Intelligent positioning method and system for short-circuit fault of power distribution system, and medium

Abstract

The invention relates to the technical field of power system fault positioning, in particular to an intelligent positioning method and system for a short-circuit fault of a power distribution system and a medium, and the method comprises the following steps: S1, monitoring a high-frequency part of a current in real time; s2, extracting a frequency spectrum index related to the short circuit event; s3, distinguishing short-circuit fault features and normal load changes; s4, combining an analysis result of the deep learning algorithm with geographic information to determine a short circuit position; s5, after the short circuit position is determined, notifying a maintenance team and providing optimal maintenance path guidance; and S6, optimizing the neural network model and the fast Fourier transform processing strategy. According to the invention, by applying high-frequency current monitoring, fast Fourier transform, convolutional neural network analysis and geographic information system technologies, rapid positioning and efficient processing of the short-circuit fault of the power distribution system are realized, and the adaptability and long-term performance of the system are improved by using a self-learning optimization method.

Description

Technical Field

[0001] The present invention relates to the technical field of power system fault location, and in particular to a method, system and medium for intelligently locating short-circuit faults in a power distribution system. Background Art

[0002] With the rapid development of industrialization and urbanization, the stable operation of the power system has an increasingly greater impact on social life and economic activities. As an important part of the power system, the stability of the distribution system is directly related to the safety and reliability of the entire power grid. Short-circuit fault is one of the common serious faults in the distribution system. It not only causes power outages, but may also cause damage to power equipment and power grid accidents. Traditional fault detection methods mainly rely on manual inspections or automatic monitoring systems based on simple thresholds. These methods are often difficult to achieve rapid fault location and effective response, especially in large-scale and complex power grid environments.

[0003] Existing fault location technologies face several major problems: first, traditional methods have limitations in data processing and fault identification accuracy, making it difficult to adapt to the rapidly changing power grid environment and emerging fault types; second, the fault response speed is slow and cannot meet the needs of rapid power restoration, affecting the stability of the power grid and user satisfaction; third, the existing technology lacks an effective self-learning mechanism, making it difficult to automatically optimize fault handling strategies over time. Therefore, the development of an intelligent system that can quickly and accurately locate and effectively respond to short-circuit faults has become an urgent need in the power industry. Summary of the invention

[0004] Based on the above objectives, the present invention provides a method, system and medium for intelligently locating short-circuit faults in a power distribution system.

[0005] A method for intelligently locating short-circuit faults in a power distribution system comprises the following steps:

[0006] S1: Deploy current monitoring devices based on optical fiber technology at predetermined nodes of the distribution system to monitor the high-frequency part of the current in real time to capture current spikes when short-circuit events occur;

[0007] S2: Perform fast Fourier transform processing on the data collected by the optical fiber monitoring device, analyze the spectrum characteristics of the current signal, and extract spectrum indicators related to the short-circuit event;

[0008] S3: Use a preset deep learning algorithm to analyze the extracted spectrum indicators and train a supervised neural network model to distinguish short-circuit fault characteristics from normal load changes;

[0009] S4: Based on the geographic and topological data of the distribution system, the analysis results of the deep learning algorithm are combined with geographic information to determine the short circuit location;

[0010] S5: Once the short circuit location is determined, the fault response program is automatically initiated to notify the maintenance team and provide guidance on the optimal repair path;

[0011] S6: During the operation of the distribution system, fault processing data is continuously collected, and the neural network model and fast Fourier transform processing strategy are optimized using self-learning methods to adapt to changes in the power grid environment and emerging fault types.

[0012] Optionally, the S1 specifically includes:

[0013] S11: Predetermined nodes of the power distribution system, including transformer outlets, important branch nodes or load concentration areas, used as deployment locations for current monitoring devices;

[0014] S12: Install a current monitoring device based on fiber Bragg grating technology at the selected nodes. The device uses the dependence of fiber Bragg grating on the wavelength of incident light to sense changes in the surrounding magnetic field, thereby accurately monitoring the high-frequency part of the current;

[0015] S13: configuring a signal processing unit connected to the current monitoring device, the unit comprising a light source, a spectrum analyzer and a data converter; the light source is used to emit a stable light wave to the optical fiber;

[0016] S14: The spectrum analyzer receives the light waves affected by the current, analyzes the wavelength change, and converts the wavelength change into the current value. By monitoring the high-frequency wavelength change, the current spike when the short circuit event occurs is captured;

[0017] S15: The data converter is used to convert the output of the spectrum analyzer into a digital signal.

[0018] Optionally, the S2 specifically includes:

[0019] S21: transmitting the current signal data collected by the optical fiber monitoring device from the signal processing unit to the central processing unit, and preprocessing the received current signal data in the central processing unit, including denoising and normalization processing;

[0020] S22: Perform fast Fourier transform on the preprocessed current signal data to convert the time domain signal into a frequency domain signal; the calculation formula is: Wherein, X(k) is the frequency domain signal, x(n) is the time domain signal, N is the length of the current signal data, k is the frequency index, and j is the imaginary unit;

[0021] S23: Analyze the spectrum after fast Fourier transform processing, identify spectrum characteristics, and extract spectrum indicators related to the short circuit event by observing the amplitude and phase changes of frequency components in the spectrum. The spectrum indicators include amplitude spectrum peak, spectrum energy and spectrum center;

[0022] S24: Save the extracted spectrum index to the database.

[0023] Optionally, the S3 specifically includes:

[0024] S31: Use the spectrum indicators extracted in S2 as input data sets to prepare training and test data sets; the training data set is used for model training, and the test data set is used to verify the accuracy of the model;

[0025] S32: Select convolutional neural network as the preset deep learning algorithm, and construct a neural network structure including an input layer, multiple convolutional layers, a pooling layer, a fully connected layer, and an output layer; the convolutional layer is used to extract local features in the spectrum indicators, the pooling layer is used to reduce the feature dimension and prevent overfitting, and the fully connected layer is used to integrate features and perform classification;

[0026] S33: Initialize the parameters of the neural network, including convolution kernel size, pooling method, and activation function;

[0027] S34: input the training data set into the neural network model and perform training using a supervised learning method; calculate the predicted value of each sample through forward propagation, and compare it with the actual value to calculate the loss function value; the loss function is a mean square error or a cross entropy;

[0028] S35: Use the back propagation algorithm to adjust the parameters of the neural network according to the loss function value, gradually reduce the loss function value, and improve the prediction accuracy of the model;

[0029] S36: After the training is completed, the test data set is input into the trained neural network model to verify the classification performance of the model and evaluate its ability to distinguish short circuit fault characteristics and normal load changes;

[0030] S37: According to the test results, if the model performance does not meet the expected standards, adjust the model structure or training parameters and retrain until a satisfactory classification effect is achieved.

[0031] Optionally, the S35 specifically includes:

[0032] S351: In the forward propagation, the input data is passed through each layer of the neural network to calculate the activation value and output value of each neuron;

[0033] S352: Calculate a loss function value L, which is used to measure the difference between the model prediction result and the actual result, and the loss function is a cross entropy loss;

[0034] S353: Perform back propagation to calculate the gradient of the loss function for each weight and bias;

[0035] S354: Update weights and biases, and adjust the parameters of each layer using gradient descent;

[0036] S355: Repeat the above steps S351-S354 until the loss function value converges or reaches the preset training rounds, and the training of the supervised neural network model is completed; the supervised neural network model finally trained is expressed as:

[0037] Model(X)=Softmax(FC(ReLU(Pool(Conv(X,W (1) , b (1) )))…)), where X is the input spectrum index data; Conv(X, W (1) , b (1) ) represents the operation of the convolutional layer, W (1) and b (1) are the weight and bias of the layer respectively; Pool() represents the pooling layer, which is used to reduce the spatial size of the feature; ReLU() is the activation function, which is used to introduce nonlinearity; FC() represents the fully connected layer, which is used to output the final features to the classifier; Softmax() is the classification layer, which outputs the final classification probability.

[0038] Optionally, the S4 specifically includes:

[0039] S41: using the short-circuit fault characteristics and spectrum indicators obtained by the deep learning algorithm analysis as input, and matching them with the geographic and topological data of the distribution system, wherein the geographic and topological data include node locations, line connection relationships, and equipment locations of distribution lines;

[0040] S42: Use geographic information system to associate spectrum index with geographic location, determine potential fault areas, and calculate the change of spectrum index ΔF of each node. i Compare with the threshold T to filter out faulty nodes. The formula is: in, is the spectrum index of the currently measured i-th node, is the baseline spectrum index of the node; when ΔF i When ≥T, the node is marked as a potential fault node;

[0041] S43: performing correlation analysis on the initially screened fault node locations and the topological data of the power distribution system to determine the fault impact range. The topological data is used to identify other devices and lines connected to the fault node and evaluate the scope of the fault. Specifically, the topological connection matrix A is used to calculate the fault propagation path of each node. The formula is: R=A×P, where R is the fault propagation path matrix, P is the initial fault node label vector, and A is the topological connection matrix of the power distribution system.

[0042] S44: Use a weighted positioning algorithm to locate the selected nodes. The weighted positioning algorithm calculates the failure probability of each node according to the spectrum index weight, geographical location and connection relationship of the node. The formula is: Among them, P i is the failure probability of the i-th node, w i is the weight of the node, ΔF i is the change of the spectrum index of the node, j is all the nodes initially screened;

[0043] S45: According to the node with the highest fault probability, combined with the actual operation data of the distribution system, the precise location of the short circuit fault is finally determined, and a fault location report is generated, which includes the geographical location of the fault node, the fault type and the impact range.

[0044] Optionally, the S5 specifically includes

[0045] S51: After determining the short circuit location, automatically triggering a fault response program, which is integrated in the power distribution management system and immediately starts a preset fault handling process by receiving the short circuit location data from step S4;

[0046] S52: The fault response program first sends the fault information to the central control room of the power distribution system; the fault information includes the fault location, fault type and the impact range of the preliminary analysis; at the same time, the system automatically sends a fault notification to the maintenance team via email, SMS or mobile application push;

[0047] S53: The fault response program uses a geographic information system and a shortest path algorithm to calculate the optimal repair path from the current location of the maintenance team to the fault location. The path calculation takes into account the actual road conditions, traffic information, and the current location of the maintenance team;

[0048] S54: The calculated optimal maintenance path, estimated arrival time, and prepared maintenance measures are synchronously sent to the mobile device of the maintenance team.

[0049] Optionally, the S6 specifically includes:

[0050] S61: During the operation of the power distribution system, continuously collect fault handling data, including spectrum indicators of short-circuit faults, fault types, fault occurrence time, and fault handling time;

[0051] S62: Update the fast Fourier transform processing strategy using the newly collected data, and dynamically adjust the fast Fourier transform window size and sampling rate by comparing the spectrum characteristics of the new data and the historical data. The adjustment formula is: Among them, N new and N old are the new and old window sizes, σ new and σold are the standard deviations of the new and old data respectively;

[0052] S63: Perform self-learning optimization on the neural network model, use the newly collected data in S61 for incremental training, add new data to the original model and retrain, the formula is: Among them, W new is the updated weight, W old is the original weight, η is the learning rate, is the gradient of the loss function;

[0053] S64: Evaluate the performance of the neural network model after incremental training through cross-validation. Cross-validation includes dividing the data into a training set and a validation set. The training set is used to train the model, and the validation set is used to evaluate the model performance.

[0054] S65: Dynamically update the fault diagnosis rules, specifically adjust the decision boundary of the neural network model according to the new fault type and spectrum characteristics, so as to adapt to the changes in the power grid environment and the newly emerging fault types.

[0055] A distribution system short-circuit fault intelligent positioning system, used to implement the above-mentioned distribution system short-circuit fault intelligent positioning method, includes the following modules:

[0056] Data acquisition module: deployed at the predetermined nodes of the power distribution system, used to monitor the high-frequency part of the current in real time to capture the current spike when a short circuit occurs, and transmit the monitored current data to the data processing module;

[0057] Data processing module: connected to the data acquisition module, receives and pre-processes the transmitted current data, performs denoising and normalization, and converts the time domain current signal into the frequency domain signal through fast Fourier transform to extract the spectrum indicators related to the short circuit event;

[0058] Deep learning analysis module: connected to the data processing module, uses the preset convolutional neural network to analyze the extracted spectrum indicators and train the supervised neural network model to distinguish short-circuit fault characteristics and normal load changes;

[0059] Fault location module: connected to the deep learning analysis module, based on the geographical and topological data of the distribution system, the analysis results of the deep learning algorithm are combined with the geographic information system to determine the short circuit location;

[0060] Fault response module: connected to the fault location module, after determining the short circuit location, automatically starts the fault response program, sends the fault information to the central control room, and notifies the maintenance team;

[0061] Self-learning optimization module: connects the deep learning analysis module and the data processing module, continuously collects fault processing data during the operation of the distribution system, and uses self-learning methods to optimize the neural network model and fast Fourier transform processing strategy to adapt to changes in the power grid environment and new types of faults.

[0062] A computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the steps of the above-mentioned method for intelligently locating short-circuit faults in a power distribution system.

[0063] Beneficial effects of the present invention:

[0064] The present invention significantly improves the positioning speed and processing efficiency of short-circuit faults in the distribution system by integrating high-frequency current monitoring, fast Fourier transform processing, convolutional neural network analysis and geographic information system technology. This method can capture abnormal changes in current in real time, quickly analyze the characteristics of the fault, and accurately determine the location of the fault. The automatically triggered fault response process greatly shortens the time from fault detection to maintenance team response, effectively reduces the downtime of the power system, and improves the continuity and stability of power supply.

[0065] The present invention continuously optimizes fault detection and response strategies by adopting a self-learning method, so that the system can adapt to changes in the power grid environment and newly emerging fault types. Through continuous learning and adjustment, the system's fault handling capability is enhanced over time, thereby improving the overall safety and operating efficiency of the power grid. This adaptive optimization mechanism ensures the efficiency and reliability of the system in long-term operation, and meets the needs of modern power systems for high-tech fault handling solutions. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] In order to more clearly illustrate the technical solutions in the present invention or the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings in the following description are only for the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0067] Figure 1 A schematic diagram of a short-circuit fault intelligent positioning method according to an embodiment of the present invention;

[0068] Figure 2 Schematic diagram of a short-circuit fault intelligent positioning system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0069] The present invention is described in detail below in conjunction with the accompanying drawings and specific embodiments. At the same time, it is explained here that in order to make the embodiments more detailed, the following embodiments are the best and preferred embodiments, and those skilled in the art may also adopt other alternatives to implement some known technologies; and the accompanying drawings are only for more specific description of the embodiments, and are not intended to specifically limit the present invention.

[0070] It should be noted that the references to "one embodiment", "an embodiment", "an exemplary embodiment", "some embodiments" and the like in the specification indicate that the embodiments described may include specific features, structures or characteristics, but not every embodiment may include the specific features, structures or characteristics. In addition, when a specific feature, structure or characteristic is described in conjunction with an embodiment, it should be within the knowledge of a person skilled in the art to implement such feature, structure or characteristic in conjunction with other embodiments (whether or not explicitly described).

[0071] In general, a term can be understood, at least in part, from its use in context. For example, depending, at least in part, on the context, the term "one or more" as used herein can be used to describe any feature, structure, or characteristic in the singular sense, or can be used to describe a combination of features, structures, or characteristics in the plural sense. Additionally, the term "based on" can be understood as not necessarily intended to convey an exclusive set of factors, but can instead, depending, at least in part, on the context, allow for the presence of other factors that are not necessarily explicitly described.

[0072] like Figure 1 As shown, a method for intelligently locating short-circuit faults in a power distribution system comprises the following steps:

[0073] S1: Deploy current monitoring devices based on optical fiber technology at predetermined nodes of the distribution system to monitor the high-frequency part of the current in real time to capture current spikes when short-circuit events occur;

[0074] S2: Perform fast Fourier transform (FFT) processing on the data collected by the optical fiber monitoring device, analyze the spectrum characteristics of the current signal, and extract spectrum indicators related to the short-circuit event;

[0075] S3: Use a preset deep learning algorithm to analyze the extracted spectrum indicators and train a supervised neural network model to distinguish short-circuit fault characteristics from normal load changes;

[0076] S4: Based on the geographic and topological data of the distribution system, the analysis results of the deep learning algorithm are combined with geographic information to determine the short circuit location;

[0077] S5: Once the short circuit location is determined, the fault response program is automatically initiated to notify the maintenance team and provide guidance on the optimal repair path to reduce repair time and impact range;

[0078] S6: During the operation of the distribution system, fault processing data is continuously collected, and the neural network model and fast Fourier transform processing strategy are optimized using self-learning methods to adapt to changes in the power grid environment and emerging fault types.

[0079] S1 specifically includes:

[0080] S11: Predetermined nodes of the power distribution system, including transformer outlets, important branch nodes or load concentration areas, are used as deployment locations for current monitoring devices to ensure that the most critical current change information in the system can be covered;

[0081] S12: Install a current monitoring device based on fiber Bragg grating (FBG) technology at the selected nodes. The device uses the dependence of fiber Bragg grating on the wavelength of incident light to sense changes in the surrounding magnetic field, thereby accurately monitoring the high-frequency part of the current;

[0082] S13: configuring a signal processing unit connected to the current monitoring device, the unit comprising a light source, a spectrum analyzer and a data converter; the light source is used to emit a stable light wave to the optical fiber, when passing through the fiber grating, the wavelength of the light wave will change due to the influence of the magnetic field generated by the current;

[0083] S14: The spectrum analyzer receives the light waves affected by the current, analyzes the wavelength change, and converts the wavelength change into the current value. By monitoring the high-frequency wavelength change, the current spike when the short circuit event occurs can be quickly captured;

[0084] S15: The data converter is used to convert the output of the spectrum analyzer into a digital signal and send it to the central processing unit in real time for further data analysis and short-circuit fault location; by deploying current monitoring devices based on fiber Bragg grating (FBG) technology at key nodes of the distribution system, a high-sensitivity and high-response speed monitoring method is provided. The use of fiber optic technology not only improves the measurement accuracy, but also enhances the stability and reliability of the system in a complex power environment due to its inherent electromagnetic interference immunity. In addition, real-time monitoring of the high-frequency part of the current can ensure that key data is captured quickly and accurately in the first time a short circuit occurs, thereby significantly improving the fault response efficiency and the overall safety of the system.

[0085] S2 specifically includes:

[0086] S21: The current signal data collected by the optical fiber monitoring device is transmitted from the signal processing unit to the central processing unit to ensure the integrity and real-time performance of the data. In the central processing unit, the received current signal data is preprocessed, including denoising and normalization. Denoising can use a bandpass filter to filter out low-frequency and high-frequency noise signals and retain useful signal parts. Normalization uses the following formula:

[0087] Wherein, x is the original current signal data, y is the normalized data, min(x) and max(x) are the minimum and maximum values ​​of the current signal data respectively;

[0088] S22: Perform fast Fourier transform (FFT) on the pre-processed current signal data to convert the time domain signal into a frequency domain signal; the calculation formula is: Wherein, X(k) is the frequency domain signal, x(n) is the time domain signal, N is the length of the current signal data, k is the frequency index, and j is the imaginary unit;

[0089] S23: Analyze the spectrum after fast Fourier transform processing, identify the spectrum characteristics, and extract the spectrum indicators related to the short circuit event by observing the amplitude and phase changes of the frequency components in the spectrum. The spectrum indicators include amplitude spectrum peak, spectrum energy and spectrum center. The calculation formulas are:

[0090] Amplitude spectrum peak: A peak = max(|X(k)|), where A peak is the peak value of the amplitude spectrum, |X(k)| is the amplitude of the frequency domain signal;

[0091] Spectral energy: Where E is the spectrum energy, |X(k)| is the amplitude of the frequency domain signal;

[0092] Spectrum Center: Where C is the center of the spectrum, k is the frequency index, and |X(k)| is the amplitude of the frequency domain signal;

[0093] S24: The extracted spectrum indicators are saved to the database for use by subsequent deep learning algorithms to distinguish short-circuit fault characteristics from normal load changes; by calculating and extracting spectrum indicators related to short-circuit events, the accuracy and reliability of fault detection can be significantly improved. This processing method not only improves the accuracy of current signal analysis, but also can capture the characteristics of short-circuit events in a timely manner, providing a high-quality data foundation for subsequent deep learning algorithms, thereby achieving more accurate short-circuit fault location and response.

[0094] Training the supervised neural network model in S3 specifically includes:

[0095] S31: Use the spectrum indicators extracted in S2 as input data sets to prepare training and test data sets; the training data set is used for model training, and the test data set is used to verify the accuracy of the model;

[0096] S32: Select convolutional neural network (CNN) as the preset deep learning algorithm, and construct a neural network structure including an input layer, multiple convolutional layers, a pooling layer, a fully connected layer, and an output layer; the convolutional layer is used to extract local features in the spectrum index, the pooling layer is used to reduce the feature dimension and prevent overfitting, and the fully connected layer is used to integrate features and perform classification;

[0097] S33: Initialize the parameters of the neural network, including the convolution kernel size, pooling method, and activation function, to ensure that the model has a good initial state;

[0098] S34: Input the training data set into the neural network model and train it using supervised learning; calculate the predicted value of each sample through forward propagation, compare it with the actual value, and calculate the loss function value; the loss function is mean square error or cross entropy;

[0099] S35: Use the back propagation algorithm to adjust the parameters of the neural network according to the loss function value, gradually reduce the loss function value, and improve the prediction accuracy of the model; back propagation achieves parameter adjustment by calculating gradients and updating weights and biases;

[0100] S36: After the training is completed, the test data set is input into the trained neural network model to verify the classification performance of the model and evaluate its ability to distinguish short-circuit fault characteristics from normal load changes. Specifically, the performance of the model is evaluated by calculating the classification accuracy, recall rate, and F1 score.

[0101] S37: According to the test results, if the model performance does not meet the expected standards, the model structure or training parameters are adjusted and retrained until a satisfactory classification effect is achieved; through the above steps, the extracted spectrum indicators are analyzed and the supervised neural network model is trained using a convolutional neural network (CNN). The present invention can effectively distinguish short-circuit fault characteristics from normal load changes. The use of convolutional neural networks improves the feature extraction capability and classification accuracy, ensuring that the model has good robustness and generalization capabilities under different working conditions. The model parameters are optimized through supervised learning and back-propagation algorithms, which improves the accuracy and response speed of short-circuit fault detection, providing reliable protection for the safe operation of the power system.

[0102] The adjustment of the parameters of the neural network in S35 specifically includes:

[0103] S351: In the forward propagation, the input data is passed through each layer of the neural network to calculate the activation value and output value of each neuron. For the convolution layer, the output value is set to 0, and the calculation formula is:

[0104] in, is the output of the lth convolutional layer, is the convolution kernel weight, is the input of the l-1th layer, b (l) is the bias, f is the activation function;

[0105] S352: Calculate the loss function value L, which is used to measure the difference between the model prediction result and the actual result. The loss function is the cross-entropy loss (Cross-EntropyLoss); the formula is:

[0106] Where N is the number of samples, y i is the true label of the i-th sample, is the predicted probability of the i-th sample;

[0107] S353: Perform back propagation to calculate the gradient of the loss function for each weight and bias, and the gradient of the convolutional layer weight W The calculation formula is: in, is the error term of the lth layer, and the calculation formula is: Where f′ is the derivative of the activation function;

[0108] S354: Update weights and biases, and use the gradient descent method to adjust the parameters of each layer; the update formula for weight W and bias b is: and Among them, η is the learning rate, which is used to control the update step size;

[0109] S355: Repeat the steps of forward propagation, loss calculation, back propagation and parameter update of S351-S354 above until the loss function value converges or reaches the preset training rounds, and the training of the supervised neural network model is completed; the supervised neural network model finally trained is expressed as:

[0110] Model(X)=Softmax(FC(ReLU(Pool(Conv(X,W (1) ,b (1) )))…)), where X is the input spectrum index data; Conv(X,W (1) ,b (1) ) represents the operation of the convolutional layer, W (1) and b (1)are the weight and bias of the layer respectively; Pool() represents the pooling layer, which is used to reduce the spatial size of the feature; ReLU() is the activation function, which is used to introduce nonlinearity; FC() represents the fully connected layer, which is used to output the final feature to the classifier; Softmax() is the classification layer, which outputs the final classification probability; through the neural network model structure described in detail in the above steps, the neural network of the present invention can ensure efficient learning and accurate classification performance through fine adjustment and optimization of each layer. The model structure uses the convolution layer to accurately extract the features of the input data, enhances the generalization ability of the model through the pooling layer and nonlinear activation, and outputs the final fault identification result through the fully connected layer and the Softmax classifier. This comprehensive model architecture ensures that the short-circuit fault characteristics and normal load changes can be accurately distinguished in practical applications, and significantly improves the detection accuracy and response speed of short-circuit faults in the power system.

[0111] S4 specifically includes:

[0112] S41: The short-circuit fault characteristics and their spectrum indicators obtained by the deep learning algorithm analysis are used as inputs to match the geographic and topological data of the distribution system, which includes the node locations, line connection relationships, and equipment locations of the distribution lines;

[0113] S42: Use Geographic Information System (GIS) to associate spectrum indicators with geographic locations to determine potential fault areas, and calculate the change in spectrum indicators ΔF for each node. i Compare with the threshold T to filter out faulty nodes. The formula is: in, is the spectrum index of the currently measured i-th node, is the baseline spectrum index of the node; when ΔF i When ≥T, the node is marked as a potential fault node;

[0114] S43: further correlation analysis is performed on the initially screened fault node locations and the topological data of the power distribution system to determine the fault impact range. The topological data is used to identify other devices and lines connected to the fault node and evaluate the scope of the fault. Specifically, the topological connection matrix A is used to calculate the fault propagation path of each node. The formula is: R = A × P, where R is the fault propagation path matrix, P is the preliminary fault node label vector, and A is the topological connection matrix of the power distribution system;

[0115] S44: Use a weighted positioning algorithm to further locate the selected nodes. The weighted positioning algorithm calculates the failure probability of each node according to the spectrum index weight, geographical location and connection relationship of the node. The formula is: Among them, P i is the failure probability of the i-th node, w iis the weight of the node, ΔF i is the change of the spectrum index of the node, j is all the nodes initially screened;

[0116] S45: Based on the node with the highest probability of failure and the actual operation data of the power distribution system, the precise location of the short-circuit fault is finally determined, and a fault location report is generated. The report includes the geographical location of the fault node, the fault type, and the possible impact range, so that the maintenance team can respond quickly.

[0117] Through the above detailed steps, the present invention can combine the analysis results of the deep learning algorithm with the geographic and topological data of the distribution system to accurately locate the short-circuit fault location. By using the geographic information system (GIS) and the weighted positioning algorithm, through a specific calculation formula, not only the accuracy of fault location is improved, but also the impact range of the fault can be evaluated, providing comprehensive information support for fault handling.

[0118] The fault response procedure in S5 specifically includes

[0119] S51: After determining the short circuit location, automatically trigger the fault response program, which is integrated in the power distribution management system and has real-time data processing and event triggering capabilities. The system immediately starts the preset fault handling process by receiving the short circuit location data from step S4;

[0120] S52: The fault response procedure first sends the fault information to the central control room of the power distribution system; the fault information includes the fault location, fault type and the impact range of the preliminary analysis. The information transmission is carried out through the built-in security network of the power distribution system to ensure the accuracy and timeliness of the information; at the same time, the system automatically sends a fault notification to the maintenance team, which is sent via email, SMS or mobile application push, depending on the actual configuration of the maintenance team. The notification includes detailed information of the fault and a specific call to the maintenance team members;

[0121] S53: The fault response program uses a geographic information system (GIS) and a shortest path algorithm to calculate the optimal maintenance path from the current location of the maintenance team to the fault location. The path calculation takes into account the actual road conditions, traffic information, and the current location of the maintenance team. The shortest path algorithm uses the Dijkstra algorithm. The specific steps include:

[0122] Step 1: Define a graph G = (V, E), where V represents a set of nodes, E represents a set of edges, and each edge has a weight w, which represents the distance or time from one node to another.

[0123] Step 2: Starting from the maintenance team’s current location node s, initialize the distance d of all nodes to infinity, and set the distance of the starting node to 0;

[0124] Step 3: Use the priority queue to select the node u with the smallest current distance, and update the distance d(v) between it and its adjacent node v. The calculation formula is: d(v) = min(d(v), d(u) + w(u, v));

[0125] Step 4: Repeat the above steps until the shortest paths of all nodes are determined, and finally the shortest path P from the current location s to the fault location t is obtained.

[0126] S54: The calculated optimal maintenance path, estimated arrival time and prepared maintenance measures are synchronously sent to the maintenance team's mobile device. The maintenance team can view the route details through a dedicated application and update the route status and fault handling progress in real time. Through the above steps, the present invention can quickly and effectively start the fault response program after the short-circuit fault location is accurately determined, ensuring the high efficiency of fault handling and the rapid recovery of the system. Through the combination of geographic information system (GIS) and the shortest path algorithm, the optimal maintenance path from the current location of the maintenance team to the fault location is calculated, which improves the response speed and the accuracy of path selection. The provision of automated fault notification and optimal maintenance path not only optimizes the use and scheduling of resources, but also reduces the maintenance time and the power outage time of users.

[0127] S6 specifically includes:

[0128] S61: During the operation of the distribution system, fault handling data is continuously collected, including the spectrum index of short-circuit faults, fault types, fault occurrence time and fault handling time. All data are stored in the central database for further analysis and processing;

[0129] S62: Update the fast Fourier transform (FFT) processing strategy using the newly collected data. By comparing the spectrum characteristics of the new data and the historical data, dynamically adjust the window size and sampling rate of the fast Fourier transform to improve the accuracy of the spectrum analysis. The adjustment formula is: Among them, N new and N old are the new and old window sizes, σ new and σ old are the standard deviations of the new and old data respectively;

[0130] S63: Perform self-learning optimization on the neural network model, use the newly collected data in S61 for incremental training, add new data to the original model and retrain, the formula is: Among them, W new is the updated weight, W old is the original weight, η is the learning rate, is the gradient of the loss function;

[0131] S64: Evaluate the performance of the neural network model after incremental training through cross-validation to ensure the prediction accuracy and generalization ability of the model on new data. Cross-validation includes dividing the data into a training set and a validation set. The training set is used to train the model, and the validation set is used to evaluate the model performance.

[0132] S65: Dynamically update fault diagnosis rules, specifically adjust the decision boundary of the neural network model according to new fault types and spectrum characteristics to adapt to changes in the power grid environment and newly emerging fault types. The adjustment of the decision boundary is achieved by maximizing the classification accuracy and minimizing the false alarm rate. Through the above detailed steps, fault processing data can be continuously collected during the operation of the distribution system, and the neural network model and fast Fourier transform (FFT) processing strategy can be dynamically optimized using a self-learning method. Through incremental training, cross-validation and dynamic adjustment, the system can adapt to changes in the power grid environment and newly emerging fault types, thereby improving the accuracy and response speed of short-circuit fault detection. This adaptive optimization mechanism ensures the long-term stability and reliability of the system and effectively improves the fault handling capability and operating efficiency of the distribution system.

[0133] like Figure 2 As shown, a power distribution system short-circuit fault intelligent positioning system is used to implement the above-mentioned power distribution system short-circuit fault intelligent positioning method, including the following modules:

[0134] Data acquisition module: deployed at the predetermined nodes of the power distribution system, used to monitor the high-frequency part of the current in real time to capture the current spike when a short circuit occurs, and transmit the monitored current data to the data processing module;

[0135] Data processing module: connected to the data acquisition module, receives and pre-processes the transmitted current data, performs denoising and normalization, and converts the time domain current signal into the frequency domain signal through fast Fourier transform to extract the spectrum indicators related to the short circuit event;

[0136] Deep learning analysis module: connected to the data processing module, uses the preset convolutional neural network to analyze the extracted spectrum indicators and train the supervised neural network model to distinguish short-circuit fault characteristics and normal load changes;

[0137] Fault location module: connected to the deep learning analysis module, based on the geographical and topological data of the distribution system, the analysis results of the deep learning algorithm are combined with the geographic information system (GIS) to determine the short circuit location;

[0138] Fault response module: connected to the fault location module, after determining the short circuit location, automatically starts the fault response program, sends the fault information to the central control room, and notifies the maintenance team; uses the geographic information system (GIS) and the shortest path algorithm to calculate the optimal maintenance path from the current location of the maintenance team to the fault location, and provides maintenance path guidance;

[0139] Self-learning optimization module: connects the deep learning analysis module and the data processing module, continuously collects fault processing data during the operation of the distribution system, and uses self-learning methods to optimize the neural network model and fast Fourier transform processing strategy to adapt to changes in the power grid environment and new types of faults.

[0140] A computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the steps of the above-mentioned method for intelligently locating short-circuit faults in a power distribution system.

[0141] The present invention covers any substitution, modification, equivalent method and scheme made on the essence and scope of the present invention. In order to make the public have a thorough understanding of the present invention, specific details are described in detail in the following preferred embodiments of the present invention, but those skilled in the art can fully understand the present invention without the description of these details. In addition, in order to avoid unnecessary confusion about the essence of the present invention, well-known methods, processes, procedures, components and circuits are not described in detail.

[0142] The above is only a preferred embodiment of the present invention. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications should also be regarded as the scope of protection of the present invention.

Claims

1. A method for intelligently locating short-circuit faults in a power distribution system, characterized in that: The following steps are involved: S1: Deploy current monitoring devices based on optical fiber technology at predetermined nodes of the distribution system to monitor the high-frequency part of the current in real time to capture current spikes when short-circuit events occur; S2: Perform fast Fourier transform processing on the data collected by the optical fiber monitoring device, analyze the spectrum characteristics of the current signal, and extract spectrum indicators related to the short-circuit event; S3: Use a preset deep learning algorithm to analyze the extracted spectrum indicators and train a supervised neural network model to distinguish short-circuit fault characteristics from normal load changes; S4: Based on the geographic and topological data of the distribution system, the analysis results of the deep learning algorithm are combined with geographic information to determine the short circuit location; S5: Once the short circuit location is determined, the fault response program is automatically initiated to notify the maintenance team and provide guidance on the optimal repair path; S6: During the operation of the distribution system, fault processing data is continuously collected, and the neural network model and fast Fourier transform processing strategy are optimized using self-learning methods to adapt to changes in the power grid environment and emerging fault types.

2. A method for intelligently locating short-circuit faults in a power distribution system according to claim 1, characterized in that: The S1 specifically includes: S11: Predetermined nodes of the power distribution system, including transformer outlets, important branch nodes or load concentration areas, used as deployment locations for current monitoring devices; S12: Install a current monitoring device based on fiber Bragg grating technology at the selected nodes. The device uses the dependence of fiber Bragg grating on the wavelength of incident light to sense changes in the surrounding magnetic field, thereby accurately monitoring the high-frequency part of the current; S13: configuring a signal processing unit connected to the current monitoring device, the unit comprising a light source, a spectrum analyzer and a data converter; the light source is used to emit a stable light wave to the optical fiber; S14: The spectrum analyzer receives the light waves affected by the current, analyzes the wavelength change, and converts the wavelength change into the current value. By monitoring the high-frequency wavelength change, the current spike when the short circuit event occurs is captured; S15: The data converter is used to convert the output of the spectrum analyzer into a digital signal.

3. The method for intelligently locating short-circuit faults in a power distribution system according to claim 1, characterized in that: The S2 specifically includes: S21: transmitting the current signal data collected by the optical fiber monitoring device from the signal processing unit to the central processing unit, and preprocessing the received current signal data in the central processing unit, including denoising and normalization processing; S22: Perform fast Fourier transform on the preprocessed current signal data to convert the time domain signal into a frequency domain signal; the calculation formula is: Wherein, X(k) is the frequency domain signal, x(n) is the time domain signal, N is the length of the current signal data, k is the frequency index, and j is the imaginary unit; S23: Analyze the spectrum after fast Fourier transform processing, identify spectrum characteristics, and extract spectrum indicators related to the short circuit event by observing the amplitude and phase changes of frequency components in the spectrum. The spectrum indicators include amplitude spectrum peak, spectrum energy and spectrum center; S24: Save the extracted spectrum index to the database.

4. A method for intelligently locating short-circuit faults in a power distribution system according to claim 3, characterized in that: The S3 specifically includes: S31: Use the spectrum indicators extracted in S2 as input data sets to prepare training and test data sets; the training data set is used for model training, and the test data set is used to verify the accuracy of the model; S32: Select convolutional neural network as the preset deep learning algorithm, and construct a neural network structure including an input layer, multiple convolutional layers, a pooling layer, a fully connected layer, and an output layer; the convolutional layer is used to extract local features in the spectrum indicators, the pooling layer is used to reduce the feature dimension and prevent overfitting, and the fully connected layer is used to integrate features and perform classification; S33: Initialize the parameters of the neural network, including convolution kernel size, pooling method, and activation function; S34: input the training data set into the neural network model and perform training using a supervised learning method; calculate the predicted value of each sample through forward propagation, and compare it with the actual value to calculate the loss function value; the loss function is a mean square error or a cross entropy; S35: Use the back propagation algorithm to adjust the parameters of the neural network according to the loss function value, gradually reduce the loss function value, and improve the prediction accuracy of the model; S36: After the training is completed, the test data set is input into the trained neural network model to verify the classification performance of the model and evaluate its ability to distinguish short circuit fault characteristics and normal load changes; S37: According to the test results, if the model performance does not meet the expected standards, adjust the model structure or training parameters and retrain until a satisfactory classification effect is achieved.

5. A method for intelligently locating short-circuit faults in a power distribution system according to claim 4, characterized in that: The S35 specifically includes: S351: In the forward propagation, the input data is passed through each layer of the neural network to calculate the activation value and output value of each neuron; S352: Calculate a loss function value L, which is used to measure the difference between the model prediction result and the actual result, and the loss function is a cross entropy loss; S353: Perform back propagation to calculate the gradient of the loss function for each weight and bias; S354: Update weights and biases, and adjust the parameters of each layer using gradient descent; S355: Repeat the above steps S351-S354 until the loss function value converges or reaches the preset training rounds, and the training of the supervised neural network model is completed; the supervised neural network model finally trained is expressed as: Model(X)=Softmax(FC(ReLU(Pool(Conv(X,W (1) ,b (1) )))…)), where X is the input spectrum index data; Conv(X,W (1) ,b (1) ) represents the operation of the convolutional layer, W (1) and b (1) are the weight and bias of the layer respectively; Pool() represents the pooling layer, which is used to reduce the spatial size of the feature; ReLU() is the activation function, which is used to introduce nonlinearity; FC() represents the fully connected layer, which is used to output the final features to the classifier; Softmax() is the classification layer, which outputs the final classification probability.

6. The method for intelligently locating short-circuit faults in a power distribution system according to claim 1, characterized in that: The S4 specifically includes: S41: using the short-circuit fault characteristics and spectrum indicators obtained by the deep learning algorithm analysis as input, and matching them with the geographic and topological data of the distribution system, wherein the geographic and topological data include node locations, line connection relationships, and equipment locations of distribution lines; S42: Use geographic information system to associate spectrum index with geographic location, determine potential fault areas, and calculate the change of spectrum index ΔF of each node. i Compare with the threshold T to filter out faulty nodes. The formula is: in, is the spectrum index of the currently measured i-th node, is the baseline spectrum index of the node; when ΔF i When ≥T, the node is marked as a potential fault node; S43: performing correlation analysis on the initially screened fault node locations and the topological data of the power distribution system to determine the fault impact range. The topological data is used to identify other devices and lines connected to the fault node and evaluate the scope of the fault. Specifically, the topological connection matrix A is used to calculate the fault propagation path of each node. The formula is: R=A×P, where R is the fault propagation path matrix, P is the initial fault node label vector, and A is the topological connection matrix of the power distribution system. S44: Use a weighted positioning algorithm to locate the selected nodes. The weighted positioning algorithm calculates the failure probability of each node according to the spectrum index weight, geographical location and connection relationship of the node. The formula is: Among them, P i is the failure probability of the i-th node, w i is the weight of the node, ΔF i is the change of the spectrum index of the node, j is all the nodes initially screened; S45: According to the node with the highest fault probability, combined with the actual operation data of the distribution system, the precise location of the short circuit fault is finally determined, and a fault location report is generated, which includes the geographical location of the fault node, the fault type and the impact range.

7. The method for intelligently locating short-circuit faults in a power distribution system according to claim 1, characterized in that: The S5 specifically includes S51: After determining the short circuit location, automatically triggering a fault response program, which is integrated in the power distribution management system and immediately starts a preset fault handling process by receiving the short circuit location data from step S4; S52: The fault response program first sends the fault information to the central control room of the power distribution system; the fault information includes the fault location, fault type and the impact range of the preliminary analysis; at the same time, the system automatically sends a fault notification to the maintenance team via email, SMS or mobile application push; S53: The fault response program uses a geographic information system and a shortest path algorithm to calculate the optimal repair path from the current location of the maintenance team to the fault location. The path calculation takes into account the actual road conditions, traffic information, and the current location of the maintenance team; S54: The calculated optimal maintenance path, estimated arrival time, and prepared maintenance measures are synchronously sent to the mobile device of the maintenance team.

8. The method for intelligently locating short-circuit faults in a power distribution system according to claim 1, characterized in that: The S6 specifically includes: S61: During the operation of the power distribution system, continuously collect fault handling data, including spectrum indicators of short-circuit faults, fault types, fault occurrence time, and fault handling time; S62: Update the fast Fourier transform processing strategy using the newly collected data, and dynamically adjust the fast Fourier transform window size and sampling rate by comparing the spectrum characteristics of the new data and the historical data. The adjustment formula is: Among them, N new and N old are the new and old window sizes, σ new and σ old are the standard deviations of the new and old data respectively; S63: Perform self-learning optimization on the neural network model, use the newly collected data in S61 for incremental training, add new data to the original model and retrain, the formula is: Among them, W new is the updated weight, W old is the original weight, η is the learning rate, is the gradient of the loss function; S64: Evaluate the performance of the neural network model after incremental training through cross-validation. Cross-validation includes dividing the data into a training set and a validation set. The training set is used to train the model, and the validation set is used to evaluate the model performance. S65: Dynamically update the fault diagnosis rules, specifically adjust the decision boundary of the neural network model according to the new fault type and spectrum characteristics, so as to adapt to the changes in the power grid environment and the newly emerging fault types.

9. A distribution system short-circuit fault intelligent positioning system, used to implement a distribution system short-circuit fault intelligent positioning method as claimed in any one of claims 1 to 8, characterized in that: Includes the following modules: Data acquisition module: deployed at the predetermined nodes of the power distribution system, used to monitor the high-frequency part of the current in real time to capture the current spike when a short circuit occurs, and transmit the monitored current data to the data processing module; Data processing module: connected to the data acquisition module, receives and pre-processes the transmitted current data, performs denoising and normalization, and converts the time domain current signal into the frequency domain signal through fast Fourier transform to extract the spectrum indicators related to the short circuit event; Deep learning analysis module: connected to the data processing module, uses the preset convolutional neural network to analyze the extracted spectrum indicators and train the supervised neural network model to distinguish short-circuit fault characteristics and normal load changes; Fault location module: connected to the deep learning analysis module, based on the geographical and topological data of the distribution system, the analysis results of the deep learning algorithm are combined with the geographic information system to determine the short circuit location; Fault response module: connected to the fault location module, after determining the short circuit location, automatically starts the fault response program, sends the fault information to the central control room, and notifies the maintenance team; Self-learning optimization module: connects the deep learning analysis module and the data processing module, continuously collects fault processing data during the operation of the distribution system, and uses self-learning methods to optimize the neural network model and fast Fourier transform processing strategy to adapt to changes in the power grid environment and new types of faults.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the steps of the method for intelligently locating short-circuit faults in a power distribution system according to any one of claims 1 to 8 are implemented.

Images