Distribution network fault section location detection method and related device based on Transformer and particle swarm hybrid

Abstract

A method and related apparatus for fault location and detection in distribution networks based on a hybrid Transformer and Particle Swarm Optimization (PSO) approach are disclosed. The method includes: constructing and initializing the upstream segment matrix, downstream segment matrix, upstream power source matrix, and downstream power source matrix at the current moment; reading the operating parameters of the regional distribution network and the configuration data of the measuring equipment; training a Transformer model using the operating parameters and the configuration data of the measuring equipment; extracting current fault information from feeder terminal units (FTUs) and using the Transformer model to complete the missing data; constructing an objective function based on the constructed matrices and the data obtained after completing the missing data, and solving the objective function to locate the fault segment. This invention introduces an intelligent algorithm into the existing FTU configuration of the distribution network to supplement missing data and uses a particle swarm optimization algorithm for solving, thereby improving the computational accuracy and robustness of fault location in the distribution network.

Description

Technical Field

[0001] This invention relates to the field of power system monitoring and data analysis, specifically a method and related device for locating and detecting fault sections in distribution networks based on a hybrid Transformer and particle swarm optimization approach. Background Technology

[0002] Distribution network condition fault detection and islanding detection are key technologies for accurately perceiving the state of the power system. They provide fundamental data for advanced applications such as relay protection systems, fault location, islanding detection, voltage control, and economic dispatch. With the large-scale integration of distributed power sources such as photovoltaics, wind turbines, and various energy storage devices, as well as flexible loads like electric vehicles on the load side, into the distribution network, the system's randomness and volatility have significantly increased. The future vision of electric vehicles as energy storage devices for the grid may become a reality as verification progresses, making the operating state of the distribution network under high-proportion inverter conditions more complex and uncertain than that of traditional distribution networks. The direction of power flow in the distribution network under high-proportion inverters has undergone a significant change in complexity compared to the unidirectional power flow in traditional distribution networks. Therefore, research on fast and accurate fault location and islanding detection algorithms for distributed distribution networks under high-proportion inverters has become particularly important.

[0003] Fault location in power grids has long been a research topic in power systems. Traditional methods for fault location in distribution networks mainly include traveling wave method, impedance method, and matrix algorithm. However, with the current integration of distributed generation into the grid, the phase of voltage and current in the distribution network changes, and impedance calculation becomes difficult due to power flow variations.

[0004] After a fault occurs in the distribution network, the relevant FTUs and RTUs transmit real-time information from the corresponding sectionalizing switches and tie switches to the control center via data communication. The fault location software then locates the fault. However, in reality, due to various reasons, information distortion or packet loss may occur in the real-time information, causing ordinary algorithms to misjudge the fault. Traditional matrix fault detection methods establish a suitable mathematical model for the distribution network by constructing evaluation functions. Different evaluation functions have different evaluation effects and redundancy. These methods are characterized by the difficulty in constructing the mathematical model, the randomness of redundancy judgment, poor robustness, and unsuitability for judging faults caused by a high proportion of inverters forming islands.

[0005] Therefore, a model with higher interpretability and robustness is needed to accurately reflect the fault state of the distribution network under conditions of signal integrity or distortion, in order to meet the current scenario of distribution networks with a high proportion of inverters and to quickly and accurately calculate the real fault range. Summary of the Invention

[0006] To address the aforementioned problems, this invention proposes a method and related device for fault location and detection in distribution networks based on a hybrid Transformer and Particle Swarm Optimization (PSO) approach. This invention introduces intelligent algorithms to supplement missing data based on the existing FTU configuration of the distribution network and uses the PSO algorithm for solving the problem, thereby improving the calculation accuracy and robustness of fault location in the distribution network.

[0007] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows:

[0008] A method for locating and detecting fault sections in a distribution network based on a hybrid Transformer and Particle Swarm Optimization approach includes the following steps:

[0009] Construct and initialize the upstream segment matrix, downstream segment matrix, upstream power source matrix, and downstream power source matrix at the current moment;

[0010] Read the operating parameters of the regional power distribution network and the configuration data of the measuring equipment;

[0011] The read operating parameters of the power distribution network and the configuration data of the measuring equipment are divided into training set, validation set and test set as datasets for training the Transformer model;

[0012] Extract current fault information from the feeder terminal unit (FTU) and assess the completeness of the data in the fault information. If data is missing, call the trained Transformer model and perform step-by-step prediction on the missing data based on historical data to gradually fill in the missing data.

[0013] Based on the constructed matrix and the data obtained after completing the missing data using the Tranformer model, an overall objective function and a special objective function for island detection are constructed. The objective function is solved using the binary particle swarm optimization algorithm to locate the fault segment.

[0014] Based on the location of the faulty section, the faulty section is analyzed, and it is determined whether the islanding status in the distribution network has changed. If an island is detected, the generated section is recorded, and island location support is provided for the island-specific objective function. If the fault has been recovered, the island-specific objective function is deleted in the next round of calculation.

[0015] Furthermore, the operating parameters of the regional distribution network include the load, equipment parameters, and feeder terminal unit (FTU) settings in the distribution network, and the configuration data of the measurement equipment includes the installation location, measurement type, and measurement data of the measurement equipment.

[0016] Furthermore, the training steps of the Transformer model include: performing random masking on the input node feature data to simulate various node data missing situations that may occur in real applications, so that the model faces diverse missing patterns during training. The Transformer model uses a self-attention mechanism to capture the correlation between known nodes and missing nodes globally. When the model receives input data containing a mask, the self-attention mechanism aggregates the information of known nodes and calculates the most likely feature estimate for the missing nodes. The input of the trained Transformer model is missing data with a random mask, and the output is complete original data.

[0017] Furthermore, the overall objective function is constructed for the entire power grid system, while the dedicated objective function for island detection is used to determine whether islands exist after the first round of operation of the power grid system. If islands exist, the objective function is applied separately to each island.

[0018] Furthermore, the construction method for the overall objective function and the island detection-specific objective function is consistent, as shown in the following equation:

[0019]

[0020] in It is the deviation between the actual location of the fault in the distribution network and the predicted value determined by the switch state function. It is a predicted value; This indicates the total number of feeder switches in the distribution network; This represents the total number of feeder sections in the distribution network. For the feeder section, the state variables are {0, 1}; This represents the total number of faulty sections. To prevent misjudgment of coefficients, the concept of a minimum set is used. exist Values; This is the status variable of the feeder switch. When it is 1, it indicates that the short-circuit current flowing through the feeder switch flows from the system power supply to the end of the line. When it is -1, it indicates that the short-circuit overcurrent flowing through the feeder switch flows from the end of the line to the system power supply side.

[0021] Furthermore, the determination of whether an island exists specifically includes:

[0022] (a) Determine the area affected by the fault

[0023] According to the location of the fault Determine the upstream segment associated with the location of the fault. and downstream section ;

[0024] Find the segment index corresponding to the fault location from the upstream segment matrix and the downstream segment matrix;

[0025] Based on the segment index, the set of upstream segments affected by the fault is set as follows. The downstream segment set is ;

[0026] (II) Power Connectivity Analysis

[0027] For the upstream segment set :

[0028] Check upstream power matrix related elements Determine the switching status of the upstream power supply connected to each upstream segment;

[0029] Based on the logical operations in the matrix formula, calculate whether each upstream segment can be connected by power supply;

[0030] For upstream sections that cannot be connected by power, they are marked as part of a potential islanding area;

[0031] For the downstream segment set :

[0032] Check downstream power matrix related elements Determine the switching status of downstream power supplies connected to each downstream segment;

[0033] Based on the logical operations in the matrix formula, calculate whether each downstream segment can be connected by power.

[0034] For downstream sections that cannot be connected by power, they are marked as part of a potential islanding area;

[0035] (III) Island Determination

[0036] Taking into account the power connection situation of both upstream and downstream sections:

[0037] If there is a segment that belongs to both the upstream and downstream potential island areas, or if there is an isolated segment that cannot be connected to power in either the upstream or downstream, then an island is determined to have been formed.

[0038] If all affected sections can be connected to a power source in some way, then it is determined that no island has been formed.

[0039] Furthermore, it also includes the step of: returning to the step of reading the operating parameters of the distribution network and the configuration data of the measuring equipment in the reading area according to the set time interval, and continuing to perform the next round of fault location and detection to ensure real-time perception and fault handling of the distribution network status.

[0040] A distribution network fault section location and detection system based on a hybrid Transformer and particle swarm optimization method includes a microprocessor and a memory interconnected thereto, wherein the microprocessor is programmed or configured to execute the aforementioned distribution network fault section location and detection method based on a hybrid Transformer and particle swarm optimization method.

[0041] A computer-readable storage medium storing a computer program for being programmed or configured by a microprocessor to execute the aforementioned method for locating and detecting fault sections in a power distribution network based on a hybrid Transformer and particle swarm optimization approach.

[0042] Compared with the prior art, the present invention has the following main advantages:

[0043] (1) Traditional fault location methods usually require manual construction of objective functions, which is time-consuming and complex. This invention is based on a hybrid algorithm of Transformer and Particle Swarm Optimization, which can automatically construct objective functions suitable for power distribution networks, greatly reducing the preliminary preparation work and improving the applicability and flexibility of the algorithm.

[0044] (2) By using the Transformer model to process distribution network data, this invention can effectively recover missing power flow parameter data, enhance the robustness and integrity of the data, and significantly improve the accuracy of fault section location. The Transformer model analyzes historical data based on a self-attention mechanism and is retrained every 72 hours, ensuring the model's high tracking ability of the system state.

[0045] (3) This invention can not only quickly locate faulty sections, but also reduce the number of objective function variables by constructing a dedicated objective function for islands, thereby improving the detection accuracy of island states and avoiding the algorithm from getting stuck in local optima or making calculation errors.

[0046] (4) Traditional fault location algorithms often rely on simple objective functions, while this invention significantly improves the algorithm's ability to interpret complex power grid scenarios by combining the predictive power of the Transformer model with the optimization power of the particle swarm optimization algorithm. At the same time, this invention can still maintain high robustness in the case of signal loss or distortion, and adapt to the complex distribution network environment with a high proportion of inverters.

[0047] (5) This invention combines neural networks and intelligent optimization algorithms, which can perceive the operating status of the distribution network in real time and solve the fault section quickly through particle swarm algorithm, greatly improving the calculation efficiency and making it suitable for online monitoring and fault location of actual distribution network systems. Attached Figure Description

[0048] Figure 1This is a flowchart of the distribution network fault section location and detection method based on the hybrid Transformer and particle swarm optimization of the present invention.

[0049] Figure 2 This is a schematic diagram of the improved IEEE 33-node model. Detailed Implementation

[0050] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0051] Please see Figure 1 This invention provides a method for locating and detecting fault sections in a power distribution network based on a hybrid Transformer and Particle Swarm Optimization approach, comprising the following steps:

[0052] Step (1): Construct and initialize the upstream segment matrix at the current time. Downstream segment matrix The upstream power source matrix and the downstream power source matrix; these matrices provide the necessary preparation for the subsequent construction of the objective function.

[0053] Step (2): Read the operating parameters of the regional distribution network and the configuration data of the measuring equipment: The operating parameters of the regional distribution network include the load, equipment parameters and the setting values ​​of the feeder terminal unit (FTU) in the distribution network. The configuration data of the measuring equipment includes the installation location of the measuring equipment, the measurement type and the measurement data of the measuring equipment (e.g., power flow detection data).

[0054] Step (3), data preparation and model training: The operating parameters of the power distribution network and the configuration data of the measuring equipment are divided into training set, validation set and test set as the dataset, and the Transformer model is trained.

[0055] In this invention, the training of the Transformer model is specifically as follows: To effectively complete missing node data in power systems, this invention introduces a random masking mechanism based on the original Transformer architecture. During the training phase, the input node feature data is randomly masked, i.e., some nodes are randomly selected according to a preset probability, and the feature vectors of these nodes are replaced with special mask markers (such as zero vectors or specific identifiers). This operation simulates various node data missing situations that may occur in real-world applications, allowing the model to face diverse missing patterns during training. The Transformer model utilizes its self-attention mechanism to capture the correlation between known nodes and missing nodes globally. When the model receives input data containing a mask, the self-attention mechanism aggregates the information of known nodes and calculates the most likely feature estimate for the missing node. By continuously training the model to predict the features of masked nodes under different masking conditions, the model gradually learns and masters the ability to infer global data based on partially visible data, ultimately achieving accurate completion of node data missing in power systems caused by sensor failures, communication interruptions, etc., improving the integrity and reliability of system data. The model takes missing data with a random mask as input and outputs the complete original data.

[0056] Step (4), Fault Information Extraction and Missing Data Prediction: Extract the current fault information from the feeder terminal unit (FTU) and evaluate the completeness of the data in the fault information. If the data is missing, call the trained Transformer model and perform step-by-step prediction on the missing data based on historical data to gradually fill in the data.

[0057] Traditional prediction methods often have limitations when dealing with data with complex temporal relationships, particularly in data completion tasks. The Transformer architecture of this invention, with its self-attention mechanism, effectively captures inter-element dependencies when processing long sequences of data, achieving significant results in fields such as natural language processing. Applying it to single-step prediction tasks can achieve more accurate data predictions.

[0058] Step (5), Objective function construction and fault location: Based on the matrix constructed in step (1) and the data obtained after the Tranformer model in step (4) completes the missing data, construct the overall objective function and the island detection-specific objective function. The construction methods of the overall and specific objective functions are completely consistent. The overall objective function is constructed for the entire power grid, and the island detection-specific objective function is constructed for the nodes in the identified islands using the same method.

[0059] The objective function, used to describe the error between the actual switch state value uploaded by the FTU and the expected value, is shown in the following formula:

[0060]

[0061] in the formula It is the deviation between the actual location of the fault in the distribution network and the predicted value determined by the switch state function. These are predicted values; the function is a multivalued function. This represents the total number of feeder switches in the distribution network.

[0062] To prevent misjudgment or omission of fault location in the distribution network, the concept of a minimum set can be used to improve the objective function. The improved objective function is shown in the following formula.

[0063]

[0064] in, This represents the total number of feeder sections in the distribution network. This represents the total number of faulty sections. To prevent misjudgment of coefficients, the concept of a minimum set is used. exist The value is chosen to prevent missed or incorrect judgments; in this embodiment, 0.5 is used.

[0065] Both a global objective function and a dedicated islanding detection objective function are used to locate faulty sections and determine the existence of new islands or island recovery phenomena. The construction methods for the global and islanding detection objective functions are consistent. The global objective function is constructed for the entire power grid system, while the islanding detection objective function is used after the first round of power grid system operation to determine if islands exist. If islands exist, a separate objective function is constructed for each island, serving a redundancy purpose. The binary particle swarm optimization algorithm is used to solve the objective function, which can accurately locate the faulty section and further determine whether new islands or island recovery phenomena exist.

[0066] Binary Particle Swarm Optimization (BPSO) is an optimization algorithm developed from the standard Particle Swarm Optimization (PSO) algorithm, primarily used to solve discrete optimization problems. Its core idea is to find the optimal solution by simulating the foraging behavior of a flock of birds. In each iteration, particles update their position and velocity based on their own experience and the experience of the group.

[0067] Regarding the selection of hyperparameters, this embodiment uses: , , ,

[0068] The binary particle swarm optimization algorithm described above can effectively solve the objective function in the graph, enabling accurate estimation of the fault location in the distribution network.

[0069] Step (6) Fault Analysis and Islanding Detection: Based on the calculation results of the objective function, analyze the faulty section and determine whether the islanding status in the distribution network has changed. If the formation or recovery of an island is detected, record the relevant data. If an island is detected, record the generated section and provide island location support for the island-specific objective function in step (6). If the fault has been recovered, delete the island-specific objective function in the next round of calculation to facilitate the establishment of a specific objective function for each island in step (5).

[0070] The objective function, after being solved using the particle swarm optimization algorithm, yields the fault status of each branch. The fault status can be used to locate the faulty branch, and the following algorithm can be used to determine whether an island has been formed.

[0071] (a) Determine the area affected by the fault

[0072] According to the location of the fault Determine the upstream segment associated with the location of the fault. and downstream section ;

[0073] Find the segment index corresponding to the fault location from the upstream segment matrix and the downstream segment matrix;

[0074] Based on the segment index, the set of upstream segments affected by the fault is set as follows. The downstream segment set is .

[0075] (II) Power Connectivity Analysis

[0076] For the upstream segment set :

[0077] Check upstream power matrix related elements Determine the switching status of the upstream power supply connected to each upstream segment;

[0078] Based on the logical operations in the matrix formula, calculate whether each upstream segment can be connected by power supply;

[0079] For upstream sections that cannot be connected by power, they are marked as part of a potential islanding area;

[0080] For the downstream segment set :

[0081] Check downstream power matrix related elements Determine the switching status of downstream power supplies connected to each downstream segment;

[0082] Based on the logical operations in the matrix formula, calculate whether each downstream segment can be connected by power.

[0083] Downstream sections that cannot be connected by power are marked as part of a potential islanded area.

[0084] (III) Island Determination

[0085] Taking into account the power connection situation of both upstream and downstream sections:

[0086] If there is a segment that belongs to both an upstream potential island area and a downstream potential island area, or if there is an isolated segment that cannot be connected to power in either the upstream or downstream, then an island is determined to have been formed.

[0087] If all affected sections can be connected to a power source in some way, then it is determined that no island has been formed.

[0088] Step (7), Polling and Repeated Detection: The system returns to step (2) at the set time interval to continue the next round of fault location and detection, ensuring real-time perception and fault handling of the power distribution network status.

[0089] This invention uses an improved IEEE 33-node model as an example. Based on the IEEE 33-node model, four distributed power sources are added after nodes 23, 24, 18, and 33, respectively. Figure 2 The symbol is DG, where G represents an infinite power source. Nodes are numbered 1-37, and the segment number to the left of each node is the node number minus 1.

[0090] In step (1), taking the improved IEEE 33-node system as an example, the system matrix is ​​read as follows:

[0091] Branch matrix:

[0092] (1-1);

[0093] The row vectors in the branch matrix B represent a set of connected nodes.

[0094] Power supply matrix:

[0095]

[0096] Each column in the power supply matrix S represents a power source.

[0097] The methods for constructing the upstream segment matrix, downstream segment matrix, upstream power source matrix, and downstream power source matrix specifically include:

[0098] (1-2) Determine the number of branches and power sources, and initialize the upstream segment matrix, downstream segment matrix, upstream power source matrix, and downstream power source matrix.

[0099] (1-3) Traverse each branch, temporarily remove the current path for each branch and analyze connectivity. Find each connecting point, determine upstream and downstream nodes, and construct the upstream segment matrix, downstream segment matrix, upstream power source matrix, and downstream power source matrix;

[0100] (1-4) The generation principles of upstream and downstream nodes are the same. Taking the upstream branch matrix as an example, the generation method is introduced. The upstream matrix generation process is as follows:

[0101] 1. Create a zero matrix Au with dimension 1. , where n1 is the number of branches.

[0102] 2. Iterate through each branch, temporarily removing the branch currently being iterated from the network to analyze the connectivity of the remaining network.

[0103] 3. Construct the construction diagram of the remaining branches.

[0104] 4. Locate the connected components in the graph.

[0105] 5. Determine the upstream connected components and find all nodes belonging to those components.

[0106] 6. Mark the positions of these nodes in the Au matrix; these nodes are the upstream nodes.

[0107] Since the matrix is ​​too large, only the top-left submatrix is ​​shown. The content is shown in the matrix below. Taking node 4 as an example, if columns 1, 2, and 3 in the 4th row of the matrix are 1, then these three branches are the upstream branches of node 4.

[0108]

[0109] The generation principles of upstream and downstream power supply matrices are similar. Taking the upstream power supply matrix as an example, the generation method is introduced as follows:

[0110] 1. Create a zero matrix Ku with dimension 1. . For the number of branches, This represents the number of power supplies.

[0111] 2. For each power source, use matrix Au to detect the connected component in which it is located.

[0112] 3. If the power supply has a heavy upstream continuity component, then set the corresponding position of Ku to 1.

[0113] Due to the large size of the matrix, only the 9-18 node matrix is ​​displayed.

[0114]

[0115] This invention uses an improved IEEE 33-node model as an example. Based on the IEEE 33-node model, four distributed power sources are added after nodes 23, 24, 18, and 33, respectively. Figure 2 The symbol is DG, where G represents an infinite power source. Nodes are numbered 1-37, and the segment number to the left of each node is the node number minus 1.

[0116] Step (2) Reading Distribution Network Parameters and Measurement Configuration: First, by reading the regional parameters and topology of the distribution network, relevant information of the Supervisory Control and Data Acquisition (SCADA) configuration is obtained, including the load, equipment parameters, and feeder terminal units (FTU) settings in the distribution network. Simultaneously, the installation location, measurement type, and measurement data of the measurement equipment are read to provide data support for subsequent steps. At the same time, an association matrix between distribution network nodes is established.

[0117] Step (3) specifically includes:

[0118] Read the voltage and power data collected by each node at the current moment, and use the data from the previous 72 hours to construct a single-step prediction dataset. Perform classification training on the n-1, n-2, n-3, and n-4 cases in the collected signals.

[0119] The model is retrained every 24 hours, and each model is retrained again after 72 hours of training.

[0120] The Transformer model handles the linear and nonlinear components of measurements through a self-attention mechanism and optimizes the state estimation process by incorporating historical state information. The self-attention formula in the Transformer model is expressed as follows:

[0121]

[0122] Mode In this matrix, Q represents the query matrix; K represents the key matrix; and V represents the value matrix. is the dimension of the key vector, used to scale the dot product to prevent the dot product from being too large and causing the gradient of the softmax function to be too small; softmax is the activation function used to calculate the weights.

[0123] Step (4): Determine whether the fault information matrix is ​​complete, and select the corresponding trained model to perform single-step prediction of the missing parameters.

[0124] The data generated by the FTU is judged. If there is data loss, the Transformer model trained in step (4) is called to perform single-step prediction. Assuming that the model is missing 4 sets of FTU data (the tuning judgment and power flow information are all missing), then n-4 prediction is used. In the subsequent prediction, it becomes n-3 model. After n-3 prediction, n-2 model is obtained. By recursively, complete data information can be obtained.

[0125] Then, based on the parameters selected during the setting, the actual value is calculated and compared with the set value to obtain the simulated FTU fault signal and the original state quantity, which is then substituted into step (6) for calculation.

[0126] If the matrix is ​​determined to be intact, skip this step.

[0127] Step (5): Construct the overall objective function and the objective function of the isolated parts.

[0128] Using the island locations obtained in the previous round, a separate objective function for island detection is constructed to improve positioning accuracy.

[0129] feeder switch status variables And when it is 1, it indicates that the current flows through the feeder switch. The short-circuit current flows from the system power supply to the end of the line; when it is -1, it indicates that it flows through the feeder switch. The short-circuit overcurrent flows from the end of the line to the system power supply side. For a feeder section, its state variables are {0, 1}. The feeder switch state function represents the state of the distribution network's missing section. The function reflects the distribution network section The running status.

[0130] Using the matrix constructed from upstream and downstream components, the switching state function of the upstream region containing the system power supply in the system is: The downstream section switch state function is Construct the following objective function:

[0131]

[0132] in

[0133] in It is the deviation between the actual location of the fault in the distribution network and the predicted value determined by the switch state function. It is a predicted value; This indicates the total number of feeder switches in the distribution network; This represents the total number of feeder sections in the distribution network. For the feeder section, the state variables are {0, 1}; This represents the total number of faulty sections. To prevent misjudgment of coefficients, the concept of a minimum set is used. exist Values; This is the status variable of the feeder switch. When it is 1, it indicates that the short-circuit current flowing through the feeder switch flows from the system power supply to the end of the line. When it is -1, it indicates that the short-circuit overcurrent flowing through the feeder switch flows from the end of the line to the system power supply side.

[0134] Finally, the binary particle swarm optimization method is used to evaluate the objective function ( The fault location can be obtained by performing calculations. The example matrix below shows the results of simulating FTU signals and particle swarm optimization.

[0135] Simulated FTU signal matrix (1-36 node one-dimensional vector):

[0136] The output is:

[0137] This indicates that faults 1, 6, and 14 are in line with expectations.

[0138] If a fault point is added or removed, determine whether the node has become an island or whether the island has been recovered. Save the data for use in step (5) of the next round of calculation.

[0139] Step (6) Fault Analysis and Islanding Detection: Based on the calculation results of the objective function, analyze the faulty section and determine whether the islanding status in the distribution network has changed. If the formation or recovery of an island is detected, record the relevant data to provide a basis for subsequent calculations. If an island is detected, record the generated section and provide island location support for the island-specific objective function in Step 5, so that Step 5 can establish a specific objective function for each island.

[0140] Step (7) Polling and Repeated Detection: The system returns to step (2) at the set time interval to continue the next round of fault location and detection, ensuring real-time perception and fault handling of the power distribution network status.

[0141] Furthermore, embodiments of the present invention also provide a distribution network fault section location and detection system based on a hybrid Transformer and particle swarm optimization, comprising a microprocessor and a memory interconnected thereto, wherein the microprocessor is programmed or configured to execute the distribution network fault section location and detection method based on a hybrid Transformer and particle swarm optimization.

[0142] Furthermore, embodiments of the present invention also provide a computer-readable storage medium storing a computer program, the computer program being programmed or configured by a microprocessor to execute the distribution network fault section location and detection method based on a hybrid Transformer and particle swarm optimization.

[0143] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for locating and detecting fault sections in a distribution network based on a hybrid Transformer and Particle Swarm Optimization approach, characterized in that, Includes the following steps: Construct and initialize the upstream segment matrix, downstream segment matrix, upstream power source matrix, and downstream power source matrix at the current moment; Read the operating parameters of the regional power distribution network and the configuration data of the measuring equipment; The read operating parameters of the power distribution network and the configuration data of the measuring equipment are divided into training set, validation set and test set as datasets for training the Transformer model; Extract current fault information from the feeder terminal unit (FTU) and assess the completeness of the data in the fault information. If data is missing, call the trained Transformer model and perform step-by-step prediction on the missing data based on historical data to gradually fill in the missing data. Based on the constructed matrix and the data obtained after completing the missing data using the Tranformer model, an overall objective function and a special objective function for island detection are constructed. The objective function is solved using the binary particle swarm optimization algorithm to locate the fault segment. Based on the location of the faulty section, the faulty section is analyzed, and it is determined whether the islanding status in the distribution network has changed. If an island is detected, the generated section is recorded, and island location support is provided for the island-specific objective function. If the fault has been recovered, the island-specific objective function is deleted in the next round of calculation. The training steps of the Transformer model include: performing random masking on the input node feature data to simulate various node data missing situations that may occur in real applications, so that the model faces diverse missing patterns during training. The Transformer model uses a self-attention mechanism to capture the correlation between known nodes and missing nodes globally. When the model receives input data containing a mask, the self-attention mechanism aggregates the information of known nodes and calculates the most likely feature estimate for the missing nodes. The input of the trained Transformer model is missing data with a random mask, and the output is complete original data. The overall objective function is constructed for the entire power grid system, while the island detection objective function is used to determine whether islands exist after the first round of operation of the power grid system. If islands exist, the objective function is applied separately to each island. The overall objective function and the island detection-specific objective function are constructed using the same method, as shown in the following formula: ； in It is the deviation between the actual location of the fault in the distribution network and the predicted value determined by the switch state function. It is a predicted value; This indicates the total number of feeder switches in the distribution network; This represents the total number of feeder sections in the distribution network. For the feeder section, the state variables are {0, 1}; This represents the total number of faulty sections. To prevent misjudgment of coefficients, the concept of a minimum set is used. exist Values; This is the status variable of the feeder switch. When it is 1, it indicates that the short-circuit current flowing through the feeder switch flows from the system power supply to the end of the line. When it is -1, it indicates that the short-circuit overcurrent flowing through the feeder switch flows from the end of the line to the system power supply side.

2. The method for locating and detecting fault sections in a distribution network based on a hybrid Transformer and Particle Swarm Optimization approach as described in claim 1, characterized in that: The operating parameters of the regional distribution network include the load, equipment parameters, and feeder terminal unit (FTU) settings in the distribution network. The configuration data of the measuring equipment includes the installation location, measurement type, and measurement data of the measuring equipment.

3. The method for locating and detecting fault sections in a distribution network based on a hybrid Transformer and Particle Swarm Optimization approach as described in claim 1, characterized in that: The determination of whether an island exists specifically includes: (a) Determine the area affected by the fault According to the location of the fault Determine the upstream segment associated with the location of the fault. and downstream section ; Find the segment index corresponding to the fault location from the upstream segment matrix and the downstream segment matrix; Based on the segment index, the set of upstream segments affected by the fault is set as follows. The downstream segment set is ; (II) Power Connectivity Analysis For the upstream segment set : Check upstream power matrix related elements Determine the switching status of the upstream power supply connected to each upstream segment; Based on the logical operations in the matrix formula, calculate whether each upstream segment can be connected by power supply; For upstream sections that cannot be connected by power, they are marked as part of a potential islanding area; For the downstream segment set : Check downstream power matrix related elements Determine the switching status of downstream power supplies connected to each downstream segment; Based on the logical operations in the matrix formula, calculate whether each downstream segment can be connected by power. For downstream sections that cannot be connected by power, they are marked as part of a potential islanding area; (III) Island Determination Taking into account the power connection situation of both upstream and downstream sections: If there is a segment that belongs to both the upstream and downstream potential island areas, or if there is an isolated segment that cannot be connected to power in either the upstream or downstream, then an island is determined to have been formed. If all affected sections can be connected to a power source in some way, then it is determined that no island has been formed.

4. The method for locating and detecting fault sections in a distribution network based on a hybrid Transformer and Particle Swarm Optimization approach as described in claim 1, characterized in that: It also includes the step of returning to the step of reading the operating parameters of the distribution network and the configuration data of the measuring equipment in the reading area according to the set time interval, and continuing to carry out the next round of fault location and detection to ensure real-time perception and fault handling of the distribution network status.

5. A distribution network fault section location and detection system based on a hybrid Transformer and particle swarm optimization, comprising a microprocessor and a memory interconnected thereto, wherein the microprocessor is programmed or configured to execute the distribution network fault section location and detection method based on a hybrid Transformer and particle swarm optimization as described in any one of claims 1-4.

6. A computer-readable storage medium storing a computer program for being programmed or configured by a microprocessor to perform the distribution network fault location and detection method based on a hybrid Transformer and particle swarm optimization method as described in any one of claims 1-4.

Images