Power transmission line fault positioning method, device and equipment and computer readable storage medium

Abstract

The invention discloses a power transmission line fault positioning method, device and equipment and a computer readable storage medium, and the method comprises the steps: collecting an electrical transient signal when a power transmission line breaks down, and carrying out the preprocessing of the electrical transient signal, and obtaining a standardized signal; performing multi-scale decomposition on the standardized signal, extracting energy features of each sub-band, and constructing a time-frequency-energy three-dimensional feature vector representing a fault; constructing a gating circulation unit network model, and performing global optimization on key hyper-parameters of the GRU network model by adopting a sparrow search algorithm SSA; and processing the time-frequency-energy three-dimensional feature vector by using a GRU network model subjected to SSA optimization so as to position the fault of the power transmission line. According to the method, the problems of low positioning precision and poor adaptability caused by noise interference, model mismatch and manual parameter adjustment limitation in a complex power grid environment can be solved, and millisecond-level, high-precision and high-robustness precise positioning of the power transmission line fault is realized.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of power transmission line fault diagnosis and positioning, and in particular to a power transmission line fault positioning method, device, equipment and computer readable storage medium. BACKGROUND

[0002] With the construction of new power systems, hybrid DC transmission technology with large capacity and long distance transmission has become an important part of the main network architecture. However, power transmission lines often pass through complex terrain and harsh environments, and faults occur frequently. In a system containing a multi-terminal hybrid topology, the coupling of each subsystem is strong, and local faults can easily trigger a chain reaction, posing a serious threat to the safe and stable operation of the power grid. Therefore, achieving rapid and accurate positioning of faults is crucial for isolating fault sections in a timely manner, shortening power outage time and improving power grid resilience.

[0003] Currently, power transmission line fault positioning mainly relies on traditional methods such as the traveling wave method and the impedance method. The traveling wave method relies on the time it takes for the transient traveling wave front generated by the fault to reach the measurement point, but its positioning accuracy is easily affected by uncertainties in line wave speed, clock synchronization errors, and high-frequency noise interference. In a complex noise environment, wave front recognition is difficult, and positioning stability is insufficient. The impedance method is based on the lumped parameter model of the line, and estimates the fault distance by calculating the impedance of the fault loop. This method is not sensitive to high-resistance faults, and because the line distribution parameter model is oversimplified, the positioning error significantly increases when dealing with long lines, complex topologies, and changes in fault transition resistance.

[0004] In recent years, with the development of artificial intelligence technology, some fault positioning methods based on machine learning or deep learning have been proposed. These methods attempt to mine the mapping relationship between fault features and location through data-driven methods. However, existing intelligent methods rely heavily on artificially designed features, and have limited ability to capture deep spatiotemporal correlation characteristics hidden in fault transient signals. In addition, the performance of the model of such methods is significantly affected by the setting of hyperparameters, and the manual parameter tuning process is inefficient and prone to local optima, resulting in insufficient robustness of the model under noise interference, especially in the case of multi-terminal coupling faults. At the same time, the model complexity and computational overhead of existing solutions are large, making it difficult to meet the dual demands of real-time and economy of fault positioning in new power systems.

[0005] Therefore, there is an urgent need in the art for an intelligent fault positioning technology that can adaptively extract deep fault features, has strong robustness, high accuracy, and is suitable for complex power grid environments, to overcome the limitations of existing methods. SUMMARY

[0006] The application provides a power transmission line fault positioning method, device, equipment and computer readable storage medium, which can solve the technical problems of low positioning accuracy and poor adaptability caused by noise interference, model mismatch and artificial parameter adjustment limitations in a complex power grid environment, and realize fast, high-precision and strong-robust precise positioning of power transmission line faults.

[0007] In a first aspect, the embodiments of the application provide a power transmission line fault positioning method, which comprises: Collecting an electrical transient signal when a fault occurs in a power transmission line, preprocessing the electrical transient signal to obtain a standardized signal; Performing multi-scale decomposition on the standardized signal, extracting energy features of each sub-band, and constructing a time-frequency-energy three-dimensional feature vector representing the fault; Constructing a gated recurrent unit (GRU) network model, and using a sparrow search algorithm (SSA) to globally optimize key hyperparameters of the GRU network model, wherein the key hyperparameters include the number of hidden layers, the number of hidden layer units, the initial learning rate and the batch size of the GRU network model; Using the GRU network model optimized by the SSA to process the time-frequency-energy three-dimensional feature vector to position the fault of the power transmission line.

[0008] By combining gated recurrent unit (GRU) time series modeling and sparrow search algorithm (SSA) intelligent optimization into a complete method, it realizes deep mining and decoupling of fault features in a complex noise environment, overcomes the defects of traditional methods relying on artificial experience and weak model generalization ability by adaptively optimizing key model parameters, thereby achieving stable, fast and accurate intelligent positioning of power transmission line faults, and having strong robustness.

[0009] In combination with the first aspect, in an implementation mode, the globally optimizing the key hyperparameters of the GRU network model using the sparrow search algorithm (SSA) comprises: Initializing a sparrow population, wherein the position vector of each sparrow individual in the sparrow population represents a set of key hyperparameter candidate combinations; Configuring each set of key hyperparameter candidate combinations in the GRU network model for training, and calculating the mean square error on the validation set as the fitness value of the sparrow individual; According to the fitness value, the sparrow population is divided into discoverers, followers and sentinels; According to the position updating rules of the discoverers, followers and sentinels, iteratively update the positions of the sparrow population; After reaching a preset number of iterations, a combination of key hyperparameters represented by a sparrow individual with the best fitness is taken as the final optimized parameters of the GRU network model.

[0010] In combination with the first aspect, in an implementation, the position updating rule of the discoverer comprises: determining an updated position of the discoverer according to a preset security threshold, a preset maximum number of iterations and a random factor. The position updating rule of the follower comprises: determining an updated position of the follower according to a fitness ranking of the follower. The position updating rule of the alarm comprises: determining an updated position of the alarm according to a random direction disturbance coefficient.

[0011] In combination with the first aspect, in an implementation, the multi-scale decomposition of the standardized signal comprises: The multi-scale decomposition of the standardized signal is performed by using wavelet packet transform (WPT).

[0012] In combination with the first aspect, in an implementation, the multi-scale decomposition of the standardized signal by using wavelet packet transform (WPT) comprises: The standardized signal is decomposed by using a wavelet basis function to obtain a plurality of sub-bands. Wavelet packet coefficient energy of each sub-band is determined to form a time-frequency-energy three-dimensional feature vector.

[0013] In combination with the first aspect, in an implementation, the electrical transient signal is a voltage or current signal; and the preprocessing comprises: Phase-mode transformation is performed on the electrical transient signal to obtain a line-mode component signal, and the line-mode component signal is standardized.

[0014] In combination with the first aspect, in an implementation, the phase-mode transformation adopts a Karenbauer transformation matrix or a Clarke transformation matrix.

[0015] The second aspect provides a power transmission line fault positioning device, which comprises: A preprocessing module is configured to collect an electrical transient signal when a fault occurs in a power transmission line, and to preprocess the electrical transient signal to obtain a standardized signal. A decomposition module is configured to perform multi-scale decomposition on the standardized signal, extract energy features of each sub-band, and construct a time-frequency-energy three-dimensional feature vector representing a fault. The optimization module is configured to construct a gated recurrent unit (GRU) network model, and perform global optimization on key hyperparameters of the GRU network model by using a sparrow search algorithm (SSA), wherein the key hyperparameters include a number of hidden layers, a number of hidden units, an initial learning rate, and a batch size of the GRU network model. The positioning module is configured to process the time-frequency-energy three-dimensional feature vector by using the GRU network model optimized by the SSA, so as to locate the fault of the power transmission line.

[0016] In a third aspect, an embodiment of the present application provides a power transmission line fault locating device, which comprises a processor, a memory, and a power transmission line fault locating program stored in the memory and executable by the processor, wherein the power transmission line fault locating program, when executed by the processor, implements the steps of the power transmission line fault locating method described above.

[0017] In a fourth aspect, an embodiment of the present application provides a computer readable storage medium, which stores a power transmission line fault locating program, wherein the power transmission line fault locating program, when executed by a processor, implements the steps of the power transmission line fault locating method described above.

[0018] The technical scheme provided by the embodiments of the present application has the following beneficial effects: 1. The completeness and robustness of fault feature extraction are improved: the wavelet packet transform is used to perform full-band multi-scale decomposition on the fault transient signal, and the time-frequency-energy three-dimensional feature vector constructed can simultaneously capture the high-frequency abrupt change characteristics of the traveling wave head and the low-frequency energy distribution of the fault steady state, overcoming the defects of incomplete feature representation of traditional single time domain or frequency domain analysis methods, and showing stronger anti-interference ability in a low signal-to-noise ratio environment.

[0019] 2. The modeling accuracy and stability of fault distance mapping are enhanced: the gated recurrent unit network is used to model the time sequence of the feature vector, and the gating mechanism effectively alleviates the gradient problem in long sequence training, which can more accurately learn the complex nonlinear mapping relationship between the fault feature and the propagation distance, and provides a reliable model basis for realizing high-precision positioning.

[0020] 3. Intelligent global optimization of key parameters of the model is realized: the sparrow search algorithm is introduced, and the performance of the verification set is used as the fitness function to automatically perform global optimization on key hyperparameters such as the number of network layers, the number of units, and the learning rate. This method overcomes the subjectivity and limitations of manual parameter tuning, enables the model to automatically converge to a better configuration, and significantly improves the generalization performance and convergence speed of the model.

[0021] 4. Significant improvement in comprehensive performance is achieved: the synergy of the above technologies enables the final fault location system to complete fault diagnosis within milliseconds, the average positioning error is significantly reduced (for example, it can reach within 0.3 kilometers), and it also shows good adaptability and reliability to high resistance faults, complex noise and new power system scenarios, meeting the comprehensive requirements of precision, speed and robustness for engineering applications.

[0022] In summary, the present application effectively solves the technical problems of one-sided feature extraction, model dependence on artificial priori, low parameter adjustment efficiency and insufficient precision under complex working conditions of traditional fault location methods, and provides an intelligent fault location solution with high precision, fast speed and strong adaptability. BRIEF DESCRIPTION OF DRAWINGS

[0023] Figure 1 It is a flowchart of an embodiment of the power transmission line fault location method of the present application. Figure 2 It is a functional module schematic diagram of an embodiment of the power transmission line fault location device of the present application. Figure 3 It is a hardware structure schematic diagram of the power transmission line fault location equipment involved in the embodiment of the present application. DETAILED DESCRIPTION

[0024] In order for those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor fall within the scope of protection of the present application.

[0025] First, some technical terms in the present application are explained and described to facilitate understanding of the present application by those skilled in the art.

[0026] In a first aspect, the embodiments of the present application provide a power transmission line fault location method.

[0027] In an embodiment, with reference to Figure 1 , Figure 1 It is a flowchart of the first embodiment of the power transmission line fault location method of the present application. As Figure 1 shown, the power transmission line fault location method comprises: Step S10: Collecting the electrical transient signal when the power transmission line fails, pre-processing the electrical transient signal to obtain a standardized signal; In some embodiments, the electrical transient signals within a preset time window (e.g., 10 milliseconds centered on the time of fault triggering) after the fault of the transmission line can be collected in real time by a potential transformer (PT) or a current transformer (CT) installed at the end of the transmission line. The electrical transient signals can be three-phase voltage or three-phase current signals. To ensure that the high-frequency components of the fault traveling wave can be captured completely, the sampling frequency can be selected to be no less than 100 kHz to meet the requirements of the Nyquist sampling theorem.

[0028] Due to the electromagnetic coupling among the three-phase lines, the directly collected phase voltage or phase current signals cannot be directly used to represent the traveling wave mode which is strongly related to the fault distance. Therefore, the three-phase electrical transient signals need to be subjected to phase-mode transformation to decouple independent modal components. Specifically, the collected three-phase voltage signal column vector is multiplied by a phase-mode transformation matrix to obtain the decoupled modal voltage signals . Among them: , , is a line-mode component, which has stable wave speed and is less affected by the environment, and is a characteristic component strongly related to fault location. is a zero-mode component.

[0029] The phase-mode transformation matrix may be a Karenbauer transformation matrix or a Clarke transformation matrix. For example, the Karenbauer transformation matrix is defined as: The transformation formula is: The line-mode components and / or are preferably extracted as input signals for subsequent processing to eliminate common-mode noise interference.

[0030] To eliminate dimensional differences, accelerate model convergence, and improve generalization ability, the decoupled line-mode component signals are subjected to standardization processing. The Z-score standardization method can be used in this embodiment. Specifically, the mean and the standard deviation of the line-mode component signal segment are calculated, and then each data point in the signal is subjected to the following transformation: After this processing, the standardized signal with a mean of 0 and a standard deviation of 1 is obtained, which provides a regular and high-quality input for the subsequent feature extraction step.

[0031] Through the above steps, the pre-processing process from the original fault electrical transient signal to the normalized and decoupled standardized signal is completed, and a reliable data foundation is laid for subsequent accurate feature extraction and intelligent analysis.

[0032] Step S20: performing multi-scale decomposition on the standardized signal, extracting energy features of each sub-band, and constructing a time-frequency-energy three-dimensional feature vector representing the fault; In some embodiments, the multi-scale decomposition of the standardized signal can include using wavelet packet transform (WPT) to perform multi-scale decomposition on the standardized signal. In other embodiments, other algorithms can also be used to perform multi-scale decomposition on the standardized signal, which is not limited in the embodiments of the present application.

[0033] In some embodiments, the multi-scale decomposition of the standardized signal using wavelet packet transform (WPT) includes: using a wavelet basis function to perform full-band decomposition on the standardized signal to obtain a plurality of sub-bands; determining the wavelet packet coefficient energy of each sub-band to form the time-frequency-energy three-dimensional feature vector.

[0034] The specific implementation process is as follows: 1. Decomposition process and parameter setting 1) Selecting a wavelet basis function: Daubechies 4 (db4) wavelet is selected as the wavelet basis function. This wavelet has tight support and certain regularity, and has good localization ability in time domain and frequency domain, which is very suitable for analyzing non-stationary fault transient signals.

[0035] Let the orthogonal scaling function corresponding to the wavelet basis function be , the wavelet function be , and the wavelet packet function be It is recursively generated by the following bi-scale equation: wherein and are the coefficients of the orthogonal low-pass filter and high-pass filter corresponding to the selected db4 wavelet, respectively, and satisfy , is a time variable, n is a discrete index (usually a serial number or number), is another discrete index (usually representing translation or sequence position), and Z is a real number set. This equation is the mathematical core of wavelet packet decomposition, which specifies how to generate two child node functions (low-frequency approximation) and (high-frequency detail) from the parent node wavelet packet function through filtering and scaling.

[0036] 2) Decomposition level determination: The normalized line-mode signal (e.g. alpha-mode component) is decomposed by J (J=5 in this example) level full-binary-tree wavelet packet decomposition.

[0037] 3) Sub-band generation: After J level decomposition, the original signal's frequency range (0 Hz to Nyquist frequency) is divided into N = 2 ^J (32 in this example) sub-bands. Each sub-band corresponds to a terminal node on the wavelet packet decomposition tree, covering a specific frequency interval. For example, in a system with a sampling frequency of 100 kHz, each sub-band covers a frequency range of about .

[0038] 2. Energy feature extraction After decomposition, the energy feature is extracted from each sub-band as the core information representing the fault: 1) Wavelet packet coefficient calculation: Let the wavelet packet coefficient sequence of the j th level and the i th sub-band (node) be denoted as , where is the length of the sub-band coefficient.

[0039] 2) Sub-band energy calculation: The energy of the sub-band is defined by the sum of the squares of its wavelet packet coefficients, and the calculation formula is: This energy value quantitatively represents the energy intensity of the original fault signal in this specific frequency interval.

[0040] 3) Traverse all sub-bands: Repeat the above calculation for all 32 terminal nodes (sub-bands) obtained by decomposition, and obtain a set of energy values .

[0041] 4) Construction of time-frequency-energy three-dimensional feature vector Arrange the above energy values in a predetermined order (e.g. frequency from low to high) to form a fixed vector: Feature vector form: Physical connotation of "three-dimensional" feature: Frequency dimension: reflected by the index order of the vector. Each index position corresponds to a pre-defined sub-band, recording the energy distribution on the frequency axis.

[0042] Energy dimension: reflected by the numerical value of each element in the vector. Directly reflects the strength of each frequency band energy.

[0043] Time dimension: guaranteed by the original data time window (such as 10 ms after the fault) from which the vector is generated and the time-frequency localization characteristics of wavelet transform. The entire vector represents the frequency energy state of the signal in the specific time period.

[0044] In a preferred embodiment of the present application, in order to improve the representation ability of the feature, the energy of all nodes (including non-terminal nodes) of the 5-layer decomposition tree is calculated, and finally a 253-dimensional high-dimensional energy spectrum feature vector is constructed, which is represented as , which can more comprehensively and meticulously depict the frequency spectrum energy distribution of the fault signal.

[0045] The "time-frequency-energy three-dimensional feature vector" constructed through the above process has the following advantages: 1. Strong decoupling: converting complex time-domain waveforms into energy distribution in frequency bands effectively decouples fault features.

[0046] 2. Good noise resistance: wavelet packet transform itself has certain noise reduction ability, and energy features are not sensitive to random noise.

[0047] 3. Clear physical meaning: each element in the vector directly corresponds to the energy of a physical frequency band, providing a highly interpretable high-quality input feature for subsequent artificial intelligence models.

[0048] Step S30: Construct a gated recurrent unit (GRU) network model, and use the sparrow search algorithm (SSA) to globally optimize key hyperparameters of the GRU network model, wherein the key hyperparameters include the number of hidden layers, the number of hidden layer units, the initial learning rate, and the batch size of the GRU network model.

[0049] In some embodiments, a GRU network architecture is constructed according to a deep learning framework (such as TensorFlow or PyTorch). The input layer dimension of the model must strictly match the dimension (e.g., 253 dimensions) of the "time-frequency-energy three-dimensional feature vector" generated in the previous step. The output layer of the model is set to a linear layer or a SoftMax layer according to whether the final task is to implement accurate regression of fault distance (output continuous numerical value) or discriminant classification of fault section (output discrete category).

[0050] One of the keys of the present application is to define the number of hidden layers, the number of hidden layer units, the initial learning rate, and the batch size of the GRU model as "key hyperparameters" to be optimized, which have a decisive influence on the performance of the model. Before optimization begins, a reasonable search space is set for each parameter, for example: the number of hidden layers is searched within [1, 4], the number of hidden layer units is searched within [64, 56], the initial learning rate is searched within the logarithmic space [1e-4, 1e-2], and the batch size is searched within [32, 128].

[0051] The core of GRU is to dynamically adjust the information flow through two gating mechanisms, which are mathematically defined as follows: (1) Reset Gate: Control the influence of historical hidden state on candidate hidden state: where: is the reset gate output (range [0, 1]); is the sigmoid activation function; and are the weight matrix and bias term of the reset gate; is the hidden state at the previous time step; is the input at the current time step.

[0052] Effect: Decide whether to ignore the past hidden state. When tends to 0, the candidate hidden state only depends on the current input, discarding historical information.

[0053] (2) Update Gate Control the degree of update of the hidden state: where: is the update gate output (range [0, 1]); is the sigmoid activation function; and are the weight matrix and bias term of the update gate; Effect: tends to 1, more historical information is retained ( ); when tends to 0, more depends on the current candidate state ( ). Among them, is the hidden state at the current time step.

[0054] (3) Candidate Hidden State Combine the current input and the reset historical state: where: is the parameter to be updated; denotes the Hadamard product, i.e., element-wise multiplication; is the activation function, which restricts the output to [-1, 1], and are the weight matrix and bias term of the candidate hidden layer.

[0055] Action: Reset gate output Filter irrelevant history information, generate new candidate state.

[0056] (4) Final hidden state (Hidden State) Update gate weighted history state and candidate state: The larger, the more inclined to retain historical information; The smaller, the more inclined to update to the candidate state.

[0057] In order to realize the efficient and automatic acquisition of the optimal combination of the above key hyperparameters, SSA is used for global optimization, and the specific process is as follows: 1. Population initialization and coding Randomly generate a population containing sparrows (such as ). The position of each sparrow represents a set of key hyperparameter candidate solutions, where the key hyperparameters are hidden layer number, hidden layer unit number, initial learning rate, and batch size.

[0058] 2. Fitness evaluation Define the fitness function as the performance indicator of the GRU model on the independent validation set. For regression tasks, the mean square error is usually used; for classification tasks, the 1-classification accuracy can be used. The evaluation process is: according to the sparrow position decoded hyperparameter configuration, instantiate and train the GRU model, and calculate its fitness value on the validation set . The lower the fitness value, the better the performance of the group of hyperparameters.

[0059] 3. Role division and iterative update Role division: in each iteration, the population is divided into three categories according to the fitness value: the top 20% individuals with the best fitness are the discoverers, responsible for global exploration; the middle 60% are followers, for local development; the last 20% are sentries, responsible for early warning and jumping out of local optimum.

[0060] Position update: The position update formula of the discoverer introduces a safety threshold. When the risk is perceived (the random number is less than the safety threshold), a large range of exploratory movement will be performed, corresponding to a wide area search in the hyperparameter space; otherwise, it will move closer to the current known optimal position for fine-tuning. Its position update formula is as follows: where is the current iteration number,​ is the maximum number of iterations, is the convergence rate adjustment factor, is a random number subject to standard normal distribution, is a unit vector, is a random number within [-1, 1]. and are the early warning value and safety threshold, respectively. When , the discoverer conducts extensive random exploration; when , the discoverer moves towards the current global optimal position .

[0061] The position update of the follower depends on the discoverer. The follower with higher fitness will conduct intensive search (local development) near the optimal discoverer, while the follower with lower fitness will fly to other areas to increase population diversity. The position update formula of the follower is: wherein, is the current worst position, is a random disturbance term. The follower with lower fitness ranking ( ) flies to the poor area for extensive search, while the follower with higher fitness ranking conducts fine development near the optimal solution.

[0062] The guard is randomly generated in a certain proportion, and its position will move towards the current global optimal position or be randomly disturbed to help the population jump out of the local optimal solution. The position of the guard is updated by random disturbance: wherein, is a random direction coefficient subject to standard normal distribution.

[0063] 4. Iteration termination and output Repeat the fitness evaluation and position update until the maximum number of iterations (e.g. ) is reached. After the iteration is completed, the historical global optimal position is output, and the optimal hyperparameter combination of the GRU model is obtained after decoding.

[0064] Through the above SSA optimization process, the hyperparameter configuration of the GRU model is automatically and intelligently determined. In a specific implementation case, the model obtained by SSA optimization (optimal parameters such as 3 layers, 196 units, learning rate 0.0007, batch size 48) has lower average positioning error on the test set compared to the model using the default parameters based on experience, effectively solving the problem of low efficiency and easy to fall into local optimum in manual parameter tuning, and significantly improving the precision and reliability of the fault location system.

[0065] Step S40: The time-frequency-energy three-dimensional feature vector is processed by using the GRU network model optimized by SSA to locate the fault of the power transmission line.

[0066] In some embodiments, this step is the final execution stage, which is to use the Gated Recurrent Unit (GRU) network model optimized by Sparrow Search Algorithm (SSA) to map the input fault feature vector to accurate fault location information.

[0067] 1. Model training and solidification Before positioning by using the model, the final training and solidification of the model need to be completed: Training data preparation: Collect historical or simulated fault recording data containing various fault types (such as single-phase grounding, two-phase short circuit, etc.), different fault locations (covering the whole line), and various working conditions (such as different loads and noise levels). Process these data according to the method of the foregoing embodiments to batch generate standardized "time-frequency-energy three-dimensional feature vectors" as input features, and label their corresponding real fault distances (for regression) or fault section numbers (for classification) to constitute the training sample set required for supervised learning.

[0068] Model training: The GRU network model is initialized and retrained using the optimal hyperparameter combination (for example: 3 layers of hidden layers, 196 units per layer, initial learning rate 0.0007, batch size 48) optimized by SSA. During training, the training sample set is used, the mean square error (for regression) or cross entropy (for classification) is used as the loss function, the Adam optimizer is used for iterative optimization until the model converges. Early stopping method can be used during training to prevent overfitting.

[0069] Model solidification: After training, the final model weights, structure and preprocessing parameters (such as standardization parameters, wavelet packet decomposition tree structure) are saved together to solidify into a deployable fault location model.

[0070] 2. Online fault location inference process When a fault occurs in the power transmission line, the system performs real-time positioning according to the following process: 1) Real-time signal processing and feature extraction: The system collects the post-fault electrical transient signals (voltage or current) in real time and immediately starts the online processing flow. This flow strictly follows the steps of the aforementioned embodiments: first, the phase-to-mode transformation and standardization are performed to generate the standardized signal; then, the same wavelet packet decomposition parameters as in the training phase (such as db4 wavelet basis, 5-level decomposition) are used to process the signal, and the energy of each sub-band is extracted, finally, a real-time fault feature vector consistent with the training sample format is constructed.

[0071] 2) Forward propagation of feature vector: The generated real-time fault feature vector is input into the solidified and loaded GRU network model optimized by SSA. The gating mechanism (update gate and reset gate) inside the model will automatically activate to perform nonlinear transformation and information filtering on the high-dimensional feature sequence, capturing the complex mapping relationship between it and the fault location.

[0072] 3) Output and interpretation of positioning results: The output layer of the GRU network produces the final positioning results according to its design: If it is a regression model: the output layer will directly give a continuous scalar value , which is the model's prediction of the physical distance (unit: kilometers) from the fault point to the measurement point. For example, if the model output value is 52.3, it means that the fault occurs 52.3 kilometers away from the measurement point.

[0073] If it is a classification model: the output layer outputs a probability distribution vector through the SoftMax function, and the class with the highest probability is determined as the fault location in the preset section. For example, if the output vector shows that the probability of the 3rd section is 95%, the system determines that the fault is located in the "3rd protection section".

[0074] 4) Result pushing: the positioning results (exact distance or section information) along with the confidence level (such as the error range for regression tasks or the probability value for classification tasks) are uploaded to the power grid dispatching center or operation and maintenance system, providing key basis for fault isolation and repair decision-making.

[0075] Through the above embodiments, end-to-end automated processing from raw fault signals to accurate positioning results is achieved.

[0076] In a second aspect, the embodiments of the present application also provide a power transmission line fault positioning device.

[0077] In an embodiment, with reference to Figure 2 , Figure 2 is a functional module schematic diagram of an embodiment of the power transmission line fault positioning device of the present application. As Figure 2 shown, the power transmission line fault positioning device comprises: The preprocessing module 01 is configured to collect an electrical transient signal generated when a power transmission line fails, and preprocess the electrical transient signal to obtain a standardized signal. The decomposition module 02 is configured to perform multi-scale decomposition on the standardized signal, extract energy features of each sub-band, and construct a time-frequency-energy three-dimensional feature vector representing the fault. The optimization module 03 is configured to construct a gated recurrent unit (GRU) network model, and perform global optimization on key hyperparameters of the GRU network model by using a sparrow search algorithm (SSA), wherein the key hyperparameters include the number of hidden layers, the number of hidden layer units, the initial learning rate, and the batch size of the GRU network model. The positioning module 04 is configured to process the time-frequency-energy three-dimensional feature vector by using the GRU network model optimized by the SSA, so as to locate the fault of the power transmission line.

[0078] Further, in an embodiment, the optimization module 03 is further configured to: initialize a sparrow population, wherein a position vector of each sparrow individual in the sparrow population represents a set of candidate combinations of the key hyperparameters; configure each set of candidate combinations of the key hyperparameters in the GRU network model for training, and calculate a mean square error on a validation set as a fitness value of the sparrow individual; divide the sparrow population into discoverers, followers, and sentinels according to the fitness values; iteratively update the positions of the sparrow population according to position update rules of the discoverers, the followers, and the sentinels; after a preset number of iterations is reached, combine the key hyperparameters represented by the sparrow individual with the optimal fitness as the final optimized parameters of the GRU network model.

[0079] Further, in an embodiment, the optimization module 03 is further configured to: the position update rule of the discoverer includes determining an updated position of the discoverer according to a preset safety threshold, a preset maximum number of iterations, and a random factor; the position update rule of the follower includes determining an updated position of the follower according to a fitness ranking of the follower; the position update rule of the sentinel includes determining an updated position of the sentinel according to a random direction perturbation coefficient.

[0080] Further, in an embodiment, the decomposition module 02 is further configured to: perform multi-scale decomposition on the standardized signal by using a wavelet packet transform (WPT).

[0081] Further, in an embodiment, the decomposition module 02 is further configured to: perform full-band decomposition on the standardized signal by using a wavelet basis function to obtain a plurality of sub-bands; determine wavelet packet coefficient energy of each of the sub-bands to form the time-frequency-energy three-dimensional feature vector.

[0082] Further, in an embodiment, the preprocessing module 01 is further configured to: perform phase-mode transformation on the electrical transient signal to obtain a line-mode component signal, and perform standardization processing on the obtained line-mode component signal.

[0083] Further, in an embodiment, the preprocessing module 01 is further configured to: The phase-mode transformation adopts a Karenbauer transformation matrix or a Clarke transformation matrix.

[0084] The functions of each module in the power transmission line fault positioning device correspond to the steps in the power transmission line fault positioning method, and the functions and implementation processes will not be repeated here.

[0085] In a third aspect, the embodiments of the present application provide a power transmission line fault positioning device 20. The power transmission line fault positioning device can be a personal computer (PC), a notebook computer, a server, or other devices with data processing functions.

[0086] Referring to Figure 3 , Figure 3 FIG. 1 is a schematic diagram of a hardware structure of a power transmission line fault positioning device according to an embodiment of the present application. In the embodiments of the present application, the power transmission line fault positioning device can include a processor 21, a memory 22, a communication interface 23, and a communication bus 24.

[0087] The communication bus can be of any type, used to interconnect the processor, the memory, and the communication interface.

[0088] The communication interface includes input / output (I / O) interfaces, physical interfaces, and logical interfaces, which are used to interconnect devices inside the power transmission line fault positioning device, and are used to interconnect the power transmission line fault positioning device with other devices (such as other computing devices or user devices). The physical interface can be an Ethernet interface, a fiber interface, an ATM interface, etc.; the user device can be a display (Display), a keyboard (Keyboard), etc.

[0089] The memory can be various types of storage media, such as random access memory (RAM), read-only memory (ROM), non-volatile RAM (NVRAM), flash memory, optical storage, hard disk, programmable ROM (PROM), erasable PROM (EPROM), electrically erasable PROM (EEPROM), etc.

[0090] The processor can be a general-purpose processor, which can invoke a power line fault locating program stored in the memory and execute the power line fault locating method provided by the embodiments of the present application. For example, the general-purpose processor can be a central processing unit (CPU). The method executed by the power line fault locating program when invoked can refer to various embodiments of the power line fault locating method of the present application, which will not be described here.

[0091] Those skilled in the art can understand that the hardware structure shown in the above-mentioned embodiments is not a limitation of the present application, and can include more or less components than the figure, or combine certain components, or different component arrangements. Figure 3

[0092] In a fourth aspect, the embodiments of the present application further provide a computer readable storage medium.

[0093] The computer readable storage medium of the present application stores a power line fault locating program, wherein the power line fault locating program, when executed by a processor, implements the steps of the power line fault locating method as described above.

[0094] The method implemented by the power line fault locating program when executed can refer to various embodiments of the power line fault locating method of the present application, which will not be described here.

[0095] It should be noted that the above-mentioned sequence numbers of the embodiments of the present application are only for description, and do not represent the advantages and disadvantages of the embodiments.

[0096] ​The terms “include,” “comprise,” “have,” and any variations thereof, in the specification and in the claims of the present application and the above-described drawings, are intended to cover a non-exclusive inclusion. For example, a process, method, system, product, or device that includes a list of steps or units is not limited to the listed steps or units, but can optionally further include steps or units not listed, or can optionally further include other steps or units inherent to such processes, methods, products, or devices. The terms “first,” “second,” and “third” and the like descriptions are used to distinguish different objects, and do not represent a sequence or limit the types of “first,” “second,” and “third.”

[0097] In the description of the embodiments of the present application, “exemplary”, “for example”, or “for instance” is used to represent an example, illustration, or description. Any embodiment or design scheme described as “exemplary”, “for example”, or “for instance” in the embodiments of the present application should not be interpreted as more preferred or more advantageous than other embodiments or design schemes. Rather, the words “exemplary”, “for example”, or “for instance” are intended to present the relevant concept in a specific manner.

[0098] In the description of the embodiments of the present application, unless otherwise specified, “ / ” represents the meaning of or, for example, A / B can represent A or B; “and / or” in the text only represents a description of the relationship between the associated objects, which means that there can be three relationships, for example, A and / or B can represent the three cases of A alone, A and B together, and B alone. In addition, in the description of the embodiments of the present application, “multiple” means two or more than two.

[0099] In some of the processes described in the embodiments of the present application, a plurality of operations or steps are included in a specific order, but it should be understood that these operations or steps can be executed or performed in parallel or in an order different from that in which they appear in the embodiments of the present application. The serial number of the operation is only used to distinguish different operations, and the serial number itself does not represent any execution order. In addition, these processes can include more or fewer operations, and these operations or steps can be executed in sequence or in parallel, and these operations or steps can be combined.

[0100] From the above description of the embodiments, those skilled in the art can clearly understand that the above-mentioned embodiment method can be realized by means of software and a general hardware platform as required, of course, it can also be realized by hardware, but in many cases the former is a better embodiment. Based on such understanding, the technical solutions of the present application can be embodied in the form of a software product, which is stored in a storage medium (such as a ROM / RAM, a magnetic disk, an optical disk) as described above, and includes a plurality of instructions for causing a terminal device to execute the methods described in the embodiments of the present application.

[0101] The preferred embodiments of the present application have been described above with the illustrated embodiments, and are not intended to limit the scope of patent protection for the present application. Any equivalent structure or equivalent process variations, which directly or indirectly apply to the content of the specification and drawings, or are directly or indirectly used in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A method for locating faults in transmission lines, characterized in that, The method for locating faults in transmission lines includes: The electrical transient signals when a fault occurs in the transmission line are collected, and the electrical transient signals are preprocessed to obtain standardized signals; The standardized signal is decomposed into multiple scales to extract the energy features of each sub-band and construct a time-frequency-energy three-dimensional feature vector characterizing the fault. A gated recurrent unit (GRU) network model is constructed, and the Sparrow Search Algorithm (SSA) is used to globally optimize the key hyperparameters of the GRU network model. The key hyperparameters include the number of hidden layers, the number of hidden units, the initial learning rate, and the batch size of the GRU network model. The time-frequency-energy three-dimensional feature vector is processed using the SSA-optimized GRU network model to locate faults in the transmission line.

2. The method for locating transmission line faults as described in claim 1, characterized in that, The process of using the Sparrow Search Algorithm (SSA) to globally optimize the key hyperparameters of the GRU network model includes: Initialize a sparrow population, wherein the position vector of each individual sparrow in the population represents a set of candidate combinations of the key hyperparameters; Each set of key hyperparameter candidate combinations is configured in the GRU network model for training, and the mean squared error is calculated on the validation set as the fitness value of the sparrow individual. Based on the fitness values, the sparrow population is divided into discoverers, followers, and vigilants; The location of the sparrow population is iteratively updated according to the location update rules of the discoverer, the followers, and the watchers; After reaching the preset number of iterations, the combination of key hyperparameters represented by the sparrow with the best fitness is used as the final optimization parameters of the GRU network model.

3. The method for locating transmission line faults as described in claim 2, characterized in that, The location update rule for the discoverer includes: determining the updated location of the discoverer based on a preset security threshold, a preset maximum number of iterations, and a random factor; The follower position update rule includes: determining the updated position of the follower based on the follower's fitness ranking; The position update rule for the vigilant includes: determining the updated position of the vigilant based on the random directional perturbation coefficient.

4. The method for locating transmission line faults as described in claim 1, characterized in that, The multi-scale decomposition of the standardized signal includes: The standardized signal is decomposed into multiple scales using wavelet packet transform (WPT).

5. The method for locating transmission line faults as described in claim 4, characterized in that, The multi-scale decomposition of the standardized signal using wavelet packet transform (WPT) includes: The standardized signal is decomposed into multiple sub-bands using wavelet basis functions. The wavelet packet coefficient energy of each sub-frequency band is determined to form the time-frequency-energy three-dimensional feature vector.

6. The method for locating transmission line faults as described in claim 1, characterized in that, The electrical transient signal is a voltage or current signal; the preprocessing includes: The electrical transient signal is subjected to phase-mode transformation to decouple it into a linear mode component signal, and the decoupled linear mode component signal is then standardized.

7. The method for locating transmission line faults as described in claim 6, characterized in that, The phase mode transformation uses either the Karenbauer transformation matrix or the Clarke transformation matrix.

8. A fault location device for transmission lines, characterized in that, The transmission line fault location device includes: The preprocessing module is used to collect electrical transient signals when a fault occurs in a transmission line, and to preprocess the electrical transient signals to obtain standardized signals. The decomposition module is used to perform multi-scale decomposition on the standardized signal, extract the energy features of each sub-frequency band, and construct a time-frequency-energy three-dimensional feature vector characterizing the fault. The optimization module is used to construct a gated recurrent unit (GRU) network model. The Sparrow Search Algorithm (SSA) is used to globally optimize the key hyperparameters of the GRU network model. The key hyperparameters include the number of hidden layers, the number of hidden units, the initial learning rate, and the batch size of the GRU network model. The positioning module is used to process the time-frequency-energy three-dimensional feature vector using the GRU network model optimized by SSA in order to locate the fault in the transmission line.

9. A fault location device for power transmission lines, characterized in that, The transmission line fault location device includes a processor, a memory, and a transmission line fault location program stored in the memory and executable by the processor, wherein when the transmission line fault location program is executed by the processor, it implements the steps of the transmission line fault location method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a transmission line fault location program, wherein when the transmission line fault location program is executed by a processor, it implements the steps of the transmission line fault location method as described in any one of claims 1 to 7.

Images