Power distribution communication network fault positioning method and system based on deep learning

Abstract

The invention discloses a power distribution communication network fault positioning method and system based on deep learning, and the method comprises the steps: carrying out the fitting processing of obtained time series data of an optical fiber link, and obtaining an actual power attenuation curve; calculating an absolute difference value between the actual power attenuation curve and the theoretical power attenuation curve to obtain standardized difference data; inputting the standardized difference data into a nonlinear attenuation model to obtain nonlinear change characteristic data; inputting the topological characteristic data and the nonlinear change characteristic data of the power distribution network into a fault identification model to obtain a fault probability value of each optical fiber link; comparing the fault probability value with a preset fault probability threshold value to obtain a potential fault link, and performing feature extraction and abrupt change node identification on the potential fault link to obtain attenuation abrupt change node data; and processing the attenuation abrupt change node data, and determining fault point data based on the obtained stress value data. According to the method provided by the invention, the loss fault in the optical fiber network can be quickly positioned.

Description

Technical Field

[0001] The present invention relates to the technical field of fault analysis, and in particular to a fault location method and system for a distribution communication network based on deep learning. Background Art

[0002] As a key infrastructure of the smart grid, the distribution communication network undertakes the core tasks of power system data transmission, equipment control, and status monitoring. Among them, the optical fiber link has become the main transmission medium of the distribution communication network due to its advantages of high bandwidth and anti-interference. However, the optical fiber link is easily affected by external force extrusion, temperature change, connector aging and other factors, resulting in the attenuation or even interruption of the optical signal power, which seriously threatens the safe and stable operation of the distribution network.

[0003] The existing optical fiber fault location methods rely on hardware devices such as optical time domain reflectometers to detect fault points by analyzing the attenuation curve of the backscattered optical signal. The single detection period of this method for fault location is as long as dozens of seconds to several minutes, while the fault evolution speed of the distribution communication network far exceeds the data update frequency, resulting in the inability to locate the fault in time when the fault occurs. Summary of the Invention

[0004] The present invention provides a fault location method and system for a distribution communication network based on deep learning to solve the technical problem of how to improve the existing fault location method for a distribution communication network, and achieve the effect of reducing the fault handling time of the distribution communication network.

[0005] To solve the above technical problem, an embodiment of the present invention provides a fault location method for a distribution communication network based on deep learning, including: Obtaining the optical power data of the optical fiber link in the target distribution network, calculating the optical power difference between adjacent sampling points in the optical power data, and generating time series data; Performing fitting processing on the time series data to obtain an actual power attenuation curve, and calculating the absolute difference between the actual power attenuation curve and the theoretical power attenuation curve to obtain standardized difference data; Inputting the standardized difference data into a pre-constructed non-linear attenuation model to obtain non-linear change characteristic data reflecting the optical signal power attenuation in the optical fiber link; Performing feature extraction on the physical topology structure of the target distribution network to obtain distribution network topology feature data, and inputting the distribution network topology feature data and the non-linear change characteristic data into a pre-constructed fault identification model to obtain the fault probability values of each optical fiber link; Comparing the fault probability value with a preset fault probability threshold to obtain potential fault links, and sequentially performing feature extraction and mutation node identification on the potential fault links to obtain attenuation mutation node data; Process the attenuation mutation node data according to the stress inversion method, and determine the fault point data in the target distribution network based on the obtained stress value data.

[0006] As one of the preferred solutions, the obtaining of the optical power data of the optical fiber link in the target distribution network, calculating the optical power difference between adjacent sampling points in the optical power data, and generating time series data includes: Real-time collect the original optical power data of the optical fiber link in the target distribution network, and preprocess the original optical power data; Perform time sorting on the preprocessed original optical power data to obtain an optical power sequence data set; Calculate the optical power difference between adjacent sampling points in the optical power sequence data set to obtain difference data; wherein if the optical power difference exceeds the preset optical power fluctuation threshold, it is marked as an abnormal data point; Screen the non-abnormal data points in the difference data to obtain time series data.

[0007] As one of the preferred solutions, before performing fitting processing on the time series data, it further includes: Perform linear regression processing on the time series data to obtain optical power trend data, and calculate the mean square error between the theoretical stress characteristic data of the optical fiber link and the optical power trend data; If the mean square error exceeds the preset error threshold, perform fitting processing on the time series data.

[0008] As one of the preferred solutions, the calculating the mean square error between the theoretical stress characteristic data of the optical fiber link and the optical power trend data includes: Generate stress distribution data according to the obtained stress distribution parameters of the optical fiber link, and perform frequency domain feature extraction on the stress distribution data by using Fourier transform to obtain the theoretical stress characteristic data; Perform linear regression processing on the time series data by using the least squares method to obtain an optical power trend function, and obtain the trend function values at each sampling time point according to the optical power trend function to constitute the optical power trend data; Perform time alignment processing on the theoretical stress characteristic data and the optical power trend data, and calculate the mean square error between the time-aligned theoretical stress characteristic data and the optical power trend data.

[0009] As one of the preferred solutions, the performing fitting processing on the time series data to obtain an actual power attenuation curve, and calculating the absolute difference between the actual power attenuation curve and the theoretical power attenuation curve to obtain standardized difference data includes: Perform curve fitting on the time series data based on the cubic spline interpolation method to obtain the actual power attenuation curve; Construct a theoretical power attenuation curve according to the physical parameters of the optical fiber link; Calculate the absolute difference at each sampling point on the time axis of the actual power attenuation curve and the theoretical power attenuation curve to obtain an absolute difference sequence reflecting the degree of attenuation anomaly; Perform normalization processing on the absolute difference sequence to obtain the standardized difference data.

[0010] As one of the preferred solutions, the physical topology structure of the target distribution network is characterized to obtain distribution network topology feature data, including: Based on the obtained physical connection relationship of the optical fiber link, construct a node adjacency matrix and a link attribute table; Sum each row of the node adjacency matrix to obtain a node degree vector reflecting the importance of nodes; Calculate link weight data according to the link attribute table, and construct a link weight matrix based on the link weight data; Concatenate the node degree vector and the link weight matrix to obtain the distribution network topology feature data.

[0011] As one of the preferred solutions, the failure probability value is compared with a preset failure probability threshold to obtain potential failure links, and feature extraction and mutation node identification are sequentially performed on the potential failure links to obtain attenuation mutation node data, including: Compare the failure probability value with the preset failure probability threshold, filter out the links whose failure probability value is greater than or equal to the preset failure probability threshold, and generate a list of potential failure links; Obtain the optical power sequence data corresponding to the links in the failure link list, and convert the optical power sequence data into a two-dimensional matrix; Concatenate the two-dimensional matrix with the standardized difference data and the non-linear change feature data to obtain a fusion feature tensor; and input the fusion feature tensor into a pre-constructed deep residual network for feature extraction to obtain attenuation mutation feature data; Perform clustering analysis on the attenuation mutation feature data to obtain the attenuation mutation node data with a changed attenuation trend.

[0012] As one of the preferred solutions, the attenuation mutation node data is processed according to the stress inversion method, and the fault point data in the target distribution network is determined based on the obtained stress value data, including: Construct a stress attenuation relationship model according to the obtained stress sensitive parameters, material stress threshold and link physical length of the optical fiber link; Extract the power attenuation amount corresponding to each of the attenuation mutation node data and the physical length of the link where it is located, and input the power attenuation amount and the physical length into the stress attenuation relationship model to obtain a stress value data set; Compare each stress value data in the stress value data set with the material stress threshold respectively, and screen the comparison results to obtain a high-stress mutation node list; Determine the fault point data in the target distribution network based on the high-stress mutation node list.

[0013] As one of the preferred solutions, the determining the fault point data in the target distribution network based on the high-stress mutation node list includes: Construct a candidate fault point set based on each node in the high-stress mutation node list and its adjacent nodes; Obtain the coordinate data of each node in the candidate fault point set, and calculate the relative fault distance between each high-stress mutation node and its adjacent nodes based on the coordinate data; Perform multi-condition screening according to the stress value and the relative fault distance of each obtained high-stress mutation node, and obtain the fault point data according to the results of the multi-condition screening.

[0014] Another embodiment of the present invention provides a fault location system for a distribution communication network based on deep learning, including: An acquisition module, configured to acquire the optical power data of the optical fiber link in the target distribution network, calculate the optical power difference between adjacent sampling points in the optical power data, and generate time series data; A calculation module, configured to perform fitting processing on the time series data to obtain an actual power attenuation curve, and calculate the absolute difference between the actual power attenuation curve and the theoretical power attenuation curve to obtain standardized difference data; A first extraction module, configured to input the standardized difference data into a pre-constructed non-linear attenuation model to obtain non-linear change characteristic data reflecting the optical signal power attenuation in the optical fiber link; A second extraction module, configured to extract features of the physical topology structure of the target distribution network to obtain distribution network topology feature data, and input the distribution network topology feature data and the non-linear change characteristic data into a pre-constructed fault identification model to obtain the fault probability values of each optical fiber link; An identification module, configured to compare the fault probability values with a preset fault probability threshold to obtain potential fault links, and perform feature extraction and mutation node identification on the potential fault links in sequence to obtain attenuation mutation node data; A positioning module, configured to process the attenuation mutation node data according to a stress inversion method, and determine the fault point data in the target distribution network based on the obtained stress value data.

[0015] Compared with the prior art, the beneficial effects of the embodiments of the present invention are at least one of the following: 1) The present invention realizes the accurate capture of non-linear fault features and multi-dimensional data fusion through a deep learning model, breaks through the dependence of traditional methods on linear attenuation features, and significantly improves the fault recognition accuracy in complex scenarios. Specifically, the present invention uses a non-linear attenuation model to deeply process the standardized difference data, effectively extracts the non-linear change features of the optical signal power attenuation, and constructs a fault recognition model in combination with the topological feature data of the distribution network, which can accurately identify progressive faults such as micro-bends and connector aging that are easily missed by traditional methods, improve the early fault detection rate by more than 40%, and at the same time reduce the misjudgment rate in multi-branch networks through topological structure analysis; 2) The present invention constructs a full-process intelligent positioning closed loop of "data acquisition - model inference - stress inversion", realizes the automatic processing from fault probability assessment to accurate positioning, and greatly improves the fault positioning efficiency and engineering practicability. Specifically, the present invention converts the attenuation mutation node data into specific stress values through a stress inversion method, accurately locks the fault point position in combination with the material stress threshold, avoids the time-consuming defect of traditional OTDR segment-by-segment detection, shortens the positioning time by more than 60% in complex topologies such as ring networks, and can adapt to network topology changes in real time, providing direct guidance for the rapid repair of the distribution communication network, and significantly reducing the operation and maintenance cost and fault handling time. Description of the Drawings

[0016] Figure 1 is a schematic flow chart of a method for fault location of a distribution communication network based on deep learning in one embodiment of the present invention; Figure 2 is a schematic diagram of a system for fault location of a distribution communication network based on deep learning in one embodiment of the present invention; Reference Signs: Among them, 11, acquisition module; 12, calculation module; 13, first extraction module; 14, second extraction module; 15, recognition module; 16, positioning module. Detailed Embodiments

[0017] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. The purpose of providing these embodiments is to make the disclosure of the present invention more thorough and comprehensive. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts belong to the scope of protection of the present invention.

[0018] In the description of the present invention, the terms "first", "second", "third", etc. are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly specifying the quantity of the indicated technical features. Thus, the features defined with "first", "second", "third", etc. may explicitly or implicitly include one or more of such features. In the description of the present invention, unless otherwise stated, the meaning of "a plurality" is two or more.

[0019] In the description of the present invention, it should be noted that unless otherwise clearly defined and limited, the terms "installed", "connected", "connected" should be understood in a broad sense. For example, it may be a fixed connection, a detachable connection, or an integral connection; it may be a mechanical connection or an electrical connection; it may be directly connected, or indirectly connected through an intermediate medium, and it may be the communication inside two components. The terms "vertical", "horizontal", "left", "right", "up", "down" and similar expressions used herein are only for the purpose of illustration, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus cannot be understood as a limitation of the present invention. The term "and / or" used herein includes any and all combinations of one or more of the related listed items. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0020] In the description of the present invention, it should be noted that unless otherwise defined, all technical and scientific terms used in the present invention have the same meaning as commonly understood by those of ordinary skill in the technical field to which this technology belongs. The terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0021] An embodiment of the present invention provides a method for fault location of a distribution communication network based on deep learning. Specifically, please refer to Figure 1 , Figure 1 which shows a schematic flow chart of the method for fault location of a distribution communication network based on deep learning in one of the embodiments of the present invention, and it includes steps S1 - step S6: S1: Obtain the optical power data of the optical fiber link in the target distribution network, calculate the optical power difference between adjacent sampling points in the optical power data, and generate time series data; Preferably, in an embodiment of the present invention, the obtaining the optical power data of the optical fiber link in the target distribution network, calculating the optical power difference between adjacent sampling points in the optical power data, and generating time series data includes: Real-time collect the original optical power data of the optical fiber link in the target distribution network, and preprocess the original optical power data; Perform time sorting on the preprocessed original optical power data to obtain an optical power sequence data set; Calculate the optical power difference between adjacent sampling points in the optical power sequence data set to obtain difference data; wherein if the optical power difference exceeds a preset optical power fluctuation threshold, it is marked as an abnormal data point; Screen the non-abnormal data points in the difference data to obtain time series data.

[0022] Among them, the optical power data refers to the monitoring data of the energy intensity of the optical signal in the optical fiber link, usually in decibel-milliwatt (dBm) as the unit, which reflects the signal transmission quality and is the core index for judging the optical fiber attenuation state. The time series data refers to a one-dimensional data sequence arranged in time order. In this embodiment, it is an ordered set of optical power differences between adjacent sampling points, which is used to depict the dynamic change trend of the optical power over time. The preprocessing includes noise filtering and outlier removal of the original optical power data to ensure data reliability. Common methods include filtering algorithms (such as median filtering, S-G filtering) and outlier detection. The preset optical power fluctuation threshold is a critical value set according to the power fluctuation range during the normal operation of the optical fiber link, which is used to identify abnormal data points that significantly deviate from the normal state (such as sudden power changes caused by sudden interference or equipment failures).

[0023] It should be noted that the original optical power data may be affected by environmental noise (such as electromagnetic interference, temperature fluctuations) and equipment errors. Direct use will lead to misjudgment in subsequent model analysis, and invalid data needs to be removed through preprocessing. Moreover, the time series of the optical power difference can reflect the dynamic change of signal attenuation and is the basis for identifying non-linear attenuation (such as mutations, periodic fluctuations). Abnormal data points will interfere with feature extraction, and threshold screening is required to ensure the effectiveness of the sequence.

[0024] Specifically, in this embodiment, optical power data of the optical fiber link is real-time collected through devices such as an optical time domain reflectometer (OTDR) and an ultra-weak fiber Bragg grating array (uwFBG), and the sampling timestamp (with a precision up to the microsecond level) and spatial position information are synchronously recorded.

[0025] The median filter (3×3 window) is used to smooth the raw data, suppress high-frequency noise, and improve the signal-to-noise ratio to more than 20dB. The mean and standard deviation of the data are calculated based on the 3σ principle, and the points whose absolute values ​​exceed the mean ±3σ are marked as abnormal; at the same time, double screening is combined with physical thresholds (such as optical power mutation exceeding ±0.5dB) to ensure that the abnormal data elimination rate reaches more than 99%.

[0026] Arrange the preprocessed data in ascending order according to the sampling timestamp to generate an optical power sequence data set; subtract the power values ​​of adjacent sampling points to obtain difference data reflecting instantaneous power changes. Set a preset fluctuation threshold (which can be dynamically adjusted according to historical data), remove abnormal data points whose difference exceeds the threshold, retain continuous and stable non-abnormal data points, and finally generate time series data for subsequent analysis.

[0027] S2: performing fitting processing on the time series data to obtain an actual power attenuation curve, and calculating an absolute difference between the actual power attenuation curve and a theoretical power attenuation curve to obtain standardized difference data; Preferably, in one embodiment of the present invention, before the fitting process is performed on the time series data, the method further includes: Performing linear regression processing on the time series data to obtain optical power trend data, and calculating the mean square error between the theoretical stress characteristic data of the optical fiber link and the optical power trend data; If the mean square error exceeds a preset error threshold, fitting processing is performed on the time series data.

[0028] Preferably, in one embodiment of the present invention, the calculating the mean square error between the theoretical stress characteristic data of the optical fiber link and the optical power trend data comprises: Generate stress distribution data according to the acquired stress distribution parameters of the optical fiber link, and extract frequency domain features of the stress distribution data using Fourier transform to obtain the theoretical stress feature data; Performing linear regression processing on the time series data using the least squares method to obtain an optical power trend function, and obtaining a trend function value at each sampling time point according to the optical power trend function to form the optical power trend data; The theoretical stress characteristic data and the optical power trend data are time-aligned, and a mean square error between the theoretical stress characteristic data and the optical power trend data after time alignment is calculated.

[0029] It should be noted that the optical power change of the optical fiber link may be affected by various factors, including the regular trend caused by normal stress (such as slow environmental temperature changes) and the abnormal mutation caused by faults. By first extracting the overall trend of the optical power through linear regression (such as long-term attenuation or slight fluctuations), and comparing it with the expected trend (theoretical stress characteristic data) under the influence of theoretical stress, it can be judged whether the trend in the current data conforms to the normal physical law. If the deviation (mean square error) between the two exceeds the preset threshold, it indicates that the actual trend may contain abnormal changes due to non-stress factors (such as poor contact, micro-bending of the optical fiber), and it is necessary to further mine more complex non-linear characteristics through fitting processing to avoid the interference of normal stress fluctuations on fault detection.

[0030] Specifically, in this embodiment, stress distribution data is generated according to the laying environment parameters of the optical fiber link (such as the preset temperature change range, allowable tensile value, etc.), and then the frequency characteristics of these stress data are analyzed through Fourier transform (such as whether there are regular fluctuations caused by periodic temperature changes) to obtain the frequency domain characteristics that the optical power change should have under the influence of theoretical stress (i.e., theoretical stress characteristic data).

[0031] The least squares method is used to process the time series data to find a straight line or a low-degree curve (trend function) that can best fit the data. The output value of this function (trend function value) is the optical power trend data, representing the long-term change trend of the data after removing short-term fluctuations.

[0032] The theoretical stress characteristic data and the optical power trend data are made one-to-one correspondence (time alignment) according to the time points, the difference between the two at each time point is calculated, and then the average of the squares of all differences (mean square error) is calculated. If this error exceeds the preset threshold, it indicates that the actual trend deviates greatly from the normal trend under the influence of theoretical stress, and more complex fitting processing (such as polynomial fitting, non-linear fitting) needs to be carried out on the time series data to capture the abnormal characteristics related to possible faults.

[0033] Preferably, in an embodiment of the present invention, the fitting process of the time series data to obtain the actual power attenuation curve and calculating the absolute difference between the actual power attenuation curve and the theoretical power attenuation curve to obtain the standardized difference data includes: Performing curve fitting on the time series data based on the cubic spline interpolation method to obtain the actual power attenuation curve; Constructing a theoretical power attenuation curve according to the physical parameters of the optical fiber link; Calculating the absolute difference between each sampling point on the time axis of the actual power attenuation curve and the theoretical power attenuation curve to obtain an absolute difference sequence reflecting the degree of attenuation abnormality; Normalizing the absolute difference sequence to obtain the standardized difference data.

[0034] Among them, the cubic spline interpolation method is a mathematical method used to construct a smooth curve based on known discrete data points (here, the optical power differences in time series data). The cubic spline interpolation method uses a cubic polynomial to fit between every two adjacent data points, ensuring that the curve has continuous first and second derivatives throughout the interval. The resulting curve is both smooth and can well approximate the original data.

[0035] The actual power attenuation curve is the curve obtained by curve fitting the time series data, reflecting the actual attenuation change of the optical power over time in the optical fiber link. The theoretical power attenuation curve is constructed based on the physical parameters of the optical fiber link (such as the material, length, and bending degree of the optical fiber) according to physical principles and theoretical models, representing the attenuation change that the optical power should have over time under ideal and fault-free conditions.

[0036] It should be noted that in the actual operation of the optical fiber link, the attenuation of the optical power will be affected by various factors, such as environmental temperature, external force extrusion, and connector loosening, resulting in a difference between the actual attenuation situation and the theoretical situation. By fitting the actual power attenuation curve and comparing it with the theoretical power attenuation curve, these differences can be found.

[0037] Although the absolute difference sequence can reflect the degree of attenuation anomaly, since its numerical range may vary for different optical fiber links or different time periods, it is not convenient for direct analysis and comparison. The standardized difference data obtained after normalization can measure the attenuation anomaly on a unified scale, providing a more reliable basis for subsequent judgment of whether there is a fault in the optical fiber link and the severity of the fault.

[0038] Specifically, in this embodiment, each data point in the time series data is regarded as a discrete point in the two-dimensional plane of time-optical power difference. Then, the cubic spline interpolation method is used to determine a cubic polynomial between every two adjacent data points. The coefficients of this cubic polynomial are calculated by satisfying certain conditions (such as the continuity of the first and second derivatives at the connection of adjacent intervals). Finally, the cubic polynomials of all adjacent intervals are combined to form a smooth curve, that is, the actual power attenuation curve. The actual power attenuation curve can well fit the time series data and show the actual change trend of the optical power attenuation over time.

[0039] Collect the physical parameters of the optical fiber link, including the type of optical fiber (such as single-mode fiber, multi-mode fiber), length, loss coefficient (related to the optical fiber material), bending radius, etc. According to the physical principles and theoretical models of optical fiber transmission (such as the attenuation of optical fiber is proportional to the length), establish a functional relationship between time and optical power attenuation. Using this functional relationship, calculate the theoretical attenuation values of optical power at different time points, and connecting these theoretical attenuation values gives the theoretical power attenuation curve. The theoretical power attenuation curve represents the attenuation of optical power over time under ideal conditions without external interference and faults.

[0040] Taking the time axis as a reference, subtract the optical power attenuation values corresponding to the actual power attenuation curve and the theoretical power attenuation curve at each sampling point. Take the absolute value of the subtraction result to form a sequence of absolute differences. Each value in the sequence of absolute differences reflects the magnitude of the difference between the actual attenuation and the theoretical attenuation at that sampling point. The greater the difference, the more obvious the attenuation anomaly at that point.

[0041] For each value in the difference sequence, perform normalization calculations. Combine all the normalized values to form standardized difference data. The standardized difference data can more intuitively reflect the degree of attenuation anomaly at each sampling point, facilitating subsequent analysis and processing.

[0042] S3: Input the standardized difference data into a pre-constructed non-linear attenuation model to obtain non-linear change characteristic data reflecting the optical signal power attenuation in the optical fiber link; The standardized difference data is the data of the degree of optical power attenuation anomaly after preprocessing and normalization, eliminating the influence of dimension and link differences. The numerical range is within [0,1] or in the interval with a mean of 0 and a standard deviation of 1, and it is the core input data reflecting the degree of deviation of the actual attenuation from the theoretical model.

[0043] The non-linear attenuation model is a neural network model based on deep learning, which can capture non-linear and non-stationary change characteristics (such as mutations, exponential decay, periodic fluctuations) in optical power attenuation. Different from the fitting of linear trends by traditional linear models (such as the least squares method), it is suitable for feature extraction in complex fault scenarios.

[0044] The purpose of this step is to automatically extract the hidden fault characteristics in the standardized difference data, avoid the one-sidedness of manually designed features, and at the same time convert the abstract numerical sequence into interpretable physical features, providing fault characterization information for subsequent topological structure analysis to improve the positioning accuracy.

[0045] Specifically, convert the standardized difference data into a format suitable for model input; then construct a non-linear attenuation model. Taking the CNN-LSTM combined model as an example, first use the CNN layer to extract local features, then use the LSTM layer to capture long-term dependence features, and finally map the features into a non-linear change feature vector containing information such as the probability of mutation position, attenuation mode label, and feature confidence through a fully connected layer; during model training, use the historical fault data set and enhance the generalization ability by artificially injecting fault simulation data, and optimize using a multi-task loss function combining cross-entropy loss and mean square error; during inference, input the real-time standardized difference data into the trained model to output non-linear change feature data such as "a certain type of mutation is detected at a certain position with a certain confidence".

[0046] S4: Extract features from the physical topology structure of the target distribution network to obtain distribution network topology feature data, and input the distribution network topology feature data and the non-linear change feature data into a pre-constructed fault identification model to obtain the fault probability values of each optical fiber link; Preferably, in an embodiment of the present invention, the extracting features from the physical topology structure of the target distribution network to obtain distribution network topology feature data includes: Based on the obtained physical connection relationship of the optical fiber link, construct a node adjacency matrix and a link attribute table; Sum each row of the node adjacency matrix to obtain a node degree vector reflecting the importance of the nodes; Calculate link weight data according to the link attribute table, and construct a link weight matrix based on the link weight data; Perform splicing processing on the node degree vector and the link weight matrix to obtain the distribution network topology feature data.

[0047] The physical topology structure of the target distribution network refers to the actual connection method and layout among various electrical devices (such as transformers, switches, lines, etc.) in the target distribution network, which describes the path and structure of power transmission in the network.

[0048] The node adjacency matrix is a two-dimensional matrix used to represent the connection relationship between each node in the distribution network. The rows and columns of the matrix correspond to the nodes in the distribution network respectively. If there is a direct connection between two nodes, the value at the corresponding position in the matrix is 1, otherwise it is 0. The link attribute table records the relevant attribute information of each link (such as transmission line) in the distribution network, such as the length, capacity, resistance, etc. of the link.

[0049] The node degree vector is obtained by summing each row of the node adjacency matrix. Each element in the vector represents the number of connections of the corresponding node with other nodes, reflecting the importance of the node in the distribution network. The more connections, the more important the node. The link weight data is calculated based on the information in the link attribute table and is a value used to measure the importance or transmission capacity, etc. of each link in the distribution network. The link weight matrix is a matrix constructed from the link weight data, which further describes the relative importance between links in the distribution network.

[0050] It should be noted that it is not comprehensive enough to judge whether an optical fiber link is faulty only relying on the non-linear change characteristic data of the optical signal power attenuation. The physical topology structure of the distribution network has an important impact on the occurrence and propagation of faults. Nodes and links in different positions have different importance in the network, and the impact on the entire distribution network when a fault occurs is also different. By extracting the topological characteristic data of the distribution network and combining it with the non-linear change characteristic data and inputting them into the fault identification model, the topological structure and the optical signal attenuation situation can be comprehensively considered to more accurately judge the fault probability of each optical fiber link and improve the accuracy and reliability of fault identification.

[0051] Specifically, in this embodiment, first, obtain the physical connection relationship of the optical fiber links in the target distribution network, which can be obtained by referring to materials such as the design drawings and equipment ledgers of the distribution network. Then, construct a node adjacency matrix according to these connection relationships. Suppose there are n nodes, then the node adjacency matrix is an n×n matrix. If there is a direct connection between node i and node j, the element values in the i-th row and j-th column and the j-th row and i-th column of the matrix are 1; if there is no connection, it is 0. At the same time, sort out the relevant attribute information of the optical fiber links, such as the length, transmission capacity, resistance, etc. of the links, and record this information in the link attribute table.

[0052] Perform a summation operation on each row of the node adjacency matrix. For example, for the i-th row, add all the elements in this row, and the sum obtained is the degree of node i. Arrange the degrees of all nodes in order to form a vector, that is, the node degree vector. This vector can intuitively reflect the importance of each node in the distribution network. The larger the degree, the more nodes the node is connected to, and the more critical it is in the network.

[0053] According to the information in the link attribute table, select a suitable method to calculate the weight of each link. For example, the weight can be comprehensively calculated according to factors such as the length and transmission capacity of the link. Links with shorter lengths and larger transmission capacities may have higher weights. Construct the calculated link weights into a matrix according to the connection relationship between nodes, that is, the link weight matrix. The element in the i-th row and j-th column of the matrix represents the weight of the link between node i and node j.

[0054] Concatenate the node degree vector and the link weight matrix. The node degree vector can be added as a new column to the link weight matrix, or other suitable concatenation methods can be adopted. The data set obtained after concatenation is the topological feature data of the distribution network, which synthesizes information such as the importance of nodes and the relative importance of links.

[0055] S5: Compare the fault probability value with a preset fault probability threshold to obtain potential fault links, and sequentially perform feature extraction and mutation node identification on the potential fault links to obtain attenuation mutation node data; Preferably, in an embodiment of the present invention, the comparing the fault probability value with a preset fault probability threshold to obtain potential fault links, and sequentially performing feature extraction and mutation node identification on the potential fault links to obtain attenuation mutation node data includes: Compare the fault probability value with a preset fault probability threshold, and filter out the links whose fault probability value is greater than or equal to the preset fault probability threshold to generate a list of potential fault links; Obtain the optical power sequence data corresponding to the links in the fault link list, and convert the optical power sequence data into a two-dimensional matrix; Concatenate the two-dimensional matrix with the standardized difference data and the non-linear change feature data to obtain a fusion feature tensor; and input the fusion feature tensor into a pre-constructed deep residual network for feature extraction to obtain attenuation mutation feature data; Perform clustering analysis on the attenuation mutation feature data to obtain the attenuation mutation node data with a changed attenuation trend.

[0056] Among them, the fault probability value is the possibility value (between 0 and 1) of each optical fiber link failing calculated by the fault identification model. The higher the value, the greater the possibility of failure. The preset fault probability threshold is a critical value (such as 0.7) manually set to judge whether a link is likely to fail, and is used to screen high-risk links to avoid indiscriminate analysis of all links.

[0057] The potential fault link is an optical fiber link whose fault probability value exceeds the preset threshold and is the object of key investigation in the follow-up. The deep residual network (ResNet) is a deep learning model that solves the problem of gradient disappearance in deep networks through residual connections and is good at extracting multi-level abstract features of data (such as short-term mutations and long-term trends). The attenuation mutation node data contains information such as node numbers, attenuation anomaly degrees, and spatial positions, and is used to identify the specific nodes where significant attenuation changes occur in the optical fiber link.

[0058] It should be noted that this step quickly filters low-probability links through threshold comparison, concentrates computing resources on potential fault links with high fault probability, avoids "needle-in-a-haystack" type troubleshooting, and improves efficiency. At the same time, it narrows the fault scope from "potential links" to "specific nodes", providing direct clues for subsequent precise positioning of fault points, forming a progressive troubleshooting logic of "probability screening → feature positioning → node identification".

[0059] Specifically, in this embodiment, the fault probability values ​​of each link output by the fault identification model (such as [0.65, 0.82, 0.51, ...]) are compared with a preset threshold (such as 0.7), and links with probability ≥ threshold (such as the second link 0.82 ≥ 0.7) are retained to generate a list of potential fault links (such as ["Link_03", "Link_17"]). The list contains basic information such as link number, region, and historical fault records, which is used to quickly locate the object of investigation, such as giving priority to "Link_03" with frequent historical faults.

[0060] The optical power data (including timestamp and power value) of each potential faulty link is arranged according to the "time-power" dimension to form a two-dimensional matrix (number of rows = number of sampling points, number of columns = 1). For example, 1000 sampling points form a 1000×1 matrix to intuitively display the power change trend over time. The two-dimensional matrix is ​​spliced ​​with the standardized difference data (reflecting the degree of abnormality) and nonlinear change features (such as mutation location probability and fault mode label) in the channel dimension to form a fused feature tensor (such as a dimension of 1000×3), so that the model can simultaneously analyze the original signal, the degree of abnormality and the fault mode.

[0061] The deep residual network of the present invention adopts 34-layer ResNet, which includes multiple residual blocks. Each residual block extracts multi-layer features through "convolution-batch normalization-ReLU-residual connection": Shallow layer: capture short-term power fluctuations (such as mutations within 1 second); Deep: Learn long-term decay trends (such as decay acceleration that lasts for 1 hour).

[0062] Input and output: Input fusion feature tensor, output attenuation mutation feature vector (such as "mutation occurrence time point" and "abnormal energy proportion" and other information), for example, identify "the attenuation mutation energy proportion at the 500th time point (corresponding to 1500 meters of the link) reaches 85%".

[0063] For the feature vectors output by the residual network, the cosine similarity is used to calculate the similarity between nodes, forming a similarity matrix (the closer the value is to 1, the more similar the node attenuation trends are). The similarity matrix is converted into a graph structure, the principal eigenvector is extracted through eigenvalue decomposition, and then the nodes are classified into categories such as "normal attenuation", "gradual change", and "mutation" by K-means clustering. The nodes in the "mutation" category are screened out to generate attenuation mutation node data (such as including node ID "Node_24", mutation occurrence distance "1520 meters", and anomaly degree "high").

[0064] S6: Process the attenuation mutation node data according to the stress inversion method, and determine the fault point data in the target distribution network based on the obtained stress value data.

[0065] Preferably, in an embodiment of the present invention, the processing the attenuation mutation node data according to the stress inversion method and determining the fault point data in the target distribution network based on the obtained stress value data includes: Construct a stress attenuation relationship model according to the obtained stress sensitive parameter, material stress threshold, and link physical length of the optical fiber link; Extract the power attenuation amount corresponding to each attenuation mutation node data and the physical length of the link where it is located, and input the power attenuation amount and the physical length into the stress attenuation relationship model to obtain a stress value data set; Compare each stress value data in the stress value data set with the material stress threshold respectively, and screen the comparison results to obtain a list of high-stress mutation nodes; Determine the fault point data in the target distribution network based on the list of high-stress mutation nodes.

[0066] Preferably, in an embodiment of the present invention, the determining the fault point data in the target distribution network based on the list of high-stress mutation nodes includes: Construct a candidate fault point set based on each node in the list of high-stress mutation nodes and its adjacent nodes; Obtain the coordinate data of each node in the candidate fault point set, and calculate the relative fault point distance between each high-stress mutation node and its adjacent nodes based on the coordinate data; Perform multi-condition screening according to the obtained stress value and relative fault point distance of each high-stress mutation node, and obtain the fault point data according to the results of the multi-condition screening.

[0067] Among them, the stress inversion method is a method of inversely deducing the stress condition of the optical fiber through information such as the power attenuation of the optical signal in the known optical fiber link. Because there is a certain relationship between the optical signal attenuation of the optical fiber and the stress it receives, this relationship can be used to infer the stress.

[0068] The stress sensitivity parameter is a parameter that reflects the sensitivity of an optical fiber to the stress it receives. Optical fibers with different materials and structures have different stress sensitivity parameters, which are an important basis for constructing a stress attenuation relationship model. The material stress threshold is the maximum stress value that the optical fiber material can withstand. When the stress on the optical fiber exceeds this threshold, the optical fiber may malfunction, such as breakage, bending, etc. The stress attenuation relationship model is a model that describes the mathematical relationship between the stress on the optical fiber and the attenuation of the optical signal power. Through this model, the corresponding stress value can be calculated based on the power attenuation amount.

[0069] The power attenuation amount is the reduction in the power of the optical signal during transmission in the optical fiber link, which is the key input data for calculating the stress value. The high-stress mutation node list is a list composed of nodes whose stress values exceed the material stress threshold, and these nodes are the key objects of concern where faults may occur. The relative distance of the fault point is the distance between the high-stress mutation node and its adjacent node, which is used to further determine the location of the fault point.

[0070] It should be noted that the attenuation mutation node data obtained in the previous steps only indicates the nodes where abnormalities may exist and cannot accurately determine the location of the fault point. The stress on the optical fiber is closely related to the occurrence of faults. Through the stress inversion method, the power attenuation information can be converted into stress information, so as to more accurately locate the fault point. By comparing the stress value with the material stress threshold and screening out the high-stress mutation nodes, the scope of fault troubleshooting can be narrowed down and focused on the nodes where faults are most likely to occur. Combining the coordinate data of the nodes and the relative distance of the fault point for multi-condition screening can further improve the accuracy of fault point location.

[0071] Specifically, in this embodiment, first, information such as the stress sensitivity parameter, material stress threshold, and physical length of the optical fiber link is obtained. This information can be obtained by referring to the product manual of the optical fiber, conducting experimental measurements, etc. Then, a stress attenuation relationship model is constructed based on these information. This model can be established based on physical principles. For example, according to the mechanical properties of the optical fiber material and the optical transmission principle, the functional relationship between stress and power attenuation is deduced. Commonly, it may be a linear or non-linear function model used to describe the corresponding relationship between the stress value and the power attenuation amount.

[0072] Extract the power attenuation amount corresponding to each node from the attenuation mutation node data, and at the same time obtain the physical length of the link where the node is located. Substitute the extracted power attenuation amount and physical length into the stress attenuation relationship model, and calculate the stress value corresponding to each node through the model calculation. Collect the stress values of all nodes to form a stress value data set.

[0073] Compare each stress value in the stress value dataset with the material stress threshold. If the stress value of a certain node exceeds the material stress threshold, it indicates that the stress on this node is too high and there may be a fault. Screen out all the nodes whose stress values exceed the material stress threshold, and organize the relevant information of these nodes into a list, namely the high-stress mutation node list. The nodes in this list are the objects that need to be focused on for investigation.

[0074] Construct a set of candidate fault points based on each node in the high-stress mutation node list and its adjacent nodes. The adjacent nodes can be determined according to the topology of the optical fiber link, usually the nodes directly connected to the high-stress mutation nodes. This can further narrow down the scope of fault investigation to these candidate nodes.

[0075] Obtain the coordinate data of each node in the set of candidate fault points. These coordinate data can be obtained through methods such as Geographic Information System (GIS). Then, calculate the relative distance of the fault point between each high-stress mutation node and its adjacent nodes according to the coordinate data. This distance information can help further judge the possible location of the fault point.

[0076] Conduct multi-condition screening based on the obtained stress value and relative distance of the fault point of each high-stress mutation node. For example, nodes with larger stress values and closer distances to adjacent nodes can be given priority as they are more likely to be the fault points. Through this multi-condition screening method, finally determine the fault point data, including the specific location of the fault point, relevant stress values and other information, providing an accurate basis for subsequent fault repair.

[0077] Through the above steps, using the stress inversion method combined with multi-condition screening, it is possible to accurately determine the fault point data in the target distribution network from the attenuation mutation node data, improving the efficiency of fault investigation and repair.

[0078] An embodiment of the present invention provides a fault location system for a distribution communication network based on deep learning. Specifically, please refer to Figure 2 , Figure 2 which shows a schematic flow chart of the fault location system for a distribution communication network based on deep learning in one embodiment of the present invention, and it includes: An acquisition module 11, configured to acquire the optical power data of the optical fiber link in the target distribution network, calculate the optical power difference between adjacent sampling points in the optical power data, and generate time series data; A calculation module 12, configured to perform fitting processing on the time series data to obtain an actual power attenuation curve, and calculate the absolute difference between the actual power attenuation curve and the theoretical power attenuation curve to obtain standardized difference data; The first extraction module 13 is configured to input the standardized difference data into a pre-constructed non-linear attenuation model to obtain non-linear change characteristic data reflecting the optical signal power attenuation in the optical fiber link; The second extraction module 14 is configured to extract features from the physical topology structure of the target distribution network to obtain distribution network topology feature data, and input the distribution network topology feature data and the non-linear change characteristic data into a pre-constructed fault identification model to obtain the fault probability values of each optical fiber link; The identification module 15 is configured to compare the fault probability values with a preset fault probability threshold to obtain potential fault links, and sequentially perform feature extraction and mutation node identification on the potential fault links to obtain attenuation mutation node data; The positioning module 16 is configured to process the attenuation mutation node data according to the stress inversion method, and determine the fault point data in the target distribution network based on the obtained stress value data.

[0079] Compared with the prior art, the beneficial effects of the embodiments of the present invention are at least one of the following: 1) The present invention realizes the accurate capture of non-linear fault characteristics and multi-dimensional data fusion through a deep learning model, breaks through the dependence of traditional methods on linear attenuation characteristics, and significantly improves the fault identification accuracy in complex scenarios. Specifically, the present invention uses the non-linear attenuation model to deeply process the standardized difference data, effectively extracts the non-linear change characteristics of the optical signal power attenuation, and constructs a fault identification model in combination with the distribution network topology feature data, which can accurately identify progressive faults such as micro-bends and connector aging that are easily missed by traditional methods, improve the early fault detection rate by more than 40%, and at the same time reduce the misjudgment rate in multi-branch networks through topological structure analysis; 2) The present invention constructs a full-process intelligent positioning closed loop of "data collection - model inference - stress inversion", realizes the automated processing from fault probability assessment to accurate positioning, and greatly improves the fault positioning efficiency and engineering practicability. Specifically, the present invention converts the attenuation mutation node data into specific stress values through the stress inversion method, and accurately locks the fault point position in combination with the material stress threshold, avoiding the time-consuming defect of traditional OTDR segment-by-segment detection. The positioning time is shortened by more than 60% in complex topologies such as ring networks, and can adapt to network topology changes in real time, providing direct guidance for the rapid repair of the distribution communication network, and significantly reducing the operation and maintenance cost and fault handling time.

[0080] The above-described embodiments merely represent several implementation manners of the present invention. The description is relatively specific and detailed, but it should not be construed as a limitation on the scope of the patent for the present invention. It should be noted that for those of ordinary skill in the art, without departing from the concept of the present invention, several modifications and improvements can still be made, and these all fall within the protection scope of the present invention. Therefore, the protection scope of the patent for the present invention shall be subject to the appended claims.

Claims

1. A method for locating faults in a power distribution communication network based on deep learning, characterized in that: include: Obtaining optical power data of an optical fiber link in a target distribution network, calculating optical power differences between adjacent sampling points in the optical power data, and generating time series data; Performing fitting processing on the time series data to obtain an actual power attenuation curve, and calculating an absolute difference between the actual power attenuation curve and a theoretical power attenuation curve to obtain standardized difference data; Inputting the standardized difference data into a pre-constructed nonlinear attenuation model to obtain nonlinear variation characteristic data reflecting the attenuation of optical signal power in the optical fiber link; Extracting features of the physical topology of the target distribution network to obtain distribution network topology feature data, inputting the distribution network topology feature data and the nonlinear change feature data into a pre-built fault identification model to obtain a fault probability value of each optical fiber link; The fault probability value is compared with a preset fault probability threshold to obtain a potential fault link, and feature extraction and mutation node identification are performed on the potential fault link in sequence to obtain attenuation mutation node data; The attenuation mutation node data is processed according to a stress inversion method, and the fault point data in the target distribution network is determined based on the obtained stress value data.

2. The method for locating faults in a power distribution communication network based on deep learning according to claim 1, characterized in that: The step of obtaining optical power data of an optical fiber link in a target distribution network, calculating optical power differences between adjacent sampling points in the optical power data, and generating time series data includes: Collecting raw optical power data of optical fiber links in the target distribution network in real time, and preprocessing the raw optical power data; Time-sorting the pre-processed original optical power data to obtain an optical power sequence data set; Calculating the optical power difference between adjacent sampling points in the optical power sequence data set to obtain difference data; wherein if the optical power difference exceeds a preset optical power fluctuation threshold, it is marked as an abnormal data point; Non-abnormal data points in the difference data are screened to obtain time series data.

3. The method for locating faults in a power distribution communication network based on deep learning according to claim 1, characterized in that: Before fitting the time series data, the method further includes: Performing linear regression processing on the time series data to obtain optical power trend data, and calculating the mean square error between the theoretical stress characteristic data of the optical fiber link and the optical power trend data; If the mean square error exceeds a preset error threshold, fitting processing is performed on the time series data.

4. The method for locating faults in a power distribution communication network based on deep learning according to claim 3, characterized in that: The calculating the mean square error between the theoretical stress characteristic data of the optical fiber link and the optical power trend data comprises: Generate stress distribution data according to the acquired stress distribution parameters of the optical fiber link, and extract frequency domain features of the stress distribution data using Fourier transform to obtain the theoretical stress feature data; Performing linear regression processing on the time series data using the least squares method to obtain an optical power trend function, and obtaining a trend function value at each sampling time point according to the optical power trend function to form the optical power trend data; The theoretical stress characteristic data and the optical power trend data are time-aligned, and a mean square error between the theoretical stress characteristic data and the optical power trend data after time alignment is calculated.

5. The method for locating faults in a power distribution communication network based on deep learning according to claim 1, characterized in that: The fitting process is performed on the time series data to obtain an actual power attenuation curve, and the absolute difference between the actual power attenuation curve and the theoretical power attenuation curve is calculated to obtain standardized difference data, including: The time series data is subjected to curve fitting based on the cubic spline interpolation method to obtain the actual power attenuation curve; Constructing a theoretical power attenuation curve according to the physical parameters of the optical fiber link; Calculating the absolute difference of each sampling point on the time axis of the actual power attenuation curve and the theoretical power attenuation curve to obtain an absolute difference sequence reflecting the degree of attenuation abnormality; The absolute difference sequence is normalized to obtain the standardized difference data.

6. The method for locating faults in a power distribution communication network based on deep learning according to claim 1, characterized in that: The feature extraction of the physical topological structure of the target distribution network to obtain distribution network topological feature data includes: Based on the acquired physical connection relationship of the optical fiber link, construct a node adjacency matrix and a link attribute table; Summing each row of the node adjacency matrix to obtain a node degree vector reflecting the importance of the node; Calculate link weight data according to the link attribute table, and construct a link weight matrix based on the link weight data; The node degree vector and the link weight matrix are concatenated to obtain the distribution network topology characteristic data.

7. The method for locating faults in a power distribution communication network based on deep learning according to claim 1, characterized in that: The comparing the fault probability value with a preset fault probability threshold to obtain a potential fault link, and sequentially performing feature extraction and mutation node identification on the potential fault link to obtain attenuation mutation node data, includes: Compare the fault probability value with a preset fault probability threshold, filter out links whose fault probability values ​​are greater than or equal to the preset fault probability threshold, and generate a list of potential fault links; Acquire optical power sequence data of links corresponding to the faulty link list, and convert the optical power sequence data into a two-dimensional matrix; The two-dimensional matrix is ​​concatenated with the standardized difference data and the nonlinear change feature data to obtain a fused feature tensor; and the fused feature tensor is input into a pre-constructed deep residual network for feature extraction to obtain attenuation mutation feature data; Cluster analysis is performed on the attenuation mutation feature data to obtain the attenuation mutation node data where the attenuation trend has changed.

8. The method for locating faults in a power distribution communication network based on deep learning according to claim 1, characterized in that: The step of processing the attenuation mutation node data according to the stress inversion method and determining the fault point data in the target distribution network based on the obtained stress value data includes: Constructing a stress attenuation relationship model according to the acquired stress-sensitive parameters of the optical fiber link, the material stress threshold and the physical length of the link; Extracting the power attenuation amount and the physical length of the link corresponding to each of the attenuation mutation node data, inputting the power attenuation amount and the physical length into the stress attenuation relationship model to obtain a stress value data set; Compare each stress value data in the stress value data set with the material stress threshold, and screen the comparison results to obtain a high stress mutation node list; Fault point data in the target distribution network is determined based on the high stress mutation node list.

9. The method for locating faults in a power distribution communication network based on deep learning according to claim 8, characterized in that: The determining of the fault point data in the target distribution network based on the high stress mutation node list includes: Constructing a candidate fault point set based on each node and its adjacent nodes in the high stress mutation node list; Obtaining coordinate data of each node in the candidate fault point set, and calculating the relative distance between each high stress mutation node and its adjacent node based on the coordinate data; A multi-condition screening is performed based on the acquired stress value of each high stress mutation node and the relative distance of the fault point, and the fault point data is obtained based on the result of the multi-condition screening.

10. A distribution communication network fault location system based on deep learning, characterized in that: include: An acquisition module is used to acquire optical power data of an optical fiber link in a target distribution network, calculate optical power differences between adjacent sampling points in the optical power data, and generate time series data; A calculation module, used for performing fitting processing on the time series data to obtain an actual power attenuation curve, and calculating an absolute difference between the actual power attenuation curve and a theoretical power attenuation curve to obtain standardized difference data; A first extraction module, used for inputting the standardized difference data into a pre-built nonlinear attenuation model to obtain nonlinear change characteristic data reflecting the attenuation of optical signal power in the optical fiber link; A second extraction module is used to extract features of the physical topology structure of the target distribution network to obtain distribution network topology feature data, input the distribution network topology feature data and the nonlinear change feature data into a pre-built fault identification model to obtain a fault probability value of each optical fiber link; An identification module, used to compare the fault probability value with a preset fault probability threshold to obtain a potential fault link, and sequentially perform feature extraction and mutation node identification on the potential fault link to obtain attenuation mutation node data; The positioning module is used to process the attenuation mutation node data according to the stress inversion method, and determine the fault point data in the target distribution network based on the obtained stress value data.

Images