Power distribution communication network optical cable fault prediction method and system based on machine learning

Abstract

The invention provides a power distribution communication network optical cable fault prediction method and system based on machine learning. The method comprises the following steps: collecting Rayleigh scattering signals of a plurality of monitoring points along an optical cable; performing unsupervised learning on the optical cable adaptive reconstruction model, and calculating a reconstruction error matrix and an abnormal cumulative metric; constructing an optical cable network diagram, calculating a fault propagation probability matrix, predicting a future optical cable health index, and obtaining a future health index matrix; establishing an environment influence matrix, and adopting an optical cable environment adaptive filtering model; and based on the health index and the fault confidence coefficient matrix after filtering optimization, optical cable maintenance priority scores are calculated and sorted, and a self-adaptive inspection and maintenance strategy is formulated. According to the invention, high-precision monitoring, fault propagation prediction and environmental adaptability optimization of the health state of the optical cable are realized, and the intelligent level of operation and maintenance of the optical cable is improved.

Description

A Machine Learning-Based Method and System for Predicting Optical Cable Faults in Power Distribution Communication Networks Technical Field

[0001] This invention belongs to the field of optical cable fault prediction technology for power distribution communication networks, and particularly relates to a method and system for optical cable fault prediction in power distribution communication networks based on machine learning. Background Technology

[0002] Optical cables in power distribution communication networks bear the responsibility of highly reliable data transmission. Faults in these cables can disrupt power dispatching, remote control, and the stable operation of smart grids. Therefore, rapid detection and early prediction of optical cable faults are crucial for ensuring the stability of power grid communication. Currently, optical cable fault detection primarily relies on optical time-domain reflectometers (OTDRs) and distributed fiber optic sensing (DTS) technologies. The core principle of these methods is to send probe signals into the optical fiber and determine the cable's condition based on changes in the intensity of reflected or scattered light. However, existing technologies still face several key challenges in practical applications, limiting the effectiveness of optical cable fault detection and failing to meet the high precision, real-time performance, and predictive capabilities required by smart grids.

[0003] First, existing OTDR methods primarily rely on intensity reflection measurements, which have limited ability to detect minute damage and early degradation. The main advantage of OTDR lies in its ability to measure the location of fiber damage points and provide precise location information when a fiber breaks. However, OTDR technology is insensitive to small-amplitude signal attenuation or aging effects in optical cables, and can only detect serious damage that has already occurred, such as fiber breaks or loose connectors. It is almost powerless to detect early problems such as minute cracks on the fiber surface or damage to the fiber core. This means that when an OTDR triggers a fault alarm, the fault has often progressed to an irreversible stage, leading to delayed maintenance and potentially even severe communication outages.

[0004] Secondly, existing methods such as OTDR and DTS lack modeling of fiber optic cable fault propagation and cannot predict potentially affected areas. Fiber optic cables in power distribution communication networks typically exhibit complex network topologies. When a segment of a cable is damaged, signal attenuation can spread to other connected fiber segments, even affecting the signal transmission quality of multiple fiber optic links. However, most current detection systems treat fiber optic cables as independent links, focusing only on detecting single-point faults and failing to analyze the correlation between faults between cables. This approach struggles to address the cascading effects of fiber optic cable faults and cannot predict which cables might be affected in the future, leading to inefficient allocation of maintenance resources.

[0005] Furthermore, existing optical cable monitoring technologies are susceptible to environmental interference, resulting in a high false alarm rate and affecting the reliability of fault detection. Optical cables in power distribution communication networks are typically deployed in complex outdoor environments, where factors such as temperature, humidity, mechanical vibration, and electromagnetic interference can all affect the transmission characteristics of optical signals. For example, temperature changes can cause the expansion or contraction of optical fiber materials, leading to false alarms; mechanical vibration can cause short-term fluctuations in optical power, which may not necessarily be caused by damage to the optical cable. Most existing detection methods rely on static threshold settings (e.g., determining a fault when optical power attenuation exceeds a certain fixed value), lacking the ability to adapt to dynamic environmental characteristics, leading to both false alarms and missed alarms.

[0006] Therefore, current optical cable fault monitoring faces three core problems: lack of detection capability for minor damage and early faults, making it impossible to achieve early warning and causing maintenance delays; lack of modeling of fault propagation mechanisms, making it impossible to predict links that may be affected in the future, resulting in inefficient allocation of maintenance resources; and sensitivity to environmental noise, resulting in a high false alarm rate and difficulty in ensuring the stability and reliability of detection results.

[0007] To address the aforementioned issues, this invention proposes a method and system for optical cable fault prediction based on fiber Rayleigh scattering and topology-time fusion. This method overcomes the limitations of existing technologies, enabling high-precision monitoring of optical cable health status, fault propagation prediction, and environmental adaptability optimization, thereby improving the level of intelligent operation and maintenance of optical cables. Summary of the Invention

[0008] The purpose of this invention is to design a machine learning-based method and system for predicting optical cable faults in power distribution communication networks. By integrating fiber Rayleigh scattering signal analysis, topology-time sequence prediction, and environmental adaptive optimization, the fault prediction system can provide early warning of optical cable damage, predict the scope of fault impact, and reduce the interference of environmental noise on the detection results.

[0009] To achieve the above objectives, the first aspect of this invention provides a machine learning-based method for predicting optical cable faults in power distribution communication networks. The method comprises: collecting Rayleigh scattering signals from multiple monitoring points along the optical cable and standardizing them to obtain a standardized Rayleigh scattering signal matrix; constructing a time-series feature matrix based on the standardized Rayleigh scattering signal matrix to train a preset adaptive optical cable reconstruction model; performing unsupervised learning on the adaptive optical cable reconstruction model; calculating a reconstruction error matrix and an anomaly accumulation metric; obtaining an optical cable health index matrix based on the anomaly accumulation metric; identifying potential anomalies based on the reconstruction error matrix; constructing an optical cable network diagram based on the optical cable health index matrix; calculating a fault propagation probability matrix; predicting future optical cable health indices to obtain a future health index matrix; establishing an environmental impact matrix; employing an adaptive optical cable environment filtering model; using the future health index matrix as input to obtain a filtered and optimized health index; and calculating a fault confidence matrix; based on the filtered and optimized health index and the fault confidence matrix, calculating and ranking optical cable maintenance priority scores; formulating adaptive inspection and maintenance strategies; constructing a dynamic feedback mechanism; outputting a maintenance priority list and optimized inspection plan; executing inspection and maintenance tasks; and dynamically updating the data.

[0010] Preferably, the standardization process specifically includes: firstly, adopting an adaptive noise reduction and reconstruction scheme, and obtaining a denoised signal by iteratively calculating a dynamic noise model; then, normalizing the denoised signal to obtain a standardized signal; finally, organizing the standardized signal into an input matrix, where each column represents the time series data of a certain monitoring point and each row represents the data at a certain time point, to obtain a standardized Rayleigh scattering signal matrix.

[0011] Preferably, the time-series feature matrix includes: optical cable... Monitoring point number at time Feature vectors, past and future Historical trends at each time step and gradient characteristics of short-term signal changes.

[0012] Preferably, the unsupervised learning of the optical cable adaptive reconstruction model specifically includes: employing a variational autoencoder structure, wherein the encoder maps the input temporal feature matrix to a latent variable distribution; the decoder attempts to reconstruct the original input from the latent variables to obtain the reconstructed temporal feature matrix; the training process minimizes the loss function of the optical cable adaptive reconstruction model; wherein the first term of the loss function is the reconstruction error, used to measure whether the model can accurately restore the normal optical cable signal pattern, and the second term is a Kullback-Leibler divergence regularization term to ensure the latent variables... The loss function follows a standard normal distribution to improve the model's generalization ability; Represented as:

[0013] in, The input is the time-series feature matrix. For the reconstructed time series feature matrix, The regularization coefficient is used to balance the reconstruction error and the degree of distribution matching; the trained optical cable adaptive reconstruction model is output for fault detection; wherein, the reconstruction error matrix is ​​determined based on the absolute value of the difference between the reconstructed temporal feature matrix and the input temporal feature matrix; the anomaly accumulation metric is obtained by analyzing the error based on the reconstruction error matrix; the optical cable health index is obtained by normalization calculation based on the anomaly accumulation metric; the reconstruction error matrix of the anomaly point is greater than a preset threshold.

[0014] Preferably, constructing the optical cable network diagram based on the optical cable health index matrix specifically includes: establishing physical topology and signal topology; and constructing the optical cable network diagram. , where the node set This includes: each optical cable monitoring point corresponds to one node. Its characteristic is the health index edge set Including: fiber optic cable and optical cable If they are physically adjacent, or their signal transmission paths are logically related, then an edge is established. and assign connection weights :

[0015] in, Indicates optical cable and The geographical distance between them, the greater the distance, the weaker the propagation effect; This indicates differences in health status; the greater the difference, the weaker the impact of transmission. The signal coupling coefficient represents the degree of influence of signals between optical cables. If the signal transmission paths of two optical cables are highly correlated, then... Otherwise, it approaches 0.

[0016] Preferably, a time-series analysis based on a neural network is performed on the optical cable network diagram to obtain a fault propagation probability matrix, which is used to estimate the changes in the health status of the optical cable at future times; a state transition analysis is performed based on the fault propagation probability matrix and the optical cable health index matrix to obtain the corresponding future optical cable health index.

[0017] Preferably, the environmental impact matrix is ​​composed of optical fiber. In time The noise quantization value is composed of the influence of the nearby optical cable environment; the optical cable In time The noise quantization value affected by the nearby optical cable environment is obtained by aggregating the node features of optical cable i.

[0018] Preferably, the step of adopting an adaptive filtering model for the optical cable environment, using the future health index matrix as input, to obtain the filtered and optimized health index, and calculating the fault confidence matrix, specifically includes: obtaining the future health index matrix; using extended Kalman filtering for adaptive noise suppression to generate the filtered and optimized health index; and using logistic regression analysis based on the filtered and optimized health index to obtain the fault confidence matrix.

[0019] Preferably, the maintenance priority scoring The calculation is as follows:

[0020] in, These are weighted parameters used to balance the impact of health index, fault confidence, and fault propagation risk. Indicates optical cable The cumulative impact in the neighboring optical cable fault propagation matrix reflects its potential impact level; The health index after filtering optimization; For optical cable In time Fault confidence; maintenance priority scoring Generate a maintenance priority list to optimize maintenance scheduling; based on the maintenance priority list, formulate adaptive inspection and maintenance strategies, including: if... , If the threshold is high, mark it as a high-priority maintenance object and dispatch a maintenance team immediately; if , If the threshold is low-risk, mark it as a medium-priority inspection item, add it to the inspection plan, and monitor its status changes regularly; if If it does not work, then we will leave it alone and continue to monitor the trend of its health index.

[0021] In a second aspect, the present invention provides a machine learning-based optical cable fault prediction system for power distribution communication networks. The system includes: a data acquisition module for acquiring Rayleigh scattering signals from multiple monitoring points along the optical cable and performing standardization processing to obtain a standardized Rayleigh scattering signal matrix; a health assessment module for constructing a time-series feature matrix based on the standardized Rayleigh scattering signal matrix, used to train a preset optical cable adaptive reconstruction model, performing unsupervised learning on the optical cable adaptive reconstruction model, calculating a reconstruction error matrix and an anomaly accumulation metric, obtaining an optical cable health index matrix based on the anomaly accumulation metric, and identifying potential anomalies based on the reconstruction error matrix; and a propagation prediction module for... The optical cable health index matrix is ​​used to construct an optical cable network diagram, calculate the fault propagation probability matrix, predict the future optical cable health index, and obtain the future health index matrix. The noise suppression module establishes an environmental impact matrix, adopts an adaptive filtering model for the optical cable environment, uses the future health index matrix as input to obtain the filtered and optimized health index, and calculates the fault confidence matrix. The decision support module, based on the filtered and optimized health index and fault confidence matrix, calculates and ranks the optical cable maintenance priority score, formulates adaptive inspection and maintenance strategies, constructs a dynamic feedback mechanism, outputs a maintenance priority list and optimized inspection plan, executes inspection and maintenance tasks, and dynamically updates the data.

[0022] The beneficial technical effects of this invention are at least as follows: Addressing the shortcomings of existing optical cable fault detection technologies, this invention proposes a fault prediction system integrating fiber Rayleigh scattering signal analysis, topology-time series prediction, and environmental adaptive optimization. This system can provide early warning of optical cable damage, predict the scope of fault impact, and reduce the interference of environmental noise on the detection results. Specific innovations are as follows: 1. Employing fiber Rayleigh scattering signal analysis to detect minute damage and early faults. Addressing the limitation of OTDRs in detecting only severe damage, this invention introduces fiber Rayleigh scattering signals for high-precision monitoring. Rayleigh scattering signals are more sensitive than OTDR reflection measurements, capable of capturing microscopic changes in optical fiber materials, including minor cracks, bending, stress accumulation, and other early damage characteristics. This invention uses unsupervised learning (VAE, Variational Autoencoder) to extract latent patterns from Rayleigh scattering signals and construct an optical cable health index, thereby modeling the degradation trend of optical cables. This method can provide early warnings days or even weeks before a fault occurs, avoiding communication interruptions and improving maintenance efficiency.

[0023] 2. Constructing a Topology-Time Series Fusion Prediction Model to Identify Fault Propagation Paths and Achieve Network-wide Fault Trend Prediction. Addressing the lack of fault propagation modeling in traditional methods, this invention proposes a fault propagation prediction method based on topology-time series fusion. A graph neural network (GNN) is used to construct the optical cable topology map of the power distribution communication network, and a time series prediction model (TCN) is combined to predict the evolution trend of the optical cable health index, thereby calculating the fault propagation probability matrix. This method can determine which optical cables may be damaged due to existing faults, enabling precise operation and maintenance planning, avoiding the limitations of single-point detection, and improving network-wide prediction capabilities.

[0024] 3. Adopting environmentally adaptive noise suppression technology to improve the stability and reliability of the detection system. Addressing the issue of traditional optical cable fault detection being susceptible to environmental interference, this invention introduces an Extended Kalman Filter (EKF) for adaptive noise suppression. The EKF dynamically adjusts detection parameters based on historical data, combining external environmental data such as temperature, humidity, and vibration to filter out signal fluctuations caused by non-fault factors, thereby reducing false alarms. Furthermore, this invention employs a dynamic threshold adjustment strategy, dynamically modifying the fault judgment criteria according to the optical cable's operating environment, thus improving the adaptability of the prediction system. Attached Figure Description

[0025] The present invention will be further described with reference to the accompanying drawings, but the embodiments in the drawings do not constitute any limitation on the present invention. For those skilled in the art, other drawings can be obtained based on the following drawings without creative effort.

[0026] Figure 1 is a flowchart of the machine learning-based optical cable fault prediction method for power distribution communication networks according to the present invention.

[0027] Figure 2 is a framework diagram of the optical cable fault prediction system for power distribution communication network based on machine learning according to the present invention. Detailed Implementation

[0028] Embodiments of the present invention are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and should not be construed as limiting the present invention.

[0029] In one or more embodiments, as shown in Figure 1, a machine learning-based method for predicting optical cable faults in power distribution communication networks is disclosed. The method includes the following steps S1-S5: S1, collecting Rayleigh scattering signals from multiple monitoring points along the optical cable and performing standardization processing to obtain a standardized Rayleigh scattering signal matrix.

[0030] Preferably, the standardization process specifically includes: firstly, adopting an adaptive noise reduction and reconstruction scheme, and obtaining a denoised signal by iteratively calculating a dynamic noise model; then, normalizing the denoised signal to obtain a standardized signal; finally, organizing the standardized signal into an input matrix, where each column represents the time series data of a certain monitoring point and each row represents the data at a certain time point, to obtain a standardized Rayleigh scattering signal matrix.

[0031] Specifically, Rayleigh scattering signals were collected from multiple monitoring points along the optical cable. ,in: : No. Monitoring point number at time The intensity of the Rayleigh scattering signal at that location.

[0032] During the collection process, each Due to the influence of external factors such as temperature, humidity, and vibration, the signal contains a lot of noise and needs to be denoised.

[0033] Furthermore, the organization For time series datasets ,in: : No. Monitoring point number at time The standardized Rayleigh scattering signal at that location.

[0034] Output: As input for subsequent steps.

[0035] Furthermore, since Rayleigh scattering signals are significantly affected by random noise, an adaptive noise reduction and reconstruction (ANRR) scheme is first adopted, which iteratively calculates a dynamic noise model. , to obtain the denoised signal :

[0036] in, The noise estimation signal is calculated using time-related constraints.

[0037] Calculation time window Local mean within:

[0038] in, For the first Monitoring point number at time The local mean at that location.

[0039] Calculate the local standard deviation:

[0040] in, For the first Monitoring point number at time The local standard deviation at that point. and calculate This ensures its time-domain stability.

[0041] Furthermore, since the signal amplitudes differ at different monitoring points, directly using the raw data may lead to difficulties in model generalization. Therefore, for Normalization is performed to obtain a standardized signal. :

[0042] in, For the first Monitoring point number at time The standardized Rayleigh scattering signal at that location. and They are the first The mean and standard deviation of monitoring point No. throughout the entire time period.

[0043] Furthermore, the preprocessed normalized Rayleigh scattering signal The data is organized into an input matrix, where each column represents the time-series data of a monitoring point and each row represents the data at a specific time point.

[0044] Output: As input to step S2 (fault identification based on unsupervised learning).

[0045] S2. Construct a time-series feature matrix based on the standardized Rayleigh scattering signal matrix to train a preset optical cable adaptive reconstruction model. Perform unsupervised learning on the optical cable adaptive reconstruction model, calculate the reconstruction error matrix and anomaly accumulation metric, obtain the optical cable health index matrix based on the anomaly accumulation metric, and identify potential anomalies based on the reconstruction error matrix.

[0046] Preferably, the time-series feature matrix includes: optical cable... Monitoring point number at time Feature vectors, past and future Historical trends at each time step and gradient characteristics of short-term signal changes.

[0047] Before an optical cable fault occurs, the Rayleigh scattering signal distribution is usually relatively stable. However, when the optical cable is damaged or stress accumulates, the signal distribution in local areas may become abnormal. Therefore, to more accurately characterize abnormal situations, we construct a time-series feature matrix. :

[0048] in, For optical cable Monitoring point number at time eigenvectors; , For use in depicting the past and future Historical trends at each time step; To enhance sensitivity to abrupt changes by using gradient features to capture short-term signal variations.

[0049] Output: Feature matrix Preferably, the unsupervised learning of the optical cable adaptive reconstruction model, as the input to the unsupervised learning model, specifically includes: Since the health status of the optical cable is affected by environmental factors (temperature, humidity, etc.), the signal pattern may fluctuate slightly at different times; therefore, an adaptive anomaly detection model is used. The goal of unsupervised learning is to learn the signal patterns of the optical cable under normal operating conditions and detect deviations.

[0050] A variational autoencoder (VAE) structure is employed, where the encoder maps the input temporal feature matrix to a latent variable distribution; the decoder attempts to reconstruct the original input from the latent variables to obtain the reconstructed temporal feature matrix; the training objective is to use a variational autoencoder (VAE) structure where the encoder... Input features Mapping to latent variable distribution:

[0051] decoder Attempting to reconstruct the original input from the latent variables:

[0052] The training process minimizes the loss function of the optical cable adaptive reconstruction model; wherein, the first term of the loss function is the reconstruction error, used to measure whether the model can accurately restore the normal optical cable signal pattern, and the second term is the Kullback-Leibler divergence regularization term, to ensure the latent variables The loss function follows a standard normal distribution to improve the model's generalization ability; Represented as:

[0053] in, The input is the time-series feature matrix. For the reconstructed time series feature matrix, is the regularization coefficient, used to balance the reconstruction error and the degree of distribution matching. During training, the model only uses data from healthy optical cables to learn the normal pattern, while the signal pattern of damaged optical cables deviates from the normal pattern. Therefore, the damaged data will produce a large error when the model reconstructs the data.

[0054] The trained optical cable adaptive reconfiguration model is output for fault detection; wherein the reconfiguration error matrix is ​​determined based on the absolute value of the difference between the reconfigured temporal feature matrix and the input temporal feature matrix; the reconfiguration error matrix is ​​calculated as follows:

[0055] in, It reflects the degree of deviation of the optical cable signal mode. The larger the error, the higher the degree of deviation between the signal mode and the healthy state at that moment.

[0056] The anomaly accumulation metric is obtained by analyzing the error based on the reconstructed error matrix. Since some environmental factors may cause short-term anomalies, but do not necessarily represent actual faults, an anomaly accumulation metric is introduced. :

[0057] in, As a measure of smoothing outomas, it can reduce the impact of random noise. This is a smoothing coefficient used to control the weight of the cumulative impact of historical anomalies (generally taken as...). ).

[0058] The optical cable health index is obtained by normalizing the accumulated anomalies; the optical cable health index is calculated as follows:

[0059] in, The lower the value, the higher the degree of damage to the optical cable. Set a health threshold. ,like If the reconstruction error matrix of the anomaly point, which indicates a potential risk of damage to the optical cable, is greater than a preset threshold, then the identified potential anomaly point is considered to be at risk of damage. This is used for subsequent analysis.

[0060] S3. Construct an optical cable network diagram based on the optical cable health index matrix, calculate the fault propagation probability matrix, predict the future optical cable health index, and obtain the future health index matrix.

[0061] Preferably, constructing the optical cable network diagram based on the optical cable health index matrix specifically includes: establishing physical topology and signal topology; and constructing the optical cable network diagram. , where the node set This includes: each optical cable monitoring point corresponds to one node. Its characteristic is the health index edge set Including: fiber optic cable and optical cable If they are physically adjacent, or their signal transmission paths are logically related, then an edge is established. and assign connection weights :

[0062] in, Indicates optical cable and The geographical distance between them, the greater the distance, the weaker the propagation effect; This indicates differences in health status; the greater the difference, the weaker the impact of transmission. The signal coupling coefficient represents the degree of influence of signals between optical cables. If the signal transmission paths of two optical cables are highly correlated, then... Otherwise, it approaches 0.

[0063] Specifically, in power distribution communication networks, optical cables not only have a geographical topology but also a logical topology for signal transmission paths. Both of these topologies affect the propagation mode of optical cable faults, thus requiring the establishment of topological relationships: Physical topology: Based on the actual geographical location of the optical cable, adjacency relationships are defined.

[0064] Signal topology: Based on the characteristics of optical fiber transmission, establish logical connections between nodes.

[0065] Preferably, a time-series analysis based on a neural network is performed on the optical cable network diagram to obtain a fault propagation probability matrix, which is used to estimate the changes in the health status of the optical cable at future times; a state transition analysis is performed based on the fault propagation probability matrix and the optical cable health index matrix to obtain the corresponding future optical cable health index.

[0066] Since optical cable faults not only depend on static topological relationships but are also closely related to historical change trends, it is necessary to construct a spatiotemporal propagation model.

[0067] Define the fault propagation probability matrix Used to estimate changes in the health status of optical cables in the future:

[0068] in, for A matrix of intervals, each element The fault is located in the fiber optic cable. spread to The probability of. The adjacency matrix of the optical cable topology characterizes the influence of spatial structure. It is a time smoothing factor that measures the balance between the influence of the current state and the past states. The environmental disturbance matrix is ​​defined as follows:

[0069] in, For optical cable The adjacent optical cable collection. The environmental noise is mainly caused by external factors such as temperature, humidity, and electromagnetic interference.

[0070] Output: The calculated propagation probability matrix It is used to predict the scope of the impact of a fault.

[0071] Combination Predicting the future health index of optical cables:

[0072] in, This represents the predicted time step. for The health index matrix at each time point reflects the future trend of fault propagation.

[0073] Output: Predicted future optical cable health index This serves as the input for the next step.

[0074] S4. Establish the environmental impact matrix, adopt the optical cable environment adaptive filtering model, take the future health index matrix as input, obtain the filtered and optimized health index, and calculate the fault confidence matrix.

[0075] Preferably, the environmental impact matrix is ​​composed of optical fiber. In time The noise quantization value is composed of the influence of the nearby optical cable environment; the optical cable In time The noise quantization value affected by the nearby optical cable environment is obtained by aggregating the node features of optical cable i.

[0076] Specifically, due to environmental factors (such as temperature, humidity, electromagnetic interference, and mechanical vibration) affecting fiber optic signals, the health index may be affected by noise interference, leading to false alarms or missed alarms. Therefore, it is necessary to establish a dynamic model of environmental noise and define an environmental impact matrix. :

[0077] in, Indicates optical cable In time Noise quantification value affected by the nearby optical cable environment. For optical cable The nearby optical cable collection, For connecting optical cables and The signal affects the weight. This is an independent noise term that simulates random external disturbances (such as temperature changes, electromagnetic interference, etc.).

[0078] Output: The calculated environmental impact matrix This is used for subsequent filter optimization.

[0079] Preferably, the step of adopting an adaptive filtering model for the optical cable environment, using the future health index matrix as input, to obtain the filtered and optimized health index, and calculating the fault confidence matrix, specifically includes: obtaining the future health index matrix; using extended Kalman filtering for adaptive noise suppression to generate the filtered and optimized health index; and using logistic regression analysis based on the filtered and optimized health index to obtain the fault confidence matrix.

[0080] Specifically, due to the dynamic nature of environmental noise, traditional filtering methods are difficult to adapt. This step employs the Cable Adaptive Filtering (CAF) model, with the Extended Kalman Filter (EKF) as its core, integrating the cable health index and environmental noise information to achieve real-time optimization: Prediction Phase:

[0081] in, The state transition matrix of the health index represents the time evolution of the optical cable's health status. This is the environmental impact matrix, representing the weight of environmental noise on the health index of optical cables.

[0082] Update phase:

[0083] in, The dynamic Kalman gain adaptively adjusts the estimated weights based on historical errors.

[0084] Output: Health Index after filtering optimization This reduces environmental noise interference and improves fault detection accuracy.

[0085] Furthermore, since the health index may still have uncertainties after noise suppression, a fault confidence level is calculated to measure the reliability of the prediction:

[0086] in, Indicates optical cable In time The fault confidence level, ranging from [0,1], is used to assist in subsequent maintenance decisions. The adjustment coefficient controls the steepness of the confidence curve. This is the fault detection threshold; a value below this threshold indicates a higher risk of failure.

[0087] Output: Fault confidence matrix It is used to assist in the decision-making of maintenance plans.

[0088] S5. Based on the filtered and optimized health index and fault confidence matrix, calculate and rank the optical cable maintenance priority score, formulate an adaptive inspection and maintenance strategy, construct a dynamic feedback mechanism, output a maintenance priority list and optimized inspection plan, execute inspection and maintenance tasks, and perform dynamic updates.

[0089] Preferably, the maintenance priority scoring The calculation is as follows:

[0090] in, These are weighted parameters used to balance the impact of health index, fault confidence, and fault propagation risk. Indicates optical cable The cumulative impact in the neighboring optical cable fault propagation matrix reflects its potential impact level; The health index after filtering optimization; For optical cable In time Fault confidence; maintenance priority scoring Generate a maintenance priority list to optimize maintenance scheduling; based on the maintenance priority list, formulate adaptive inspection and maintenance strategies, including: if... , If the threshold is high, mark it as a high-priority maintenance object and dispatch a maintenance team immediately; if , If the threshold is low-risk, mark it as a medium-priority inspection item, add it to the inspection plan, and monitor its status changes regularly; if If it does not work, then we will leave it alone and continue to monitor the trend of its health index.

[0091] Furthermore, since the condition of optical cables changes over time, maintenance plans need to be adjusted in real time. Therefore, a dynamic feedback mechanism is constructed: collecting inspection and maintenance feedback data and updating... and Recalculate .

[0092] If the health index significantly improves after repair ( If so, its maintenance priority will be reduced.

[0093] If additional faults are found during the inspection, replace the nearby fiber optic cables. Optimize the propagation prediction model.

[0094] Output: Dynamically adjusted maintenance priority and inspection plan This ensures that maintenance strategies are optimized in real time.

[0095] Furthermore, based on Conduct inspections and, according to Dispatch a maintenance team to repair high-priority faulty optical cables. Inspection data is fed back to the system in real time, updating the health index. This ensures the continuous optimization of the intelligent maintenance system.

[0096] In one or more embodiments, as shown in Figure 2, a machine learning-based optical cable fault prediction system for power distribution communication networks is disclosed. The system includes: a data acquisition module 1: acquiring Rayleigh scattering signals from multiple monitoring points along the optical cable and performing standardization processing to obtain a standardized Rayleigh scattering signal matrix; a health assessment module 2: constructing a time-series feature matrix based on the standardized Rayleigh scattering signal matrix to train a preset optical cable adaptive reconstruction model, performing unsupervised learning on the optical cable adaptive reconstruction model, calculating a reconstruction error matrix and an anomaly accumulation metric, obtaining an optical cable health index matrix based on the anomaly accumulation metric, and identifying potential anomalies based on the reconstruction error matrix; and a propagation prediction module. 3. Based on the optical cable health index matrix, construct an optical cable network diagram, calculate the fault propagation probability matrix, predict the future optical cable health index, and obtain the future health index matrix; Noise suppression module 4: Establish an environmental impact matrix, adopt an optical cable environment adaptive filtering model, take the future health index matrix as input, obtain the filtered and optimized health index, and calculate the fault confidence matrix; Decision support module 5: Based on the filtered and optimized health index and fault confidence matrix, calculate and rank the optical cable maintenance priority score, formulate adaptive inspection and maintenance strategies, construct a dynamic feedback mechanism, output a maintenance priority list and optimized inspection plan, execute inspection and maintenance tasks, and perform dynamic updates.

[0097] It is worth noting that the specific workflow of the machine learning-based optical cable fault prediction system for power distribution communication networks provided in this embodiment is the same as that of the machine learning-based optical cable fault prediction method for power distribution communication networks described in the above embodiments, and will not be repeated here.

[0098] Compared with existing technologies, the machine learning-based optical cable fault prediction system for power distribution communication networks provided in this embodiment of the invention collects Rayleigh scattering signals from multiple monitoring points along the optical cable and performs standardization processing to obtain a standardized Rayleigh scattering signal matrix. A time-series feature matrix is ​​constructed based on the standardized Rayleigh scattering signal matrix to train a preset optical cable adaptive reconstruction model. Unsupervised learning is performed on the optical cable adaptive reconstruction model to calculate the reconstruction error matrix and anomaly accumulation metric. An optical cable health index matrix is ​​obtained based on the anomaly accumulation metric, and potential anomalies are identified based on the reconstruction error matrix. An optical cable network diagram is constructed based on the optical cable health index matrix, and a fault propagation probability matrix is ​​calculated to predict the future optical cable health index, resulting in a future health index matrix. An environmental impact matrix is ​​established, and an optical cable environmental adaptive filtering model is used, with the future health index matrix as input, to obtain a filtered and optimized health index, and a fault confidence matrix is ​​calculated. Based on the filtered and optimized health index and the fault confidence matrix, an optical cable maintenance priority score is calculated and ranked, an adaptive inspection and maintenance strategy is formulated, a dynamic feedback mechanism is constructed, a maintenance priority list and optimized inspection plan are output, inspection and maintenance tasks are executed, and dynamic updates are performed.

[0099] This invention also provides a machine learning-based optical cable fault prediction device for power distribution communication networks, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor. When the processor executes the computer program, it implements the steps as described in the above embodiments of the machine learning-based optical cable fault prediction method for power distribution communication networks, such as steps S1 to S5 in Figure 1; or, when the processor executes the computer program, it implements the functions of each module in the above system embodiments.

[0100] For example, the computer program may be divided into one or more modules, which are stored in the memory and executed by the processor to complete the present invention. The one or more modules may be a series of computer program instruction segments capable of performing specific functions, which describe the execution process of the computer program in the machine learning-based power distribution communication network optical cable fault prediction device.

[0101] The machine learning-based optical cable fault prediction device for power distribution communication networks can be a computing device such as a desktop computer, laptop, handheld computer, or cloud server. This device may include, but is not limited to, processors and memory. Those skilled in the art will understand that the machine learning-based optical cable fault prediction device may also include input / output devices, network access devices, buses, etc.

[0102] The processor can be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor can be a microprocessor or any conventional processor. This processor is the control center of the machine learning-based power distribution communication network optical cable fault prediction device, connecting all parts of the device via various interfaces and lines.

[0103] The memory can be used to store the computer programs and / or modules. The processor implements various functions of the machine learning-based power distribution communication network optical cable fault prediction device by running or executing the computer programs and / or modules stored in the memory and calling the data stored in the memory. The memory may mainly include a program storage area and a data storage area. The program storage area may store the operating system, at least one application program required for a function, etc.; the data storage area may store data created based on the operation of the air conditioner controller, etc. In addition, the memory may include high-speed random access memory, and may also include non-volatile memory, such as hard disk, memory, plug-in hard disk, smart memory card (SMC), secure digital card (SD), flash memory card, at least one disk storage device, flash memory device, or other volatile solid-state storage devices.

[0104] The integrated module of the machine learning-based power distribution communication network optical cable fault prediction device, if implemented as a software functional unit and sold or used as an independent product, can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signals, telecommunication signals, and software distribution media, etc.

[0105] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0106] The above description represents the preferred embodiments of the present invention. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A machine learning-based method for predicting optical cable faults in power distribution communication networks, characterized in that, The method includes: collecting Rayleigh scattering signals from multiple monitoring points along the optical cable and standardizing them to obtain a standardized Rayleigh scattering signal matrix; constructing a time-series feature matrix based on the standardized Rayleigh scattering signal matrix to train a preset optical cable adaptive reconstruction model; performing unsupervised learning on the optical cable adaptive reconstruction model to calculate a reconstruction error matrix and anomaly accumulation metric; obtaining an optical cable health index matrix based on the anomaly accumulation metric; identifying potential anomalies based on the reconstruction error matrix; constructing an optical cable network diagram based on the optical cable health index matrix; calculating a fault propagation probability matrix; predicting future optical cable health indices to obtain a future health index matrix; establishing an environmental impact matrix; using an optical cable environmental adaptive filtering model with the future health index matrix as input to obtain a filtered and optimized health index; and calculating a fault confidence matrix; calculating and ranking optical cable maintenance priority scores based on the filtered and optimized health index and fault confidence matrix; formulating adaptive inspection and maintenance strategies; constructing a dynamic feedback mechanism; outputting a maintenance priority list and optimized inspection plan; executing inspection and maintenance tasks; and dynamically updating the data.

2. The machine learning-based optical cable fault prediction method for power distribution communication networks according to claim 1, characterized in that, The standardization process specifically includes: first, adopting an adaptive noise reduction and reconstruction scheme, and obtaining a denoised signal by iteratively calculating a dynamic noise model; then, normalizing the denoised signal to obtain a standardized signal; finally, organizing the standardized signal into an input matrix, where each column represents the time series data of a certain monitoring point and each row represents the data at a certain time point, to obtain a standardized Rayleigh scattering signal matrix.

3. The method for predicting optical cable faults in power distribution communication networks based on machine learning according to claim 1, characterized in that, The time-series feature matrix includes: optical cable number... Monitoring point number at time Feature vectors, past and future Historical trends at each time step and gradient characteristics of short-term signal changes.

4. The machine learning-based optical cable fault prediction method for power distribution communication networks according to claim 3, characterized in that, The unsupervised learning of the optical cable adaptive reconstruction model specifically includes: employing a variational autoencoder structure, where the encoder maps the input temporal feature matrix to a latent variable distribution; the decoder attempts to reconstruct the original input from the latent variables to obtain the reconstructed temporal feature matrix; the training process minimizes the loss function of the optical cable adaptive reconstruction model; wherein the first term of the loss function is the reconstruction error, used to measure whether the model can accurately restore the normal optical cable signal pattern, and the second term is a Kullback-Leibler divergence regularization term to ensure the latent variables... The loss function follows a standard normal distribution to improve the model's generalization ability; Represented as: in, The input is the time-series feature matrix. For the reconstructed time series feature matrix, The regularization coefficient is used to balance the reconstruction error and the degree of distribution matching; the trained optical cable adaptive reconstruction model is output for fault detection; wherein, the reconstruction error matrix is ​​determined based on the absolute value of the difference between the reconstructed temporal feature matrix and the input temporal feature matrix; the anomaly accumulation metric is obtained by analyzing the error based on the reconstruction error matrix; the optical cable health index is obtained by normalization calculation based on the anomaly accumulation metric; the reconstruction error matrix of the anomaly point is greater than a preset threshold.

5. The machine learning-based optical cable fault prediction method for power distribution communication networks according to claim 1, characterized in that, The construction of the optical cable network diagram based on the optical cable health index matrix specifically includes: establishing the physical topology and signal topology; and constructing the optical cable network diagram. , where the node set This includes: each optical cable monitoring point corresponds to one node. Its characteristic is the health index edge set Including: fiber optic cable and optical cable If they are physically adjacent, or their signal transmission paths are logically related, then an edge is established. and assign connection weights ： in, Indicates optical cable and The geographical distance between them, the greater the distance, the weaker the propagation effect; This indicates differences in health status; the greater the difference, the weaker the impact of transmission. The signal coupling coefficient represents the degree of influence of signals between optical cables. If the signal transmission paths of two optical cables are highly correlated, then... Otherwise, it approaches 0.

6. The machine learning-based optical cable fault prediction method for power distribution communication networks according to claim 5, characterized in that, Based on the optical cable network diagram, a time series analysis driven by a neural network is performed to obtain a fault propagation probability matrix, which is used to estimate the changes in the optical cable health status at future times. Based on the fault propagation probability matrix and the optical cable health index matrix, a state transition analysis is performed to obtain the corresponding future optical cable health index.

7. The method for predicting optical cable faults in power distribution communication networks based on machine learning according to claim 1, characterized in that, The environmental impact matrix is ​​composed of optical cables. In time The noise quantization value is composed of the influence of the nearby optical cable environment; the optical cable In time The noise quantization value affected by the nearby optical cable environment is obtained by aggregating the node features of optical cable i.

8. The machine learning-based optical cable fault prediction method for power distribution communication networks according to claim 7, characterized in that, The method employs an adaptive filtering model for the optical cable environment, using the future health index matrix as input to obtain the filtered and optimized health index, and calculates the fault confidence matrix. Specifically, this includes: obtaining the future health index matrix; using extended Kalman filtering for adaptive noise suppression to generate the filtered and optimized health index; and performing logistic regression analysis based on the filtered and optimized health index to obtain the fault confidence matrix.

9. The machine learning-based optical cable fault prediction method for power distribution communication networks according to claim 8, characterized in that, The maintenance priority score The calculation is as follows: in, These are weighted parameters used to balance the impact of health index, fault confidence, and fault propagation risk. Indicates optical cable The cumulative impact in the neighboring optical cable fault propagation matrix reflects its potential impact level; The health index after filtering optimization; For optical cable In time Fault confidence; maintenance priority scoring Generate a maintenance priority list to optimize maintenance scheduling; based on the maintenance priority list, formulate adaptive inspection and maintenance strategies, including: if... ， If the threshold is high, mark it as a high-priority maintenance object and dispatch a maintenance team immediately; if ， If the threshold is low-risk, mark it as a medium-priority inspection item, add it to the inspection plan, and monitor its status changes regularly; if If it does not work, then we will leave it alone and continue to monitor the trend of its health index.

10. A machine learning-based optical cable fault prediction system for power distribution communication networks, characterized in that, The system includes: a data acquisition module: acquiring Rayleigh scattering signals from multiple monitoring points along the optical cable and standardizing them to obtain a standardized Rayleigh scattering signal matrix; a health assessment module: constructing a time-series feature matrix based on the standardized Rayleigh scattering signal matrix to train a preset optical cable adaptive reconstruction model, performing unsupervised learning on the optical cable adaptive reconstruction model, calculating a reconstruction error matrix and anomaly accumulation metric, obtaining an optical cable health index matrix based on the anomaly accumulation metric, and identifying potential anomalies based on the reconstruction error matrix; a propagation prediction module: constructing an optical cable network diagram based on the optical cable health index matrix, calculating a fault propagation probability matrix, predicting the future optical cable health index, and obtaining a future health index matrix; a noise suppression module: establishing an environmental impact matrix, using an optical cable environmental adaptive filtering model, taking the future health index matrix as input, obtaining a filtered and optimized health index, and calculating a fault confidence matrix; and a decision support module: calculating and ranking optical cable maintenance priority scores based on the filtered and optimized health index and fault confidence matrix, formulating adaptive inspection and maintenance strategies, constructing a dynamic feedback mechanism, outputting a maintenance priority list and optimized inspection plan, executing inspection and maintenance tasks, and dynamically updating the data.

Images