A method and system for early warning of optical cable interruption based on spatiotemporal correlation analysis

Abstract

This application relates to the field of transmission status early warning technology, and in particular to a method and system for early warning of optical cable interruptions based on spatiotemporal correlation analysis. The method includes: adaptively adjusting the window length according to the coupling growth rate between optical cable segments in historical periods, extracting optical power attenuation values ​​to obtain an attenuation sequence and performing differential processing; calculating the coupling coefficient and propagation delay of the differential sequence, constructing directed edge weights based on this and updating the coupling growth rate; extracting the trend features of the attenuation sequence, combining the directed edge weights to perform weighted aggregation on adjacent optical cable segments to obtain fusion features, and then combining the coupling growth rate to obtain the risk probability; when the risk probability is greater than the early warning threshold, calculating the remaining safe time based on the attenuation change rate and issuing an early warning work order. The technical solution of this application can issue early warnings and provide emergency repair guidance in a timely manner before the optical cable transmission link is interrupted.

Description

Technical Field

[0001] This application relates to the field of transmission status early warning technology, and in particular to a method and system for early warning of optical cable interruption based on spatiotemporal correlation analysis. Background Technology

[0002] Fiber optic transmission networks carry massive amounts of core business data, making the security of their transmission links crucial. In practical engineering, multiple fiber optic cables are often laid in parallel within the same underground conduit. When encountering external disturbances such as ground subsidence or construction excavation, multiple fiber optic cables in the same spatial area will successively exhibit precursor signals of increased optical power attenuation. Therefore, an early warning method is needed that can monitor the status of fiber optic cables in real time and accurately capture abnormal signals before the fiber optic transmission link is interrupted, in order to guide emergency repair teams to intervene in advance and avoid service interruptions.

[0003] To meet the daily monitoring needs of fiber optic links, the industry currently typically uses a single-point independent monitoring method based on fixed thresholds. This method mainly relies on the operation and maintenance platform to set a uniform and fixed alarm threshold for the optical power attenuation values ​​reported by each acquisition terminal. When processing data, the real-time acquisition values ​​of each fiber optic cable segment are analyzed in isolation, and the absolute values ​​are compared point by point with the fixed alarm threshold. Once the real-time attenuation value of a single fiber optic cable exceeds the limit, the platform immediately generates an alarm work order and dispatches maintenance personnel to the site for emergency repair.

[0004] However, due to the use of a single fixed threshold for static comparison, the system cannot adaptively extract the attenuation acceleration trend of a single optical cable over a continuous time period, nor can it capture the temporal propagation patterns of multiple adjacent optical cables within the same conduit caused by shared disturbances. This isolated judgment method, lacking joint spatiotemporal analysis across optical cables, results in the dispersion and filtering of precursor signals that emerge successively on adjacent optical cables, failing to trigger effective early warnings before cable blockage and leading to severe warning delays. Summary of the Invention

[0005] To address the technical problems of lacking cross-optical cable spatiotemporal joint analysis and severely delayed early warning, this application provides an optical cable interruption early warning method and system based on spatiotemporal correlation analysis, which can issue early warnings and provide emergency repair guidance in a timely manner before the optical cable transmission link is interrupted.

[0006] In a first aspect, this application provides a method for early warning of optical cable outages based on spatiotemporal correlation analysis. The method includes: adaptively adjusting the window length of each optical cable segment according to the coupling growth rate between segments in a historical period; truncating the optical power attenuation value of each optical cable segment in the current period according to the window length to obtain an attenuation sequence; performing differential processing on the attenuation sequence to obtain a differential sequence; calculating the coupling coefficient and propagation delay of the differential sequences between adjacent optical cable segments using an autocorrelation algorithm, and determining the attenuation change rate of each optical cable segment based on the mean of the differential sequences; and determining the attenuation change rate of each optical cable segment based on the coupling coefficient. The directed edge weights between each optical cable segment are constructed based on the propagation delay; the coupling growth rate is updated based on the coupling coefficient of the current cycle and the historical cycle for the next cycle adjustment; the trend features of the attenuation sequence are extracted, and the trend features of adjacent optical cable segments are weighted and aggregated based on the directed edge weights to obtain the fused features; the risk probability of each optical cable segment is obtained based on the fused features and the updated coupling growth rate; in response to any optical cable segment having a risk probability greater than the warning threshold, the remaining safe time of the optical cable segment is calculated based on the optical power attenuation value and the attenuation change rate, and a warning work order is issued.

[0007] By differentially processing the optical cable attenuation sequence and utilizing the spatiotemporal correlation characteristics of directed network topology aggregation, static background attenuation interference is effectively eliminated, and risk assessment is performed along the actual propagation path of external construction damage, significantly improving the accuracy and sensitivity of optical cable segment crack early warning in complex underground pipeline environments.

[0008] Preferably, the adaptive adjustment of the window length of each optical cable segment based on the coupling growth rate between each optical cable segment in the historical period includes: obtaining the average value of the coupling growth rate of all adjacent optical cable segments of any optical cable segment in the historical period; obtaining the ratio of the absolute value of the average value to a quantization threshold, and rounding the ratio up to obtain an adjustment coefficient; obtaining a reference length, subtracting the product of the adjustment coefficient, the adjustment step size, and the sign function of the average value to obtain a preselected length; responding to the preselected length being between the lower and upper bounds of the length, using the preselected length as the window length of the arbitrary optical cable segment; otherwise, using the boundary value closest to the preselected length as the window length of the arbitrary optical cable segment.

[0009] Based on the average coupling growth rate of adjacent optical cable segments, the window length is dynamically adjusted, allowing the data capture span to flexibly expand and contract with the intensity of external disturbances. When facing construction excavation, the window is contracted to focus on abrupt signals, and when experiencing slow settlement, it is extended to ensure estimation stability, thus solving the problem of information loss caused by a fixed window.

[0010] Preferably, the step of calculating the coupling coefficient and propagation delay of the differential sequences between adjacent optical cable segments using the autocorrelation algorithm includes: adjusting the delay offset of the differential sequences between adjacent optical cable segments one by one; calculating the Pearson cross-correlation coefficient of the differential sequences between adjacent optical cable segments under each delay offset; taking the maximum value of the Pearson cross-correlation coefficient as the coupling coefficient between the adjacent optical cable segments, and taking the delay offset corresponding to the maximum value as the propagation delay between the adjacent optical cable segments.

[0011] Preferably, constructing the directed edge weights between each optical cable segment based on the coupling coefficient and the propagation delay includes: determining whether the propagation delay between any adjacent optical cable segments is greater than zero; in response to the propagation delay being greater than zero, using the coupling coefficient between any adjacent optical cable segments as the directed edge weight between any adjacent optical cable segments; otherwise, resetting the directed edge weight between any adjacent optical cable segments to zero.

[0012] By determining the positive or negative direction of propagation delay to activate or deactivate the directed edge weights between adjacent optical cable segments, it is ensured that the information flow is transmitted unidirectionally along the real disturbance path in the network topology, thus blocking the erroneous feedback of random noise from lagging optical cable segments to leading optical cable segments.

[0013] Preferably, updating the coupling growth rate based on the coupling coefficients of the current period and historical periods includes: obtaining the historical coupling coefficients of a preset historical interval period; calculating the difference between the coupling coefficient of the current period and the historical coupling coefficients, and using the difference as the updated coupling growth rate.

[0014] Preferably, extracting the trend features of the decay sequence includes: using a recurrent neural network to extract features from the decay sequence to obtain trend features.

[0015] Preferably, the weighted aggregation of trend features of adjacent optical cable segments in combination with the directed edge weights includes: obtaining the target trend features of any optical cable segment and the upstream trend features of the upstream optical cable segments with non-zero directed edge weights; using the target trend features as a first feature vector; performing a weighted summation of each upstream trend feature according to the directed edge weights to obtain a second feature vector; and adding the first feature vector and the second feature vector and inputting them into an activation function to obtain the fused features of the arbitrary optical cable segment.

[0016] Preferably, obtaining the risk probability of each optical cable segment based on the fusion characteristics and the updated coupling growth rate includes: calculating the average value of the updated coupling growth rate of each optical cable segment, and concatenating the average value with the fusion characteristics to obtain a feature vector; using the target weight vector to perform a weighted summation of the feature vector, and inputting the summation result into a logistic regression function to obtain the risk probability of each optical cable segment.

[0017] By combining the average growth rate, which reflects the intensity of coupling, with spatiotemporal fusion features, the model is given the ability to perceive the suddenness of external disturbances. This enables it to clearly distinguish between the different degrees of urgency caused by rapid construction damage and slow temperature drift, thereby ensuring the sensitivity of early warning.

[0018] Preferably, calculating the remaining safe duration of each optical cable segment based on the optical power attenuation value and the attenuation change rate includes: obtaining a preset interruption attenuation threshold; calculating the difference between the interruption attenuation threshold and the optical power attenuation value; multiplying the ratio of the difference to the attenuation change rate by the acquisition cycle duration to obtain the remaining safe duration of each optical cable segment.

[0019] By using the difference between the optical power attenuation value and the interruption threshold, and combining it with the attenuation change rate, the remaining safe time of the optical cable can be calculated. This transforms the abstract risk probability into an intuitive time window, providing clear reference information for the operation and maintenance team before the optical cable experiences a substantial break.

[0020] In a second aspect, this application also provides a fiber optic cable outage early warning system based on spatiotemporal correlation analysis, including a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement the fiber optic cable outage early warning method based on spatiotemporal correlation analysis according to the first aspect of this application.

[0021] The technical solution of this application has the following beneficial technical effects: By dynamically adjusting the analysis window length of each optical cable segment through coupling speed increase, the propagation delay is transformed into a directed weight. Then, by combining spatiotemporal characteristics, the risk probability and safe duration are obtained, accurately matching the transmission process of construction disturbances, issuing early warning information in a timely manner before the optical cable transmission link is blocked, and providing reliable repair time. Attached Figure Description

[0022] Figure 1 This is a flowchart of a fiber optic cable interruption early warning method based on spatiotemporal correlation analysis according to an embodiment of this application.

[0023] Figure 2 This is a structural block diagram of an optical cable interruption early warning system based on spatiotemporal correlation analysis according to an embodiment of this application. Detailed Implementation

[0024] According to the first aspect of this application, this application provides a method for early warning of optical cable interruptions based on spatiotemporal correlation analysis. In the daily operation and maintenance scenario of urban backbone optical fiber transmission networks, multiple optical cable segments are laid in parallel within underground ducts. An optical time domain reflectometer (OTDR) monitoring terminal continuously collects the optical power attenuation values ​​of each optical cable segment according to a fixed acquisition cycle. When construction vehicles are excavating or geological subsidence occurs on the ground of that section, the optical cable segment closest to the disturbance source will be the first to show a precursor signal of increased attenuation. Subsequently, adjacent optical cable segments sharing the same duct space will also successively show a similar increasing attenuation phenomenon. Based on the above scenario, the method provided in this application implements early warning through the following steps.

[0025] Figure 1 This is a flowchart of a fiber optic cable interruption early warning method based on spatiotemporal correlation analysis according to an embodiment of this application. Figure 1 As shown, the optical cable interruption early warning method based on spatiotemporal correlation analysis includes steps S101 to S105, which are described in detail below.

[0026] S101, adaptively adjust the window length of each optical cable segment according to the coupling growth rate between each optical cable segment in the historical cycle; and extract the optical power attenuation value of each optical cable segment in the current cycle according to the window length to obtain the attenuation sequence.

[0027] In one embodiment, in order to overcome the shortcomings of a fixed time span in adapting to different stages of external disturbance development, it is necessary to dynamically adjust the number of data samples that need to be back-analyzed based on the rate of change of the correlation between adjacent optical cable segments, so as to balance the capture sensitivity of recent mutation signals with the smoothness and stability of historical normal data.

[0028] Accordingly, the adaptive adjustment of the window length of each optical cable segment based on the coupling growth rate between each optical cable segment in the historical period includes: obtaining the average value of the coupling growth rate of all adjacent optical cable segments of any optical cable segment in the historical period; obtaining the ratio of the absolute value of the average value to the quantization threshold, and rounding the ratio up to obtain the adjustment coefficient; obtaining the reference length, subtracting the product of the adjustment coefficient, the adjustment step size and the sign function of the average value to obtain the preselected length; responding to the preselected length being between the lower and upper bounds of the length, using the preselected length as the window length of the arbitrary optical cable segment; otherwise, using the boundary value closest to the preselected length as the window length of the arbitrary optical cable segment.

[0029] The pre-selected length of any optical cable segment and the final window length satisfy the following relationship: In the formula, The pre-selected length corresponding to the current cycle; The window length for the current period; The preferred length is 12 acquisition cycles, and its value is determined by increasing the statistical average of the longest duration of historical disturbance precursor signals by 20%. To adjust the step size, a value of 2 acquisition cycles is used, and the value is determined based on the average duration of the acceleration phase preceding historical interruption events. It is a symbolic function; The rounding up symbol; The average coupling growth rate of all adjacent optical cable segments in any given optical cable segment over a historical period is used. If historical period data is lacking during the initial startup phase, the final window length is directly assigned as the baseline length. The quantification threshold is set to 0.1, which is determined based on one time the standard deviation of the normal fluctuations in the coupling growth rate in historical data. The lower bound for the length is set to 6 collection cycles to ensure that the subsequent correlation calculation has the minimum number of sample points required for stable estimation. The upper bound of the length is set at 16 collection cycles, serving as an empirical upper limit to avoid introducing too many historical data segments that are irrelevant to the current disturbance. and Function combinations are used to restrict a preselected length between a lower bound and an upper bound.

[0030] Understandably, when the average coupling growth rate across the optical cable segment in the previous cycle is greater than zero, it indicates that the coupling effect between the optical cable segments is rapidly increasing. At this time, the window length is shortened compared to the baseline length, thereby significantly focusing on recent abrupt signals and improving the response sensitivity to rapid disturbances. When the average coupling growth rate is less than or equal to zero, it indicates that the coupling tends to stabilize or decrease. At this time, the window length is lengthened, increasing the sample size on which subsequent data processing depends and improving the estimation stability of slow geological subsidence disturbances.

[0031] After adaptively adjusting the window length, in order to transform the discrete acquisition points into a continuous data structure suitable for time-series analysis, the optical power attenuation values ​​of each optical cable segment in the current period are truncated based on the window length. The constructed attenuation sequence satisfies the following relationship: In the formula, This represents the attenuation sequence of any optical cable segment in the current period. This is the optical power attenuation value for the current period, which is uploaded in real time by the optical time domain reflectometer at a sampling period of 1 minute. This is the first optical power attenuation value traced back to history based on the window length.

[0032] Thus, when urban underground pipelines are subjected to external disturbances such as excavation by construction vehicles or geological changes, the window length of the current period is dynamically adjusted based on the coupling growth rate of the historical cycle. Within this window length, the continuously collected optical power attenuation values ​​are truncated to generate an attenuation sequence. This ensures that the extracted data span always accurately matches the disturbance stage, avoiding the signal dilution problem caused by a single fixed time span, and providing a data foundation for subsequent time-series correlation analysis of multiple optical cable segments.

[0033] S102, perform differential processing on the attenuation sequence to obtain a differential sequence; calculate the coupling coefficient and propagation delay of the differential sequence between adjacent optical cable segments, and determine the attenuation change rate of each optical cable segment based on the mean of the differential sequence.

[0034] In one embodiment, to eliminate the differences in the basic background attenuation values ​​of each optical cable segment, it is necessary to transform the non-stationary absolute value sequence into a stationary sequence of changes, thereby accurately reflecting the dynamic changes in the attenuation trend. The attenuation sequence is differentially processed, and the constructed differential sequence satisfies the following relationship: In the formula, This is the difference sequence for the current period; This represents the optical power attenuation value for the current cycle. This is the first optical power attenuation value traced back to history based on the window length; This is the second optical power attenuation value traced back to history; This represents the optical power attenuation value from the previous cycle. This represents the window length for the current period.

[0035] After obtaining a stationary differential sequence, in order to evaluate the order and correlation strength of different optical cable segments within the same underground pipeline affected by the same external disturbance, the time delay offset of the differential sequence between adjacent optical cable segments is adjusted one by one; the Pearson cross-correlation coefficient of the differential sequence between the adjacent optical cable segments is calculated under each time delay offset; the maximum value of the Pearson cross-correlation coefficient is taken as the coupling coefficient between the adjacent optical cable segments, and the time delay offset corresponding to the maximum value is taken as the propagation time delay between the adjacent optical cable segments.

[0036] The Pearson cross-correlation coefficient is calculated using the standard Pearson cross-correlation function, and the maximum value and the corresponding time delay offset satisfy the following relationship: In the formula, For adjacent optical cable segments The coupling coefficient between them ranges from zero to one; For transmission delay; and Adjacent optical cable segments The corresponding difference sequences between them; This is the time delay offset; Delay offset Lower difference sequence and The standard Pearson cross-correlation function; For the offset boundary, the preferred value is 5 acquisition cycles, and its value is determined based on the maximum time delay of the disturbance in the pipeline from one end to the other. To find the maximum value function; The function that finds the maximum value of the independent variable.

[0037] Understandably, a larger coupling coefficient indicates a more similar pattern between the differential sequences of two adjacent optical cable segments within the current window length, and a higher likelihood of being affected by the same external disturbance, such as construction excavation. The sign of the propagation delay indicates the direction of the disturbance's propagation; for example, when construction excavation begins from the optical cable segment... Starting from one side, the fiber optic cable segment The attenuation sequence will be earlier than the optical cable segment. An increasing trend appears after two to three acquisition cycles. If the propagation delay is greater than zero, it indicates that the decay change of the former is ahead of the latter on the time axis, clearly reflecting the process of the disturbance propagating from the former to the latter.

[0038] Furthermore, to preserve the key feature reflecting the average rate of attenuation change in the differential sequence, the attenuation change rate of each optical cable segment is determined based on the mean of the differential sequence. Specifically, this attenuation change rate satisfies the following relationship: In the formula, This represents the rate of decay change in the current period; To find the mean function; This is the difference sequence for the current period. The rate of decay change is used as the difference sequence to support the subsequent assessment of the safe duration.

[0039] Thus, by eliminating static interference in the absolute value sequence through differential processing and using correlation scanning under multiple time delay offsets, the variation characteristics and average attenuation rate of the optical cable segment itself are accurately extracted. The complex cross-optical cable segment temporal correlation in the underground pipeline is transformed from a qualitative description into two quantifiable values: coupling coefficient and propagation delay. These values ​​can characterize the propagation path of external disturbances in the pipeline space, providing a data foundation for the subsequent construction of directed network topology.

[0040] S103, construct the directed edge weights between each optical cable segment based on the coupling coefficient and the propagation delay; update the coupling speedup based on the coupling coefficient of the current cycle and the historical cycle for use in the next cycle adjustment.

[0041] In one embodiment, after obtaining the coupling coefficient and propagation delay between each adjacent optical cable segment, in order to encode these data reflecting the real-time coupling strength and disturbance propagation direction into the topology of the spatiotemporal graph network and ensure that network information is transmitted only along the real disturbance path, constructing the directed edge weights between each optical cable segment based on the coupling coefficient and the propagation delay includes: determining whether the propagation delay between any adjacent optical cable segment is greater than zero; in response to the propagation delay being greater than zero, using the coupling coefficient between any adjacent optical cable segment as the directed edge weight between any adjacent optical cable segment; otherwise, resetting the directed edge weight between any adjacent optical cable segment to zero.

[0042] The directed edge weights between any two adjacent optical cable segments satisfy the following relationship: In the formula, For optical cable segment Pointing to adjacent optical cable segments The weight of the directed edge; The coupling coefficient for the current cycle; For direction indication function, when At the time of its establishment, ,on the contrary ; This is to account for the transmission delay.

[0043] Understandably, when the propagation delay is greater than zero, it indicates that the disturbance was indeed caused by the optical cable segment. Transmitted to adjacent optical cable segments At this time, the direction indicator function takes the value of 1, the directed edge in this direction is activated and the weight is directly equal to the coupling coefficient. The larger the coupling coefficient, the stronger the subsequent aggregation of the precursor signal. When the propagation delay is less than or equal to zero, the direction indicator function takes the value of 0, the directed edge in this direction is forcibly closed and the weight is reset to zero, thereby avoiding the introduction of noise errors from the lagging optical cable segment that has not yet produced a precursor signal into the feature calculation of the leading optical cable segment.

[0044] Furthermore, considering that unforeseen events such as construction and excavation can cause a rapid increase in the coupling strength between optical cable segments within underground pipelines, in order to accurately capture this dynamic change rate to differentiate the urgency of the disturbance and provide a closed-loop feedback signal for window control in the next cycle, it is also necessary to update the coupling growth rate based on the coupling coefficients of the current cycle and historical cycles. Specifically, this includes: obtaining the historical coupling coefficients for a preset historical interval; calculating the difference between the coupling coefficient of the current cycle and the historical coupling coefficients; and using the difference as the updated coupling growth rate.

[0045] It should be noted that the updated coupling growth rate satisfies the following relation: In the formula, The updated coupling speed; The coupling coefficient for the current cycle; The historical coupling coefficient is the preset historical interval period; the number 3 in the subscript represents the number of historical interval periods. Preferably, the interval is 3 collection periods to ensure that the difference can effectively reflect the recent acceleration trend of the coupling strength.

[0046] Understandably, the coupling growth rate reflects the rate of change in the coupling relationship between two adjacent optical cable segments. A larger coupling growth rate indicates that the synchronous impact of external disturbances such as construction and excavation on the two optical cable segments is rapidly increasing, and the event is highly unpredictable. The updated coupling growth rate not only characterizes the urgency of external disturbances to participate in subsequent risk assessments, but also serves as a basis for adaptive adjustment of the window length in the next acquisition cycle.

[0047] Thus, when urban underground pipelines face complex disturbances such as external construction and excavation, by jointly encoding the coupling coefficient and propagation delay into dynamic directed edge weights, the network topology can be updated in real time with each collection cycle, ensuring that feature information flows unidirectionally only along the actual propagation path of the disturbance; at the same time, the calculated coupling growth rate quantifies the suddenness of the disturbance outbreak and improves the perception accuracy of the early warning method for sudden disturbances.

[0048] S104, extract the trend features of the attenuation sequence, combine the directed edge weights to perform weighted aggregation of the trend features of adjacent optical cable segments to obtain fused features; obtain the risk probability of each optical cable segment based on the fused features and the updated coupling growth rate.

[0049] In one embodiment, after obtaining the attenuation sequence, in order to accurately capture the continuous increase in optical power attenuation value in the early stage of construction excavation or geological subsidence from the time dimension, a recurrent neural network is used to extract features from the attenuation sequence, extract the attenuation change pattern of each optical cable segment, and then obtain trend features.

[0050] The recurrent neural network can be an existing recurrent neural network such as LSTM or Transformer, and this application does not impose any restrictions.

[0051] Subsequently, in the spatial dimension, the local single-point precursor signal is correlated and diffused across optical cable segments along the actual disturbance propagation path. The weighted aggregation of trend features of adjacent optical cable segments, combined with the directed edge weights, includes: obtaining the target trend feature of any optical cable segment and the upstream trend feature of the upstream optical cable segment with non-zero directed edge weights; using the target trend feature as a first feature vector; weighting and summing each upstream trend feature according to the directed edge weights to obtain a second feature vector; and adding the first and second feature vectors and inputting the sum into an activation function to obtain the fused feature of the arbitrary optical cable segment.

[0052] It should be noted that any optical cable segment is referred to as an optical cable segment. For example, the above weighted aggregation operation satisfies the following relation: In the formula, This is the first eigenvector; The target trend characteristics of any optical cable segment; This is the second feature vector; The set of all upstream optical cable segments that satisfy the condition that the directed edge weight is non-zero; The weight of the directed edge corresponding to the upstream optical cable segment pointing to the arbitrary optical cable segment; This corresponds to the upstream trend characteristics; For example, the fused feature is mapped as a 32-dimensional vector; The activation function is either the Sigmoid function or the ReLU function.

[0053] Understandably, the larger the weight of the directed edge of the upstream optical cable segment, the stronger the physical correlation between the construction and excavation disturbance and the transmission from the upstream optical cable segment to the target optical cable segment. The higher the contribution of the upstream trend feature to the second feature vector in the weighted summation process, the more accurate the directional flow of feature information within the underground pipeline spatial topology is achieved.

[0054] Furthermore, in order for the model to distinguish between sudden bursts of rapid construction damage and long-term stable slow temperature drift, the average value of the updated coupling growth rate of each optical cable segment is calculated, and the average value is concatenated with the fused features to obtain a feature vector; the feature vector is weighted and summed using the target weight vector, and the summation result is input into the logistic regression function to obtain the risk probability of each optical cable segment.

[0055] Among them, optical cable section For example, the risk probability satisfies the following relationship: In the formula, The concatenated feature vector; Features of fusion; This represents the average of the calculated growth rates of each coupling. For optical cable segment The probability of risk; For logistic regression function; This is the transpose of the target weight vector, which represents the learnable parameters of the model.

[0056] It should be noted that both the recurrent neural network and the target weight vector can be obtained through supervised training using historical operation and maintenance data. Preferably, samples from 30 minutes prior to the occurrence of the historical optical cable interruption event are used as positive samples. The loss function is the binary cross-entropy loss function, and gradient descent is used to update the recurrent neural network and the target weight vector. The trained recurrent neural network and target weight vector are obtained when the maximum number of updates is reached or the value of the binary cross-entropy loss function is less than a preset value. The maximum number of updates can be set to 150, and the preset value can be set to 0.001.

[0057] Thus, when dealing with complex external disturbances to urban underground pipelines, the attenuation variation pattern of a single optical cable segment is extracted through a recurrent neural network. Then, the spatial information of multiple optical cable segments is aggregated through a graph convolutional network along the real disturbance propagation path determined by the directed edge weights. Finally, the output bias of the logistic regression function is corrected by introducing the average value of the coupled growth rate that reflects the suddenness of the event. This makes the final output risk probability integrate the leading trend in the time dimension, the transmission correlation in the spatial dimension, and the suddenness and urgency of the event dimension, thereby improving the accuracy of the risk probability.

[0058] S105, in response to the risk probability of any optical cable segment being greater than the warning threshold, calculate the remaining safe time of the optical cable segment based on the optical power attenuation value and the attenuation change rate, and issue a warning work order.

[0059] In one embodiment, the early warning method further includes obtaining the early warning threshold. The method for obtaining the early warning threshold includes: obtaining the 10th percentile of the risk probability in the positive sample distribution and the 90th percentile of the risk probability in the negative sample distribution based on historical operation and maintenance data; and determining the early warning threshold within the numerical interval formed by the 10th percentile and the 90th percentile. In this embodiment, the early warning threshold is set to 0.6.

[0060] After dynamically setting the warning threshold, the risk probability of each optical cable segment in the underground pipeline at the current moment is determined. When the risk probability is less than the warning threshold, no action is triggered and the system waits for the next cycle update; when the risk probability is greater than the warning threshold, the remaining safe time is provided to the repair team as a reference for scheduling.

[0061] Specifically, calculating the remaining safe duration of the optical cable segment based on the optical power attenuation value and the attenuation change rate includes: obtaining a preset interruption attenuation threshold; calculating the difference between the interruption attenuation threshold and the optical power attenuation value; and multiplying the ratio of the difference to the attenuation change rate by the acquisition cycle duration to obtain the remaining safe duration of each optical cable segment.

[0062] It should be noted that the calculation of the remaining safe time is based on a linear prediction of the current attenuation rate, using the optical cable segment as an example. For example, the remaining safe duration satisfies the following relationship: In the formula, For optical cable segment The remaining safe time, in minutes; The preset interruption attenuation threshold is, for example, obtained by querying the historical alarm database to obtain the statistical average of the attenuation values ​​when the corresponding model of optical cable experienced a physical break. For the current period of optical cable segment Optical power attenuation value; For optical cable segment The rate of change of decay; This represents the data collection period duration, measured in minutes. To avoid calculation errors, when... When the value is less than or equal to 0, it indicates that the attenuation trend of the optical cable is improving or remaining stable. At this time, it is determined that there is no immediate risk of interruption of the optical cable, and the remaining safe time can be determined to be infinite or set to the preset safe time upper limit value.

[0063] Finally, an early warning work order containing the remaining safe time is sent to the operation and maintenance platform. For example, the early warning work order contains the trigger time, the corresponding optical cable segment number, the current risk probability, and the estimated remaining safe time, thereby assisting the repair team in making the optimal scheduling decision on when to arrive at the site.

[0064] According to a second aspect of this application, this application also provides an optical cable interruption early warning system based on spatiotemporal correlation analysis. Figure 2 This is a structural block diagram of a fiber optic cable interruption early warning system based on spatiotemporal correlation analysis, according to an embodiment of this application. Figure 2 As shown, the system 50 includes a processor and a memory. The memory stores computer program instructions, which, when executed by the processor, implement the optical cable interruption early warning method based on spatiotemporal correlation analysis according to the first aspect of this application. The system also includes other components well-known to those skilled in the art, such as a communication bus and a communication interface. Their configurations and functions are known in the art and will not be described further here.

[0065] It should be noted that, for those skilled in the art, several modifications and improvements can be made without departing from the concept of this application, and these all fall within the scope of protection of this application.

Claims

1. A method for early warning of optical cable interruption based on spatiotemporal correlation analysis, characterized in that, The early warning method includes: The window length of each optical cable segment is adaptively adjusted based on the coupling growth rate between each optical cable segment in the historical cycle; the optical power attenuation value of each optical cable segment in the current cycle is truncated based on the window length to obtain the attenuation sequence, and the coupling growth rate reflects the rate of change of the coupling relationship between two adjacent optical cable segments; The attenuation sequence is differentially processed to obtain a differential sequence; the coupling coefficient and propagation delay of the differential sequences between adjacent optical cable segments are calculated using an autocorrelation algorithm, and the attenuation change rate of each optical cable segment is determined based on the mean of the differential sequences. The larger the coupling coefficient, the more similar the shapes of the differential sequences of two adjacent optical cable segments are within the current window length. The directed edge weights between each optical cable segment are constructed based on the coupling coefficient and the propagation delay; the coupling speedup is updated based on the coupling coefficient of the current cycle and the historical cycle for use in the next cycle adjustment. The trend features of the attenuation sequence are extracted, and the trend features of adjacent optical cable segments are weighted and aggregated in combination with the directed edge weights to obtain the fused features; the risk probability of each optical cable segment is obtained based on the fused features and the updated coupling growth rate. If the risk probability of any optical cable segment exceeds the warning threshold, the remaining safe time of the optical cable segment is calculated based on the optical power attenuation value and the attenuation change rate, and a warning work order is issued.

2. The optical cable interruption early warning method based on spatiotemporal correlation analysis according to claim 1, characterized in that, The adaptive adjustment of the window length of each optical cable segment based on the coupling growth rate between each optical cable segment in the historical cycle includes: Obtain the average coupling growth rate of all adjacent optical cable segments of any given optical cable segment during the historical period; Obtain the ratio of the absolute value of the average value to the quantization threshold, and round the ratio up to obtain the adjustment coefficient; Obtain the baseline length, subtract the product of the adjustment coefficient, the adjustment step size, and the sign function of the average value to obtain the preselected length; In response to the preselected length being between the lower and upper bounds of the length, the preselected length is used as the window length of the arbitrary optical cable segment; otherwise, the boundary value closest to the preselected length is used as the window length of the arbitrary optical cable segment.

3. The optical cable interruption early warning method based on spatiotemporal correlation analysis according to claim 1, characterized in that, The calculation of the coupling coefficient and propagation delay of the differential sequence between adjacent optical cable segments using the autocorrelation algorithm includes: Adjust the time delay offset of the differential sequence between adjacent optical cable segments one by one; Calculate the Pearson cross-correlation coefficient of the differential sequence between adjacent optical cable segments under each time delay offset; take the maximum value of the Pearson cross-correlation coefficient as the coupling coefficient between the adjacent optical cable segments, and take the time delay offset corresponding to the maximum value as the propagation time delay between the adjacent optical cable segments.

4. The optical cable interruption early warning method based on spatiotemporal correlation analysis according to claim 1, characterized in that, Constructing the directed edge weights between each optical cable segment based on the coupling coefficient and the propagation delay includes: determining whether the propagation delay between any two adjacent optical cable segments is greater than zero; In response to the propagation delay being greater than zero, the coupling coefficient between any two adjacent optical cable segments is used as the weight of the directed edge between the two adjacent optical cable segments; otherwise, the weight of the directed edge between the two adjacent optical cable segments is reset to zero.

5. The optical cable interruption early warning method based on spatiotemporal correlation analysis according to claim 1, characterized in that, The step of updating the coupling growth rate based on the coupling coefficient of the current cycle and the historical cycle includes: obtaining the historical coupling coefficient of a preset historical interval period; Calculate the difference between the coupling coefficient in the current period and the historical coupling coefficient, and use the difference as the updated coupling growth rate.

6. The optical cable interruption early warning method based on spatiotemporal correlation analysis according to claim 1, characterized in that, The extraction of trend features from the decay sequence includes: The trend features are obtained by extracting features from the decay sequence using a recurrent neural network.

7. The optical cable interruption early warning method based on spatiotemporal correlation analysis according to claim 1, characterized in that, The weighted aggregation of trend features of adjacent optical cable segments based on the directed edge weights includes: obtaining the target trend features of any optical cable segment, and the upstream trend features of the upstream optical cable segment with non-zero directed edge weights. The target trend feature is used as the first feature vector; The second feature vector is obtained by weighting and summing the upstream trend features according to the directed edge weights. The first feature vector and the second feature vector are added together and then input into the activation function to obtain the fused features of the arbitrary optical cable segment.

8. The optical cable interruption early warning method based on spatiotemporal correlation analysis according to claim 1, characterized in that, Obtaining the risk probability of each optical cable segment based on the fusion characteristics and the updated coupling growth rate includes: calculating the average value of the updated coupling growth rate of each optical cable segment, and concatenating the average value with the fusion characteristics to obtain a feature vector; The feature vector is weighted and summed using the target weight vector, and the summation result is input into the logistic regression function to obtain the risk probability of each optical cable segment.

9. The optical cable interruption early warning method based on spatiotemporal correlation analysis according to claim 1, characterized in that, The remaining safe time for each optical cable segment is calculated based on the optical power attenuation value and the attenuation change rate, including: Obtain the preset interrupt decay threshold; Calculate the difference between the interruption attenuation threshold and the optical power attenuation value; The remaining safe time for each optical cable segment is obtained by multiplying the ratio of the difference to the attenuation change rate by the acquisition period.

10. A fiber optic cable breakage early warning system based on spatiotemporal correlation analysis, characterized in that, It includes a processor and a memory, the memory storing computer program instructions, which, when executed by the processor, implement the optical cable interruption early warning method based on spatiotemporal correlation analysis according to any one of claims 1 to 9.

Images