Backbone network line attenuation analysis early warning method and system based on deep learning

Abstract

The invention discloses a backbone network line attenuation analysis early warning method and system based on deep learning, and belongs to the technical field of line attenuation analysis. Acquiring channels in a link, and constructing a channel database; generating a data acquisition node sequence, and setting a time window for collecting receiving and transmitting optical power data of a channel; establishing a receiving and transmitting light power data set, and analyzing and quantifying a data acquisition period by changing the size of a time window; the overall loss fault degree of the links is analyzed and evaluated, and the correlation degree between the different links is calculated; presetting an overall loss fault degree threshold value, and analyzing and outputting an early warning; according to the invention, through power data analysis and fluctuation monitoring based on the time window, the data acquisition period is optimized, and the network health state identification capability is improved; by evaluating the overall loss fault degree of the links and calculating the correlation degree between the links, potential faults are found in time, a reliable basis is provided for network maintenance, and the operation stability and reliability of the network are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of line loss analysis, and specifically to a backbone network line loss analysis early warning method and system based on deep learning. Background Art

[0002] With the rapid development of communication technology and network scale, backbone networks, as the core components of modern communication networks, carry a large amount of data and information transmission tasks; the stability of backbone network lines and the quality of data transmission directly affect the operating efficiency and user experience of the entire network; in this context, fiber-optic communication technology has become the main transmission means of backbone networks; fiber-optic communication lines have the advantages of high bandwidth, long transmission distance, and strong anti-interference, but at the same time, fiber-optic lines will produce attenuation phenomena under long-term use or due to external factors, resulting in reduced transmission signal strength and increased data bit error rate, and even communication interruption in severe cases; at present, traditional line detection and early warning methods mostly rely on periodic manual detection and fixed threshold monitoring, which are difficult to cope with complex network environments and dynamically changing attenuation characteristics; in addition, existing methods often have problems such as heavy computational burden and poor real-time performance when processing large amounts of data, and cannot meet the efficient operation requirements of modern large-scale backbone networks.

[0003] Although the existing technology can evaluate the operating status of the optical fiber line through the channel's transmitted and received optical power data, it has many shortcomings when accurately analyzing data fluctuations and link losses. Most traditional methods perform data analysis based on a fixed time window or a single sampling period, and fail to fully consider the fluctuation characteristics of the optical power signal over time and environmental changes, resulting in low detection sensitivity and easy miss of early fault warning signals. The existing attenuation analysis method fails to effectively combine the correlation between each link and channel, making it difficult to provide accurate warnings when multiple links are attenuated or correlated. Summary of the invention

[0004] The purpose of the present invention is to provide a backbone network line loss analysis and early warning method and system based on deep learning to solve the problems raised in the above background technology.

[0005] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0006] A backbone network line attenuation analysis and early warning method based on deep learning, the method includes the following steps: uniformly numbering the links in the backbone network and building a link database; obtaining the channels in the links and building a channel database; coordinating the data acquisition nodes, generating a data acquisition node sequence, and setting a time window to collect the channel's transmit and receive optical power data; based on the time window, establishing a transmit and receive optical power data set, and analyzing and quantifying the data acquisition cycle by changing the scale of the time window; analyzing and evaluating the overall loss fault degree of the link, and calculating the correlation between different links; presetting the overall loss fault degree threshold, analyzing and outputting early warnings.

[0007] As a preferred solution of the backbone network line attenuation analysis and early warning method based on deep learning of the present invention, the links in the backbone network are uniformly numbered to construct a link database, which is recorded as LS={LS a |a∈[1,A]}, where LS a represents the ath link, and A represents the total number of links.

[0008] Get Link LS a The channels in the channel database are constructed, denoted as CH = {CH ab |a∈[1,A],b∈[1,B]}, where CH ab Indicates link LS a The bth channel in , B represents the total number of channels; the channel represents a path for data transmission on the link.

[0009] As a preferred solution of the backbone network line loss analysis and early warning method based on deep learning of the present invention, the time window is set as follows: in units of minutes, a data acquisition node sequence is constructed, denoted as Y={Y t |t∈[1,T]}, where Y t represents the tth data collection node, T represents the total number of data collection nodes, and one data collection node collects the transmitted and received optical power data once.

[0010] Taking i data acquisition nodes as a step length, the data acquisition node sequence is divided into ceil{T / i} continuous time windows, where ceil{} represents a rounding function; the nth time window is denoted as S n , where n∈[1,ceil{T / i}], if n≠ceil{T / i}, then the time window S n The time window S records the received and received optical power data collected from the i(n-1)+1th data collection node to the i·nth data collection node. If n=ceil{T / i}, then the time window S n The data recorded therein contains the transmitted and received optical power data collected from the i(n-1)+1th data collection node to the Tth data collection node.

[0011] As a preferred solution of the backbone network line loss analysis and early warning method based on deep learning described in the present invention, the specific implementation process of analyzing and quantifying the data collection cycle includes:

[0012] Let i=i+1, return to step 2, and perform deep learning iterative analysis of the time window. The analysis process for each i-th iterative analysis is as follows:

[0013] Based on the time window, the SDN controller controls the channel CH ab Perform status monitoring and construct a data set of received and received optical power EPD(i) = {EPD ab (S n )|a∈[1,A],b∈[1,B]}, where EPD ab (S n ) indicates that in the time window S n Inner channel CH ab Collected transmit and receive optical power data.

[0014] Based on the received and received optical power data set EPD(i), the data difference fluctuation value of the time window is calculated using the following formula:

[0015]

[0016]

[0017] in, It represents the mean value of the received and received optical power data collected by different channels in different time windows, FV(S n |i) represents the data difference fluctuation value of the time window analyzed in the i-th iteration, μ represents the mean value of the received and received optical power data collected from different channels in different time windows under different links, and σ 2 It represents the variance of the transmitted and received optical power data collected from different channels in different time windows under different links.

[0018] In the present invention, μ refers to the average value of the optical power data collected on all links, which is used to measure the overall operating status of the system, and the variance σ 2 It is used to reflect the degree of data fluctuation. The larger the variance, the greater the data fluctuation in the system, which may indicate that the link status is unstable. It reflects the distribution characteristics of the data around the mean μ. For the power data of each link and channel, the "volatility" of its distribution is reflected by the weight of this formula; the formula is calculated by the data in different time windows S n Fluctuation value FV(S n |i) Determine whether the link has abnormal fluctuations.

[0019] When i=T, stop the operation of i=i+1, that is, stop the iteration, and conduct a comprehensive analysis of all data difference fluctuation values ​​in the iteration process to obtain the step length i corresponding to the minimum data difference fluctuation value, as follows:

[0020] FV(S n |i) min =min i∈[1,T] [FV(S n |i)];

[0021] Among them, FV(S n |i) min Indicates the minimum data difference fluctuation value.

[0022] In the present invention, by calculating the data fluctuation value FV(S n |i), find the step length i that minimizes the fluctuation value, that is, FV(S n |i) min , which means finding an optimal step size so that the system can maintain sufficient sensitivity in the data collection period without increasing the calculation burden due to frequent collection; this formula helps the system automatically select the most suitable data collection frequency (step size), thereby reducing the amount of calculation while ensuring the accuracy and real-time nature of the data. In this way, the system can more effectively detect line loss and issue early warnings.

[0023] As a preferred solution of the backbone network line loss analysis and early warning method based on deep learning described in the present invention, the step length i corresponding to the minimum data difference fluctuation value is used as the data collection period, the received and received optical power data is collected, and the link LS is analyzed and evaluated. a The overall loss failure degree is as follows:

[0024]

[0025] Among them, OD(LS a ) indicates link LS a The overall loss failure rate.

[0026] In the present invention, the link loss fault degree OD (LS a ) is an important indicator to measure the health of the optical fiber link. High loss may lead to a decline in the quality of data transmission or even cause communication interruption. In order to evaluate the loss of the link, it is necessary to combine the data of the transmitted and received optical power and its volatility (variance) for calculation; this formula provides a quantitative loss fault degree indicator for the link. By calculating the loss fault degree OD (LS a ), the system can monitor the health status of each link in real time. The higher the loss fault degree, the more serious the link loss is, and it may need to be repaired or replaced in time.

[0027] Link-based LS a The overall loss failure degree of the link LS is calculated a With Link LS c The difference in the loss failure degree between the two is calculated as follows: dv(LS a ,LS c )=|OD(LS a )-OD(LS c )|, where dv(LS a ,LS c ) indicates link LS a With Link LS c Based on the loss fault degree difference value dv(LS a ,LS c ), calculate the link LS a With Link LS c The correlation between them is calculated as follows:

[0028]

[0029] Among them, AD(LS a ,LS c ) indicates link LS a With Link LS c The correlation between Indicates the maximum loss fault degree difference value between all links.

[0030] In the present invention, is the maximum difference coefficient among all links, which is used for normalization so that the correlation range is between 0 and 1, and AD(LS a ,LS c ) ranges from 0 to 1, where the link LS a With Link LS c The correlation between AD(LS a ,LS c ) is closer to 1, indicating that the link is LS a With Link LS c The more similar the loss patterns of the links, the more similar the a With Link LS c The correlation between AD(LS a ,LS c ) is closer to 0, the greater the difference in loss patterns between links.

[0031] Preset correlation threshold θ, if AD(LS a ,LS c )>θ, then the link LS is determined aWith Link LS c There is a correlation between them.

[0032] Preset the overall loss fault threshold τ, if OD(LS a )>τ, then the link LS is determined a If there is a risk of loss, the link LS a All links with associated relationships are warned.

[0033] A backbone network line attenuation analysis and early warning system based on deep learning. The system includes: a link and channel database construction module, a time window division and data acquisition module, a data fluctuation analysis and iterative calculation module, and a loss fault analysis and early warning module.

[0034] The link and channel database construction module: uniformly numbers the links in the backbone network to construct a link database; obtains the channels in the links to construct a channel database.

[0035] The time window division and data collection module coordinates the data collection nodes, generates a data collection node sequence, and sets a time window to collect the transmit and receive optical power data of the channel.

[0036] The data fluctuation analysis and iterative calculation module: establishes a data set of received and received light power based on a time window, and analyzes and quantifies the data collection period by changing the scale of the time window.

[0037] The loss fault analysis and early warning module: analyzes and evaluates the overall loss fault degree of the link, calculates the correlation between different links; presets the overall loss fault degree threshold, analyzes and outputs early warnings.

[0038] Furthermore, the link and channel database construction module includes a link number and database construction unit and a channel number and database construction unit.

[0039] The link numbering and database building unit: uniformly numbers the links in the backbone network and builds a link database, which is recorded as LS={LS a |a∈[1,A]}, where LS a represents the ath link, and A represents the total number of links.

[0040] The channel number and database construction unit: obtain link LS a The channels in the channel database are constructed, denoted as CH = {CH ab |a∈[1,A],b∈[1,B]}, where CH ab Indicates link LS a The bth channel in , B represents the total number of channels.

[0041] Furthermore, the time window division and data collection module includes a data collection node sequence construction unit and a time window division and step size setting unit.

[0042] The data acquisition node sequence construction unit: constructs a data acquisition node sequence in units of minutes, denoted as Y={Y t |t∈[1,T]}, where Y t represents the tth data collection node, T represents the total number of data collection nodes, and one data collection node collects the transmitted and received optical power data once.

[0043] The time window division and step size setting unit: taking i data acquisition nodes as a step size, divides the data acquisition node sequence into ceil{T / i} continuous time windows, where ceil{} represents a rounding function; the nth time window is recorded as S n , where n∈[1,ceil{T / i}], if n≠ceil{T / i}, then the time window S n The time window S records the received and received optical power data collected from the i(n-1)+1th data collection node to the i·nth data collection node. If n=ceil{T / i}, then the time window S n The data recorded therein contains the transmitted and received optical power data collected from the i(n-1)+1th data collection node to the Tth data collection node.

[0044] Furthermore, the data fluctuation analysis and iterative calculation module includes a received and received light power data set construction unit and a data difference fluctuation value calculation and step size optimization unit.

[0045] The received and received optical power data set construction unit: let i=i+1, return to step 2, and perform deep learning iterative analysis of the time window. The analysis process of each i-th iterative analysis is as follows:

[0046] Based on the time window, the SDN controller controls the channel CH ab Perform status monitoring and construct a data set of received and received optical power EPD(i) = {EPD ab (S n )|a∈[1,A],b∈[1,B]}, where EPD ab (S n ) indicates that in the time window S n Inner channel CH ab Collected transmit and receive optical power data.

[0047] The data difference fluctuation value calculation and step size optimization unit calculates the data difference fluctuation value of the time window based on the received and received light power data set EPD(i). When i=T, the operation of i=i+1 is stopped, i.e., iteration is stopped. A comprehensive analysis is performed on all data difference fluctuation values ​​in the iteration process to obtain the step size i corresponding to the minimum data difference fluctuation value.

[0048] Furthermore, the loss fault analysis and early warning module includes a link loss fault degree evaluation unit and a link correlation degree calculation and early warning unit.

[0049] The link loss fault degree evaluation unit: uses the step length i corresponding to the minimum data difference fluctuation value as the data collection period to collect the received and received optical power data, analyze and evaluate the link LS a The overall loss failure rate.

[0050] The link correlation calculation and warning unit: based on the link LS a The overall loss failure degree of the link LS is calculated a With Link LS c Based on the difference in the loss fault degree, the link LS is calculated. a With Link LS c The correlation between AD(LS a ,LS c ).

[0051] Preset correlation threshold θ, if AD(LS a ,LS c )>θ, then the link LS is determined a With Link LS c There is a correlation between them.

[0052] Preset the overall loss fault threshold τ, if OD(LS a )>τ, then the link LS is determined a If there is a risk of loss, the link LS a All links with associated relationships are warned.

[0053] Compared with the prior art, the beneficial effects achieved by the present invention are as follows: in the backbone network line attenuation analysis and early warning method and system based on deep learning provided by the present invention, the backbone network links are uniformly numbered and a link and channel database is constructed, thereby improving the efficiency of data management and retrieval; in the data collection process, by setting time windows and collection node sequences, systematic and flexible data collection is achieved, ensuring the timeliness and continuity of the data; power data analysis and fluctuation monitoring based on time windows optimize the data collection cycle and improve the ability to identify the health status of the network; finally, by evaluating the overall loss fault degree of the link and calculating the correlation between links, potential faults are discovered in a timely manner, providing a reliable basis for network maintenance, thereby significantly improving the operational stability and reliability of the network. BRIEF DESCRIPTION OF THE DRAWINGS

[0054] The accompanying drawings are used to provide further understanding of the present invention and constitute a part of the specification. They are used to explain the present invention together with the embodiments of the present invention and do not constitute a limitation of the present invention.

[0055] Figure 1 It is a schematic diagram of the steps of the backbone network line loss analysis and early warning method based on deep learning of the present invention;

[0056] Figure 2 It is a structural schematic diagram of the backbone network line attenuation analysis and early warning system based on deep learning of the present invention. DETAILED DESCRIPTION

[0057] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0058] See also Figure 1 In the first embodiment of the present invention, a backbone network line loss analysis and early warning method based on deep learning is provided, and the method comprises the following steps:

[0059] Step S1: uniformly number the links in the backbone network and construct a link database; obtain the channels in the links and construct a channel database.

[0060] Specifically, the links in the backbone network are uniformly numbered to construct a link database, which is denoted as LS = {LS a |a∈[1,A]}, where LS a represents the ath link, and A represents the total number of links.

[0061] Further, obtain the link LS aThe channels in the channel database are constructed, denoted as CH = {CH ab |a∈[1,A],b∈[1,B]}, where CH ab Indicates link LS a The bth channel in , B represents the total number of channels; the channel represents a path for data transmission on the link.

[0062] Step S2: Coordinate the data collection nodes, generate a data collection node sequence, and set a time window to collect the channel's transmit and receive optical power data.

[0063] Specifically, a data collection node sequence is constructed in units of minutes, denoted as Y = {Y t |t∈[1,T]}, where Y t represents the tth data collection node, T represents the total number of data collection nodes, and one data collection node collects the transmitted and received optical power data once.

[0064] Furthermore, with i data acquisition nodes as a step length, the data acquisition node sequence is divided into ceil{T / i} continuous time windows, where ceil{} represents a rounding function; the nth time window is recorded as S n , where n∈[1,ceil{T / i}], if n≠ceil{T / i}, then the time window S n The time window S records the received and received optical power data collected from the i(n-1)+1th data collection node to the i·nth data collection node. If n=ceil{T / i}, then the time window S n The data recorded therein contains the transmitted and received optical power data collected from the i(n-1)+1th data collection node to the Tth data collection node.

[0065] Step S3: Based on the time window, a data set of transmitted and received optical power is established, and the data collection period is analyzed and quantified by changing the scale of the time window.

[0066] Specifically, let i=i+1, return to step 2, and perform deep learning iterative analysis of the time window. The analysis process for each i-th iterative analysis is as follows:

[0067] Based on the time window, the SDN controller controls the channel CH ab Perform status monitoring and construct a data set of received and received optical power EPD(i) = {EPD ab (S n )|a∈[1,A],b∈[1,B]}, where EPD ab (S n ) indicates that in the time window S n Inner channel CH ab Collected transmit and receive optical power data.

[0068] Based on the received and received optical power data set EPD(i), the data difference fluctuation value of the time window is calculated using the following formula:

[0069]

[0070] in, It represents the mean value of the received and received optical power data collected by different channels in different time windows, FV(S n |i) represents the data difference fluctuation value of the time window analyzed in the i-th iteration, μ represents the mean value of the received and received optical power data collected from different channels in different time windows under different links, and σ 2 It represents the variance of the transmitted and received optical power data collected from different channels in different time windows under different links.

[0071] For example, suppose A is 2, B is 2, T is 10, i is 5, ceil{T / i} is 2, EPD 11 (S1) is 2.5, EPD 11 (S2) is 3, EPD 12 (S1) is 2, EPD 12 (S2) is 2.5, EPD 21 (S1) is 3, EPD 21 (S2) is 3.5, EPD 22 (S1) is 4, EPD 22 (S2) is 3, and we can get is 1, μ is 1.18, σ 2 is 0.0613, and FV(S1|5) is

[0072] Furthermore, when i=T, the operation of i=i+1 is stopped, that is, the iteration is stopped, and a comprehensive analysis is performed on all data difference fluctuation values ​​during the iteration process to obtain the step length i corresponding to the minimum data difference fluctuation value, as follows:

[0073] FV(S n |i) min =min i∈[1,T] [FV(S n |i)];

[0074] Among them, FV(S n |i) min Indicates the minimum data difference fluctuation value.

[0075] For example, it is known that When i is 6, assuming Then FV(S1|5)>FV(S1|6), and the step length corresponding to the minimum data difference fluctuation value is 5.

[0076] Step S4: Analyze and evaluate the overall loss failure degree of the link, calculate the correlation between different links; preset the overall loss failure degree threshold, analyze and output warning.

[0077] Specifically, the step length i corresponding to the minimum data difference fluctuation value is used as the data collection period to collect the received and received optical power data, analyze and evaluate the link LS a The overall loss failure degree is as follows:

[0078]

[0079] Among them, OD(LS a ) indicates link LS a The overall loss failure rate.

[0080] For example, the overall loss fault degree of link LS1 is

[0081] Link-based LS a The overall loss failure degree of the link LS is calculated a With Link LS c The difference in the loss failure degree between the two is calculated as follows: dv(LS a ,LS c )=|OD(LS a )-OD(LS c )|, where dv(LS a ,LS c ) indicates link LS a With Link LS c Based on the loss fault degree difference value dv(LS a ,LS c ), calculate the link LS a With Link LS c The correlation between them is calculated as follows:

[0082]

[0083] Among them, AD(LS a ,LS c ) indicates link LS a With Link LS c max[dv] represents the maximum loss fault degree difference between all links.

[0084] For example, the overall loss fault degree of link LS2 is The difference in the loss fault degree between link LS1 and link LS2 is |4.05-5.46|=1.41, so the correlation between link LS1 and link LS2 is

[0085] Preset correlation threshold θ, if AD(LS a ,LS c )>θ, then the link LS is determined a With Link LS c There is a correlation between them.

[0086] Preset the overall loss fault threshold τ, if OD(LS a )>τ, then the link LS is determined a If there is a risk of loss, the link LS a All links with associated relationships are warned.

[0087] For example, assuming that the preset correlation threshold θ is 0.6 and the overall loss fault threshold τ is 2, then AD(LS1, LS2)>θ, and it is determined that there is a correlation relationship between link LS1 and link LS2, and OD(LS1)>τ, then it is determined that link LS1 has a loss risk, and an early warning is issued for link LS2.

[0088] See also Figure 2 In the second embodiment of the present invention, a backbone network line attenuation analysis and early warning system based on deep learning is provided, the system comprising: a link and channel database construction module, a time window division and data acquisition module, a data fluctuation analysis and iterative calculation module, and a loss fault analysis and early warning module.

[0089] The link and channel database construction module: uniformly numbers the links in the backbone network to construct a link database; obtains the channels in the links to construct a channel database.

[0090] The time window division and data acquisition module coordinates the data acquisition nodes, generates a data acquisition node sequence, and sets a time window to collect the transmit and receive optical power data of the channel.

[0091] The data fluctuation analysis and iterative calculation module: establishes a data set of received and received light power based on a time window, and analyzes and quantifies the data collection period by changing the scale of the time window.

[0092] The loss fault analysis and early warning module: analyzes and evaluates the overall loss fault degree of the link, calculates the correlation between different links; presets the overall loss fault degree threshold, analyzes and outputs early warnings.

[0093] Furthermore, the link and channel database construction module includes a link number and database construction unit and a channel number and database construction unit.

[0094] The link numbering and database building unit: uniformly numbers the links in the backbone network and builds a link database, which is recorded as LS={LS a |a∈[1,A]}, where LS a represents the ath link, and A represents the total number of links.

[0095] The channel number and database construction unit: obtain link LS a The channels in the channel database are constructed, denoted as CH = {CH ab |a∈[1,A],b∈[1,B]}, where CH ab Indicates link LS a The bth channel in , B represents the total number of channels.

[0096] Furthermore, the time window division and data collection module includes a data collection node sequence construction unit and a time window division and step size setting unit.

[0097] The data acquisition node sequence construction unit: constructs a data acquisition node sequence in units of minutes, denoted as Y={Y t |t∈[1,T]}, where Y t represents the tth data collection node, T represents the total number of data collection nodes, and one data collection node collects the transmitted and received optical power data once.

[0098] The time window division and step size setting unit: taking i data acquisition nodes as a step size, divides the data acquisition node sequence into ceil{T / i} continuous time windows, where ceil{} represents a rounding function; the nth time window is recorded as S n , where n∈[1,ceil{T / i}], if n≠ceil{T / i}, then the time window S n The time window S records the received and received optical power data collected from the i(n-1)+1th data collection node to the i·nth data collection node. If n=ceil{T / i}, then the time window S n The data recorded therein contains the transmitted and received optical power data collected from the i(n-1)+1th data collection node to the Tth data collection node.

[0099] Furthermore, the data fluctuation analysis and iterative calculation module includes a received and received light power data set construction unit and a data difference fluctuation value calculation and step size optimization unit.

[0100] The received and received optical power data set construction unit: let i=i+1, return to step 2, and perform deep learning iterative analysis of the time window. The analysis process of each i-th iterative analysis is as follows:

[0101] Based on the time window, the SDN controller controls the channel CH ab Perform status monitoring and construct a data set of received and received optical power EPD(i) = {EPD ab (S n )|a∈[1,A],b∈[1,B]}, where EPD ab (S n ) indicates that in the time window S n Inner channel CH ab Collected transmit and receive optical power data.

[0102] The data difference fluctuation value calculation and step size optimization unit calculates the data difference fluctuation value of the time window based on the received and received light power data set EPD(i). When i=T, the operation of i=i+1 is stopped, i.e., iteration is stopped. A comprehensive analysis is performed on all data difference fluctuation values ​​in the iteration process to obtain the step size i corresponding to the minimum data difference fluctuation value.

[0103] Furthermore, the loss fault analysis and early warning module includes a link loss fault degree evaluation unit and a link correlation degree calculation and early warning unit.

[0104] The link loss fault degree evaluation unit: uses the step length i corresponding to the minimum data difference fluctuation value as the data collection period to collect the received and received optical power data, analyze and evaluate the link LS a The overall loss failure rate.

[0105] The link correlation calculation and warning unit: based on the link LS a The overall loss failure degree of the link LS is calculated a With Link LS c Based on the difference in the loss fault degree, the link LS is calculated. a With Link LS c The correlation between AD(LS a ,LS c ).

[0106] Preset correlation threshold θ, if AD(LS a ,LS c )>θ, then the link LS is determined a With Link LS c There is a correlation between them.

[0107] Preset the overall loss fault threshold τ, if OD(LS a )>τ, then the link LS is determineda If there is a risk of loss, the link LS a All links with associated relationships are warned.

[0108] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0109] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A backbone network line loss analysis and early warning method based on deep learning, characterized in that: The method comprises the following steps: Step S1: uniformly number the links in the backbone network and construct a link database; obtain the channels in the links and construct a channel database; Step S2: Coordinate the data collection nodes, generate a data collection node sequence, and set a time window to collect the channel's transmit and receive optical power data; Step S3: Based on the time window, a data set of transmitted and received optical power is established, and the data collection period is analyzed and quantified by changing the scale of the time window; Step S4: Analyze and evaluate the overall loss failure degree of the link, calculate the correlation between different links; preset the overall loss failure degree threshold, analyze and output warning.

2. The backbone network line loss analysis and early warning method based on deep learning according to claim 1 is characterized in that: The specific implementation process of step S1 includes: Uniformly number the links in the backbone network and build a link database, denoted as LS = {LS a |a∈[1,A]}, where LS a represents the ath link, and A represents the total number of links; Get Link LS a The channels in the channel database are constructed, denoted as CH = {CH ab |a∈[1,A],b∈[1,B]}, where CH ab Indicates link LS a The bth channel in , B represents the total number of channels.

3. The backbone network line loss analysis and early warning method based on deep learning according to claim 2 is characterized in that: The time window is set as follows: In units of minutes, a data collection node sequence is constructed, denoted as Y = {Y t |t∈[1,T]}, where Y t represents the tth data collection node, T represents the total number of data collection nodes, and one data collection node corresponds to collecting the transmitted and received optical power data once; Taking i data acquisition nodes as a step length, the data acquisition node sequence is divided into ceil{T / i} continuous time windows, where ceil{} represents a rounding function; the nth time window is denoted as S n , where n∈[1,ceil{T / i}], if n≠ceil{T / i}, then the time window S n The time window S records the received and received optical power data collected from the i(n-1)+1th data collection node to the i·nth data collection node. If n=ceil{T / i}, then the time window S n The data recorded therein contains the transmitted and received optical power data collected from the i(n-1)+1th data collection node to the Tth data collection node.

4. The backbone network line loss analysis and early warning method based on deep learning according to claim 3 is characterized in that: The specific implementation process of analyzing and quantifying the data collection cycle includes: Let i=i+1, return to step 2, and perform deep learning iterative analysis of the time window. The analysis process for each i-th iterative analysis is as follows: Based on the time window, the SDN controller controls the channel CH ab Perform status monitoring and construct a data set of received and received optical power EPD(i) = {EPD ab (S n )|a∈[1,A],b∈[1,B]}, where EPD ab (S n ) indicates that in the time window S n Inner channel CH ab Collected transmitted and received optical power data; Based on the received and received optical power data set EPD(i), the data difference fluctuation value of the time window is calculated using the following formula: in, It represents the mean value of the received and received optical power data collected by different channels in different time windows, FV(S n |i) represents the data difference fluctuation value of the time window analyzed in the i-th iteration, μ represents the mean value of the received and received optical power data collected from different channels in different time windows under different links, and σ 2 It represents the variance of the received and received optical power data collected from different channels in different time windows under different links; When i=T, stop the operation of i=i+1, that is, stop the iteration, and conduct a comprehensive analysis of all data difference fluctuation values ​​in the iteration process to obtain the step length i corresponding to the minimum data difference fluctuation value, as follows: FV(S n |i) min =min i∈[1,T] [FV(S n |i)]; Among them, FV(S n |i) min Indicates the minimum data difference fluctuation value.

5. The backbone network line loss analysis and early warning method based on deep learning according to claim 4 is characterized in that: The specific implementation process of step S4 includes: The step length i corresponding to the minimum data difference fluctuation value is used as the data collection period to collect the received and received optical power data, analyze and evaluate the link LS a The overall loss failure degree is as follows: Among them, OD(LS a ) indicates link LS a The overall loss failure rate; Link-based LS a The overall loss failure degree of the link LS is calculated a Link LS c The difference in the loss failure degree between the two is calculated as follows: dv(LS a ,LS c )=|OD(LS a )-OD(LS c )|, where dv(LS a ,LS c ) indicates link LS a With Link LS c Based on the loss fault degree difference value dv(LS a ,LS c ), calculate the link LS a With Link LS c The correlation between them is calculated as follows: Among them, AD(LS a ,LS c ) indicates link LS a Link LS c The correlation degree between them, max[dv] represents the maximum loss fault degree difference value between all links; Preset correlation threshold θ, if AD(LS a ,LS c )>θ, then the link LS is determined a Link LS c There is a correlation between them; Preset the overall loss fault threshold τ, if OD(LS a )>τ, then the link LS is determined a If there is a risk of loss, the link LS a All links with associated relationships are warned.

6. The backbone network line loss analysis and early warning system based on deep learning is characterized by: The system includes: a link and channel database construction module, a time window division and data acquisition module, a data fluctuation analysis and iterative calculation module, and a loss fault analysis and early warning module; The link and channel database construction module: uniformly number the links in the backbone network to construct a link database; obtain the channels in the links to construct a channel database; The time window division and data acquisition module: coordinates the data acquisition nodes, generates a data acquisition node sequence, and sets a time window for collecting the transmit and receive optical power data of the channel; The data fluctuation analysis and iterative calculation module: establishes a data set of received and received optical power based on a time window, and analyzes and quantifies the data collection period by changing the scale of the time window; The loss fault analysis and early warning module: analyzes and evaluates the overall loss fault degree of the link, calculates the correlation between different links; presets the overall loss fault degree threshold, analyzes and outputs early warnings.

7. The backbone network line loss analysis and early warning system based on deep learning according to claim 6 is characterized in that: The link and channel database construction module includes a link number and database construction unit and a channel number and database construction unit; The link numbering and database building unit: uniformly numbers the links in the backbone network and builds a link database, which is recorded as LS={LS a |a∈[1,A]}, where LS a represents the ath link, and A represents the total number of links; The channel number and database construction unit: obtain link LS a The channels in the channel database are constructed, denoted as CH = {CH ab |a∈[1,A],b∈[1,B]}, where CH ab Indicates link LS a The bth channel in , B represents the total number of channels.

8. The backbone network line loss analysis and early warning system based on deep learning according to claim 7 is characterized in that: The time window division and data acquisition module includes a data acquisition node sequence construction unit and a time window division and step size setting unit; The data acquisition node sequence construction unit: constructs a data acquisition node sequence in units of minutes, denoted as Y={Y t |t∈[1,T]}, where Y t represents the tth data collection node, T represents the total number of data collection nodes, and one data collection node corresponds to collecting the transmitted and received optical power data once; The time window division and step size setting unit: taking i data acquisition nodes as a step size, divides the data acquisition node sequence into ceil{T / i} continuous time windows, where ceil{} represents a rounding function; the nth time window is recorded as S n , where n∈[1,ceil{T / i}], if n≠ceil{T / i}, then the time window S n The time window S records the received and received optical power data collected from the i(n-1)+1th data collection node to the i·nth data collection node. If n=ceil{T / i}, then the time window S n The data recorded therein contains the transmitted and received optical power data collected from the i(n-1)+1th data collection node to the Tth data collection node.

9. The backbone network line loss analysis and early warning system based on deep learning according to claim 8 is characterized in that: The data fluctuation analysis and iterative calculation module includes a receiving and transmitting light power data set construction unit and a data difference fluctuation value calculation and step size optimization unit; The received and received optical power data set construction unit: let i=i+1, return to step 2, and perform deep learning iterative analysis of the time window. The analysis process of each i-th iterative analysis is as follows: Based on the time window, the SDN controller controls the channel CH ab Perform status monitoring and construct a data set of received and received optical power EPD(i) = {EPD ab (S n )|a∈[1,A],b∈[1,B]}, where EPD ab (S n ) indicates that in the time window S n Inner channel CH ab Collected transmitted and received optical power data; The data difference fluctuation value calculation and step size optimization unit calculates the data difference fluctuation value of the time window based on the received and received light power data set EPD(i). When i=T, the operation of i=i+1 is stopped, i.e., iteration is stopped. A comprehensive analysis is performed on all data difference fluctuation values ​​in the iteration process to obtain the step size i corresponding to the minimum data difference fluctuation value.

10. The backbone network line loss analysis and early warning system based on deep learning according to claim 9 is characterized in that: The loss fault analysis and early warning module includes a link loss fault degree evaluation unit and a link correlation degree calculation and early warning unit; The link loss fault degree evaluation unit: uses the step length i corresponding to the minimum data difference fluctuation value as the data collection period to collect the received and received optical power data, analyze and evaluate the link LS a The overall loss failure rate; The link correlation calculation and warning unit: based on the link LS a The overall loss failure degree of the link LS is calculated a Link LS c The difference value of loss failure degree between them; Calculate the link LS based on the loss fault degree difference value a With Link LS c The correlation between AD(LS a ,LS c ); Preset correlation threshold θ, if AD(LS a ,LS c )>θ, then the link LS is determined a With Link LS c There is a correlation between them; Preset the overall loss fault threshold τ, if OD(LS a )>τ, then the link LS is determined a If there is a risk of loss, the link LS a All links with associated relationships are warned.

Images