Power distribution network fault intelligent positioning method and device

Abstract

The invention discloses a power distribution network fault intelligent positioning method and a model training method, and the method comprises the steps: carrying out the data collection through a multi-mode sensor of a power distribution network, and obtaining the traveling wave data when the power distribution network breaks down; inputting the traveling wave data into a traveling wave speed model to obtain a traveling wave speed predicted value and a traveling wave time difference predicted value, defining a fault position, the traveling wave speed predicted value and the traveling wave time difference predicted value as state vectors, and defining a traveling wave time difference measured value as an observation vector; using a Kalman filtering method to obtain a fault position estimation value according to the state vector and the observation vector; for model training of a traveling wave speed model, a multi-channel convolutional network and a time sequence diagram neural network are utilized to construct a model architecture, and through channel allocation of multi-modal training data and dynamic capture of node correlation by the time sequence diagram neural network, a complex time sequence relationship of traveling wave data can be processed more flexibly.

Description

A method and device for intelligent fault location in power distribution networks Technical Field

[0001] This invention pertains to power grid safety monitoring technology, and particularly relates to a method and device for intelligent fault location in distribution networks. Background Technology

[0002] Faults in the power distribution network can cause economic losses and social impacts. Therefore, it is necessary to locate power distribution network faults quickly and accurately.

[0003] Current methods for fault location in power distribution networks mainly include impedance methods and traveling wave methods. Impedance methods, based on the impedance relationship between the fault point and the measurement point, are simple and easy to implement, but their accuracy is significantly affected by factors such as line parameters and load changes, especially in complex, multi-branch networks where errors are substantial. Traveling wave methods utilize the propagation characteristics of traveling waves generated when a fault occurs in power lines for location. Traditional traveling wave location methods mainly rely on measuring the arrival time of the traveling wave front and calculating the fault distance through the time difference. However, their accuracy is low when various types of traveling wave reflections and refractions exist. Furthermore, existing technologies lack effective fusion methods for multi-frequency traveling wave information, making it difficult to fully utilize the advantages of different frequency bands to improve location accuracy. Summary of the Invention

[0004] Based on this, the present invention aims to propose an intelligent fault location method and device for distribution networks. It uses a traveling wave velocity model to predict the traveling wave velocity and traveling wave time difference based on historical traveling wave data, and then uses the Kalman filter method to estimate the fault location of the distribution network line, so as to improve the accuracy of fault location.

[0005] In a first aspect, the present invention provides a method for intelligent fault location in a power distribution network, comprising:

[0006] Collect traveling wave data of power distribution network faults. The traveling wave data includes the travel wave time difference measurement.

[0007] Input the traveling wave data into the traveling wave velocity model to obtain the predicted values ​​of the traveling wave velocity and the traveling wave time difference;

[0008] The fault location, predicted traveling wave velocity, and predicted traveling wave time difference are defined as the state vector, and the measured traveling wave time difference is defined as the observation vector. The fault location estimate is obtained by using the Kalman filter method based on the state vector and the observation vector.

[0009] Furthermore, the fault location estimate obtained using the Kalman filter method based on the state vector and observation vector includes:

[0010] The traveling wave data is divided into several frequency bands, and a Kalman filter is constructed for each frequency band.

[0011] The estimated fault location of each frequency band is calculated using the Kalman filter corresponding to each frequency band based on the state vector and observation vector of each frequency band.

[0012] The fault location estimate is calculated based on the fault location estimate for each frequency band.

[0013] Furthermore, the calculation of fault location estimates based on fault location estimates for each frequency band includes:

[0014] Define the frequency band fault location estimate and its confidence level as an evidence body, determine the evidence body corresponding to each frequency band, wherein the confidence level of the frequency band fault location estimate is determined according to the output of the Kalman filter;

[0015] By using Dempster's combination rule to fuse the various evidence bodies, the estimated fault location and its confidence level are obtained.

[0016] Furthermore, the traveling wave data includes at least one of the following: the arrival time of the initial traveling wave front, the arrival time of the reflected traveling wave front, the traveling wave amplitude, waveform details, polarity, waveform polarity reversal time, line length, and line wave impedance.

[0017] In a second aspect, the present invention provides a model training method for training a traveling wave velocity model as described in the first aspect, comprising:

[0018] Historical traveling wave data, geographical data, and historical line data of distribution network faults were collected as the training set.

[0019] The initial network for constructing the traveling wave velocity model is specifically composed of a data input layer, a physical model layer, a data-driven layer, a model fusion layer, a monitoring feedback layer, and an output layer.

[0020] The data input layer includes a multi-channel temporal convolutional network, used to allocate channels to the geographic data and historical route data in the training set to obtain geographic features and route features respectively. The physical model layer includes a perturbation correction propagation velocity model, used to output a dynamic velocity matrix based on geographic features and route features. The data-driven layer includes a temporal graph neural network, used to output a dynamic node feature matrix based on geographic features and route features. The model fusion layer is used to fuse the outputs of the physical model layer and the data-driven layer to obtain a fused velocity matrix. The monitoring feedback layer is used to calculate the confidence level based on the fused velocity matrix and historical traveling wave data. The output layer is used to output the predicted values ​​of traveling wave velocity and traveling wave time difference based on the fused velocity matrix and historical traveling wave data.

[0021] The initial network is trained to obtain a trained traveling wave velocity model.

[0022] Furthermore, the data-driven layer includes the TGAT network, which outputs a dynamic node feature matrix based on geographical and route characteristics, including:

[0023] Using geographical data and fault labels from training set samples as nodes in a graph structure, and historical route data as edges, the graph structure is used as input to the TGAT network to obtain a dynamic node feature matrix.

[0024] Furthermore, the monitoring feedback layer is used to calculate the confidence level based on the fused velocity matrix and historical traveling wave data, including:

[0025] The confidence level was calculated using a Gaussian mixture model (GMM) based on the fused velocity matrix and historical traveling wave data.

[0026] Thirdly, the present invention provides an intelligent fault location device for a power distribution network, comprising:

[0027] The traveling wave data acquisition module is used to acquire traveling wave data of distribution network faults. The traveling wave data includes the traveling wave time difference measurement value.

[0028] The traveling wave velocity modeling module takes traveling wave data as input into the traveling wave velocity model to obtain predicted values ​​for traveling wave velocity and traveling wave time difference.

[0029] The fault location estimation module is used to define the fault location, the predicted value of the traveling wave velocity, and the predicted value of the traveling wave time difference as the state vector, and the measured value of the traveling wave time difference as the observation vector. The Kalman filter method is used to obtain the fault location estimate based on the state vector and the observation vector.

[0030] Fourthly, the present invention provides a model training apparatus for training a traveling wave velocity model as described in the first aspect, comprising:

[0031] The training set acquisition module is used to collect historical traveling wave data, geographical data, and historical line data of distribution network faults as the training set.

[0032] The model building module is used to build the initial network for the traveling wave velocity model. The initial network specifically includes a data input layer, a physical model layer, a data-driven layer, a model fusion layer, a monitoring feedback layer, and an output layer.

[0033] The data input layer includes a multi-channel temporal convolutional network, used to allocate channels to the geographic data and historical route data in the training set to obtain geographic features and route features respectively. The physical model layer includes a perturbation correction propagation velocity model, used to output a dynamic velocity matrix based on geographic features and route features. The data-driven layer includes a temporal graph neural network, used to output a dynamic node feature matrix based on geographic features and route features. The model fusion layer is used to fuse the outputs of the physical model layer and the data-driven layer to obtain a fused velocity matrix. The monitoring feedback layer is used to calculate the confidence level based on the fused velocity matrix and historical traveling wave data. The output layer is used to output the predicted values ​​of traveling wave velocity and traveling wave time difference based on the fused velocity matrix and historical traveling wave data.

[0034] The model training module is used to train the initial network to obtain the trained traveling wave velocity model.

[0035] Thirdly, the present invention provides an electronic device including a memory storing computer-executable instructions and a processor, wherein when the computer-executable instructions are executed by the processor, the device performs the steps of the intelligent fault location method for power distribution networks provided in the first aspect, and / or the model training method provided in the second aspect.

[0036] Fourthly, the present invention provides a readable storage medium storing a computer-executable program that, when executed, can implement the various steps of the intelligent fault location method for distribution networks provided in the first aspect, and / or the model training method provided in the second aspect.

[0037] Compared with the prior art, the present invention has the following beneficial effects:

[0038] This invention provides an intelligent fault location method and model training method for distribution networks. The intelligent fault location method collects data from multi-modal sensors in the distribution network to obtain traveling wave data at the time of a fault. The traveling wave data is input into a traveling wave velocity model to obtain predicted values ​​of traveling wave velocity and traveling wave time difference. Then, the Kalman filter method is used to obtain the fault location estimate, which can effectively improve the accuracy of fault location, especially for the fusion estimation of multi-frequency traveling wave data. For the model training of the traveling wave velocity model, a multi-channel convolutional network and a time-series graph neural network are used to construct the model architecture. By allocating channels for multi-modal training data and dynamically capturing node correlations through the time-series graph neural network, the complex temporal relationships of traveling wave data can be handled more flexibly. Attached Figure Description

[0039] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0040] Figure 1 is a flowchart of the intelligent fault location method for power distribution networks provided in an embodiment of the present invention.

[0041] Figure 2 is a flowchart of the model training method provided in an embodiment of the present invention;

[0042] Figure 3 is a structural diagram of the traveling wave velocity model provided in an embodiment of the present invention;

[0043] Figure 4 is a schematic diagram of the intelligent fault location device for power distribution networks provided in an embodiment of the present invention.

[0044] Figure 5 is a schematic diagram of the model training device provided in an embodiment of the present invention;

[0045] Figure 6 is an electronic device architecture diagram provided in an embodiment of the present invention. Detailed Implementation

[0046] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0047] Referring to Figure 1, an embodiment of the present invention provides a method for intelligent fault location in a power distribution network, comprising the following steps:

[0048] Step S110. Collect traveling wave data of distribution network faults. The traveling wave data includes the travel wave time difference measurement value.

[0049] The traveling wave data acquisition in this step is accomplished through traveling wave sensors. Specifically, a distributed multimodal traveling wave sensor network is set up at key nodes of the distribution network, line branch points, and substation outgoing lines to acquire traveling wave data of different frequency bands when a distribution network fault occurs.

[0050] Furthermore, when configuring the sensor network, multi-modal sensors of electric field, magnetic field, current, and voltage types are deployed at key nodes, line branch points, and substation outgoing lines in the distribution network. Each sensor synchronously collects traveling wave data of different modes. The arrival times of the initial traveling wavefront and the reflected wavefront are identified using waveform abrupt change detection algorithms (such as zero-crossing detection and energy abrupt change detection). The high-frequency components of the waveform are analyzed using wavelet transform or Fourier transform to extract detailed waveform features (such as glitches and oscillations). The polarity of the traveling wave is determined by detecting the positive and negative directions of the waveform, and the moment of polarity reversal is recorded. In addition, the line length can be obtained through pre-measured line information or a GIS system.

[0051] For example, consider a simple distribution network consisting of a substation, a main line, and two branch lines. Assume a single-phase ground fault occurs on line 2. Multimodal sensors are deployed at the substation outlet, the start and end points of the main line, and the start points of the branch lines to synchronously collect traveling wave data generated by the fault.

[0052] The data collected by sensor 1 (located at the substation outgoing line) includes: the arrival time t1 of the initial traveling wave front and the arrival time of the reflected traveling wave front. Traveling wave amplitude A1, waveform details d1 (few high-frequency components), polarity (positive polarity), waveform polarity reversal time (no reversal).

[0053] The data collected by sensor 2 (starting point of the main line) includes: the arrival time of the initial traveling wave front. Time of arrival of the reflected traveling wave front , traveling wave amplitude A2, waveform details d2 (more high-frequency components), polarity (positive polarity), waveform polarity reversal time (no reversal).

[0054] The data collected by sensor 3 (the end point of the main line) includes: the arrival time of the initial traveling wave front t3, and the arrival time of the reflected traveling wave front. , traveling wave amplitude A3, waveform details d3 (more high-frequency components), polarity (positive polarity), waveform polarity reversal time (no reversal).

[0055] The data collected by sensor 4 (starting point of line 1) includes: arrival time of the initial traveling wave front t4, arrival time of the reflected traveling wave front (none), traveling wave amplitude A4, waveform details d4 (few high-frequency components), polarity (positive polarity), and waveform polarity reversal time (no reversal).

[0056] The data collected by sensor 5 (the starting point of line 2) includes: the arrival time t5 of the initial traveling wave front and the arrival time of the reflected traveling wave front. Traveling wave amplitude A5, waveform details d5 (more high-frequency components), polarity (negative polarity), waveform polarity reversal time. .

[0057] By comparing the initial traveling wave front arrival time differences of sensors 1, 2, 3, and 5, and combining this with the line length and propagation speed, it can be preliminarily determined that the fault occurs on branch line 2. The polarity reversal information from sensor 5 indicates a single-phase ground fault. By combining the initial traveling wave front arrival time, the reflected wave front arrival time, and the line length from sensor 5, the fault location can be precisely pinpointed.

[0058] In one alternative approach, the distributed multimodal traveling wave sensor network includes electric field type, magnetic field type, current type and voltage type sensors for acquiring electric field, magnetic field, current and voltage traveling wave information when a fault occurs.

[0059] In this embodiment, an electric field sensor (capacitive or dipole type) measures changes in the electric field intensity in space, capturing electric field disturbances caused by traveling waves. A magnetic field sensor (coil or Hall effect type) measures changes in the magnetic field intensity in space, capturing magnetic field disturbances caused by traveling waves. A current sensor measures changes in the current in a conductor, capturing current abrupt changes caused by traveling waves propagating in the conductor. A voltage sensor measures changes in the voltage across a conductor, capturing voltage abrupt changes caused by traveling waves propagating in the conductor. For example, a set of multi-mode sensors is deployed near each key tower of a 110kV transmission line. The electric field sensor is installed on the tower insulator, and the magnetic field sensor is installed near the tower.

[0060] The traveling wave data collected in this embodiment includes the arrival time of the initial traveling wave front, the arrival time of the reflected traveling wave front, the traveling wave amplitude, waveform details, polarity, waveform polarity reversal time, line length, and line impedance. Among these, the arrival time of the initial traveling wave front is the basis for traveling wave localization; the time difference of arrival of the initial wave front received by different sensors can preliminarily determine the fault location. Reflected traveling waves are generated due to factors such as line mismatch and branch points; the arrival time of the reflected wave front can help determine the location of the fault point and the reflection point. The traveling wave amplitude is related to the severity of the fault and the line impedance. Waveform details reflect the local state of the line and the fault type; for example, high-frequency components in the waveform can indicate local changes at the fault point. Polarity changes indicate the fault type; for example, a single-phase ground fault will cause a polarity reversal in the voltage traveling wave. The waveform polarity reversal time helps distinguish different types of faults; for example, the polarity reversal characteristics of phase-to-phase faults and ground faults are different. Line length is a fundamental parameter for calculating the traveling wave propagation time. Line impedance affects the reflection and transmission of the traveling wave during propagation.

[0061] Step S120. Input the traveling wave data into the traveling wave velocity model to obtain the predicted value of the traveling wave velocity and the predicted value of the traveling wave time difference.

[0062] Specifically, the traveling wave velocity model used in this step is trained based on historical traveling wave data, geographical data, and historical line data during distribution network faults. The trained traveling wave velocity model is trained by taking the traveling wave inputs of different frequency bands collected in real time by the traveling wave sensor network at the time of the fault, and outputs the predicted values ​​of traveling wave velocity and traveling wave time difference for each frequency band, which are then used for subsequent fault location using the Kalman filtering method.

[0063] Step S130. Define the fault location, the predicted value of the traveling wave velocity, and the predicted value of the traveling wave time difference as the state vector, and the measured value of the traveling wave time difference as the observation vector. Use the Kalman filter method to obtain the estimated value of the fault location based on the state vector and the observation vector.

[0064] This step defines the initial state estimation. Initial covariance matrix The observation noise covariance matrix R, the transition noise covariance matrix Q, the state transition matrix F, the control input matrix B, and the observation matrix H are all included.

[0065] In the prediction step, the state transition equation is used to predict the state estimate for the next time step. Covariance Matrix .

[0066] In the update step, the Kalman gain is calculated. Then use the observation vector Update state estimation Covariance Matrix ,calculate At that time, it was actually to and Multiply the differences and add them to the product. , ,in, , which represents the difference between the observed value and the predicted value.

[0067] The updated state estimate is obtained through recursive updating of the Kalman filter. Covariance Matrix This reflects the best estimate of the fault location, traveling wave velocity, and time difference of arrival given the observation data.

[0068] In one alternative approach, Kalman filtering is performed separately for different frequency bands, and step S130 includes the following steps:

[0069] Step S131. Divide the traveling wave data into several frequency bands and construct a Kalman filter for each frequency band;

[0070] Step S132. Calculate the estimated fault location of each frequency band using the Kalman filter corresponding to each frequency band based on the state vector and observation vector of each frequency band;

[0071] Step S133. Calculate the fault location estimate based on the fault location estimate for each frequency band.

[0072] Independent Kalman filters are used to estimate the location of the fault point for different frequency bands. The Kalman filters corresponding to different frequency bands have independent measurement model parameters, including the process noise covariance matrix and the measurement noise covariance matrix.

[0073] Specifically, traveling wave signals in different frequency bands are subject to varying degrees of attenuation and distortion during transmission. High-frequency components attenuate faster but have higher positioning accuracy; low-frequency components have a longer transmission distance but lower positioning accuracy. Using independent Kalman filters to optimize for the characteristics of each frequency band and selecting an appropriate measurement noise covariance matrix for each frequency band can more effectively suppress the impact of noise on positioning accuracy.

[0074] For example, wavelet transform can be used to extract traveling wave signals in three frequency bands—low frequency (LF), medium frequency (MF), and high frequency (HF)—from the transmission lines of a power distribution network.

[0075] For the LF band, the measurement model is as follows: ,in, It is measurement noise, and its covariance matrix is For the MF band, the measurement model is as follows: ,in, It is measurement noise, and its covariance matrix is For the HF band, the measurement model is as follows: ,in It is measurement noise, and its covariance matrix is .

[0076] The measured noise in each frequency band is independent of each other and usually meets the following requirements. Kalman filtering iterations were performed on each frequency band to obtain the fault location estimate output by the LF filter. The fault location estimate output by the MF filter Fault location estimate output by the HF filter The estimated fault locations for the three frequency bands are weighted and averaged, with the weights determined based on the estimated variance of each frequency band.

[0077] Furthermore, for high-frequency traveling wave signals, the diagonal elements of their measurement noise covariance matrix are increased, reducing the weight of high-frequency measurements in the filtering process and minimizing the impact of noise on fault location, thereby improving the reliability of high-frequency signal location results. For low-frequency traveling wave signals, the diagonal elements of their measurement noise covariance matrix are decreased, increasing the weight of low-frequency measurements in the filtering process, leveraging the advantages of low-frequency signals to improve location accuracy. By dynamically adjusting the size of the process noise covariance matrix, the Kalman filter adapts to changes in power grid conditions, enabling the fusion of data from different frequency bands.

[0078] The power grid state is not static; it changes dynamically due to factors such as load variations and switching operations, causing changes in the propagation speed and arrival time of traveling wave signals. By dynamically adjusting the size of the process noise covariance matrix, the Kalman filter can adapt to changes in the power grid state more quickly, improving its state tracking capability and ensuring the accuracy of fault location estimation.

[0079] In one optional implementation, calculating the fault location estimate based on the fault location estimate for each frequency band includes:

[0080] Define the frequency band fault location estimate and its confidence level as an evidence body, determine the evidence body corresponding to each frequency band, wherein the confidence level of the frequency band fault location estimate is determined according to the output of the Kalman filter;

[0081] By using Dempster's combination rule to fuse the various evidence bodies, the estimated fault location and its confidence level are obtained.

[0082] Specifically, due to noise, interference, or model errors, the positioning results of different frequency bands may conflict, and the fault location estimates output by the Kalman filter may be uncertain. Combining the estimates and confidence levels as evidence (e.g., if the fault location is within the 1-100km range, a segment near each estimate can be used as the focus set) can more comprehensively represent the positioning results. The Dempster combination rule uses the Basic Probability Assignment Function (BPA) to represent the confidence level of the evidence. Unlike probability theory, BPA allows confidence levels to be assigned to uncertain or unknown propositions, which can better handle incomplete information. The combination rule can fuse the BPAs of multiple evidence segments (e.g., the fused BPA is mainly concentrated in the [50,52]km interval, corresponding to BPA m([50,52]). When the positioning results of a certain frequency band are interfered with, the positioning results of other frequency bands provide supplementary information, reducing the possibility of misjudgment.

[0083] The conflict coefficient is used to characterize the degree of conflict between fault location estimates in different frequency bands. If the conflict system is higher than the preset threshold, the validity of the estimation results of that frequency band is re-evaluated and the estimates of each frequency band are readjusted and then weighted and fused.

[0084] The formula for weight fusion can be expressed as:

[0085]

[0086] in, This represents the basic probability assignment value of the fault location by the Kalman filter from frequency band n. This represents the intersection of events for all Kalman filters across all frequency bands.

[0087] Referring to Figure 2, one embodiment of the present invention provides a model training method for training a traveling wave velocity model, comprising:

[0088] Step S210. Collect historical traveling wave data, geographical data, and historical line data of distribution network faults as training set.

[0089] This step involves collecting historical traveling wave data of distribution network faults through a traveling wave sensor network deployed in the distribution network, pre-extracting features from the historical traveling wave data, and using the obtained traveling wave detail features as input to the traveling wave velocity model for model training.

[0090] Specifically, historical traveling wave data includes the arrival time of the initial traveling wave front, the arrival time of the reflected traveling wave front, the traveling wave amplitude, waveform details, polarity, waveform polarity reversal time, line length, and line wave impedance. Feature pre-extraction can be completed through wavelet transform to determine features that are relevant to fault location.

[0091] In a more preferred implementation, Morlet or Meyer wavelet basis functions are selected to perform wavelet transform on the traveling wave signal, and the wavelet coefficients of the traveling wave signal are calculated to obtain the signal representation at different frequency levels. From the wavelet coefficients at different frequency levels, detailed features of the traveling wave signal (such as arrival time, amplitude, frequency, etc.) are extracted, and the statistical measure (such as mutual information) of the correlation between each feature and fault location is calculated. Features with correlation exceeding a preset correlation threshold are selected.

[0092] This step collects geographic data including the location coordinates of the traveling wave sensor, line length, line path (a sequence of geographic coordinates describing the line's direction), vegetation type (forest, grassland), building density, topography (mountainous area, plain), and humidity level; historical line data includes historical load data (such as hourly load curves), meteorological data (temperature, humidity, wind speed, rainfall, light intensity), line type (overhead line, cable), line materials, line operating time, and maintenance records.

[0093] Step S220. Construct the initial network for the traveling wave velocity model. The initial network specifically includes a data input layer, a physical model layer, a data-driven layer, a model fusion layer, a monitoring feedback layer, and an output layer.

[0094] The data input layer includes a multi-channel temporal convolutional network, used to allocate channels to the geographic data and historical route data in the training set to obtain geographic features and route features respectively. The physical model layer includes a perturbation correction propagation velocity model, used to output a dynamic velocity matrix based on geographic features and route features. The data-driven layer includes a temporal graph neural network, used to output a dynamic node feature matrix based on geographic features and route features. The model fusion layer is used to fuse the outputs of the physical model layer and the data-driven layer to obtain a fused velocity matrix. The monitoring feedback layer is used to calculate the confidence level based on the fused velocity matrix and historical traveling wave data. The output layer is used to output the predicted values ​​of traveling wave velocity and traveling wave time difference based on the fused velocity matrix and historical traveling wave data.

[0095] Specifically, as shown in Figure 3, it illustrates the model architecture provided in this embodiment:

[0096] Data input layer: A multi-channel one-dimensional dilated convolutional layer is used to distribute data of different modalities to different channels and add timestamp information to the data in each channel.

[0097] Geographic data and historical route data are input into the data input layer, respectively, and a multi-channel time-series convolutional network is used to allocate and form a geographic information channel and a historical route data channel. The geographic information channel extracts features from the geographic coordinate sequence of the route, learns the shape and direction of the route, and encodes environmental information along the route (vegetation type, building density) into vectors, concatenating them with other features to obtain geographic features. The historical route data channel uses a time-series model to learn patterns from historical load and meteorological data to obtain route features.

[0098] Physical Model Layer: Employing a perturbation-corrected propagation velocity model, this layer outputs a dynamic velocity matrix based on the temporal sequence of environmental parameters and line status characteristics. Specifically, it outputs the dynamic velocity matrix based on geographical and line characteristics, with dimensions representing either the number of frequency bands or the number of line segments, indicating the corrected velocity of different frequency bands on each line segment.

[0099] The data-driven layer employs the TGAT network to aggregate features of neighboring nodes based on calculated attention weights and update the features of the current node. The time-series graph represents the power distribution network as a graph structure, where nodes represent sensor locations or key nodes, and edges represent line information. Specifically, the TGAT network is input with graph-structured data. Nodes in the graph structure include sensor location coordinates from geographic data and fault labels from training samples, while edges include line segment materials and lengths from historical line data. The TGAT network outputs a dynamic node feature matrix containing spatiotemporally dependent node representations, with the matrix dimension equal to the number of nodes.

[0100] The attention weights of the TGAT network are represented as follows:

[0101]

[0102] in, This represents the attention weight from node i to node j at time step t; Let be the energy function of node pair (i,j) at time step t; ; Here is the weight matrix; || denotes the vector concatenation operation; To process the historical feature vector of node i using a convolutional neural network (CNN) Encode the time series feature vector; .

[0103] The data-driven layer uses temporal graph neural networks to dynamically capture the correlations between nodes, rather than relying on a fixed adjacency matrix, thus enabling more flexible handling of complex temporal relationships.

[0104] Model fusion layer: The outputs of the physical model layer and the data-driven layer are fused to obtain the fused velocity matrix.

[0105] Monitoring and feedback layer: Input the fusion velocity matrix and the traveling wave arrival time difference tensor. The traveling wave arrival time difference tensor can be calculated based on historical traveling wave data. The fusion velocity distribution is modeled using a Gaussian mixture model (GMM), the prediction confidence is calculated, and the confidence index and feedback control signal are output. When the confidence is lower than the preset threshold, the monitoring and feedback mechanism is triggered.

[0106] Output layer: Outputs dynamic traveling wave velocity field (traveling wave velocity distribution map with spatiotemporal resolution) and traveling wave time difference matrix. Specifically, it uses the fused velocity output by the model fusion layer as the traveling wave velocity of the current line, and calculates the traveling wave time difference between different sensors based on the traveling wave velocity.

[0107] In a further implementation, when training the traveling wave velocity model, the line dielectric loss, the electromagnetic coupling effect coefficient of adjacent lines, and the propagation velocity variation caused by the line dielectric loss are also considered. The propagation velocity variation caused by electromagnetic coupling is corrected by the electromagnetic coupling coefficient of the adjacent lines.

[0108] Specifically, real transmission lines suffer from dielectric loss, which causes signal attenuation and affects propagation speed. By incorporating dielectric loss into the model, the signal propagation characteristics can be more accurately reflected. Electromagnetic coupling exists between adjacent lines, affecting signal propagation speed. Introducing an electromagnetic coupling effect coefficient improves the accuracy of predictions.

[0109] Step S230. Train the initial network to obtain the trained traveling wave velocity model.

[0110] The above embodiments provide a method for intelligent fault location and model training in a distribution network. The intelligent fault location method collects data from multi-modal sensors in the distribution network to obtain traveling wave data at the time of a fault. The traveling wave data is input into a traveling wave velocity model to obtain predicted values ​​of traveling wave velocity and traveling wave time difference. Then, the Kalman filter method is used to obtain the fault location estimate, which can effectively improve the accuracy of fault location, especially for the fusion estimation of multi-frequency traveling wave data. For the model training of the traveling wave velocity model, a multi-channel convolutional network and a time-series graph neural network are used to construct the model architecture. By allocating channels for multi-modal training data and dynamically capturing node correlations using the time-series graph neural network, the complex temporal relationships of traveling wave data can be handled more flexibly.

[0111] The disclosed method can be implemented using various types of devices. Therefore, the present invention also discloses an apparatus corresponding to the above method, and specific embodiments are given below for detailed description.

[0112] As shown in Figure 4, one embodiment of the present invention provides a power distribution network fault intelligent location device, comprising:

[0113] The traveling wave data acquisition module 402 is used to acquire traveling wave data of distribution network faults. The traveling wave data includes the traveling wave time difference measurement value.

[0114] The traveling wave velocity modeling module 404 inputs traveling wave data into the traveling wave velocity model to obtain the predicted values ​​of traveling wave velocity and traveling wave time difference.

[0115] The fault location estimation module 406 is used to define the fault location, the predicted value of the traveling wave velocity, and the predicted value of the traveling wave time difference as state vectors, and the measured value of the traveling wave time difference as observation vectors. The fault location is estimated based on the state vector and the observation vector using the Kalman filter method.

[0116] As shown in Figure 5, one embodiment of the present invention provides a power distribution network fault intelligent location device, comprising:

[0117] A model training apparatus is provided for training a traveling wave velocity model of a first aspect, comprising:

[0118] The training set acquisition module 502 is used to collect historical traveling wave data, geographical data and historical line data of distribution network faults as a training set.

[0119] Model building module 504 is used to build the initial network of the traveling wave velocity model. The initial network specifically includes a data input layer, a physical model layer, a data-driven layer, a model fusion layer, a monitoring feedback layer, and an output layer.

[0120] The data input layer includes a multi-channel temporal convolutional network, used to allocate channels to the geographic data and historical route data in the training set to obtain geographic features and route features respectively. The physical model layer includes a perturbation correction propagation velocity model, used to output a dynamic velocity matrix based on geographic features and route features. The data-driven layer includes a temporal graph neural network, used to output a dynamic node feature matrix based on geographic features and route features. The model fusion layer is used to fuse the outputs of the physical model layer and the data-driven layer to obtain a fused velocity matrix. The monitoring feedback layer is used to calculate the confidence level based on the fused velocity matrix and historical traveling wave data. The output layer is used to output the predicted values ​​of traveling wave velocity and traveling wave time difference based on the fused velocity matrix and historical traveling wave data.

[0121] The model training module 506 is used to train the initial network to obtain the trained traveling wave velocity model.

[0122] The device provided in this application embodiment has the same implementation principle and technical effect as the aforementioned method embodiment. For the sake of brevity, any parts not mentioned in the device embodiment can be referred to the corresponding content in the aforementioned method embodiment.

[0123] The methods and related apparatus mentioned in the above embodiments are described with reference to the method flowcharts and / or structural diagrams provided in the embodiments of this application. Specifically, each block of the method flowcharts and / or structural diagrams, as well as combinations of blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing device, create means for implementing the functions specified in one or more blocks of the flowcharts and / or one or more blocks of the structural diagrams. These computer program instructions can also be stored in a computer-readable storage medium capable of directing a computer or other programmable data processing device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means that implement the functions specified in one or more blocks of the flowcharts and / or one or more blocks of the structural diagrams. These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, such that the instructions, which execute on the computer or other programmable apparatus, provide steps for implementing the functions specified in one or more flowcharts and / or one or more blocks in the structural diagram.

[0124] The following embodiments illustrate the application of this method to a computer device. It is understood that the computer device can be any device with computing and processing capabilities, including but not limited to servers or personal laptops. In one embodiment, the computer device can be an application server, which can be a server used to run the application under test.

[0125] Referring to Figure 6, a hardware block diagram of an electronic device is shown. This electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the present application described and / or claimed herein.

[0126] As shown in Figure 6, the electronic device includes: at least one processor 1, at least one communication interface 2, at least one memory 3, and at least one communication bus 4;

[0127] In this embodiment of the application, the number of processor 1, communication interface 2, memory 3, and communication bus 4 is at least one, and processor 1, communication interface 2, and memory 3 communicate with each other through communication bus 4;

[0128] Processor 1 may be a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement embodiments of the present invention.

[0129] Memory 3 may include high-speed RAM, and may also include non-volatile memory, such as at least one disk storage device;

[0130] The memory stores a program, which the processor can call. The program is used to implement the intelligent fault location method for power distribution networks provided in the foregoing embodiments, and / or the various processing flows of the model training method.

[0131] This invention also provides a readable storage medium storing a computer program thereon. When the computer program is executed by a processor, it implements the intelligent fault location method for distribution networks and / or the various processing flows of the model training method provided in any possible implementation of the above embodiments and / or in combination with the embodiments.

[0132] The invention has been described in particular detail above with respect to possible scenarios, and those skilled in the art will recognize that the invention can be practiced through other embodiments. Specific naming of components, capitalization of terms, attributes, data structures, or any other programming or structural aspects are not mandatory or important, and the mechanisms or features of implementing the invention may have different names, forms, or procedures. The system can be implemented through a combination of hardware and software (as described), entirely through hardware elements, or entirely through software elements. The specific division of functions among the various system components described herein is merely exemplary and not mandatory; rather, the functions performed by a single system component can be performed by multiple components, or the functions performed by multiple components can be performed by a single component.

[0133] Those skilled in the art should understand that the various steps of the disclosed methods can be implemented using general-purpose computing devices. They can be centralized on a single computing device or distributed across a network of multiple computing devices. Optionally, they can be implemented using device-executable program code, which can then be stored in a storage device for execution by the computing device. Alternatively, they can be fabricated as separate integrated circuit modules, or multiple modules or steps can be fabricated as a single integrated circuit module. Therefore, the embodiments disclosed in this invention are not limited to any specific hardware and software combination.

[0134] The programs (also referred to as programs, software, software applications, or code) executable by these computing devices include machine instructions of a programmable processor and can be implemented using high-level procedural and / or object-oriented programming languages, and / or assembly / machine languages. As used herein, the terms “machine-readable medium” and “computer-readable medium” refer to any computer program product, device, and / or apparatus (e.g., disk, optical disk, memory, programmable logic device (PLD)) used to provide machine instructions and / or data to a programmable processor, including machine-readable media that receive machine instructions as machine-readable signals. The term “machine-readable signal” refers to any signal used to provide machine instructions and / or data to a programmable processor.

[0135] Certain aspects of this invention include the process steps and instructions described herein in algorithmic form. It should be noted that the process steps and instructions of this invention can be implemented in software, firmware, and / or hardware, and when implemented in software, they can be downloaded, stored on various operating systems and operated from said platforms.

[0136] Those skilled in the art will understand that the structures shown in the figures are merely block diagrams of some structures related to the present application and do not constitute a limitation on the terminal device to which the present application is applied. Specific terminal devices may include more or fewer components than those shown in the figures, or combine certain components, or have different component arrangements.

[0137] In the description of this specification, the use of terms such as "one embodiment," "some embodiments," "example," "specific example," or "possible design," etc., refers to a specific feature, structure, material, or characteristic described in connection with that embodiment or example, which is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in a suitable manner in any one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0138] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0139] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for intelligent fault location in a distribution network, characterized in that, include: Collect traveling wave data of distribution network faults, the traveling wave data including traveling wave time difference measurements; Input the traveling wave data into the traveling wave velocity model to obtain the predicted values ​​of the traveling wave velocity and the traveling wave time difference; The fault location, predicted traveling wave velocity, and predicted traveling wave time difference are defined as a state vector, and the measured traveling wave time difference is defined as an observation vector. The fault location estimate is obtained using the Kalman filter method based on the state vector and the observation vector.

2. The method according to claim 1, characterized in that, The step of obtaining the fault location estimate based on the state vector and observation vector using the Kalman filter method includes: dividing the traveling wave data into several frequency bands and constructing a Kalman filter corresponding to each frequency band; calculating the fault location estimate of each frequency band based on the state vector and observation vector of each frequency band using the Kalman filter corresponding to each frequency band; and calculating the fault location estimate based on the fault location estimate of each frequency band.

3. The method according to claim 2, characterized in that, The calculation of the fault location estimate based on the fault location estimate of each frequency band includes: defining the fault location estimate of each frequency band and its confidence level as an evidence body; determining the evidence body corresponding to each frequency band; the confidence level of the fault location estimate of each frequency band is determined according to the output of the Kalman filter; and fusing the evidence bodies using the Dempster combination rule to obtain the fault location estimate and its confidence level.

4. A model training method, characterized in that, The method for training the traveling wave velocity model as described in any one of claims 1 to 3 includes: collecting historical traveling wave data, geographical data, and historical line data of distribution network faults as a training set; constructing an initial network for the traveling wave velocity model, wherein the initial network specifically includes a data input layer, a physical model layer, a data-driven layer, a model fusion layer, a monitoring feedback layer, and an output layer; the data input layer includes a multi-channel temporal convolutional network for channel allocation of geographical data and historical line data in the training set to obtain geographical features and line features respectively; the physical model layer includes a disturbance correction propagation velocity model for outputting a dynamic velocity matrix based on the geographical features and line features; the data-driven layer includes a temporal graph neural network for outputting a dynamic node feature matrix based on the geographical features and line features; the model fusion layer is used to fuse the outputs of the physical model layer and the data-driven layer to obtain a fused velocity matrix; the monitoring feedback layer is used to calculate confidence based on the fused velocity matrix and historical traveling wave data; the output layer is used to output predicted values ​​of traveling wave velocity and predicted values ​​of traveling wave time difference based on the fused velocity matrix and historical traveling wave data; and training the initial network to obtain a trained traveling wave velocity model.

5. The method according to claim 1, characterized in that, The data-driven layer includes the TGAT network. The step of outputting a dynamic node feature matrix based on the geographical features and route features includes: using geographical data and fault labels of training set samples as nodes of a graph structure, historical route data as edges of a graph structure, and using the graph structure as input to the TGAT network to obtain the dynamic node feature matrix.

6. The method according to claim 1, characterized in that, The monitoring feedback layer is used to calculate the confidence level based on the fusion velocity matrix and historical traveling wave data, including: using a Gaussian mixture model (GMM) to calculate the confidence level based on the fusion velocity matrix and historical traveling wave data.

7. A smart fault location device for power distribution networks, characterized in that, include: The traveling wave data acquisition module is used to acquire traveling wave data of distribution network faults, and the traveling wave data includes traveling wave time difference measurements. The traveling wave velocity modeling module inputs the traveling wave data into the traveling wave velocity model to obtain the predicted values ​​of the traveling wave velocity and the traveling wave time difference. The fault location estimation module is used to define the fault location, the predicted value of the traveling wave velocity, and the predicted value of the traveling wave time difference as the state vector, and the measured value of the traveling wave time difference as the observation vector. The Kalman filter method is used to obtain the fault location estimate based on the state vector and the observation vector.

8. A model training device, characterized in that, The system for training a traveling wave velocity model includes: a training set acquisition module for collecting historical traveling wave data, geographical data, and historical line data from distribution network faults as the training set; a model building module for constructing an initial network for the traveling wave velocity model, specifically comprising a data input layer, a physical model layer, a data-driven layer, a model fusion layer, a monitoring feedback layer, and an output layer; the data input layer includes a multi-channel temporal convolutional network for channel allocation of the geographical data and historical line data in the training set to obtain geographical features and line features respectively; the physical model layer includes a disturbance correction propagation velocity model for... The system outputs a dynamic velocity matrix based on geographical and route features. The data-driven layer includes a time-series graph neural network, which outputs a dynamic node feature matrix based on the geographical and route features. The model fusion layer fuses the outputs of the physical model layer and the data-driven layer to obtain a fused velocity matrix. The monitoring feedback layer calculates confidence based on the fused velocity matrix and historical traveling wave data. The output layer outputs predicted traveling wave velocity and predicted traveling wave time difference based on the fused velocity matrix and historical traveling wave data. The model training module trains the initial network to obtain a trained traveling wave velocity model.

9. An electronic device, characterized in that, It includes a memory storing computer-executable instructions and a processor, which, when executed by the processor, causes the device to perform the intelligent fault location method for distribution networks as described in any one of claims 1 to 3, and / or the model training method as described in any one of claims 4 to 6.

10. A readable storage medium, characterized in that, It stores a computer-executable program that, when executed, can implement the intelligent fault location method for distribution networks as described in any one of claims 1 to 3, and / or the model training method as described in any one of claims 4 to 6.

Images