on fitness. Replication is used to completely replicate one or more selected model individuals to the next generation. Mutation is used to randomly adjust the parameters or structure of a model individual. Insertion is used to insert new gene fragments into a model individual, including inserting any fragment from one individual into any position within that individual, or inserting any fragment from one individual into any position within another individual. Recombination is used to exchange and combine gene fragments from multiple model individuals, including splitting two or more individuals at one or more sites and then reassembling or recombinating them.

[0064] Interference in medium-voltage power grids is typically multi-source, including high-frequency harmonic interference, electromagnetic radiation noise, and zero-sequence current measurement errors. The characteristic parameters (frequency, amplitude, attenuation coefficient) of each interference source influence each other, making individual modeling difficult. Evolutionary algorithms can simultaneously optimize parameters such as the frequency component weights of harmonic interference, the amplitude attenuation coefficient of noise, and the coupling coefficient, effectively escaping local optima and finding a globally optimal solution that balances the influence of each interference source. Furthermore, interference signals from medium-voltage grounding faults often exhibit strong nonlinear coupling with line equipment and loads (such as frequency converters, transformers, and nonlinear loads), line parameters (such as distributed capacitance and impedance), and operating conditions (such as load changes). Evolutionary algorithms do not require pre-defined linearity or stationarity of the model; they can automatically learn the nonlinear relationships between interference signals and multiple influencing factors (line length, line equipment, line load, etc.) through population iteration (selection, crossover, mutation), constructing a more realistic interference model. Moreover, evolutionary algorithms do not require prior knowledge of the interference signal's generation mechanism; only historical traveling wave data needs to be input to automatically uncover the hidden characteristics of the interference signal through population evolution. This allows for a more accurate quantification of the impact of interference on fault characteristics, thereby improving the anti-interference capability and accuracy of fault location.

[0065] Next, step S103 is executed: the real-time line-mode voltage traveling wave signal is determined, and the traveling wave interference prediction signal is determined using the interference signal prediction model.

[0066] Next, step S104 is executed: based on the difference between the real-time line-mode voltage traveling wave signal and the traveling wave interference prediction signal, the real-time clean traveling wave signal is determined.

[0067] Specifically, in this embodiment of the invention, acquisition points are deployed at both ends of the medium-voltage bus or feeder in the power grid to continuously acquire real-time voltage signals. The real-time voltage signals are then subjected to phase-mode transformation to obtain real-time line-mode voltage traveling wave signals. Finally, an interference signal prediction model is used to determine the predicted traveling wave interference signal. For example, at time... Based on the interference signal prediction model, the following is calculated: The predicted traveling wave interference signal for the time period is denoted as . Then, in [ During the specified time period, the voltage signal was continuously monitored, and the result was calculated. Real-time line-mode voltage traveling wave signal over a time period .Will[ Real-time line-mode voltage traveling wave signal over a time period With traveling wave interference prediction signal Do the difference, and you will get [ Real-time clean traveling wave signal over a time period .

[0068] Next, proceed to step S105: determine whether a medium-voltage grounding fault has occurred.

[0069] Furthermore, determining whether a medium-voltage grounding fault has occurred includes: training a grounding fault detection model; and inputting the real-time pure traveling wave signal into the grounding fault detection model to determine whether a medium-voltage grounding fault has occurred.

[0070] Further, the training of the ground fault detection model includes: performing a phase-mode transformation on the historical raw voltage signal to obtain a historical line-mode voltage traveling wave signal; wherein the historical raw voltage signal is collected from both fault and non-fault periods; using the interference signal prediction model to determine a historical traveling wave interference prediction signal corresponding to the historical line-mode voltage traveling wave signal; determining a historical clean traveling wave signal based on the difference between the historical line-mode voltage traveling wave signal and its corresponding historical traveling wave interference prediction signal; assigning a fault state label to the historical clean traveling wave signal based on the fault state of the historical raw voltage signal to which the historical clean traveling wave signal is based; wherein the fault state label includes a fault label and a non-fault label; constructing a detection sample dataset based on the historical clean traveling wave signal and its corresponding fault state label, and using the detection sample dataset to perform supervised training on a deep neural network to construct the ground fault detection model.

[0071] Furthermore, the deep neural network adopts a network architecture that integrates temporal neural networks and convolutional neural networks.

[0072] Specifically, in this embodiment of the invention, historical raw voltage signals from both fault and non-fault periods are collected. A phase-mode transformation is performed on the historical raw voltage signals to obtain historical line-mode voltage traveling wave signals. An interference signal prediction model is used to determine the corresponding historical traveling wave interference prediction signal for the same period. Based on the difference between the historical line-mode voltage traveling wave signal and its corresponding historical traveling wave interference prediction signal, a historical clean traveling wave signal is determined. The corresponding historical clean traveling wave signal is labeled according to whether a medium-voltage grounding fault occurred during the historical raw voltage signal collection period. The historical clean traveling wave signal and its corresponding fault status label are used as training examples. For example, training examples... 1 i in These are the d time series values ​​of the historical pure traveling wave signal; The label values ​​for whether a fault occurred during this process include two values, representing whether a medium-voltage grounding fault occurred or not, resulting in the training sample dataset R= Then, a deep neural network is constructed, employing a network architecture that integrates temporal and convolutional neural networks. This deep neural network is then trained under supervised supervision using a training sample dataset to obtain the ground fault detection model M.

[0073] The deep neural network constructed in this embodiment can be formally expressed using the following core calculation formula:

[0074] ;

[0075] ;

[0076] ;

[0077] ;

[0078] ;

[0079] in, In time step t The output of the input gate represents the weight of the new information; σ It is the sigmoid activation function; W xi It is the input weight matrix of the input gate; X t In time step t Input data; W hi It is the hidden state weight matrix of the input gate; H t 1 is in time step t The hidden state of 1; W ci It is the cell state weight matrix of the input gate; C t 1 is in time step t 1. Cell state; b i It is the bias term of the input gate; Convolution operation; The Hadamard product operation refers to multiplying the elements at corresponding positions. f t In time step t The output of the forget gate represents the state from the cell state. C t Information retained in 1; W xf It is the input weight matrix of the forget gate; W hf It is the hidden state weight matrix of the forget gate; W cf It is the cell state weight matrix of the forgetting gate; b f It is the bias term of the forget gate; tanh refers to the hyperbolic tangent activation function; W xc It is the weight matrix of the input data; W hc It is the weight matrix of the hidden states; b c It is a bias term for cell state; o t In time step t The output of the output gate; W xo It is the input weight matrix of the output gate; W ho It is the hidden state weight matrix of the output gate; W co It is the cell state weight matrix of the output gate; C t In time step t cellular state; b o It is the bias term of the output gate; hidden state H t It is an output gate o t and cell state C t go through The Hadamard product after activation function.

[0080] After obtaining the real-time clean traveling wave signal, it is input into the ground fault detection model M. The model output is used to determine whether a medium-voltage ground fault has occurred. If a medium-voltage ground fault is determined to have occurred, the location of the fault is determined using the real-time clean traveling wave signal.

[0081] The medium-voltage grounding fault location method provided by this invention can quickly and accurately determine whether a medium-voltage grounding fault has occurred. It not only significantly improves the accuracy and anti-interference capability of fault detection and reduces false alarms and missed alarms, but also realizes the intelligent and automated detection process, possessing good real-time performance and environmental adaptability.

[0082] Finally, step S106 is executed: if a medium-voltage grounding fault is determined to have occurred, the location of the medium-voltage grounding fault is determined using a real-time pure traveling wave signal.

[0083] Furthermore, the step of determining the location of a medium-voltage grounding fault using a real-time clean traveling wave signal includes: performing wavelet transform on the real-time clean traveling wave signal to determine the wavelet transform modulus maxima; using the wavelet transform modulus maxima to determine the arrival time of the fault traveling wave front; and using the arrival time of the fault traveling wave front to determine the location of the medium-voltage grounding fault.

[0084] Specifically, in this embodiment of the invention, a multi-resolution wavelet transform is performed on the real-time clean traveling wave signal to extract the wavelet transform coefficients at each decomposition scale. The modulus maxima are calculated based on these coefficients, and effective modulus maxima points are selected through the correlation between modulus maxima at adjacent scales. Based on the distribution of effective modulus maxima points on the time axis, and combined with the Lipschitz exponent criterion, the arrival time of the fault traveling wave front is accurately determined. Based on the propagation speed of the traveling wave in the transmission line and the arrival time of the fault traveling wave front, the location of the medium-voltage grounding fault is calculated using either a double-ended traveling wave location algorithm (utilizing the time difference between the two ends) or a single-ended traveling wave location algorithm (utilizing the time difference of the reflected wave).

[0085] According to the method provided by the present invention, the fault traveling wave can be accurately reconstructed and the fault can be detected and located accordingly, thereby improving the anti-interference capability of medium-voltage grounding fault detection and location and increasing the accuracy of detection and location.

[0086] Please refer to Figure 2The second aspect of the present invention provides a medium-voltage grounding fault location device, the medium-voltage grounding fault location device comprising: an interference signal extraction and sample set preparation module, used to obtain a time-series traveling wave interference signal based on historical raw voltage signals and form a sample dataset; wherein the sample dataset is established based on the time-series traveling wave interference signal and its corresponding acquisition time; an interference signal prediction model construction module, used to construct an interference signal prediction model based on the sample dataset using an evolutionary algorithm; an interference signal prediction module, used to determine the real-time line-mode voltage traveling wave signal and use the interference signal prediction model to determine the traveling wave interference prediction signal; a real-time pure traveling wave signal calculation module, used to determine the real-time pure traveling wave signal based on the difference between the real-time line-mode voltage traveling wave signal and the traveling wave interference prediction signal; a fault determination module, used to determine whether a medium-voltage grounding fault has occurred; and a fault location module, used to determine the location of the medium-voltage grounding fault using the real-time pure traveling wave signal when it is determined that a medium-voltage grounding fault has occurred.

[0087] Furthermore, the step of obtaining the time-series traveling wave interference signal based on the historical original voltage signal and forming a sample dataset includes: performing a phase-mode transformation on the historical original voltage signal to obtain a historical line-mode voltage traveling wave signal; performing a Fourier expansion on the historical line-mode voltage traveling wave signal to separate historical traveling wave interference signals of different frequencies; merging the historical traveling wave interference signals of different frequencies according to the acquisition time of the corresponding historical original voltage signal to obtain the time-series traveling wave interference signal; and obtaining the sample dataset based on the mapping between the acquisition time of the historical original voltage signal corresponding to the time-series traveling wave interference signal and the time-series traveling wave interference signal.

[0088] Further, the construction of the interference signal prediction model based on the sample dataset using an evolutionary algorithm includes the following steps: 1) determining the evolution termination condition; 2) generating an initial model population; wherein the initial model population includes multiple model individuals; 3) determining the fitness of each model individual based on the sample dataset; wherein the fitness of a model individual is used to characterize the prediction performance of that model individual; 4) determining whether the current model population meets the evolution termination condition; 5) if it is determined that the current model population does not meet the evolution termination condition, performing an evolutionary operation on the current model population to generate the next generation model population, and returning to step 3) to continue iteration; 6) if it is determined that the current model population meets the evolution termination condition, identifying the model individual with the highest fitness in the current model population as the interference signal prediction model.

[0089] Furthermore, the evolution termination condition includes: the fitness of at least one model individual in the model population is greater than a set fitness; or the number of iterations is greater than a set number.

[0090] Further, the generation of the initial model population includes: constructing a function symbol set and a terminal symbol set; wherein the variables in the terminal symbol set are selected from the data collection time dataset; generating multiple individual models based on the function symbol set and the terminal symbol set to obtain the initial model population; wherein each individual model consists of two parts, a first sequence and a second sequence, wherein the elements in the first sequence are selected from the function symbol set and the terminal symbol set, and the elements in the second sequence are selected from the terminal symbol set.

[0091] Furthermore, each individual model satisfies the following constraints: Where n is the length of the first sequence in the model individual, and m is the length of the second sequence in the model individual. This represents the maximum number of parameters processed by the function symbols in the function symbol set.

[0092] Furthermore, the fitness of individual models is determined in the following way: ;in, Let be the fitness of the i-th model individual; This represents the number of samples in the sample dataset. The predicted value of the interference signal is calculated by the i-th model individual based on the acquisition time included in the j-th sample. Let be the value of the interference signal contained in the j-th sample.

[0093] Furthermore, the evolutionary operation includes at least one of the following operations: retention operation, selection operation, replication operation, mutation operation, insertion operation, and recombination operation.

[0094] Furthermore, the interference signal prediction model is an explicit calculation model.

[0095] Furthermore, determining whether a medium-voltage grounding fault has occurred includes: training a grounding fault detection model; and inputting the real-time pure traveling wave signal into the grounding fault detection model to determine whether a medium-voltage grounding fault has occurred.

[0096] Further, the training of the ground fault detection model includes: performing a phase-mode transformation on the historical raw voltage signal to obtain a historical line-mode voltage traveling wave signal; wherein the historical raw voltage signal is collected from both fault and non-fault periods; using the interference signal prediction model to determine a historical traveling wave interference prediction signal corresponding to the historical line-mode voltage traveling wave signal; determining a historical clean traveling wave signal based on the difference between the historical line-mode voltage traveling wave signal and its corresponding historical traveling wave interference prediction signal; assigning a fault state label to the historical clean traveling wave signal based on the fault state of the historical raw voltage signal to which the historical clean traveling wave signal is based; wherein the fault state label includes a fault label and a non-fault label; constructing a detection sample dataset based on the historical clean traveling wave signal and its corresponding fault state label, and using the detection sample dataset to perform supervised training on a deep neural network to construct the ground fault detection model.

[0097] Furthermore, the deep neural network adopts a network architecture that integrates temporal neural networks and convolutional neural networks.

[0098] Furthermore, the step of determining the location of a medium-voltage grounding fault using a real-time clean traveling wave signal includes: performing wavelet transform on the real-time clean traveling wave signal to determine the wavelet transform modulus maxima; using the wavelet transform modulus maxima to determine the arrival time of the fault traveling wave front; and using the arrival time of the fault traveling wave front to determine the location of the medium-voltage grounding fault.

[0099] A third aspect of the present invention provides a medium-voltage grounding fault location system, the system comprising: the medium-voltage grounding fault location device described above, used for fault determination and / or fault location; and a processing device, used for receiving fault determination results and / or fault location results from the medium-voltage grounding fault location device, and for monitoring, analyzing or processing the power distribution network, power plant, etc. in the power system based on the fault determination results and / or fault location results.

[0100] Specifically, in this embodiment of the invention, the processing device can be a single edge-side device, a cloud-side facility, or a combination of both. Its specific functions may include fault analysis, fault section isolation, fault clearing, or distribution network self-healing control. The medium-voltage grounding fault location system may further include: a data acquisition device for real-time acquisition of electrical data from distribution network lines and buses, and transmitting the acquired data to the medium-voltage grounding fault location device; and a communication network for transmitting data and instructions between the data acquisition device, the medium-voltage grounding fault location device, and the processing device.

[0101] A fourth aspect of the present invention provides a machine-readable storage medium storing instructions that, when executed by a processor, configure the processor to perform the medium-voltage grounding fault location method described above.

[0102] The fifth aspect of the present invention provides a terminal device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the medium-voltage grounding fault location method described above.

[0103] A sixth aspect of the present invention provides a cloud facility including a server and a computer program running on the server, wherein the server executes the computer program to implement the medium-voltage grounding fault location method described above.

[0104] It is worth noting that the medium-voltage grounding fault location method, device, system, medium and terminal equipment of this application are applicable not only to medium voltage, but also to fault detection and location on lines and in substations of other voltage levels in the power system, depending on the actual application scenario.

[0105] The preferred embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the specific details of the above embodiments. Within the scope of the technical concept of the present invention, various simple modifications can be made to the technical solution of the present invention, and these simple modifications all fall within the protection scope of the present invention.

[0106] It should also be noted that the various specific technical features described in the above specific embodiments can be combined in any suitable manner without contradiction. In order to avoid unnecessary repetition, the present invention will not describe the various possible combinations separately.

[0107] Furthermore, various different embodiments of the present invention can be combined in any way, as long as they do not violate the spirit of the present invention, they should also be regarded as the content disclosed by the present invention.

Claims

1. A method of medium voltage earth fault location, characterized in that, The medium-voltage grounding fault location method includes: The process involves obtaining a time-series traveling wave interference signal based on historical raw voltage signals and forming a sample dataset. This includes: performing a phase-mode transformation on the historical raw voltage signals to obtain historical line-mode voltage traveling wave signals; performing Fourier expansion on the historical line-mode voltage traveling wave signals to separate historical traveling wave interference signals of different frequencies; merging the historical traveling wave interference signals of different frequencies according to the acquisition time of their corresponding historical raw voltage signals to obtain a time-series traveling wave interference signal; and obtaining the sample dataset based on the mapping between the acquisition time of the historical raw voltage signals corresponding to the time-series traveling wave interference signals and the time-series traveling wave interference signals. Based on the sample dataset, an interference signal prediction model is constructed using an evolutionary algorithm, including: generating an initial model population containing multiple model individuals; determining the fitness of each model individual based on the sample dataset; iteratively selecting the model individual with the highest fitness as the interference signal prediction model through evolutionary operations; wherein, the interference signal prediction model is an explicit computation model, consisting of one or more analytical expressions, and inputting time into the analytical expressions can calculate and output the traveling wave interference prediction signal. The real-time line-mode voltage traveling wave signal is determined, and the traveling wave interference prediction signal is determined using the interference signal prediction model. The real-time clean traveling wave signal is determined based on the difference between the real-time line-mode voltage traveling wave signal and the traveling wave interference prediction signal. Determine if a medium-voltage grounding fault has occurred; In cases where a medium-voltage grounding fault is confirmed, the location of the medium-voltage grounding fault is determined using a real-time pure traveling wave signal.

2. The method of medium voltage ground fault location according to claim 1, characterized in that, The method of constructing an interference signal prediction model based on a sample dataset and using an evolutionary algorithm also includes the following steps: Determine the conditions for the end of evolution; Determine whether the current model population meets the evolutionary termination condition; If it is determined that the current model population does not meet the evolution termination condition, an evolutionary operation is performed on the current model population to generate the next generation of model population, and the iteration continues; If the current model population meets the evolution termination condition, the model individual with the highest fitness in the current model population is identified as the interference signal prediction model.

3. The method of medium voltage ground fault location according to claim 2, characterized in that, The conditions for terminating the evolution include: At least one model in the model population has a fitness greater than the set fitness; or The number of iterations is greater than the set number.

4. The method of medium voltage ground fault location of claim 2, wherein, The initial generation of model population includes: Construct a function symbol set and a terminal symbol set; wherein the variables in the terminal symbol set are selected from the acquisition time dataset; Multiple individual models are generated based on the function symbol set and the terminal symbol set to obtain the initial model population; wherein each individual model consists of two parts, a first sequence and a second sequence, the elements in the first sequence are selected from the function symbol set and the terminal symbol set, and the elements in the second sequence are selected from the terminal symbol set.

5. The method of medium voltage ground fault location according to claim 4, characterized in that, Each model individual satisfies the following constraints: ； where n is the length of the first sequence in the model individual, m is the length of the second sequence in the model individual, is the maximum number of parameters processed for a function symbol in the set of function symbols.

6. The method of medium voltage ground fault location of claim 2, wherein, The fitness of individual models is determined in the following way: ； wherein, is the fitness of the i-th model individual; is the number of samples in the sample data set; is the predicted value of the interference signal calculated based on the acquisition time comprised by the i-th model individual for the j-th sample; is the value of the interference signal comprised by the j-th sample.

7. The method of medium voltage ground fault location of claim 2, wherein, The evolutionary operation includes at least one of the following operations: The operations include retain, select, copy, mutate, insert, and recombine.

8. The method of medium voltage ground fault location of claim 1, wherein, The determination of whether a medium-voltage grounding fault has occurred includes: Training grounding fault detection model; The real-time pure traveling wave signal is input into the ground fault detection model to determine whether a medium-voltage ground fault has occurred.

9. The method of medium voltage ground fault location according to claim 8, characterized in that, The training ground fault detection model includes: The historical original voltage signal is subjected to phase-mode transformation to obtain the historical line-mode voltage traveling wave signal; wherein, the historical original voltage signal is collected from the fault period and the non-fault period; The interference signal prediction model is used to determine the historical traveling wave interference prediction signal corresponding to the historical line-mode voltage traveling wave signal; Based on the difference between the historical line-mode voltage traveling wave signal and its corresponding historical traveling wave interference prediction signal, the historical pure traveling wave signal is determined. Based on the fault state of the historical raw voltage signal to which the historical pure traveling wave signal is based, a fault state label is assigned to the historical pure traveling wave signal; wherein, the fault state label includes a fault label and a non-fault label. A detection sample dataset is constructed based on historical pure traveling wave signals and their corresponding fault state labels, and the deep neural network is trained under supervision using the detection sample dataset to construct the ground fault detection model.

10. The method of medium voltage ground fault location according to claim 9, wherein, The deep neural network adopts a network architecture that integrates temporal neural networks and convolutional neural networks.

11. The medium-voltage grounding fault location method according to claim 1, characterized in that, The method of determining the location of a medium-voltage grounding fault using real-time pure traveling wave signals includes: Perform wavelet transform on the real-time pure traveling wave signal to determine the wavelet transform modulus maxima; The arrival time of the fault traveling wavefront is determined by using the modulus maxima of wavelet transform. The location of a medium-voltage grounding fault can be determined by the arrival time of the traveling wave front of the fault.

12. A medium-voltage grounding fault location device, characterized in that, The medium-voltage grounding fault location device includes: An interference signal extraction and sample set preparation module is used to obtain time-series traveling wave interference signals based on historical raw voltage signals and form a sample dataset. This includes: performing a phase-mode transformation on the historical raw voltage signals to obtain historical line-mode voltage traveling wave signals; performing Fourier expansion on the historical line-mode voltage traveling wave signals to separate historical traveling wave interference signals of different frequencies; merging the historical traveling wave interference signals of different frequencies according to the acquisition time of the corresponding historical raw voltage signals to obtain time-series traveling wave interference signals; and obtaining the sample dataset based on the mapping between the acquisition time of the historical raw voltage signals corresponding to the time-series traveling wave interference signals and the time-series traveling wave interference signals. An interference signal prediction model construction module is used to construct an interference signal prediction model based on a sample dataset using an evolutionary algorithm. This includes: generating an initial model population containing multiple individual models; determining the fitness of each individual model based on the sample dataset; and iteratively selecting the individual model with the highest fitness as the interference signal prediction model through evolutionary operations. The interference signal prediction model is an explicit computational model, consisting of one or more analytical expressions. Inputting time into the analytical expressions allows for the calculation and output of the traveling wave interference prediction signal. An interference signal prediction module is used to determine the real-time line-mode voltage traveling wave signal and to determine the traveling wave interference prediction signal using the interference signal prediction model. The real-time clean traveling wave signal calculation module is used to determine the real-time clean traveling wave signal based on the difference between the real-time line-mode voltage traveling wave signal and the traveling wave interference prediction signal. The fault determination module is used to determine whether a medium-voltage grounding fault has occurred. The fault location module is used to determine the location of a medium-voltage grounding fault by using real-time clean traveling wave signals when a medium-voltage grounding fault is confirmed to have occurred.

13. The medium-voltage grounding fault location device according to claim 12, characterized in that, The method of constructing an interference signal prediction model based on a sample dataset and using an evolutionary algorithm also includes the following steps: Determine the conditions for the end of evolution; Determine whether the current model population meets the evolutionary termination condition; If it is determined that the current model population does not meet the evolution termination condition, an evolutionary operation is performed on the current model population to generate the next generation of model population, and the iteration continues; If the current model population meets the evolution termination condition, the model individual with the highest fitness in the current model population is identified as the interference signal prediction model.

14. The medium-voltage grounding fault location device according to claim 13, characterized in that, The conditions for terminating the evolution include: At least one model in the model population has a fitness greater than the set fitness; or The number of iterations is greater than the set number.

15. The medium-voltage grounding fault location device according to claim 13, characterized in that, The initial generation of model population includes: Construct a function symbol set and a terminal symbol set; wherein the variables in the terminal symbol set are selected from the acquisition time dataset; Multiple individual models are generated based on the function symbol set and the terminal symbol set to obtain the initial model population; wherein each individual model consists of two parts, a first sequence and a second sequence, the elements in the first sequence are selected from the function symbol set and the terminal symbol set, and the elements in the second sequence are selected from the terminal symbol set.

16. The medium-voltage grounding fault location device according to claim 15, characterized in that, Each model individual satisfies the following constraints: ； Where n is the length of the first sequence in the model individual, and m is the length of the second sequence in the model individual. The maximum number of parameters processed by the function symbols in the function symbol set.

17. The medium-voltage grounding fault location device according to claim 13, characterized in that, The fitness of individual models is determined in the following way: ； in, Let be the fitness of the i-th model individual; This represents the number of samples in the sample dataset. The predicted value of the interference signal is calculated by the i-th model individual based on the acquisition time included in the j-th sample. Let be the value of the interference signal contained in the j-th sample.

18. The medium-voltage grounding fault location device according to claim 13, characterized in that, The evolutionary operation includes at least one of the following operations: The operations include retain, select, copy, mutate, insert, and recombine.

19. The medium-voltage grounding fault location device according to claim 12, characterized in that, The determination of whether a medium-voltage grounding fault has occurred includes: Training grounding fault detection model; The real-time pure traveling wave signal is input into the ground fault detection model to determine whether a medium-voltage ground fault has occurred.

20. The medium-voltage grounding fault location device according to claim 19, characterized in that, The training ground fault detection model includes: The historical original voltage signal is subjected to phase-mode transformation to obtain the historical line-mode voltage traveling wave signal; wherein, the historical original voltage signal is collected from the fault period and the non-fault period; The interference signal prediction model is used to determine the historical traveling wave interference prediction signal corresponding to the historical line-mode voltage traveling wave signal; Based on the difference between the historical line-mode voltage traveling wave signal and its corresponding historical traveling wave interference prediction signal, the historical pure traveling wave signal is determined. Based on the fault state of the historical raw voltage signal to which the historical pure traveling wave signal is based, a fault state label is assigned to the historical pure traveling wave signal; wherein, the fault state label includes a fault label and a non-fault label. A detection sample dataset is constructed based on historical pure traveling wave signals and their corresponding fault state labels, and the deep neural network is trained under supervision using the detection sample dataset to construct the ground fault detection model.

21. The medium-voltage grounding fault location device according to claim 20, characterized in that, The deep neural network adopts a network architecture that integrates temporal neural networks and convolutional neural networks.

22. The medium-voltage grounding fault location device according to claim 12, characterized in that, The method of determining the location of a medium-voltage grounding fault using real-time pure traveling wave signals includes: Perform wavelet transform on the real-time pure traveling wave signal to determine the wavelet transform modulus maxima; The arrival time of the fault traveling wavefront is determined by using the modulus maxima of wavelet transform. The location of a medium-voltage grounding fault can be determined by the arrival time of the traveling wave front of the fault.

23. A medium-voltage grounding fault location system, characterized in that, The system includes: The medium-voltage grounding fault location device according to any one of claims 12-22 is used for fault determination and / or fault location; The processing device is used to receive fault determination results and / or fault location results from the medium-voltage grounding fault location device, and to monitor, analyze or process the power system based on the fault determination results and / or fault location results.

24. A machine-readable storage medium storing instructions thereon, characterized in that, When executed by a processor, this instruction causes the processor to be configured to perform the medium-voltage grounding fault location method as described in any one of claims 1-11.

25. A terminal device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the medium-voltage grounding fault location method according to any one of claims 1-11.

Images