Cable abnormity discrimination system and method with intelligent operation and maintenance

Abstract

The invention discloses a cable abnormity discrimination system and method with intelligent operation and maintenance, and belongs to the technical field of cable abnormity discrimination. The method comprises the steps of determining preset parameters of pulse laser and generating the pulse laser; calculating a light intensity ratio and a temperature measurement position; constructing a training data set, training the convolutional neural network model, obtaining the actual transmission speed of light in the optical fiber, and correcting the temperature measurement position; based on the timestamp, combining the corrected temperature measurement position with the corresponding temperature to obtain a temperature characteristic matrix; cable physical indexes are obtained, a graph structure is built, and the cable physical indexes are bound to all hierarchical elements of the graph structure; carrying out standardization processing on the temperature characteristic matrix, and carrying out association mapping with graph elements; setting a basic anomaly judgment rule based on the temperature characteristic matrix, and judging an abnormal node; calculating a current matching degree; performing space and link affiliation association on the abnormal node and the abnormal current position; and tracing the abnormal propagation path along the current transmission direction to obtain an abnormal source node set.

Description

Technical Field

[0001] This invention relates to the field of cable anomaly detection technology, specifically a cable anomaly detection system and method with intelligent operation and maintenance capabilities. Background Technology

[0002] As power systems develop towards intelligence and high density, cables, as the core carrier of power transmission, directly determine the reliability of power grid supply. Currently, cables are widely used in urban power distribution networks, industrial parks, offshore wind power, and other scenarios. The laying environment is becoming increasingly complex, and they operate under high load conditions for extended periods. They are prone to abnormal temperatures due to issues such as insulation aging, loose joints, external damage, and overload. If these issues are not identified and addressed in a timely manner, they may lead to cable burnout and line tripping.

[0003] Current DTS technology typically uses a fixed light transmission speed in optical fibers when calculating temperature measurement locations, failing to consider the impact of changes in the physical characteristics of the optical fibers during actual operation. This makes it impossible to accurately locate minor faults and increases the difficulty of maintenance and troubleshooting. Current technology does not structurally correlate cable physical properties with monitoring data, only displaying isolated temperature anomalies without clarifying their hierarchical affiliation within the cable link. Furthermore, it does not integrate current transmission logic to trace the anomaly propagation path, only identifying the location of the anomaly without determining its origin. Current technology only outputs alarm information for temperature anomalies, without correlating anomaly nodes with cable physical indicators or maintenance history, making it impossible to determine the severity and type of the anomaly, or trace the set of source nodes. Summary of the Invention

[0004] The purpose of this invention is to provide a cable anomaly detection system and method with intelligent operation and maintenance capabilities to solve the problems raised in the prior art.

[0005] To achieve the above objectives, the present invention provides the following technical solution:

[0006] Firstly, this application provides a cable anomaly detection method with intelligent operation and maintenance capabilities, comprising the following steps:

[0007] Based on the fiber optic parameters of the cable and the monitoring requirements, the preset parameters of the pulsed laser are determined, and the pulsed laser is generated. When the pulsed laser generates backscattered light during transmission in the fiber, the Stokes light and anti-Stokes light are received and separated, useless signals are filtered out, and the signal is converted into a digital signal to calculate the intensity ratio. The temperature is calculated based on the intensity ratio, and the temperature measurement location is calculated in combination with the echo time.

[0008] A training dataset containing physical environment parameters, fiber characteristic parameters, raw measurement data, and real location labels is constructed and preprocessed for feature reconstruction. A convolutional neural network model is constructed and trained to obtain the actual transmission speed of light in the fiber and correct the temperature measurement position. Based on the timestamp, the corrected temperature measurement position is combined with the corresponding temperature to obtain the temperature feature matrix.

[0009] Obtain cable physical indicators, build a graph structure according to global, regional and local layers, and bind the cable physical indicators to the elements of each level of the graph structure; standardize the temperature feature matrix and associate it with the graph elements; set basic anomaly detection rules based on the temperature feature matrix to identify abnormal nodes.

[0010] The system acquires real-time current and dynamic current carrying capacity data of the links associated with abnormal nodes and calculates the current matching degree. Based on the hierarchical association relationship of the graph structure, it spatially and link-wise associates the abnormal nodes with the abnormal current locations. It traces the abnormal propagation path along the current transmission direction to obtain the set of abnormal source nodes.

[0011] In conjunction with the first aspect, in the first embodiment of the first aspect of this application, the step of determining the preset parameters of the pulsed laser based on the fiber optic parameters of the cable and the monitoring requirements, and generating the pulsed laser, includes:

[0012] The optical fiber parameters include type, core diameter, length, refractive index distribution, loss coefficient, and nonlinear threshold. The monitoring requirements include spatial resolution, temperature measurement range, temperature measurement accuracy, maximum monitoring distance, and data update rate. For the type and core diameter of the optical fiber, a pulse with a corresponding pulse width range is matched. Based on the length and loss coefficient of the optical fiber, the minimum pulse power is calculated while not exceeding the optical fiber nonlinear threshold. The optical transmission speed is estimated by combining the refractive index distribution.

[0013] Based on monitoring requirements, the pulse width is determined according to spatial resolution requirements; the peak pulse power is set according to the temperature measurement range and accuracy; the pulse repetition frequency is determined based on the maximum monitoring distance and data update rate; and a wavelength that matches the low-loss window of the optical fiber is selected.

[0014] Based on wavelength-selective lasers, pulse width and repetition frequency are controlled by pulse generators and modulators to generate pulse signals with the required timing. Peak power is adjusted to a preset value by a power amplifier. Finally, the output pulsed laser is injected into the optical fiber via a coupler to complete the generation and emission of the pulsed laser.

[0015] In conjunction with the first aspect, in a second embodiment of the first aspect of this application, when backscattered light is generated during the transmission of pulsed laser light in the optical fiber, receiving and separating Stokes light and anti-Stokes light, filtering out unwanted signals, converting them into digital signals, and calculating the light intensity ratio includes:

[0016] Backscattered light generated during transmission is separated from the main transmission path by a coupler and introduced into the receiving optical path. A focusing lens is used to converge the scattered light onto the photosensitive surface. Based on the wavelength difference between Stokes light and anti-Stokes light, a wavelength selection device is used to split the converged scattered light, realizing the physical separation of the two target lights. A bandpass filter is used to further suppress ambient stray light outside the bandwidth. A low-noise preamplifier is used to amplify the weak electrical signal, while a filter circuit filters out high-frequency noise. The scattered signals of multiple consecutive pulses are accumulated and averaged to reduce random noise interference.

[0017] The filtered Stokes and anti-Stokes light signals are input into a photodetector to convert the optical signal into an electrical signal. The electrical signal is then converted into a digital signal by a high-speed analog-to-digital converter, with the sampling rate matched to the repetition frequency of the pulsed laser to preserve the characteristics of light intensity variation over time. The digitized Stokes and anti-Stokes light signals are synchronized and time-aligned, and the peak or integral intensity of the two signals is extracted within the same time window. The ratio of the anti-Stokes light intensity to the Stokes light intensity is calculated as the light intensity ratio.

[0018] In conjunction with the first aspect, in the third embodiment of the first aspect of this application, the step of calculating the temperature based on the light intensity ratio and calculating the temperature measurement location based on the echo time includes:

[0019] Based on the physical principle of Raman scattering, a preset temperature-intensity ratio correlation model is invoked to determine the mapping relationship between the intensity ratio and the absolute temperature. The temperature-intensity ratio correlation model includes Raman scattering physical parameters, physical environment parameters, and fiber characteristic parameters. A fiber segment with room temperature at the cable start end is selected as a reference segment with a known temperature in the fiber. The actual temperature and corresponding intensity ratio of this reference segment are obtained and substituted into the temperature-intensity ratio correlation model to correct the proportionality constant in the model. The intensity ratio of each spatial sampling point is substituted into the calibrated model, and the absolute temperature value of the sampling point is calculated by inverse calculation formula.

[0020] Based on the current refractive index of the monitored optical fiber, the propagation speed of light in the fiber is calculated using the formula v=c / n, where v is the propagation speed, c is the speed of light, and n is the refractive index of the optical fiber. When physical deformation occurs in the optical fiber, the value of v is dynamically adjusted using pre-stored correction coefficients. The timestamp of the laser emission moment is extracted. The timestamp of receiving the backscattered light at that sampling point Calculate echo time difference Corrected through hardware calibration data This eliminates the time deviation caused by non-fiber optic transmission and obtains the true round-trip time of light in the fiber optic cable. According to the formula Where L is the temperature measurement position; when there is a preset marker point in the optical fiber, the calculated L is compared with the position of the preset marker point to correct the position deviation caused by the optical fiber length error; compare the pre-stored optical fiber splice point and connector position information, when the position corresponding to a certain echo signal coincides with the marker point and there is no temperature abnormality at that position, it is determined to be a false echo, and the erroneous calculation result at that position is discarded; when the false echo is superimposed with the normal echo, the effective echo is separated by the difference in signal strength.

[0021] In conjunction with the first aspect, in the fourth embodiment of the first aspect of this application, the step of constructing and training a convolutional neural network model to obtain the actual transmission speed of light in the optical fiber and correcting the temperature measurement position includes:

[0022] The input layer of the convolutional neural network model receives a preprocessed feature matrix, which is organized according to spatial sampling points and feature dimensions. The output layer is a single neuron that outputs the actual transmission speed of light in the optical fiber corresponding to the sampling point. The convolutional layer uses a combination of multiple convolutions and pooling. The fully connected regression layer flattens the feature vectors output by the convolutional layer and outputs the actual transmission speed. Dropout layers are added between the fully connected layers, and L2 regularization is applied to the weights of the convolutional and fully connected layers to suppress overfitting.

[0023] Using the corrected position error as the core loss, a composite loss function is designed in conjunction with physical constraints. The base term is the mean square error between the predicted and actual velocities, and the penalty term is an increasing penalty value when the predicted velocity exceeds the physical reasonable range, ensuring that the output conforms to the physical laws of fiber optic transmission. The training dataset is divided into training, validation, and test sets proportionally, and the Adam optimizer is used, with the validation set loss monitored during training. Feature matrices and corresponding actual velocity labels are input in batches, and network weights are updated through backpropagation. After each round of training, the loss and prediction accuracy are evaluated on the validation set, and the optimal model parameters are recorded. After training, the mean absolute error and root mean square error of velocity prediction are calculated on the test set for model evaluation. Real-time collected feature data is input into the trained model, outputting the actual transmission velocity of each sampling point, recalculating the temperature measurement position, and replacing the original temperature measurement position.

[0024] In conjunction with the first aspect, in the fifth embodiment of the first aspect of this application, the step of obtaining cable physical indicators, constructing a graph structure according to the global layer, regional layer, and local layer, and binding the cable physical indicators to elements at each level of the graph structure includes:

[0025] The physical indicators of the cable include basic attributes, spatial path and key nodes; with the backbone link and core control equipment of the cable network as the skeleton, the global layer elements are defined, including backbone cable segments, core equipment and the start and end points of the backbone link; according to the actual transmission path of the backbone cable, the core elements are connected, the backbone cable segments are represented by lines and the core equipment is represented by icons, and the connection order of each backbone cable segment is marked to form the overall connection framework of the global layer.

[0026] The region is divided into regions based on the physical area of ​​cable laying, with each region corresponding to an independent subgraph structure. Within each region, the branch cable segments, the power distribution equipment, and the boundary endpoints of the region are extracted. The elements within the region are connected according to the actual route of the branch cables. The branch cable segments are represented by differentiated lines, and the power distribution equipment is represented by icons. The connection relationship between the branch cables and the trunk cables is marked to form an independent topology structure for each region. The region layer subgraph is then associated with the corresponding trunk cable branch point in the global layer.

[0027] On the global layer trunk cable segment and the regional layer branch cable segment, locate key nodes, including joints, fusion splices, terminal boxes and monitoring sensors; assign a unique identifier to each key node and mark the physical location of the node; map the key nodes to the corresponding cable segments, mark the location of key nodes on the cable segment lines of the global layer and the regional layer, mark the relationship between the node and the cable segment, and form the node distribution structure of the local layer.

[0028] Bind the physical properties of the cable to each level of the graph structure.

[0029] In conjunction with the first aspect, in the sixth embodiment of the first aspect of this application, the step of determining abnormal nodes based on setting basic anomaly detection rules according to the temperature feature matrix includes:

[0030] Based on the temperature feature matrix, temperature feature dimensions for anomaly detection are determined, and an adaptation threshold is set for each temperature feature dimension. When the real-time temperature of a node exceeds the safe temperature threshold for this type of cable, and the duration of the exceedance exceeds a preset duration, it is determined to be an abnormal node with sustained temperature exceedance. When the node's temperature rise rate exceeds a set threshold, and there is no sudden increase in load in the associated current data, it is determined to be an abnormal node with rapid temperature rise. When the difference between the node's temperature and the average temperature of the same link segment exceeds a set threshold, and there are no abnormal fluctuations in the temperature of surrounding nodes, it is determined to be an abnormal node with local high temperature. When the difference between the node's current temperature and the historical average temperature under the same operating conditions over the past month exceeds a set threshold, it is determined to be an abnormal node with temperature baseline drift. Anomaly detection conditions for different temperature feature dimensions are combined.

[0031] Based on the location coordinates of the sampling points in the temperature feature matrix, the corresponding nodes in the associated graph structure are identified; the nodes are traversed to perform rule checks, and the nodes that trigger the rules are marked in the graph structure. The marking styles are distinguished according to the severity of the anomaly to determine the abnormal nodes.

[0032] In conjunction with the first aspect, in the seventh embodiment of the first aspect of this application, the hierarchical association based on the graph structure, which spatially and link-wise associates the abnormal nodes with the locations of abnormal currents, includes:

[0033] Based on the spatial coordinate range of nodes in the graph structure, a coordinate overlap threshold is set. When the threshold condition is met, it is determined that there is a spatial correlation between the abnormal node and the abnormal current location. The spatial coordinate ranges of the two are compared one by one, and the abnormal current location is classified and marked according to the degree of coordinate overlap. The abnormal current location is specifically the location where there is an abnormal current matching degree.

[0034] Based on the hierarchical attribution logic of the graph structure, a link attribution matching rule is established between abnormal nodes and abnormal current locations. The link segment attributes of the abnormal node and the link segment attributes of the abnormal current location are compared, and the current collector ID bound to the abnormal node is matched with the current collector ID corresponding to the abnormal current location.

[0035] In conjunction with the first aspect, in the eighth embodiment of the first aspect of this application, the step of tracing the abnormal propagation path along the current transmission direction to obtain the set of abnormal source nodes includes:

[0036] Starting from the power source and ending at the load, determine the current transmission direction of the global layer backbone link and the regional layer branch link; select typical abnormal nodes distributed in different sub-links as the starting point for tracing; check the nodes on the reverse path of current transmission segment by segment in the order of starting point, upstream adjacent node, upstream branch node, and backbone link node; during the reverse tracing process, if an upstream node is abnormal, but only the starting point downstream of the node is abnormal and other sub-link nodes are not abnormal, check whether the node is caused by a local load. If it is caused by a local load, it is not included as a potential source of abnormality, and the backbone link node is traced upwards; for potential source nodes obtained by reverse tracing, check all downstream sub-link nodes along the forward current direction; when a downstream node is not abnormal during forward tracing, but its subsequent nodes are abnormal, check the physical properties of the node at the breakpoint and mark the breakpoint of abnormal propagation.

[0037] Summarize all the anomaly source nodes obtained through reverse tracing, and sort them according to priority based on the earliest anomaly start time, the widest forward propagation range, and the most significant anomaly intensity to obtain the set of anomaly source nodes.

[0038] Secondly, this application provides a cable anomaly detection system with intelligent operation and maintenance capabilities, including:

[0039] The light intensity ratio and temperature measurement position calculation module includes: a pulsed laser generation unit, a light intensity ratio calculation unit, and a temperature measurement position calculation unit. The pulsed laser generation unit determines the preset parameters of the pulsed laser based on the fiber optic parameters of the cable and the monitoring requirements, and generates the pulsed laser. The light intensity ratio calculation unit receives and separates the Stokes light and anti-Stokes light when the pulsed laser generates backscattered light during transmission in the optical fiber, filters out useless signals, converts the signal into a digital signal, and calculates the light intensity ratio. The temperature measurement position calculation unit calculates the temperature based on the light intensity ratio and combines it with the echo time to calculate the temperature measurement position.

[0040] The temperature feature matrix generation module includes: a training dataset construction unit, a temperature measurement position correction unit, and a temperature feature matrix generation unit. The training dataset construction unit builds a training dataset containing physical environment parameters, fiber optic characteristic parameters, raw measurement data, and real-world location labels, and performs preprocessing and feature reconstruction. The temperature measurement position correction unit constructs and trains a convolutional neural network model to obtain the actual transmission speed of light in the fiber optic cable and corrects the temperature measurement position. The temperature feature matrix generation unit combines the corrected temperature measurement position with the corresponding temperature based on a timestamp to obtain the temperature feature matrix.

[0041] The abnormal node determination module includes a graph structure building unit, an association mapping unit, and an abnormal node determination unit. The graph structure building unit acquires the physical indicators of the cable, builds a graph structure according to the global layer, regional layer, and local layer, and binds the physical indicators of the cable to the elements of each level of the graph structure. The association mapping unit standardizes the temperature feature matrix and associates it with the graph elements. The abnormal node determination unit sets basic anomaly detection rules based on the temperature feature matrix and determines abnormal nodes.

[0042] The abnormal source node generation module includes: a current matching degree unit, a spatial and link attribution association unit, and an abnormal source node generation unit. The current matching degree unit acquires real-time current and dynamic current carrying capacity data of the links associated with abnormal nodes and calculates the current matching degree. The spatial and link attribution association unit associates abnormal nodes with abnormal current locations based on the hierarchical relationship of the graph structure. The abnormal source node generation unit traces the abnormal propagation path along the current transmission direction to obtain a set of abnormal source nodes.

[0043] Compared with the prior art, the beneficial effects of the present invention are:

[0044] 1. This invention achieves dynamic correction of transmission speed through a data-driven model. It constructs a training dataset containing physical environment parameters, fiber characteristic parameters, and raw measurement data. After preprocessing and feature reconstruction, a convolutional neural network model is trained. This model can learn the mapping relationship between changes in fiber physical characteristics and actual transmission speed, and output real-time and accurate actual transmission speed, replacing the traditional fixed-speed calculation of temperature measurement position.

[0045] 2. This invention constructs a multi-level structured graph structure and link tracing logic. First, it obtains the physical indicators of the cable, builds a graph structure according to the global layer, regional layer, and local layer, binds the physical indicators to the elements of each level, realizes the deep association between monitoring data and link structure, and clarifies the hierarchical affiliation of abnormal nodes in the link. Then, it traces the abnormal propagation path along the current transmission direction, clearly presents where the abnormality comes from, and finally locks the set of abnormal source nodes.

[0046] 3. This invention constructs a multi-dimensional correlation analysis system, which deeply correlates abnormal nodes with cable physical indicators and operation and maintenance history, and combines temperature feature matrix and current matching degree to comprehensively determine the severity of abnormality and fault type. Attached Figure Description

[0047] Figure 1 This is a schematic diagram illustrating the steps of a cable anomaly detection method with intelligent operation and maintenance according to the present invention;

[0048] Figure 2 This is a system structure diagram of a cable anomaly detection system with intelligent operation and maintenance according to the present invention. Detailed Implementation

[0049] 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.

[0050] Example: Figures 1-2 As shown, the present invention provides a technical solution:

[0051] like Figure 1 As shown, this application provides a cable anomaly detection method with intelligent operation and maintenance, including the following steps:

[0052] Step S100: Based on the fiber optic parameters of the cable and the monitoring requirements, determine the preset parameters of the pulsed laser and generate the pulsed laser; when the pulsed laser generates backscattered light during transmission in the fiber optic cable, receive and separate the Stokes light and anti-Stokes light, filter out useless signals, convert them into digital signals, and calculate the light intensity ratio; calculate the temperature based on the light intensity ratio, and calculate the temperature measurement location by combining the echo time.

[0053] Specifically, the optical fiber parameters include type, core diameter, length, refractive index distribution, loss coefficient, and nonlinear threshold; the monitoring requirements include spatial resolution, temperature measurement range, temperature measurement accuracy, maximum monitoring distance, and data update rate; for the type and core diameter of the optical fiber, a pulse with a corresponding pulse width range is matched; based on the length and loss coefficient of the optical fiber, the minimum pulse power is calculated while not exceeding the optical fiber nonlinear threshold; and the optical transmission speed is estimated by combining the refractive index distribution.

[0054] Based on monitoring requirements, the pulse width is determined according to spatial resolution requirements; the peak pulse power is set according to the temperature measurement range and accuracy; the pulse repetition frequency is determined based on the maximum monitoring distance and data update rate; and a wavelength that matches the low-loss window of the optical fiber is selected.

[0055] Based on wavelength-selective lasers, pulse width and repetition frequency are controlled by pulse generators and modulators to generate pulse signals with the required timing. Peak power is adjusted to a preset value by a power amplifier. Finally, the output pulsed laser is injected into the optical fiber via a coupler to complete the generation and emission of the pulsed laser.

[0056] Furthermore, the backscattered light generated during transmission is separated from the main transmission path by a coupler and introduced into the receiving optical path. A focusing lens is used to converge the scattered light onto the photosensitive surface. Based on the wavelength difference between Stokes light and anti-Stokes light, a wavelength selection device is used to split the converged scattered light, realizing the physical separation of the two target lights. A bandpass filter is used to further suppress ambient stray light outside the bandwidth. A low-noise preamplifier is used to amplify the weak electrical signal, while a filter circuit is used to filter out high-frequency noise. The scattered signals of multiple consecutive pulses are accumulated and averaged to reduce random noise interference.

[0057] The filtered Stokes and anti-Stokes light signals are input into a photodetector to convert the optical signal into an electrical signal. The electrical signal is then converted into a digital signal by a high-speed analog-to-digital converter, with the sampling rate matched to the repetition frequency of the pulsed laser to preserve the characteristics of light intensity variation over time. The digitized Stokes and anti-Stokes light signals are synchronized and time-aligned, and the peak or integral intensity of the two signals is extracted within the same time window. The ratio of the anti-Stokes light intensity to the Stokes light intensity is calculated as the light intensity ratio.

[0058] Furthermore, based on the physical principle of Raman scattering, a preset temperature-intensity ratio correlation model is invoked to determine the mapping relationship between the intensity ratio and the absolute temperature. The temperature-intensity ratio correlation model includes Raman scattering physical parameters, physical environment parameters, and fiber characteristic parameters. The fiber segment at the cable start end in a room temperature environment is selected as a reference segment with a known temperature in the fiber. The actual temperature and corresponding intensity ratio of the reference segment are obtained and substituted into the temperature-intensity ratio correlation model to correct the proportionality constant in the model. The intensity ratio of each spatial sampling point is substituted into the calibrated model, and the absolute temperature value of the sampling point is calculated by inverse calculation formula.

[0059] Based on the current refractive index of the monitored optical fiber, the propagation speed of light in the fiber is calculated using the formula v=c / n, where v is the propagation speed, c is the speed of light, and n is the refractive index of the optical fiber. When physical deformation occurs in the optical fiber, the value of v is dynamically adjusted using pre-stored correction coefficients. The timestamp of the laser emission moment is extracted. The timestamp of receiving the backscattered light at that sampling point Calculate echo time difference Corrected through hardware calibration data This eliminates the time deviation caused by non-fiber optic transmission and obtains the true round-trip time of light in the fiber optic cable. According to the formula Where L is the temperature measurement position; when there is a preset marker point in the optical fiber, the calculated L is compared with the position of the preset marker point to correct the position deviation caused by the optical fiber length error; compare the pre-stored optical fiber splice point and connector position information, when the position corresponding to a certain echo signal coincides with the marker point and there is no temperature abnormality at that position, it is determined to be a false echo, and the erroneous calculation result at that position is discarded; when the false echo is superimposed with the normal echo, the effective echo is separated by the difference in signal strength.

[0060] In one specific embodiment, a 50 / 125μm multimode optical fiber built into the cable is selected with the following parameters: length 8km, refractive index 1.465, loss coefficient of 0.4dB / km at 1310nm wavelength, and nonlinear threshold of 180mW; the monitoring requirements are set as spatial resolution of 0.4m, temperature range of -10℃ to 110℃, and data update rate of 1 time / second.

[0061] The calculated pulse width is based on a spatial resolution of 0.4m and an optical transmission speed of... The calculated pulse width is approximately 3.9 ns, but it is actually set to 4 ns; Pulse peak power: 8km total loss 3.2dB (power attenuation to 1 / 2.08 of the initial value), set to 40mW (below the threshold); Repetition frequency: optical round-trip time. Set to 12kHz; select wavelength 1310nm (low loss window).

[0062] The laser was generated using a 1310nm DFB laser, modulated with a 4ns pulse width and a 12kHz repetition frequency, and the power amplifier was adjusted to 40mW. The laser was then injected into the fiber via a coupler, with an actual injected power of 36mW.

[0063] A 1×2 coupler (8:2 splitting ratio, 2% received scattered light) and a focusing lens (8mm focal length) converge the light onto the APD photosensitive surface. A narrow-band filter separates the Stokes light at 1323nm (8nm bandwidth, 88% transmittance) and the anti-Stokes light at 1297nm (8nm bandwidth, 88% transmittance), while blocking Rayleigh scattered light. The stray light is attenuated by 35dB by a bandpass filter (4nm bandwidth), and the nA-level signal is amplified to the mA level by a preamplifier (55dB gain). An RC low-pass filter (cutoff 80MHz) reduces noise to 1.8mV. After averaging 100 pulses, the signal-to-noise ratio increases from 18dB to 38dB.

[0064] The Stokes light incident APD has a power of 9nW and an output current of 7.2nA (responsivity 0.8A / W), which is 8.8mA after amplification. The anti-Stokes light incident APD has a power of 7nW and an output current of 5.6nA, which is 6.9mA after amplification. ADC (sampling rate 800MSps, 16-bit) conversion: 8.8mA corresponds to a digital value of 28672, and 6.9mA corresponds to 22144. After time alignment, the peak intensities within a 10ns window are 28672 and 22144, respectively, and the intensity ratio is approximately 22144 / 28672≈0.77.

[0065] Temperature calculation: The light intensity ratio of the reference segment (0m, room temperature 24℃=297.15K) is 0.72. Substituting it into the model "R / R0=exp(ΔE / (kT0)-ΔE / (kT))", we can find that the light intensity ratio of a certain sampling point is 0.77, which corresponds to a temperature of 28℃ (deviation 0.3℃).

[0066] Location calculation: Curvature correction factor at 4km: 1.0008, v'≈2.05m / s; Transmission timestamp: 1,000,000μs, Reception timestamp: 10,000,39.1μs, Δt=39.1μs, Hardware correction: 0.3μs, Δt'=38.8μs; Compared with the 4km marker (4002.1m), the corrected value is 4002.1m.

[0067] Step S200: Construct a training dataset containing physical environment parameters, fiber characteristic parameters, raw measurement data and real location labels, and perform preprocessing and feature reconstruction; construct and train a convolutional neural network model to obtain the actual transmission speed of light in the fiber and correct the temperature measurement position; based on the timestamp, combine the corrected temperature measurement position with the corresponding temperature to obtain the temperature feature matrix.

[0068] Specifically, the input layer of the convolutional neural network model receives a preprocessed feature matrix, which is organized according to spatial sampling points and feature dimensions; the output layer is a single neuron that outputs the actual transmission speed of light in the optical fiber corresponding to the sampling point; the convolutional layer uses a combination of multiple convolutions and pooling, and the fully connected regression layer flattens the feature vector output by the convolutional layer to output the actual transmission speed; a Dropout layer is added between the fully connected layers, and L2 regularization is applied to the weights of the convolutional and fully connected layers to suppress overfitting;

[0069] Using the corrected position error as the core loss, a composite loss function is designed in conjunction with physical constraints. The base term is the mean square error between the predicted and actual velocities, and the penalty term is an increasing penalty value when the predicted velocity exceeds the physical reasonable range, ensuring that the output conforms to the physical laws of fiber optic transmission. The training dataset is divided into training, validation, and test sets proportionally, and the Adam optimizer is used, with the validation set loss monitored during training. Feature matrices and corresponding actual velocity labels are input in batches, and network weights are updated through backpropagation. After each round of training, the loss and prediction accuracy are evaluated on the validation set, and the optimal model parameters are recorded. After training, the mean absolute error and root mean square error of velocity prediction are calculated on the test set for model evaluation. Real-time collected feature data is input into the trained model, outputting the actual transmission velocity of each sampling point, recalculating the temperature measurement position, and replacing the original temperature measurement position.

[0070] In one specific embodiment, a dataset is constructed by selecting 5000 spatial sampling points (2m intervals) along a 10km cable link: physical environmental parameters include ambient temperature (15-35℃), humidity (40%-80%), and vibration intensity (0-0.5g); fiber characteristic parameters include fiber bending degree (0-5° / m) and fusion splice loss (0.01-0.03dB) at each sampling point; raw measurement data includes echo time (40-100μs) and light intensity ratio (0.6-0.9); and the actual location tags are measured by a laser rangefinder (error ≤0.05m).

[0071] During preprocessing, all parameters are normalized to the [0,1] interval (e.g., ambient temperature 15℃→0, 35℃→1), 50 abnormal samples are deleted (e.g., light intensity ratio>1.0), and a 12-dimensional feature matrix is ​​formed after feature reconstruction (3 environment + 2 fiber + 2 measurement + 5 derived features).

[0072] The CNN model's input layer receives a 12×5000 feature matrix (12-dimensional features × 5000 sampling points); it has 3 convolutional layers: the first layer has 3×3 kernels (32) + ReLU + 2×2 max pooling, the second layer has 3×3 kernels (64) + ReLU + 2×2 max pooling, and the third layer has 5×5 kernels (128) + ReLU + 2×2 max pooling; it has 2 fully connected layers: 128 → 64 neurons, 64 → 1 neuron (output actual transmission speed); Dropout (scale 0.3) is applied between fully connected layers, and L2 regularization (coefficient 1e-4) is applied to the weights of the convolutional and fully connected layers.

[0073] The dataset was divided into a 7:1.5:1.5 ratio for training (3500 sampling points), validation (750 sampling points), and test (750 sampling points). The composite loss function consisted of MSE as the base term and a penalty term set to "add 0.5 penalty points for every 1% deviation of the prediction speed from the theoretical value within ±3%". The Adam optimizer was used (initial learning rate 0.001, decaying by 10% every 10 epochs), with a batch size of 32 and 30 training epochs. During training, the validation loss converged to 0.002 in the 25th epoch, and training stopped if it did not decrease further. The model parameters at this point were recorded (e.g., the average weight of the first convolutional kernel was 0.12).

[0074] On the test set, the mean absolute error (MAE) of velocity prediction was 0.3% and the root mean square error (RMSE) was 0.45%, meeting the requirement of "error ≤ 1%"; in extreme scenarios (curvature of 5° / m), the MAE was still ≤ 0.5%, indicating that the model's generalization ability was satisfactory.

[0075] Real-time acquisition of feature data from a sampling point is input into the model, and the actual transmission speed is output. (Theoretical value) The original temperature measurement location was calculated to be 5000m (echo time 48.9μs), and the corrected location... The error between the actual location of 4992.5m and the actual location is 0.1m. Based on a 1-second timestamp, the corrected locations of 5000 sampling points (e.g., 4992.6m) are combined with the corresponding temperatures (e.g., 28℃) to form a 5000×2 temperature feature matrix (each row contains location and temperature).

[0076] Step S300: Obtain cable physical indicators, build a graph structure according to global, regional and local layers, and bind the cable physical indicators to the elements of each level of the graph structure; standardize the temperature feature matrix and associate it with the graph elements; set basic anomaly detection rules based on the temperature feature matrix to determine abnormal nodes;

[0077] Specifically, the physical indicators of the cable include basic attributes, spatial path and key nodes; with the backbone link and core control equipment of the cable network as the skeleton, global layer elements are defined, including backbone cable segments, core equipment and the start and end points of the backbone link; according to the actual transmission path of the backbone cable, the core elements are connected, the backbone cable segments are represented by lines and the core equipment is represented by icons, and the connection order of each backbone cable segment is marked to form the overall connection framework of the global layer.

[0078] The region is divided into regions based on the physical area of ​​cable laying, with each region corresponding to an independent subgraph structure. Within each region, the branch cable segments, the power distribution equipment, and the boundary endpoints of the region are extracted. The elements within the region are connected according to the actual route of the branch cables. The branch cable segments are represented by differentiated lines, and the power distribution equipment is represented by icons. The connection relationship between the branch cables and the trunk cables is marked to form an independent topology structure for each region. The region layer subgraph is then associated with the corresponding trunk cable branch point in the global layer.

[0079] On the global layer trunk cable segment and the regional layer branch cable segment, locate key nodes, including joints, fusion splices, terminal boxes and monitoring sensors; assign a unique identifier to each key node and mark the physical location of the node; map the key nodes to the corresponding cable segments, mark the location of key nodes on the cable segment lines of the global layer and the regional layer, mark the relationship between the node and the cable segment, and form the node distribution structure of the local layer.

[0080] Bind the physical properties of the cable to each level of the graph structure.

[0081] Furthermore, based on the temperature feature matrix, temperature feature dimensions for anomaly detection are determined, and an adaptation threshold is set for each temperature feature dimension. When the real-time temperature of a node exceeds the safe temperature threshold for this type of cable, and the duration of the exceedance exceeds a preset duration, it is determined to be an abnormal node with continuous temperature exceedance. When the node's temperature rise rate exceeds a set threshold, and there is no sudden increase in load in the associated current data, it is determined to be an abnormal node with rapid temperature rise. When the difference between the node temperature and the average temperature of the same link segment exceeds a set threshold, and there are no abnormal fluctuations in the temperature of surrounding nodes, it is determined to be an abnormal node with local high temperature. When the difference between the current temperature of a node and the historical average temperature under the same operating conditions over the past month exceeds a set threshold, it is determined to be an abnormal node with temperature baseline drift. Anomaly detection conditions for different temperature feature dimensions are combined.

[0082] Based on the location coordinates of the sampling points in the temperature feature matrix, the corresponding nodes in the associated graph structure are identified; the nodes are traversed to perform rule checks, and the nodes that trigger the rules are marked in the graph structure. The marking styles are distinguished according to the severity of the anomaly to determine the abnormal nodes.

[0083] In one specific embodiment, a 10kV cable line is selected, with the following physical specifications: the main cable model is YJV22-10kV-3×250, the length is 5km, and the laying method is underground pipe gallery; the branch cable model is YJV22-10kV-3×120, with two branches each 1.5km long, and the laying areas are "East Zone Workshop" and "West Zone Warehouse" respectively; the key nodes include 5 joints (J1-J5) and 3 fusion splices (R1-R3), J1 is located 1km from the main line, and J2 is located 0.5km from the East Zone branch, all of which are crimped joints.

[0084] The diagram structure is as follows: The global layer uses "main distribution room → trunk cable → regional distribution center" as the framework. The trunk cable is represented by a thick black line, and the model and length are marked. The regional layer is divided into "East Zone Workshop" and "West Zone Warehouse" sub-diagrams. Branch cables are distinguished by thin blue and green lines, and the laying area is marked. The local layer marks a red dot (J1) at 1km of the trunk and a red dot (J2) at 0.5km of the East Zone branch. Hovering the mouse displays "J1: Crimped connector, 3 years of operation".

[0085] The temperature feature matrix contains 500 sampling points (interval 10m). Standardization processing: the 0-120℃ temperature is normalized to the 0-1 range, and the temperature rise rate (0-5℃ / min) is normalized to 0-1. Anomaly detection thresholds are set: the safe temperature of 10kV cable is 90℃ (corresponding to a normalized value of 0.75), and the duration of continuous exceedance is 5min; the temperature rise rate threshold is 1℃ / min (normalized value 0.2); the regional temperature difference threshold is 10℃ (normalized value 0.08); and the historical temperature difference threshold is 3℃ (normalized value 0.025).

[0086] Associate the matrix sampling points with the graph structure nodes: J1 corresponds to the sampling point 1km from the main trunk, and J2 corresponds to the sampling point 0.5km from the eastern branch. Check the node data: J1 real-time temperature 95℃ (5℃ above the threshold), continuously exceeding the limit for 6 minutes, temperature rise rate 0.8℃ / min, regional temperature difference 8℃, historical temperature difference 2℃—triggered the "continuous temperature exceeding limit" rule, marked as a red flashing dot in the graph; J2 real-time temperature 88℃, temperature rise rate 1.2℃ / min (exceeding the threshold), no sudden current increase, regional temperature difference 3℃, historical temperature difference 1℃—triggered the "rapid temperature rise" rule, marked as a yellow solid dot; the temperatures of the remaining nodes are all between 75-85℃, with no abnormalities.

[0087] Step S400: Obtain real-time current and dynamic current carrying capacity data of the links associated with the abnormal nodes, and calculate the current matching degree; based on the hierarchical association relationship of the graph structure, spatially and link-wise associate the abnormal nodes with the abnormal current locations; trace the abnormal propagation path along the current transmission direction to obtain the set of abnormal source nodes.

[0088] Specifically, based on the spatial coordinate range of nodes in the graph structure, a coordinate overlap threshold is set. When the threshold condition is met, it is determined that there is a spatial correlation between the abnormal node and the abnormal current location. The spatial coordinate ranges of the two are compared one by one, and the abnormal current location is classified and marked according to the degree of coordinate overlap. The abnormal current location is specifically the location where there is an abnormal current matching degree.

[0089] Based on the hierarchical attribution logic of the graph structure, a link attribution matching rule is established between abnormal nodes and abnormal current locations. The link segment attributes of the abnormal node and the link segment attributes of the abnormal current location are compared, and the current collector ID bound to the abnormal node is matched with the current collector ID corresponding to the abnormal current location.

[0090] Furthermore, starting from the power supply end and ending at the load end, the current transmission direction of the global layer backbone link and the regional layer branch link is determined; typical abnormal nodes distributed in different sub-links are selected as the starting point for tracing; nodes on the reverse path of current transmission are checked segment by segment in the order of starting point, upstream adjacent nodes, upstream branch nodes, and backbone link nodes; during the reverse tracing process, when an upstream node is abnormal, but only the starting point downstream of that node is abnormal and other sub-link nodes are not abnormal, it is checked whether the abnormality of the node is caused by local load. If it is caused by local load, it is not included in the potential abnormality source, and the backbone link nodes are traced upwards; for the potential abnormality source nodes obtained by reverse tracing, all downstream sub-link nodes are checked along the forward current direction; when a downstream node is not abnormal in the forward tracing, but its subsequent nodes are abnormal, the physical attributes of the breakpoint node are checked, and the abnormality propagation breakpoint is marked;

[0091] Summarize all the anomaly source nodes obtained through reverse tracing, and sort them according to priority based on the earliest anomaly start time, the widest forward propagation range, and the most significant anomaly intensity to obtain the set of anomaly source nodes.

[0092] In one specific embodiment, the associated links of abnormal nodes J1 (1km on the main trunk) and J2 (0.5km on the eastern branch) are selected: the real-time current of the main trunk link where J1 is located is 450A, and the dynamic current carrying capacity is calculated to be 400A based on the real-time temperature (95℃) of the link (lower than the rated current carrying capacity of 450A, due to derating at high temperature), and the current matching degree = 450 / 400×100% = 112.5% ​​(exceeding 100%, abnormal); the real-time current of the eastern branch where J2 is located is 300A, and the dynamic current carrying capacity is 320A (derating value at 88℃), and the current matching degree = 300 / 320×100% = 93.75%, but the current fluctuation range within 5 minutes is 15% (exceeding the 10% threshold, judged as abnormal).

[0093] Set a coordinate overlap threshold of 50%: J1 spatial coordinates 1000-1000.2m, abnormal current location (matching degree 112.5%) coordinates 999.8-1000.3m, overlap length 0.2m, overlap degree 80% (≥50%), spatial association; in terms of link affiliation, J1 belongs to the global layer backbone link, the abnormal current location also belongs to the backbone link, and the current collector IDs bound to both are C001, link association confirmed. J2 spatial coordinates 500-500.2m, abnormal current location (abnormal fluctuation location) coordinates 499.9-500.3m, overlap length 0.2m, overlap degree 70%, spatial association; both links belong to the regional layer east branch, collector IDs are both C002, link association confirmed.

[0094] Current transmission direction: Main distribution room (power supply end) → Global layer backbone link → Regional layer East branch (load end). Using J1 and J2 as the tracing starting points:

[0095] Tracing back to J1: Check the 0.8km node upstream from J1 (1km). The abnormal timestamp of this node is 15 minutes earlier than J1 (J1 abnormal 10:00, 0.8km node 09:55), and the current matching degree is 115% (abnormal); continue to the upstream 0.5km node, the current matching degree is 98% (normal), stop tracing, and the 0.8km node is the potential source of abnormality.

[0096] Tracing back to J2: Check the node 0.3km upstream from J2 (0.5km) (current is normal), then check the node 1km upstream on the main trunk (J1, abnormal time 09:58, 2 minutes earlier than J2). J1 is the potential source of the abnormality of J2.

[0097] Forward verification of the 0.8km node: the downstream 1km (J1) and 1.2km nodes are both abnormal (propagation range 0.8-1.2km); the 1.5km node is an isolation joint (breakpoint), and the downstream 1.6km node is not abnormal. The breakpoint does not affect the source determination.

[0098] Potential anomaly sources (0.8km node and J1) were summarized and sorted by priority: the 0.8km node had the earliest anomaly start time (09:55), the widest forward propagation range (0.4km), and the highest current matching degree (115%); the J1 anomaly started later (09:58), and its propagation range was limited to the eastern branch (0.5km). The final set of anomaly source nodes was "main trunk 0.8km node and main trunk 1km node (J1)".

[0099] like Figure 2 As shown, this application provides a cable anomaly detection system with intelligent operation and maintenance capabilities, including:

[0100] The light intensity ratio and temperature measurement position calculation module includes: a pulsed laser generation unit, a light intensity ratio calculation unit, and a temperature measurement position calculation unit. The pulsed laser generation unit determines the preset parameters of the pulsed laser based on the fiber optic parameters of the cable and the monitoring requirements, and generates the pulsed laser. The light intensity ratio calculation unit receives and separates the Stokes light and anti-Stokes light when the pulsed laser generates backscattered light during transmission in the optical fiber, filters out useless signals, converts the signal into a digital signal, and calculates the light intensity ratio. The temperature measurement position calculation unit calculates the temperature based on the light intensity ratio and combines it with the echo time to calculate the temperature measurement position.

[0101] The temperature feature matrix generation module includes: a training dataset construction unit, a temperature measurement position correction unit, and a temperature feature matrix generation unit. The training dataset construction unit builds a training dataset containing physical environment parameters, fiber optic characteristic parameters, raw measurement data, and real-world location labels, and performs preprocessing and feature reconstruction. The temperature measurement position correction unit constructs and trains a convolutional neural network model to obtain the actual transmission speed of light in the fiber optic cable and corrects the temperature measurement position. The temperature feature matrix generation unit combines the corrected temperature measurement position with the corresponding temperature based on a timestamp to obtain the temperature feature matrix.

[0102] The abnormal node determination module includes a graph structure building unit, an association mapping unit, and an abnormal node determination unit. The graph structure building unit acquires the physical indicators of the cable, builds a graph structure according to the global layer, regional layer, and local layer, and binds the physical indicators of the cable to the elements of each level of the graph structure. The association mapping unit standardizes the temperature feature matrix and associates it with the graph elements. The abnormal node determination unit sets basic anomaly detection rules based on the temperature feature matrix and determines abnormal nodes.

[0103] The abnormal source node generation module includes: a current matching degree unit, a spatial and link attribution association unit, and an abnormal source node generation unit. The current matching degree unit acquires real-time current and dynamic current carrying capacity data of the links associated with abnormal nodes and calculates the current matching degree. The spatial and link attribution association unit associates abnormal nodes with abnormal current locations based on the hierarchical relationship of the graph structure. The abnormal source node generation unit traces the abnormal propagation path along the current transmission direction to obtain a set of abnormal source nodes.

[0104] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.

Claims

1. A method for identifying cable anomalies with intelligent operation and maintenance capabilities, characterized in that, Includes the following steps: Based on the fiber optic parameters of the cable and monitoring requirements, the preset parameters for the pulsed laser are determined, and the pulsed laser is generated. When a pulsed laser generates backscattered light as it propagates in an optical fiber, the Stokes light and anti-Stokes light are received and separated. Useless signals are filtered out, and the signal is converted into a digital signal to calculate the intensity ratio. The temperature is calculated based on the intensity ratio, and the temperature measurement location is calculated by combining the echo time. A training dataset containing physical environment parameters, fiber optic characteristic parameters, raw measurement data, and real location labels is constructed, and preprocessed and reconstructed. Convolutional neural network model is constructed and trained to obtain the actual transmission speed of light in optical fiber and correct the temperature measurement position; Based on the timestamp, the corrected temperature measurement location is combined with the corresponding temperature to obtain the temperature feature matrix; Obtain cable physical indicators, build a graph structure according to global, regional and local layers, and bind the cable physical indicators to the elements of each level of the graph structure; standardize the temperature feature matrix and associate it with the graph elements; set basic anomaly detection rules based on the temperature feature matrix to identify abnormal nodes. Obtain real-time current and dynamic current carrying capacity data of the links associated with abnormal nodes, and calculate the current matching degree; based on the hierarchical association relationship of the graph structure, spatially and link-wise associate the abnormal nodes with the abnormal current locations; By tracing the abnormal propagation path along the current transmission direction, the set of abnormal source nodes is obtained.

2. The cable anomaly detection method with intelligent operation and maintenance as described in claim 1, characterized in that, The step of determining preset parameters for the pulsed laser based on the fiber optic parameters of the cable and monitoring requirements, and generating the pulsed laser, includes: The optical fiber parameters include type, core diameter, length, refractive index distribution, loss coefficient, and nonlinear threshold. The monitoring requirements include spatial resolution, temperature measurement range, temperature measurement accuracy, maximum monitoring distance, and data update rate. For the type and core diameter of the optical fiber, a pulse with a corresponding pulse width range is matched. Based on the length and loss coefficient of the optical fiber, the minimum pulse power is calculated while not exceeding the optical fiber nonlinear threshold. The optical transmission speed is estimated by combining the refractive index distribution. Based on monitoring requirements, the pulse width is determined according to spatial resolution requirements; the peak pulse power is set according to the temperature measurement range and accuracy; the pulse repetition frequency is determined based on the maximum monitoring distance and data update rate; and a wavelength that matches the low-loss window of the optical fiber is selected. Based on wavelength-selective lasers, pulse width and repetition frequency are controlled by pulse generators and modulators to generate pulse signals with the required timing. Peak power is adjusted to a preset value by a power amplifier. Finally, the output pulsed laser is injected into the optical fiber via a coupler to complete the generation and emission of the pulsed laser.

3. The cable anomaly detection method with intelligent operation and maintenance as described in claim 1, characterized in that, When the pulsed laser generates backscattered light during transmission in the optical fiber, the process involves receiving and separating the Stokes beam and the anti-Stokes beam, filtering out unwanted signals, converting the signal into a digital signal, and calculating the intensity ratio, including: Backscattered light generated during transmission is separated from the main transmission path by a coupler and introduced into the receiving optical path. A focusing lens is used to converge the scattered light onto the photosensitive surface. Based on the wavelength difference between Stokes light and anti-Stokes light, a wavelength selection device is used to split the converged scattered light, realizing the physical separation of the two target lights. A bandpass filter is used to further suppress ambient stray light outside the bandwidth. A low-noise preamplifier is used to amplify the weak electrical signal, while a filter circuit filters out high-frequency noise. The scattered signals of multiple consecutive pulses are accumulated and averaged to reduce random noise interference. The filtered Stokes and anti-Stokes light signals are input into a photodetector to convert the optical signal into an electrical signal. The electrical signal is then converted into a digital signal by a high-speed analog-to-digital converter, with the sampling rate matched to the repetition frequency of the pulsed laser to preserve the characteristics of light intensity variation over time. The digitized Stokes and anti-Stokes light signals are synchronized and time-aligned, and the peak or integral intensity of the two signals is extracted within the same time window. The ratio of the anti-Stokes light intensity to the Stokes light intensity is calculated as the light intensity ratio.

4. The cable anomaly detection method with intelligent operation and maintenance as described in claim 1, characterized in that, The method of calculating temperature based on light intensity ratio and combining it with echo time to calculate temperature measurement location includes: Based on the physical principle of Raman scattering, a preset temperature-intensity ratio correlation model is invoked to determine the mapping relationship between the intensity ratio and the absolute temperature. The temperature-intensity ratio correlation model includes Raman scattering physical parameters, physical environment parameters, and fiber characteristic parameters. A fiber segment with room temperature at the cable start end is selected as a reference segment with a known temperature in the fiber. The actual temperature and corresponding intensity ratio of this reference segment are obtained and substituted into the temperature-intensity ratio correlation model to correct the proportionality constant in the model. The intensity ratio of each spatial sampling point is substituted into the calibrated model, and the absolute temperature value of the sampling point is calculated by inverse calculation formula. Based on the current refractive index of the monitored optical fiber, the propagation speed of light in the fiber is calculated using the formula v=c / n, where v is the propagation speed, c is the speed of light, and n is the refractive index of the optical fiber. When physical deformation occurs in the optical fiber, the value of v is dynamically adjusted using pre-stored correction coefficients. The timestamp of the laser emission moment is extracted. The timestamp of receiving the backscattered light at that sampling point Calculate echo time difference Corrected through hardware calibration data This eliminates the time deviation caused by non-fiber optic transmission and obtains the true round-trip time of light in the fiber optic cable. According to the formula Where L is the temperature measurement position; when there is a preset marker point in the optical fiber, the calculated L is compared with the position of the preset marker point to correct the position deviation caused by the optical fiber length error; compare the pre-stored optical fiber splice point and connector position information, when the position corresponding to a certain echo signal coincides with the marker point and there is no temperature abnormality at that position, it is determined to be a false echo, and the erroneous calculation result at that position is discarded; when the false echo is superimposed with the normal echo, the effective echo is separated by the difference in signal strength.

5. The cable anomaly detection method with intelligent operation and maintenance as described in claim 1, characterized in that, The construction and training of the convolutional neural network model to obtain the actual propagation speed of light in the optical fiber and to correct the temperature measurement position includes: The input layer of the convolutional neural network model receives a preprocessed feature matrix, which is organized according to spatial sampling points and feature dimensions. The output layer is a single neuron that outputs the actual transmission speed of light in the optical fiber corresponding to the sampling point. The convolutional layer uses a combination of multiple convolutions and pooling. The fully connected regression layer flattens the feature vectors output by the convolutional layer and outputs the actual transmission speed. Dropout layers are added between the fully connected layers, and L2 regularization is applied to the weights of the convolutional and fully connected layers to suppress overfitting. Using the corrected position error as the core loss, a composite loss function is designed in conjunction with physical constraints. The base term is the mean square error between the predicted and actual velocities, and the penalty term is an increasing penalty value when the predicted velocity exceeds the physical reasonable range, ensuring that the output conforms to the physical laws of fiber optic transmission. The training dataset is divided into training, validation, and test sets proportionally, and the Adam optimizer is used, with the validation set loss monitored during training. Feature matrices and corresponding actual velocity labels are input in batches, and network weights are updated through backpropagation. After each round of training, the loss and prediction accuracy are evaluated on the validation set, and the optimal model parameters are recorded. After training, the mean absolute error and root mean square error of velocity prediction are calculated on the test set for model evaluation. Real-time collected feature data is input into the trained model, outputting the actual transmission velocity of each sampling point, recalculating the temperature measurement position, and replacing the original temperature measurement position.

6. The cable anomaly detection method with intelligent operation and maintenance according to claim 1, characterized in that, The process of acquiring cable physical indicators involves building a graph structure based on global, regional, and local layers, and binding the cable physical indicators to elements at each level of the graph structure, including: The physical indicators of the cable include basic attributes, spatial path and key nodes; with the backbone link and core control equipment of the cable network as the skeleton, the global layer elements are defined, including backbone cable segments, core equipment and the start and end points of the backbone link; according to the actual transmission path of the backbone cable, the core elements are connected, the backbone cable segments are represented by lines and the core equipment is represented by icons, and the connection order of each backbone cable segment is marked to form the overall connection framework of the global layer. The region is divided into regions based on the physical area of ​​cable laying, with each region corresponding to an independent subgraph structure. Within each region, the branch cable segments, the power distribution equipment, and the boundary endpoints of the region are extracted. The elements within the region are connected according to the actual route of the branch cables. The branch cable segments are represented by differentiated lines, and the power distribution equipment is represented by icons. The connection relationship between the branch cables and the trunk cables is marked to form an independent topology structure for each region. The region layer subgraph is then associated with the corresponding trunk cable branch point in the global layer. On the global layer trunk cable segment and the regional layer branch cable segment, locate key nodes, including joints, fusion splices, terminal boxes and monitoring sensors; assign a unique identifier to each key node and mark the physical location of the node; map the key nodes to the corresponding cable segments, mark the location of key nodes on the cable segment lines of the global layer and the regional layer, mark the relationship between the node and the cable segment, and form the node distribution structure of the local layer. Bind the physical properties of the cable to each level of the graph structure.

7. The cable anomaly detection method with intelligent operation and maintenance according to claim 1, characterized in that, The method of setting basic anomaly detection rules based on the temperature feature matrix to determine abnormal nodes includes: Based on the temperature feature matrix, temperature feature dimensions for anomaly detection are determined, and an adaptation threshold is set for each temperature feature dimension. When the real-time temperature of a node exceeds the safe temperature threshold for this type of cable, and the duration of the exceedance exceeds a preset duration, it is determined to be an abnormal node with sustained temperature exceedance. When the node's temperature rise rate exceeds a set threshold, and there is no sudden increase in load in the associated current data, it is determined to be an abnormal node with rapid temperature rise. When the difference between the node's temperature and the average temperature of the same link segment exceeds a set threshold, and there are no abnormal fluctuations in the temperature of surrounding nodes, it is determined to be an abnormal node with local high temperature. When the difference between the node's current temperature and the historical average temperature under the same operating conditions over the past month exceeds a set threshold, it is determined to be an abnormal node with temperature baseline drift. Anomaly detection conditions for different temperature feature dimensions are combined. Based on the location coordinates of the sampling points in the temperature feature matrix, the corresponding nodes in the associated graph structure are identified; the nodes are traversed to perform rule checks, and the nodes that trigger the rules are marked in the graph structure. The marking styles are distinguished according to the severity of the anomaly to determine the abnormal nodes.

8. The cable anomaly detection method with intelligent operation and maintenance as described in claim 1, characterized in that, The graph-based hierarchical association relationship spatially and by link to the location of abnormal nodes and abnormal currents includes: Based on the spatial coordinate range of nodes in the graph structure, a coordinate overlap threshold is set. When the threshold condition is met, it is determined that there is a spatial correlation between the abnormal node and the abnormal current location. The spatial coordinate ranges of the two are compared one by one, and the abnormal current location is classified and marked according to the degree of coordinate overlap. The abnormal current location is specifically the location where there is an abnormal current matching degree. Based on the hierarchical attribution logic of the graph structure, a link attribution matching rule is established between abnormal nodes and abnormal current locations. The link segment attributes of the abnormal node and the link segment attributes of the abnormal current location are compared, and the current collector ID bound to the abnormal node is matched with the current collector ID corresponding to the abnormal current location.

9. The cable anomaly detection method with intelligent operation and maintenance as described in claim 1, characterized in that, The process of tracing the abnormal propagation path along the current transmission direction yields a set of abnormal source nodes, including: Starting from the power source and ending at the load, determine the current transmission direction of the global layer backbone link and the regional layer branch link; select typical abnormal nodes distributed in different sub-links as the starting point for tracing; check the nodes on the reverse path of current transmission segment by segment in the order of starting point, upstream adjacent node, upstream branch node, and backbone link node; during the reverse tracing process, if an upstream node is abnormal, but only the starting point downstream of the node is abnormal and other sub-link nodes are not abnormal, check whether the node is caused by a local load. If it is caused by a local load, it is not included as a potential source of abnormality, and the backbone link node is traced upwards; for potential source nodes obtained by reverse tracing, check all downstream sub-link nodes along the forward current direction; when a downstream node is not abnormal during forward tracing, but its subsequent nodes are abnormal, check the physical properties of the node at the breakpoint and mark the breakpoint of abnormal propagation. Summarize all the anomaly source nodes obtained through reverse tracing, and sort them according to priority based on the earliest anomaly start time, the widest forward propagation range, and the most significant anomaly intensity to obtain the set of anomaly source nodes.

10. A cable anomaly detection system with intelligent operation and maintenance, using the cable anomaly detection method with intelligent operation and maintenance as described in any one of claims 1-9, characterized in that, include: The light intensity ratio and temperature measurement position calculation module includes: a pulsed laser generation unit, a light intensity ratio calculation unit, and a temperature measurement position calculation unit. The pulsed laser generation unit determines the preset parameters of the pulsed laser based on the fiber optic parameters of the cable and the monitoring requirements, and generates the pulsed laser. The light intensity ratio calculation unit receives and separates the Stokes light and anti-Stokes light when the pulsed laser generates backscattered light during transmission in the optical fiber, filters out useless signals, converts the signal into a digital signal, and calculates the light intensity ratio. The temperature measurement position calculation unit calculates the temperature based on the light intensity ratio and combines it with the echo time to calculate the temperature measurement position. The temperature feature matrix generation module includes: a training dataset construction unit, a temperature measurement position correction unit, and a temperature feature matrix generation unit. The training dataset construction unit builds a training dataset containing physical environment parameters, fiber optic characteristic parameters, raw measurement data, and real-world location labels, and performs preprocessing and feature reconstruction. The temperature measurement position correction unit constructs and trains a convolutional neural network model to obtain the actual transmission speed of light in the fiber optic cable and corrects the temperature measurement position. The temperature feature matrix generation unit combines the corrected temperature measurement position with the corresponding temperature based on a timestamp to obtain the temperature feature matrix. The abnormal node determination module includes a graph structure building unit, an association mapping unit, and an abnormal node determination unit. The graph structure building unit acquires the physical indicators of the cable, builds a graph structure according to the global layer, regional layer, and local layer, and binds the physical indicators of the cable to the elements of each level of the graph structure. The association mapping unit standardizes the temperature feature matrix and associates it with the graph elements. The abnormal node determination unit sets basic anomaly detection rules based on the temperature feature matrix and determines abnormal nodes. The abnormal source node generation module includes: a current matching degree unit, a spatial and link attribution association unit, and an abnormal source node generation unit. The current matching degree unit acquires real-time current and dynamic current carrying capacity data of the links associated with abnormal nodes and calculates the current matching degree. The spatial and link attribution association unit associates abnormal nodes with abnormal current locations based on the hierarchical relationship of the graph structure. The abnormal source node generation unit traces the abnormal propagation path along the current transmission direction to obtain a set of abnormal source nodes.

Images