A distribution network fault identification and location system based on wide-area measurement technology

Abstract

The present invention relates to the technical field of power system fault diagnosis, and discloses a distribution network fault identification and positioning system based on wide-area measurement technology, comprising: a key measurement point optimization configuration module, which determines the optimal measurement point layout position and identifies high-risk weak links in the distribution network; a virtual-real fusion measurement network construction module, which uses a generative adversarial network to generate blind area virtual measurement data that conforms to physical laws and integrates actual measurement data with virtual measurement data; an active fault excitation implementation module, which implements safe and controllable small disturbance injection and actively detects the response characteristics of weak links; a state estimation enhanced fault location module, which constructs a health status assessment model and calculates the deviation of the transfer function from the health baseline; and a resilient fault location execution module, which extracts key information dimensions and ensures the reliability of fault location. The present invention uses virtual-real fusion measurement technology and, on the basis of limited actual measurement equipment, realizes efficient monitoring of the entire distribution network.

Description

Technical Field

[0001] The present invention relates to the technical field of power system fault diagnosis, and more particularly to a distribution network fault identification and positioning system based on wide-area measurement technology. Background Art

[0002] With the continuous expansion of distribution network scale and the increasing complexity of network topology, traditional fault identification and location technologies can no longer meet the requirements of reliable operation of modern distribution networks. Existing technologies mainly rely on real-time monitoring or fault recording data for fault identification and location, but in practical application, they face three major technical difficulties:

[0003] First, although key weak links in the distribution network (such as important interconnecting lines and substation equipment) are high-prone to faults, economic constraints make it impossible to achieve full network measurement coverage, resulting in insufficient monitoring data. The measurement coverage of traditional distribution networks is usually only 30% to 40%, and a large number of nodes lack real-time monitoring, making it difficult to capture characteristic information at the initial stage of a fault. There are inevitably measurement blind spots in the distribution network, and the fault status in these areas is difficult to directly observe, which increases the uncertainty of fault location. The existence of measurement blind spots can cause the fault location error to increase by 3 to 5 times, especially for underground cables and complex branch lines, where the positioning error may reach several kilometers. The measurement equipment itself may fail due to faults. In the case of missing measurement data, the accuracy and reliability of traditional fault location methods will be reduced. In the case of partial failure of the measurement equipment (about 30%), the fault location accuracy of traditional methods will drop below 50%, seriously affecting the fault recovery speed of the distribution network.

[0004] The above problems limit the application effect of existing technologies in complex distribution networks. There is an urgent need for a fault identification and location method that can improve fault warning capabilities under limited measurement resources and maintain high accuracy even when some measurement equipment fails. Summary of the Invention

[0005] The present invention provides a distribution network fault identification and positioning system based on wide-area measurement technology, which solves the technical problems in related technologies such as insufficient distribution network measurement coverage, blind spots, easy equipment failure, and low fault positioning accuracy.

[0006] The present invention provides a distribution network fault identification and location system based on wide-area measurement technology, comprising:

[0007] Key measurement point optimization configuration module, which uses a sensitivity assessment algorithm to determine the optimal measurement point layout and identify high-risk weak links in the distribution network that require key monitoring;

[0008] The virtual-real fusion measurement network construction module uses a generative adversarial network to generate virtual measurement data in blind areas that conform to physical laws based on the optimal measurement point layout. It then fuses the actual measurement data with the virtual measurement data using a dynamic weighting method to form a hybrid measurement matrix covering the entire network.

[0009] Active fault stimulation implementation module implements safe and controllable small disturbance injection for high-risk weak links, actively detects the response characteristics of weak links, and obtains equipment health status information;

[0010] The state estimation enhanced fault location module builds a health status assessment model based on the equipment health status information and calculates the deviation of the transfer function from the health baseline;

[0011] The resilient fault location execution module, based on the hybrid measurement matrix, extracts key information dimensions to ensure the reliability of fault location when there is uncertainty in the measurement data or failure of some measurement points.

[0012] Furthermore, determining the optimal measurement point arrangement position by using a sensitivity evaluation algorithm includes:

[0013] Apply a multi-dimensional weak link identification algorithm to assess the risk level of each node in the distribution network and calculate a risk score for each node. The risk score comprehensively considers topological vulnerability, historical fault frequency, electrical stress, and node importance.

[0014] Analyze the contribution of each candidate measurement point to network observability, construct the observation equation and calculate the observation sensitivity matrix of each candidate measurement point to obtain the measurement point sensitivity index;

[0015] Based on the risk score and sensitivity index, a measurement point configuration optimization model is constructed, and the optimization model is solved to determine the final measurement point configuration plan.

[0016] Furthermore, the method of using a generative adversarial network to generate virtual blind spot measurement data that conforms to physical laws includes:

[0017] Using electrical topology and power flow constraints, a physical correlation model between measurement points and blind zone nodes is constructed;

[0018] The generator and discriminator of the generative adversarial network are trained. The generator receives random noise and electrical quantity information of surrounding measured points as input to generate virtual measurement data, and the discriminator determines the authenticity of the data.

[0019] Physical constraint regularization terms are introduced into the generator's loss function, including Kirchhoff's voltage law constraint term, Kirchhoff's current law constraint term and power flow equation constraint term.

[0020] Furthermore, the implementation of safe and controllable micro-disturbance injection for high-risk weak links includes:

[0021] Construct a perturbation signal for active fault excitation. The perturbation signal is in the form of a decaying sine wave, and the amplitude is set within a safe range that does not affect the normal operation of the system.

[0022] Through the power electronic interface equipment, the perturbation signal is safely injected into the key weak links of the distribution network, and the injection process is controlled by a closed-loop control method;

[0023] Monitor key system parameters in real time, set warning thresholds and emergency thresholds, and trigger corresponding protection actions when the monitored parameters exceed the thresholds.

[0024] Furthermore, constructing a health status assessment model based on the device health status information includes:

[0025] Record the device's response to the perturbation signal and obtain the system response transfer function;

[0026] Calculating the deviation of the current transfer function from the healthy baseline transfer function;

[0027] The health status score is evaluated by multi-feature fusion. When the health status score is lower than the preset threshold, the device is judged to have a potential failure risk and enters the warning state;

[0028] The time domain and frequency domain feature analysis methods are used to identify the fault precursor features from the response signal and construct the fault precursor feature vector.

[0029] Furthermore, the step of extracting key information dimensions when there is uncertainty in the measurement data or some measurement points fail includes:

[0030] Extract the most valuable fault features from the mixed measurement data, perform principal component analysis on the original high-dimensional features, and obtain the feature matrix after dimensionality reduction;

[0031] The contribution of each principal component to the classification of fault types was evaluated using the Fisher discriminant criterion;

[0032] According to the contribution value, several principal components with the greatest discriminative power are selected to form a set of key information dimensions;

[0033] Combine the credibility weights of multi-source data to evaluate the reliability of each measurement data. Calculate the credibility of the actual measurement data based on the device status and measurement residuals. For virtual measurement data, directly use its generation reliability as the credibility.

[0034] Furthermore, ensuring the reliability of fault location includes:

[0035] Based on the extracted key information dimensions and data credibility, a resilient fault location algorithm is constructed;

[0036] The resilient fault location algorithm uses a weighted voting fusion method to identify the fault type. The credibility weight of each piece of evidence is multiplied by the probability estimate of the fault type based on the evidence and then the sum is calculated.

[0037] Construct a probability distribution map of the fault location, multiply the credibility weight of each piece of evidence by the location probability estimate based on the evidence, and then divide the sum by the sum of the credibility weights;

[0038] Output fault type, fault location and corresponding credibility to support operation and maintenance personnel in decision-making.

[0039] Furthermore, the resilient fault location algorithm includes an evidence collection layer, an evidence preprocessing layer, a local inference layer, and an evidence weighted fusion layer, specifically including:

[0040] The evidence collection layer processes five types of evidence sources: actual measurement data, virtual measurement data, active stimulation response characteristics, historical failure modes, and topology constraint information;

[0041] The evidence preprocessing layer applies specific preprocessing methods to each type of evidence;

[0042] The local inference layer performs independent analysis on each type of pre-processed evidence to generate preliminary fault type and location estimates;

[0043] The evidence weighted fusion layer combines the credibility of each evidence source and uses the Bayesian network framework to fuse the local inference results to form the final fault judgment result.

[0044] Furthermore, the micro-disturbance injection adopts a safe injection control system, which includes:

[0045] Based on a hybrid control architecture of DSP and FPGA, DSP is responsible for upper-level control strategy and safety monitoring, while FPGA realizes high-speed waveform generation and real-time control;

[0046] Closed-loop control parameter adaptive adjustment unit automatically optimizes controller parameters according to system impedance characteristics;

[0047] Multi-layer safety protection unit monitors key parameters, including voltage, current, frequency, and harmonic content, and triggers corresponding protection actions when the monitored key parameters exceed the threshold.

[0048] The present invention provides a storage medium comprising a memory and one or more processors, wherein the memory stores executable code, and when the one or more processors execute the executable code, they are used to implement the above-mentioned distribution network fault identification and positioning system based on wide-area measurement technology.

[0049] The beneficial effects of the present invention are: through virtual-reality fusion measurement technology, on the basis of limited actual measurement equipment, efficient monitoring of the entire distribution network is achieved, especially for key weak links and measurement blind spots. The measurement coverage is improved compared with traditional methods, reducing the risk of missed faults caused by monitoring blind spots.

[0050] Detect potential faults in advance: Based on the active fault excitation technology of this application, it is possible to detect abnormal characteristics of equipment before a fault occurs, and the accuracy of early warning is improved. By identifying fault precursors in advance, power outages are avoided, power supply reliability is improved, and power outage losses to users are reduced.

[0051] By integrating optimized measurement point configuration and virtual measurement data, the fault location accuracy is improved at the same measurement point coverage. Even when some measurement equipment fails, a high fault location accuracy can still be maintained, reducing the time and scope of fault inspections.

[0052] Without the need to add a large number of physical measurement equipment, system performance is improved through software algorithms and perturbation technology, reducing the monitoring cost of each distribution network node and improving the input-output ratio;

[0053] In the event of measurement equipment failure, network topology changes, etc., the system of this application can still maintain efficient operation, the positioning algorithm calculation time is reduced, the system's resilience and real-time response capabilities are improved, and it can adapt to various complex operating environments. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] Figure 1 This is a module diagram of a distribution network fault identification and location system based on wide-area measurement technology in the present invention. DETAILED DESCRIPTION

[0055] The subject matter described herein will now be discussed with reference to example embodiments. It should be understood that these embodiments are discussed solely to enable those skilled in the art to better understand and implement the subject matter described herein, and that the functions and arrangements of the elements discussed may be varied without departing from the scope of this specification. Various examples may omit, substitute, or add various processes or components as needed. Furthermore, features described in some examples may be combined in other examples.

[0056] At least one embodiment of the present invention discloses a distribution network fault identification and location system based on wide area measurement technology, such as Figure 1 Shown, including:

[0057] Key measurement point optimization configuration module, which uses a sensitivity assessment algorithm to determine the optimal measurement point layout and identify high-risk weak links in the distribution network that require key monitoring;

[0058] The sensitivity evaluation algorithm is used to analyze the distribution network topology and historical operating data to determine the optimal measurement point layout, which includes the following sub-steps:

[0059] Step 1.1, multi-dimensional weak link identification;

[0060] The multi-dimensional weak link identification algorithm provided in this application evaluates the risk level of each node in the distribution network, and calculates each node according to the following risk scoring formula Risk score:

[0061] ;

[0062] in Representation node The comprehensive risk score of For nodes The topological vulnerability index is obtained by analyzing the electrical topology map; It is a historical fault frequency indicator, based on statistical analysis of historical operating data; It is an electrical stress index, calculated by the degree to which electrical parameters deviate from rated values; is the node importance index, considering the importance of the node power supply load;

[0063] They are the weight coefficients of topological vulnerability index, historical fault frequency index, electrical stress index, and node importance index respectively.

[0064] Optionally, in different distribution network environments, the weight coefficients of each indicator can be adjusted according to actual conditions. For example, in urban core areas where power supply reliability requirements are high, the node importance index can be increased. Weight To 0.4-0.5; in areas where equipment is severely aged, the electrical stress index can be improved Weight To 0.3-0.4.

[0065] Topological vulnerability index It can be further broken down into several sub-indicators:

[0066] ;

[0067] in For nodes The topological vulnerability index of is the node degree index, which indicates the relationship between The number of directly connected nodes; is the node betweenness index, which means The number of shortest paths; is the node proximity index, indicating that the node The inverse of the average distance to all other nodes; is the node proximity index, indicating that the node The inverse of the average distance to all other nodes; They are the weight coefficients of node degree index, node betweenness index and node proximity index respectively.

[0068] After the risk score calculation is completed, the system sorts the scores from high to low and identifies the high-risk weak links in the distribution network that need to be monitored.

[0069] Step 1.2, measurement point sensitivity evaluation;

[0070] Using the measurement point sensitivity evaluation algorithm, the contribution of each candidate measurement point to the network observability is analyzed. First, the observation equation is constructed:

[0071] ;

[0072] in is the measurement vector, which contains the measurement values of all measurement points; is a state variable Nonlinear function of is the measurement error vector, which represents the random error of the measurement value; is the system state variable.

[0073] Then calculate each candidate measurement point The observation sensitivity matrix :

[0074] ;

[0075] in is the observation sensitivity matrix, which indicates the sensitivity of the measured value to the state variable; is the partial derivative of the measurement function with respect to the state variable; is the initial estimate of the state variables or the operating point;

[0076] means that the partial derivatives are calculated when the state variables are equal to the initial estimates, is the system state variable;

[0077] is a state variable nonlinear function.

[0078] Through analysis Eigenvalues and condition numbers of , evaluate adding measurement points Improve the observability of the system and obtain the sensitivity index of the measurement point .

[0079] Step 1.3, optimizing the arrangement of measurement points;

[0080] Based on risk scores and sensitivity indicators, a measurement point configuration optimization model is constructed:

[0081] ;

[0082] Budget constraints:

[0083] ;

[0084] High-risk node coverage constraints:

[0085] ;

[0086] in represents the set of locations where measurement equipment can be installed; Indicates whether to install the measuring equipment at the candidate location. It is a 0-1 decision variable, 1 means at the location Install the measuring device, 0 means not installed; For measuring points Sensitivity index; represents the maximization objective function; Indicates the location the cost of installing the measurement equipment; Indicates the maximum budget available for installing measurement equipment; Indicates that the node can be monitored All measurement points of the state; A collection of high-risk nodes; stands for "for all", meaning that the condition is satisfied for every element in the set; Represents the summation operator.

[0087] By solving this optimization model, the final measurement point configuration scheme is determined to achieve optimal coverage of key weak links under a limited budget.

[0088] The virtual-real fusion measurement network construction module uses a generative adversarial network to generate virtual measurement data in blind areas that conform to physical laws based on the optimal measurement point layout. It then fuses the actual measurement data with the virtual measurement data using a dynamic weighting method to form a hybrid measurement matrix covering the entire network.

[0089] The following sub-steps are included:

[0090] Step 2.1, physical correlation model construction;

[0091] Using electrical topology and power flow constraints, a physical association model between measurement points and blind zone nodes is constructed. Based on Kirchhoff's laws for node voltage and branch current, the relationship between node state variables and surrounding measurement quantities is established:

[0092] ;

[0093] ;

[0094] in For nodes The branch current between nodes Flow Node Current; is the corresponding admittance, indicating the node and Admittance value between ; and Represents nodes respectively and nodes The complex voltage of and Represents nodes respectively Injected active and reactive power; Representation node complex conjugate of voltage; Representation node and The complex conjugate of the admittance between; For the node The number of connected nodes; is an imaginary unit, satisfying Used to indicate plural; Represents the summation operator.

[0095] Based on the typical characteristics of the distribution network, a simplified DC power flow model can be used to establish the physical association:

[0096] ;

[0097] in Represents a slave node Flow Node Active power; Represents nodes respectively and nodes The voltage amplitude; Indicates the reference voltage value of the system; For nodes and The reactance value between Represents nodes respectively and nodes The voltage phase angle; represents the sine function; Indicates approximately equal to.

[0098] Optionally, for low voltage distribution network or line impedance ratio For larger cases, the improved Distflow equation can be used:

[0099] ;

[0100] ;

[0101] in and Node Active power, reactive power and voltage amplitude; Node Active power, reactive power and voltage amplitude; Branch The resistance and reactance of the connection node and nodes Line parameters; and For nodes Active and reactive loads; Representation node and nodes The square of the voltage amplitude.

[0102] For example, in a 10kV urban distribution network application scenario, the system needs to estimate the status of 28 blind-spot nodes based on 15 actual measurement points. For the trunk line area, a DC power flow model is used to establish physical correlations; for the terminal low-voltage branches, a more accurate model is constructed using the Distflow equation. This hybrid physical modeling approach provides a reliable physical basis for generating virtual measurement data under normal operating conditions.

[0103] Therefore, through these physical constraint equations, the electrical state of the blind zone nodes can be preliminarily estimated, providing a physical basis for the subsequent generation of virtual measurement data.

[0104] Step 2.2, virtual measurement data generation based on generative adversarial network;

[0105] The method provided in this application uses an improved generative adversarial network (GAN) algorithm to generate virtual blind spot measurement data that conforms to physical laws. The GAN algorithm includes a generator G and a discriminator D, and the optimization objective is:

[0106] ;

[0107] in represents the probability distribution of the actual measured values; represents the random noise distribution of the generator input; is the random noise input, which represents the random vector input to the generator; represents input noise Output of the generator map; is the probability of the discriminator judging the authenticity of the data; Represents the expectation operator; express Obey the distribution ; express Obey the distribution represents the natural logarithm function; represents the minimization operation on the generator G; Represents the maximization operation on the discriminator D; represents the value function of the generative adversarial network.

[0108] It should be noted that in order to ensure that the generated virtual measurement data conforms to the physical laws of the power grid, a physical constraint regularization term is introduced into the loss function of the generator G:

[0109] ;

[0110] in Represents the objective function that needs to be minimized during the generator training process; represents the ability of the generator to deceive the discriminator; The physical constraint loss is used to penalize the generated data that violates the physical laws of the power grid; To balance the weight coefficient of GAN loss and physical constraints, the importance of physical constraints in the total loss is controlled.

[0111] In this application's improved GAN model implementation, the generator G consists of a multi-layer neural network. The input layer receives random noise and electrical quantity information from surrounding measured points. The middle layer contains three fully connected hidden layers, each with 128, 256, and 128 neurons, respectively. The activation function uses LeakyReLU to enhance the model's ability to learn nonlinear features. The output layer uses the tanh activation function to ensure that the generated electrical quantities, such as voltage and current, are within a reasonable range.

[0112] The discriminator D also uses a multi-layer neural network structure. The input layer receives the measurement data (real or generated), contains two hidden layers, each with 128 neurons, uses the LeakyReLU activation function, and the output layer is a single neuron, using the sigmoid activation function to output the probability value of the data authenticity.

[0113] Physical constraint loss This is the innovative part of this application, and its specific calculation method is:

[0114] ;

[0115] in Loss of physical restraint; is the Kirchhoff voltage law constraint, which ensures that the closed loop voltage sum is zero and represents the degree to which the generated data violates the KVL law; is the Kirchhoff current law constraint, which ensures that the algebraic sum of the node current is zero and represents the degree to which the generated data violates the KCL law; is the power flow equation constraint, ensuring that the generated data satisfies the power flow equation and indicating the degree to which the generated data violates the power flow equation; They are and The weight coefficient of .

[0116] In a virtual measurement scenario involving a 10kV feeder blind spot in distribution network area B, the improved GAN of this application was first applied to train the model using historical data from the previous six months of normal operation. The trained model was then used to generate virtual measurement data for four distribution network nodes that lacked measurement points. Compared to traditional interpolation methods, the virtual data generated by this application's model more closely matches the physical characteristics of the power grid. In particular, under conditions of sudden load changes and small disturbances, the average relative error between the virtual measurement data and the true value was reduced, providing a more reliable data foundation for subsequent fault location.

[0117] Step 2.3, fusion of virtual and real measurement data;

[0118] Through the dynamic weighting method, the actual measurement data is fused with the virtual measurement data to form a hybrid measurement matrix covering the entire network:

[0119] ;

[0120] in is a hybrid measurement matrix covering the entire network, with the dimension of number of nodes × number of time points, where each row represents a node and each column represents a time point; is the actual measurement data matrix, with the dimension of the number of actual measurement points × the number of time points, containing the measurement data of all actual measurement points; is the virtual measurement data matrix with the dimension of number of virtual measurement points × number of time points, which contains the generated data of all virtual measurement points.

[0121] In addition, for each virtual measurement point Assign weights based on their generation reliability

[0122] ;

[0123] in Virtual measuring point The weight of the virtual measurement point The credibility of the data, the value range is usually [0, 1]; The distance to the nearest measured point, indicating electrical distance or topological distance, is used to evaluate the distance relationship between the virtual measurement point and the actual measurement point; The local network topology complexity reflects the complexity of the local network structure and is usually calculated by the number of node connections or network structure characteristics; is the consistency index between historical generated data and measured data, with a value range of [0, 1]; It is a weight calculation function that determines the weight of the virtual measurement point based on multiple factors.

[0124] It can be seen that the weight This will be used in subsequent fault location algorithms to guide the degree of trust in data from different sources during the decision-making process.

[0125] Active fault stimulation implementation module implements safe and controllable small disturbance injection for high-risk weak links, actively detects the response characteristics of weak links, and obtains equipment health status information;

[0126] Implement safe and controllable active fault stimulation for identified key weak links to detect potential fault characteristics in advance, including the following sub-steps:

[0127] Step 3.1, perturbation signal construction;

[0128] In the implementation scheme of this application, a perturbation signal for active fault excitation is constructed. The perturbation signal must meet two requirements: one is to effectively excite the potential fault characteristics, and the other is not to affect the normal operation of the system. This application adopts the form of attenuated sine wave:

[0129] ;

[0130] in is the perturbation signal, indicating that The disturbance signal injected into the system; is the perturbation signal amplitude, which represents the maximum amplitude of the perturbation signal; For the perturbation signal frequency, select a frequency band that can effectively stimulate the response characteristics of the device; is the phase of the perturbation signal, which controls the initial phase of the signal; is the perturbation signal attenuation coefficient, which controls the decay rate and duration of the signal amplitude; is the time variable; represents the sine function; It is twice the value of pi and is used to convert frequency into angular frequency; is an exponential decay term, where is the base of natural logarithms.

[0131] In addition, this application constructs a perturbation signal library for different types of equipment and potential failure modes:

[0132] ;

[0133] in Represents the first, second, and A perturbation signal, is the number of signals in the perturbation signal library, indicating the total number of preset perturbation signals of different types;

[0134] Each signal is best suited to trigger a specific type of potential fault.

[0135] The perturbation signal construction algorithm of this application is specifically implemented as follows: First, through historical fault data analysis, the precursor characteristic spectrum characteristics of different types of faults are identified, and the sensitive frequency band of each fault type is determined. Then, for each fault type, a perturbation signal with different parameters is constructed, where the amplitude Determined by binary search method, it is as large as possible without causing system abnormality, usually set to 0.5% to 3% of the rated value; frequency Select the sensitive frequency corresponding to the fault type; Phase Dynamically adjust according to the current operating status of the system to avoid resonance with the system's inherent oscillation frequency; attenuation coefficient Set it according to the signal duration requirement.

[0136] In the application scenario of potential fault detection of 35kV distribution line equipment in a 220kV substation, this application constructs perturbation signals for three potential faults: insulation aging, abnormal contact resistance, and partial discharge. For example, for insulation aging fault, select (0.8% of rated voltage), A perturbation signal was constructed and injected, successfully stimulating a slight change in the tangent value of the insulation medium loss angle, thus discovering an insulation degradation point in advance and avoiding possible equipment breakdown failure.

[0137] Step 3.2, security injection control;

[0138] It should be noted that the perturbation signal is safely injected into the key weak links of the distribution network through the power electronic interface device. The injection process adopts a closed-loop control method:

[0139] ;

[0140] in express The signal that is actually injected into the system at any moment; express The ideal perturbation signal that is expected to be injected at all times; and Represent the proportional control coefficient and the integral control coefficient respectively; represents the integral operator, which performs time integration on the error signal; represents the time differential element and the integral variable; is the error signal, that is The ideal perturbation signal expected to be injected at any time and The signal actually injected into the system at any moment The difference.

[0141] At the same time, this application monitors key system parameters in real time and sets multi-level safety thresholds:

[0142] Warning threshold: Adjust the injection amplitude;

[0143] Emergency Threshold: Stop the injection immediately.

[0144] in It is the key parameter of the system; The warning threshold is usually set at about ±3% of the normal value; The emergency threshold is triggered when the system parameters reach a dangerous level and is usually set to about ±5% of the normal value; Indicates the absolute value or deviation of the parameter, which is used to compare with the threshold.

[0145] The specific implementation of the security injection control algorithm includes three main parts:

[0146] First, the perturbation signal injection device adopts a hybrid control architecture based on DSP and FPGA. The DSP is responsible for upper-level control strategy and safety monitoring, while the FPGA implements high-speed waveform generation and real-time control. The sampling frequency reaches 20kHz, ensuring precise control of the perturbation signal.

[0147] The second is the closed-loop control parameter adaptive adjustment, the controller parameters and Automatically optimize according to system impedance characteristics, The range is usually 0.05-0.2, The range is 0.01-0.05, and it is dynamically adjusted according to the tracking error during the injection process. The third is a multi-layer safety protection mechanism that monitors 10 key parameters such as voltage, current, frequency, and harmonic content, and the warning threshold Set to ±3% of normal value, emergency threshold If it is set to ±5%, any parameter exceeding the threshold will trigger the corresponding protection action.

[0148] Step 3.3, response feature extraction;

[0149] Record the device's response to the perturbation signal and extract characteristic parameters to construct the health status fingerprint. First, obtain the system response:

[0150] ;

[0151] in is the transfer function of the measured object, which describes the relationship between the input and output of the system. It is a complex frequency domain variable, which is directly related to its health status; Indicates that the system is at time Response output to perturbation signal; Indicates time The disturbance signal injected into the system.

[0152] Then, the transfer function characteristic parameters are calculated using the small signal analysis method:

[0153] Amplitude-frequency response characteristics Indicates the gain of the system at different frequencies;

[0154] Phase-frequency response characteristics Indicates the phase change of the system at different frequencies;

[0155] Damping coefficient Indicates how quickly the system oscillation decays;

[0156] Natural frequency Indicates the inherent oscillation frequency of the system.

[0157] in is the imaginary unit, is the angular frequency, and Indicates frequency; The transfer function at frequency The amplitude at The transfer function at frequency The phase angle at .

[0158] As can be seen, these parameters constitute the device health fingerprint, which is used for subsequent health status assessment:

[0159] ;

[0160] in The health status fingerprint contains a set of characteristic parameters that describe the health status of the device; The transfer function at frequency The amplitude at The transfer function at frequency The phase angle at is an imaginary unit; represents the damping coefficient; represents the natural frequency.

[0161] The state estimation enhanced fault location module builds a health status assessment model based on the equipment health status information and calculates the deviation of the transfer function from the health baseline;

[0162] Combining optimized measurement point configuration with virtual measurement data, and using state estimation technology to complete missing information, this approach enables early identification and precise fault location. This involves the following sub-steps:

[0163] Step 4.1, hybrid measurement model construction;

[0164] The method provided in this application is based on the virtual-real fusion measurement network constructed in the virtual-real fusion measurement network construction module to establish a hybrid measurement model:

[0165] ;

[0166] in It is a mixed measurement vector, including actual measurement values and virtual measurement values, and its dimension is the total number of measurement points; are the actual measurement value vector and the virtual measurement value vector respectively; and The measurement functions representing the actual measurement points and the virtual measurement points respectively map the system states to the measurement values; They represent the actual measurement error and virtual measurement error, respectively, reflecting the uncertainty of the measurement value; is the system state variable, which represents the state quantities of the power grid such as voltage amplitude and phase angle.

[0167] Considering that the reliability of virtual measurement data is lower than that of actual measurement data, this application introduces the weight matrix

[0168] ;

[0169] in is a weight matrix, which assigns different weights to the measurement values of different measurement points and is a diagonal matrix; and are the weight sub-matrices for actual measurement and virtual measurement respectively; The elements in are weighted by the weights calculated in step 2.3 set up; represents a zero matrix, indicating that there is no cross-weight between the actual measurement and the virtual measurement.

[0170] Step 4.2, health status assessment;

[0171] It should be noted that the equipment health status assessment model is constructed based on the active excitation response characteristics obtained in the active fault excitation implementation module. First, the deviation between the current transfer function and the health benchmark is calculated:

[0172] ;

[0173] in is the current transfer function, which represents the frequency domain response characteristics of the device in its current state; is the health state benchmark transfer function, which represents the frequency domain response characteristics of the equipment in a healthy state; It is the L2 norm, Euclidean norm, used to calculate the length of vector or matrix size; is the transfer function deviation, which represents the distance between the current transfer function and the healthy benchmark transfer function.

[0174] Optionally, different norms can be used to calculate the transfer function deviation according to the specific application scenario. For example, for devices that are sensitive to changes in specific frequency points in the frequency domain (such as transformers), the L1 norm can be used:

[0175] ;

[0176] in A set of frequencies of interest, including frequencies that are particularly important for evaluating the health status of equipment; Represents the sum operator, which sums all elements in the collection; Represents the absolute value operator, which calculates the modulus of a complex number or the absolute value of a real number; is the current transfer function, which represents the frequency domain response characteristics of the device in its current state; is the health state benchmark transfer function, which represents the frequency domain response characteristics of the equipment in a healthy state; is the transfer function deviation after adopting the L1 norm, Indicates the angular frequency.

[0177] For devices that need to comprehensively consider the characteristics of multiple frequency bands, the weighted Euclidean norm can be used:

[0178] ;

[0179] in For the The weight coefficient of each frequency point reflects the The importance of each frequency point to the health status assessment can be determined by setting different weights based on the equipment failure-sensitive frequency points. Represents the square root operator, which calculates the square root of an expression; Represents the square of the complex modulus or the square of a real number; is the transfer function deviation after adopting the weighted Euclidean norm; is the current transfer function, which represents the frequency domain response characteristics of the device in its current state; is the health state benchmark transfer function, which represents the frequency domain response characteristics of the equipment in a healthy state.

[0180] The health status score is then evaluated by multi-feature fusion:

[0181] ;

[0182] in The health status score is a quantitative score of the overall health status of the device, usually in the range of [0, 1], where 1 indicates complete health. and are the deviations of the damping coefficient and natural frequency, respectively, indicating the difference between the current damping coefficient and the healthy state damping coefficient, and the difference between the current natural frequency and the healthy state natural frequency; A health status scoring function is a function that maps the transfer function deviation to a health status score; is the transfer function deviation.

[0183] In some embodiments, the health status scoring function A fuzzy logic inference system can be used to map the characteristic deviations to the interval [0, 1], and the final health status score can be calculated through a set of fuzzy rules. For example, the following fuzzy rules can be established:

[0184] like For "big" and For "righteousness", then is "low";

[0185] like is "small" and is "close to zero", then is "high";

[0186] Optionally, for complex equipment, a multi-level scoring strategy can be used for health status assessment. First, the health status score of each key component of the equipment (such as the transformer core, winding, insulation system, etc.) is calculated separately. , and then obtain the overall health status score by weighted average or taking the minimum value:

[0187] ;

[0188] or

[0189] ;

[0190] in is the number of key components, indicating the total number of key components included in the equipment; It is the index of the key components of the equipment, indicating the number of a specific component; is the weight of each component, indicating the contribution of different components to the overall health status; Represents the minimum operator, which returns the minimum value in the parameter list; Indicates from The sum operation of represent the health status scores calculated using the minimum strategy and weighted average strategy respectively; Represents the first, second, and The health status score of each key component.

[0191] For example, in a 35kV distribution equipment health assessment application at a 110kV substation, the system performed an active excitation test on the circuit breaker and then evaluated the status of each component using the aforementioned multi-level scoring strategy. The system detected a significant deviation in the operating mechanism transfer function in the 55 to 65Hz frequency band. Combined with the increased damping coefficient, this decreased the health score and generated an early warning. Subsequent inspection confirmed poor lubrication within the operating mechanism, and timely maintenance prevented a potential refusal to operate failure.

[0192] In addition, when When the device is judged to have potential failure risk, it enters the early warning state. The health status warning threshold is below which a health status warning is triggered.

[0193] Step 4.3, fault precursor identification;

[0194] Using time domain and frequency domain feature analysis methods, fault precursor features are identified from the response signal. For time domain analysis, the statistical characteristics of the response waveform are calculated:

[0195] ;

[0196] in is a time domain statistical feature set, including statistical features extracted from the time domain waveform; is the maximum value of the response waveform, the peak value of the time domain waveform; is the minimum value of the response waveform, the valley value of the time domain waveform; is the standard deviation, a statistic that indicates the degree of data dispersion; is the skewness, a statistic that indicates the degree of asymmetry of the data distribution; is the kurtosis, a statistic that indicates the degree of peak in the data distribution; To measure the response signal.

[0197] For frequency domain analysis, multi-scale features are extracted through wavelet transform:

[0198] ;

[0199] in is the frequency domain feature set, which contains the features extracted from the frequency domain analysis; For the The energy of a frequency band represents the energy density of the signal within a specific frequency band; is the frequency band index, indicating the number of a specific frequency band;

[0200] is the adjacent band energy ratio, which is the ratio of the energy of the current band to the adjacent band; Represents the second and The energy of each frequency band; Represents the second and The energy ratio of each frequency band; is the total number of frequency bands.

[0201] Therefore, combining the time domain and frequency domain features, this application constructs the fault precursor feature vector Contains a set of features that may indicate a fault has occurred, used for fault type identification and location.

[0202] The resilient fault location execution module, based on a hybrid measurement matrix, extracts key information dimensions to ensure the reliability of fault location when measurement data is uncertain or some measurement points fail.

[0203] The following sub-steps are included:

[0204] Step 5.1, extract key information dimensions;

[0205] The method provided in this application uses a key information dimension extraction algorithm to extract the most valuable fault features from mixed measurement data. First, principal component analysis is performed on the original high-dimensional features:

[0206] ;

[0207] in is the original feature matrix, which contains the original features of all samples and has the dimension of number of samples × number of features; is the principal component matrix, containing the eigenvectors of the principal components, with the dimension of number of features × number of features; It is the feature matrix after dimensionality reduction, the low-dimensional representation after principal component analysis, and its dimension is the number of samples × the number of selected principal components; Representation matrix The transpose of .

[0208] The contribution of each principal component to the fault type distinction is then evaluated using the Fisher discriminant criterion:

[0209] ;

[0210] in Respectively The inter-class dispersion and intra-class dispersion of the principal components; is the principal component index, which indicates the number of a specific principal component; is the Fisher discriminant criterion value, indicating the The contribution of the principal components to the classification of fault types.

[0211] according to Value, this application selects the most distinguishing principal components, forming a set of key information dimensions, where The number of principal components selected is the number of principal components retained.

[0212] Step 5.2, data credibility assessment;

[0213] It should be noted that the reliability of each measurement data is evaluated by combining the credibility weights of multi-source data. For actual measurement data, the credibility is calculated based on the device status and measurement residuals:

[0214] ;

[0215] in is the credibility of the actual measurement data, indicating that the node The reliability of the measurement data, with a value range of [0, 1]; To measure the status indicators of the equipment, reflecting the node The working status of the measuring equipment; To measure residuals, the difference between the measured and estimated values; is a credibility calculation function, which determines the credibility of the measurement data based on multiple factors; It is the node index, which indicates the number of a specific node.

[0216] In addition, for the virtual measurement data, the weights calculated in step 2.3 are directly used As credibility.

[0217] Step 5.3, fault type identification and location;

[0218] This application builds a resilient fault location algorithm based on the key information dimensions and data credibility extracted in the previous steps, using a weighted voting fusion method:

[0219] ;

[0220] in is the final identified fault type, indicating the fault type finally determined by the algorithm; For the The credibility weight of each piece of evidence; Indicates given evidence The fault type under the condition is The conditional probability of Indicates the total amount of evidence used for fault identification; Indicates from The sum operation of Indicates the type of fault that maximizes the expression .

[0221] It can be seen that fault location also uses a weighted method to construct a fault location probability distribution map:

[0222] ;

[0223] in The fault location is The probability of the fault location is probability; Evidence-based The position probability estimate of The fault location under the condition is The conditional probability of is the fault location variable; For a specific location point; The credibility weight of each piece of evidence; Indicates the total amount of evidence used for fault identification.

[0224] Therefore, the system ultimately outputs the fault type, fault location, and corresponding credibility to support the decision-making of operation and maintenance personnel.

[0225] The resilient fault location algorithm of the present application adopts a multi-level, multi-modal fault inference structure in its implementation. The first layer is the evidence collection layer, which processes five types of evidence sources at the same time: actual measurement data, virtual measurement data, active excitation response characteristics, historical fault modes and topological constraint information. The second layer is the evidence preprocessing layer, which applies specific preprocessing methods to each type of evidence, such as noise filtering, anomaly detection and normalization of measurement data; wavelet transform and spectrum analysis of active excitation characteristics; cluster analysis of historical fault modes, etc. The third layer is the local inference layer, which trains dedicated classifiers (such as support vector machines, random forests or neural networks) for each type of preprocessed evidence to generate preliminary fault type and location estimates. The fourth layer is the evidence weighted fusion layer, which combines the credibility of each evidence source and uses the Bayesian network framework to fuse the local inference results to form the final fault judgment result.

[0226] The algorithm in this application also employs a progressive inference strategy, adjusting the precision of inferences based on the quantity and quality of available evidence. When evidence is insufficient or low-quality, the system initially provides a rough estimate of the fault's extent (e.g., the fault segment or region). As more evidence is collected or processed, the system gradually refines the fault's location. Furthermore, this method incorporates an adaptive confidence threshold: the system assigns a confidence score to each inference result. When the confidence score falls below the threshold, the system recommends that operators obtain additional evidence or initiate alternative location methods, thus avoiding wasted resources due to misjudgments.

[0227] A storage medium includes a memory and one or more processors, wherein the memory stores executable code, and when the one or more processors execute the executable code, it is used to implement the above-mentioned distribution network fault identification and positioning system based on wide-area measurement technology.

[0228] Here, the present invention provides an implementation example:

[0229] This implementation method was put to practical use in a distribution network located in the urban-rural fringe of a certain city. This distribution network encompasses two 110kV substations, five 35kV substations, and 23 10kV distribution substations, covering approximately 76 square kilometers and serving approximately 320,000 people. The distribution network in this area has the following characteristics: First, equipment age varies, with some lines and equipment operating for 15 to 20 years, posing potential failure risks. Second, measurement coverage is uneven, with relatively sufficient measurement points in core areas and sparser in peripheral areas. Third, the frequency of failures is high, with an average of 26 failures per year, of which over 40% are due to aging equipment.

[0230] Before implementing this method, the average time to locate distribution network faults in this region was 3.7 hours, with the longest locating time reaching 7.5 hours. Fault early warning capabilities were weak, with 95% of faults not detected until after they occurred, leading to frequent large-scale power outages. Furthermore, existing measurement equipment coverage was only 33%, and most areas lacked real-time monitoring capabilities, further complicating fault location.

[0231] Based on the above situation, the goal of this embodiment is to improve the fault warning capability and fault location accuracy of the distribution network, and at the same time, expand the monitoring coverage through virtual measurement technology under limited budget conditions.

[0232] First, a multi-dimensional weak link identification algorithm was applied to analyze the structural characteristics and historical operating data of the distribution network. Based on the results, the system identified 15 high-risk weak links, primarily concentrated in distribution lines with severely aged equipment and in critical load-intensive areas.

[0233] Based on sensitivity assessments of measurement points and budget constraints, 23 new measurement points were optimized and added to the existing 47 measurement points across the network. Twelve of these were located at the outgoing lines of major substations, eight at key nodes along important lines, and three at high-value user access points. This optimized configuration of measurement points increased monitoring coverage in high-risk areas from 51% to 91%.

[0234] A physical correlation model was constructed for the remaining measurement blind spots, particularly the six 10kV distribution transformer stations and 12 branch lines in the marginal areas. A DC power flow model was used for the trunk lines, and an improved Distflow equation was used for the terminal branches.

[0235] On this basis, an improved generative adversarial network was trained. This network used six months of historical operating data as its training set. The generator consisted of three hidden layers, and the discriminator had two hidden layers. The weight coefficient of the physical constraint regularization term was set to 0.45 to ensure that the generated virtual data conformed to the physical characteristics of the power grid.

[0236] Through this network, virtual measurement data was successfully generated for 43 nodes in blind spots, increasing network-wide measurement coverage from 33% to 92%. The average error rate of the virtual measurement data was kept below 1.3% under normal operating conditions.

[0237] Safe and controllable active fault excitation was implemented for the 15 high-risk weak links identified, especially five old lines and three distribution transformers that have been operating at high load for a long time.

[0238] For example, a power electronic injection device installed on the line injected a perturbation signal with an amplitude of 0.8% of the rated voltage, a frequency of 120 Hz, and a duration of 0.5 seconds into the switchgear on pole 3 of the 10kV Fenghua Line. During the injection process, the system monitored 10 key parameters, including voltage, current, and harmonic content, in real time to ensure the safety of the injection process.

[0239] By analyzing the response signal, the system detected an abnormal shift in the switchgear's phase-frequency response characteristics between 55 and 70 Hz. The damping coefficient also increased by 35%, and the health score dropped to 0.68 (below the warning threshold of 0.75). Subsequent inspections revealed poor contact in the switchgear, and proactive repairs averted a potential segmented fault.

[0240] In this example, the system built a hybrid measurement model that integrated data from 70 real measurement points and 43 virtual measurement points. The weight of the virtual measurement data was dynamically adjusted based on its generation reliability, ranging from 0.4 to 0.8.

[0241] Based on the device health fingerprints obtained through active stimulation, the system assesses the health status of all network-wide equipment. During implementation, the system identified seven potential fault points: three due to aging transformer insulation, two due to overheating cable connectors, and two due to poor switch contact. These potential fault points were flagged and placed into early warning mode, allowing operations and maintenance personnel to plan maintenance accordingly.

[0242] To verify the system's resilience, we simulated three partial failures of measurement points, randomly causing 30% of the 21 measurement points to fail. Under these conditions, the system still maintained a high level of fault location accuracy by extracting key information dimensions and assessing data credibility.

[0243] In a real-world fault case, a fallen tree caused a short circuit on a 10kV branch line. Simultaneously, a communication failure resulted in missing data from three nearby measurement points. The system leveraged a multi-level fault inference architecture, combining residual measurement data, virtual measurement data, historical fault patterns, and topology constraint information to successfully locate the fault within 800 meters of its actual location. This reduced the location range by approximately 78% compared to traditional methods, significantly reducing the workload for line patrol personnel.

[0244] One year after implementing this method in the distribution network, statistical analysis of operational data verified the following two key technical effects:

[0245] Using proactive fault stimulation and health assessment technology, the system successfully alerted 23 potential faults, accounting for 82% of all faults that year. The average lead time for system alerts was 18 days, with the longest lead time reaching 32 days. The accuracy rate of these alerts was 91.3% (21 of the 23 alerts were subsequently verified).

[0246] The distribution of potential fault warning times: 4 cases occurred within 5 to 10 days, 12 cases occurred within 11 to 20 days, 5 cases occurred within 21 to 30 days, and 2 cases occurred within more than 30 days. Warning effectiveness varied by fault type: Warnings for equipment aging faults occurred an average of 25 days in advance, for poor contact faults an average of 16 days in advance, and for insulation degradation faults an average of 22 days in advance.

[0247] By proactively detecting and addressing these potential faults, 18 potential power outages were avoided, reducing losses by approximately 2.96 million yuan and benefiting 85,000 users. Compared to the pre-implementation situation where 95% of faults were discovered only after they occurred, this method achieved an 82% early detection rate for potential faults, demonstrating technological progress.

[0248] The system located all 28 faults that occurred throughout the year (including five sudden faults that were not warned and 23 faults that were warned and observed). The average location accuracy reached 89.3%, and the average location time was reduced to 0.6 hours, a reduction of 83.8% compared to the 3.7 hours before implementation.

[0249] By integrating virtual and real measurement technology, the system increased measurement coverage from 33% to 92%. Even in areas without actual measurement points, virtual measurement data provided effective support for fault location. For faults occurring near six 10kV distribution transformer stations in marginal areas, the average location error was kept within 1.2 kilometers, a 68.4% improvement over the 3.8 kilometers achieved with traditional methods.

[0250] The system's resilience was also demonstrated. In a test simulating a 30% failure of measurement points, the system maintained an 86.7% fault location accuracy, compared to only 41.2% for traditional methods under the same conditions. Furthermore, in four real-world cases of partial communication system failure, the system maintained an average location error of less than 1.5 kilometers.

[0251] The above technical effect verification data show that this implementation method improves the distribution network fault warning capability and fault location accuracy, especially under conditions of insufficient measurement coverage and partial failure of measuring equipment, it can still maintain stable performance, fully reflecting the technical innovation and practical value of this application.

[0252] The above describes an embodiment of the present invention, but this embodiment is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Ordinary technicians in this field can also make more forms of equivalent embodiments based on the inspiration of this embodiment, all of which are protected by this embodiment.

Claims

1. A distribution network fault identification and location system based on wide-area measurement technology, characterized in that: include: Key measurement point optimization configuration module, which uses a sensitivity assessment algorithm to determine the optimal measurement point layout and identify high-risk weak links in the distribution network that require key monitoring; The virtual-real fusion measurement network construction module uses a generative adversarial network to generate virtual measurement data in blind areas that conform to physical laws based on the optimal measurement point layout. It then fuses the actual measurement data with the virtual measurement data using a dynamic weighting method to form a hybrid measurement matrix covering the entire network. Active fault stimulation implementation module implements safe and controllable small disturbance injection for high-risk weak links, actively detects the response characteristics of weak links, and obtains equipment health status information. Specifically, it records the equipment's response to the perturbation signal and obtains the system response transfer function; The state estimation enhanced fault location module constructs a health status assessment model based on the equipment health status information. The construction of the health status assessment model based on the equipment health status information specifically includes: Calculating the deviation of the current transfer function from the healthy baseline transfer function; Based on the deviation, a health status score is evaluated by multi-feature fusion. When the health status score is lower than a preset threshold, it is determined that the device has a potential failure risk and enters a warning state; The resilient fault location execution module, based on the hybrid measurement matrix, extracts key information dimensions to ensure the reliability of fault location when there is uncertainty in the measurement data or failure of some measurement points.

2. A distribution network fault identification and location system based on wide area measurement technology according to claim 1, characterized in that: Determining the optimal measurement point arrangement position by using a sensitivity evaluation algorithm includes: Apply a multi-dimensional weak link identification algorithm to assess the risk level of each node in the distribution network and calculate a risk score for each node. The risk score comprehensively considers topological vulnerability, historical fault frequency, electrical stress, and node importance. Analyze the contribution of each candidate measurement point to network observability, construct the observation equation and calculate the observation sensitivity matrix of each candidate measurement point to obtain the measurement point sensitivity index; Based on the risk score and sensitivity index, a measurement point configuration optimization model is constructed, and the optimization model is solved to determine the final measurement point configuration plan.

3. The distribution network fault identification and positioning system based on wide-area measurement technology according to claim 1, characterized in that: The method of using a generative adversarial network to generate virtual blind spot measurement data that conforms to physical laws includes: Using electrical topology and power flow constraints, a physical correlation model between measurement points and blind zone nodes is constructed; The generator and discriminator of the generative adversarial network are trained. The generator receives random noise and electrical quantity information of surrounding measured points as input to generate virtual measurement data, and the discriminator determines the authenticity of the data. Physical constraint regularization terms are introduced into the generator's loss function, including Kirchhoff's voltage law constraint term, Kirchhoff's current law constraint term and power flow equation constraint term.

4. The distribution network fault identification and location system based on wide area measurement technology according to claim 1, characterized in that: The safe and controllable injection of small disturbances targeting high-risk weak links includes: Construct a perturbation signal for active fault excitation. The perturbation signal is in the form of a decaying sine wave, and the amplitude is set within a safe range that does not affect the normal operation of the system. Through the power electronic interface equipment, the perturbation signal is safely injected into the key weak links of the distribution network, and the injection process is controlled by a closed-loop control method; Monitor key system parameters in real time, set warning thresholds and emergency thresholds, and trigger corresponding protection actions when the monitored parameters exceed the thresholds.

5. The distribution network fault identification and positioning system based on wide area measurement technology according to claim 1, characterized in that: The step of extracting key information dimensions when there is uncertainty in the measurement data or some measurement points fail includes: Extract the most valuable fault features from the mixed measurement data, perform principal component analysis on the original high-dimensional features, and obtain the feature matrix after dimensionality reduction; The contribution of each principal component to the classification of fault types was evaluated using the Fisher discriminant criterion; According to the contribution value, several principal components with the greatest discriminative power are selected to form a set of key information dimensions; The reliability of each measurement data is evaluated by combining the credibility weights of multi-source data. The credibility of the actual measurement data is calculated based on the device status and measurement residuals. For the virtual measurement data, its generation reliability is directly used as the credibility.

6. The distribution network fault identification and location system based on wide area measurement technology according to claim 5, characterized in that: Ensuring the reliability of fault location includes: Based on the extracted key information dimensions and data credibility, a resilient fault location algorithm is constructed; The resilient fault location algorithm uses a weighted voting fusion method to identify the fault type. The credibility weight of each piece of evidence is multiplied by the probability estimate of the fault type based on the evidence and then the sum is calculated. Construct a probability distribution map of the fault location, multiply the credibility weight of each piece of evidence by the location probability estimate based on the evidence, and then divide the sum by the sum of the credibility weights; Output fault type, fault location and corresponding credibility to support operation and maintenance personnel in decision-making.

7. The distribution network fault identification and positioning system based on wide-area measurement technology according to claim 6, characterized in that: The resilient fault location algorithm includes an evidence collection layer, an evidence preprocessing layer, a local inference layer, and an evidence weighted fusion layer, specifically including: The evidence collection layer processes five types of evidence sources: actual measurement data, virtual measurement data, active stimulation response characteristics, historical failure modes, and topology constraint information; The evidence preprocessing layer applies specific preprocessing methods to each type of evidence; The local inference layer performs independent analysis on each type of pre-processed evidence to generate preliminary fault type and location estimates; The evidence weighted fusion layer combines the credibility of each evidence source and uses the Bayesian network framework to fuse the local inference results to form the final fault judgment result.

8. The distribution network fault identification and location system based on wide area measurement technology according to claim 1, characterized in that: The micro-disturbance injection adopts a safe injection control system, which includes: Based on a hybrid control architecture of DSP and FPGA, DSP is responsible for upper-level control strategy and safety monitoring, while FPGA realizes high-speed waveform generation and real-time control; Closed-loop control parameter adaptive adjustment unit automatically optimizes controller parameters according to system impedance characteristics; Multi-layer safety protection unit monitors key parameters, including voltage, current, frequency and harmonic content, and triggers corresponding protection actions when the monitored key parameters exceed the threshold.

Images