Configuration method and configuration device for optical communication networks

By integrating physical layer and service layer data in optical communication networks, a link state and time-series traffic matrix is ​​constructed. An improved sequence probability ratio test method is used to dynamically adjust the decision threshold group, which solves the problems of low bandwidth allocation efficiency, high spectrum fragmentation rate and low fault detection sensitivity in optical communication networks, and realizes high reliability and low latency fault detection and resource scheduling.

CN123123768APending Publication Date: 2026-11-17INSPUR SUZHOU INTELLIGENT TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611485730.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-09-22
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

Existing optical communication networks suffer from problems such as low bandwidth allocation efficiency, high spectrum fragmentation rate, and low fault detection sensitivity.

Method used

By acquiring physical layer and service layer data of optical communication networks, a link state matrix and a time-series traffic matrix are constructed, fused feature vectors are extracted, and an improved sequence probability ratio test method is used to determine the cumulative log-likelihood ratio. The decision threshold group is dynamically adjusted, and configuration instructions are generated to achieve high reliability, low latency fault detection, and resource scheduling.

Benefits of technology

It achieves efficient bandwidth utilization, reduces spectrum fragmentation rate, improves the sensitivity and accuracy of fault detection, and meets the rapid adaptive reconfiguration requirements of optical communication networks in dynamic service scenarios.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123123768A_ABST
    Figure CN123123768A_ABST
Patent Text Reader

Abstract

This application discloses a configuration method and apparatus for an optical communication network, relating to the field of communication network technology. The method includes acquiring raw data during the first observation of the optical communication network; the raw data includes physical layer data and service layer data of the optical communication network; determining a link state matrix and a time-series traffic matrix based on the raw data; determining a fused feature vector based on the link state matrix and the time-series traffic matrix; the fused feature vector includes spatial features of the link state matrix and temporal features of the time-series traffic matrix; determining a cumulative log-likelihood ratio based on the fused feature vector; determining a decision threshold set based on an initialization threshold set, the decision threshold set including a first decision threshold and a second decision threshold; determining the state of the optical communication network based on at least one of the first decision threshold and the second decision threshold, and the cumulative log-likelihood ratio; and determining the optical communication network as being in a normal state if the cumulative log-likelihood ratio is less than or equal to the second decision threshold.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This application relates to the field of communication network technology, and in particular to a configuration method and configuration device for an optical communication network. Background Technology

[0002] With the development of emerging services such as 5G / 6G communication, cloud computing, computing power networks, and quantum communication, optical interconnection network scenarios such as Data Center Interconnection (DCI) and Elastic Optical Network (EON) place significantly higher demands on the bandwidth dynamism and real-time fault recovery capabilities of optical communication networks. To adapt to highly dynamic service requirements, related technologies have proposed a mainstream architecture of Software Defined Networking (SDN) and incorporated artificial intelligence technology for resource scheduling and fault detection in optical communication networks. However, the technical solutions provided by these technologies suffer from problems such as low bandwidth allocation efficiency, high spectrum fragmentation rate, and low fault detection sensitivity. Summary of the Invention

[0003] This application provides a configuration method and apparatus for an optical communication network to at least solve the problems of low bandwidth allocation efficiency, high spectrum fragmentation rate, and low fault detection sensitivity in related technologies.

[0004] This application provides a configuration method for an optical communication network, including: configuring the optical communication network in the first... In this observation, raw data is acquired; the raw data includes the physical layer data and the service layer data of the optical communication network. The input is a positive integer greater than 0; a link state matrix and a time-series traffic matrix are determined based on the original data; a fused feature vector is determined based on the link state matrix and the time-series traffic matrix; the fused feature vector includes the spatial features of the link state matrix and the temporal features of the time-series traffic matrix; a cumulative log-likelihood ratio is determined based on the fused feature vector; a decision threshold group is determined based on an initialization threshold group, the decision threshold group including a first decision threshold and a second decision threshold; the state of the optical communication network is determined based on at least one of the first decision threshold and the second decision threshold, and the cumulative log-likelihood ratio; if the cumulative log-likelihood ratio is less than or equal to the second decision threshold, the optical communication network is determined to be in a normal state.

[0005] This application also provides a configuration apparatus for an optical communication network, including: a data acquisition module, used for acquiring data from the optical communication network at a certain time interval. In this observation, raw data is acquired; the raw data includes the physical layer data and the service layer data of the optical communication network. The input is a positive integer greater than 0; a preprocessing module is used to determine the link state matrix and the time-series traffic matrix based on the original data; a fusion feature extraction module is used to determine a fusion feature vector based on the link state matrix and the time-series traffic matrix; the fusion feature vector includes the spatial features of the link state matrix and the temporal features of the time-series traffic matrix; a log-likelihood ratio calculation module is used to determine the cumulative log-likelihood ratio based on the fusion feature vector; a decision threshold generation module is used to determine a decision threshold group based on an initialization threshold group, the decision threshold group including a first decision threshold and a second decision threshold; a judgment module is used to determine the state of the optical communication network based on at least one of the first decision threshold and the second decision threshold, and the cumulative log-likelihood ratio; if the cumulative log-likelihood ratio is less than or equal to the second decision threshold, the optical communication network is determined to be in a normal state.

[0006] This application also provides an electronic device, including: a memory for storing a computer program; and a processor for implementing the steps of any of the above configuration methods when executing the computer program.

[0007] This application also provides a computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the steps of any of the above configuration methods.

[0008] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of any of the above configuration methods.

[0009] The configuration method provided in this application incorporates multimodal raw data from both the physical and business layers, resulting in a more comprehensive state awareness dimension and avoiding missed detections when a single data source is insufficient for fault characterization. The link state matrix and time-series traffic matrix determined based on the raw data enable subsequent processing to focus on structured spatial snapshots and time series, reducing the complexity of data fusion.

[0010] Furthermore, by fusing physical layer data and service layer data of optical communication networks, a fused feature vector containing spatial and temporal features is extracted. This vector combines the correlation of physical states between optical communication network nodes and / or links with the evolution pattern of traffic over time, making subtle anomalies caused by faults more separable in the feature space.

[0011] Furthermore, by continuously accumulating subtle anomalies from multiple observations to determine the cumulative log-likelihood ratio, a significant statistical bias can be observed in the early stages of a fault. This ensures detection sensitivity while naturally suppressing false triggers caused by random noise using a probabilistic framework. The configuration method can also adjust the decision threshold set according to the real-time conditions of the optical communication network, allowing the decision boundary to dynamically adapt to different service and / or channel environments. This overcomes the shortcomings of fixed thresholds, which are prone to false alarms during peak hours and missed alarms during off-peak hours, achieving highly reliable and low-latency fault detection. Attached Figure Description

[0012] To more clearly illustrate the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 This is a flowchart illustrating a configuration method for an optical communication network according to some embodiments; Figure 2 This is a flowchart illustrating another method for configuring an optical communication network according to some embodiments; Figure 3 This is a flowchart illustrating another method for configuring an optical communication network according to some embodiments; Figure 4 This is a flowchart illustrating another method for configuring an optical communication network according to some embodiments; Figure 5 This is a flowchart illustrating another method for configuring an optical communication network according to some embodiments; Figure 6 This is a flowchart illustrating another method for configuring an optical communication network according to some embodiments; Figure 7 This is a flowchart illustrating another method for configuring an optical communication network according to some embodiments. Detailed Implementation

[0014] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the protection scope of this application.

[0015] It should be noted that, in the description of this application, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. The terms "first," "second," etc., in this application are used to distinguish similar objects and are not used to describe a specific order or sequence.

[0016] To enable those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0017] With the continuous development of communication technology, the requirements for bandwidth dynamism and real-time fault recovery capabilities in optical communication networks have increased significantly. To adapt to highly dynamic service demands, some embodiments propose a mainstream architecture for software-defined networks, incorporating artificial intelligence technology for resource scheduling and fault diagnosis in optical communication networks. Figure 1 As shown, Figure 1 This is a flowchart illustrating a configuration method for an optical communication network according to some embodiments. The configuration method includes the following steps: S101. Acquire raw data. For example, collect physical layer status parameters such as link optical power and bit error rate (BER) of the optical communication network, as well as service layer indicators such as wavelength utilization and traffic data, through a coherent receiver and optical performance monitoring (OPM) module. Link optical power includes the optical power of the nodes in the optical communication network and the optical power of the transmission medium between nodes.

[0018] S102. Determine the static state matrix and time-series traffic data of the optical communication network based on the raw data.

[0019] S103. Extract the spatial feature vector of the static state matrix and the temporal feature vector of the time-series flow data.

[0020] S104. Determine the state of the optical communication network based on the static likelihood ratio and decision threshold set.

[0021] The static likelihood ratio is determined based on preset fault probability distribution parameters, the decision threshold group is a fixed value, and fault detection and / or fault judgment are performed based on the classic Sequential Probability Ratio Test (SPRT) method.

[0022] S105 and SDN generate configuration instructions based on the state of the optical communication network. These configuration instructions include routing paths and bandwidth configuration policies. The configuration instructions are output to hardware such as optical switches and wavelength selective switches through an electronic processing architecture, controlling the actions of these hardware devices to complete link switching and resource configuration.

[0023] Figure 1 The configuration methods shown rely on preset fixed rules or heuristic bandwidth allocation algorithms to configure the optical communication network. This fails to address the current spectrum fragmentation or real-time traffic fluctuations of the optical communication network, resulting in low EON resource utilization and difficulty in meeting dynamic bandwidth requirements based on multi-service priorities. Furthermore, the classic SPRT relies on prior distribution assumptions, and its bit error rate increases significantly in the presence of non-stationary noise or undefined fault modes in the optical communication network.

[0024] Furthermore, the separation of spatial and temporal feature vectors means that subsequent parameters determined based on either vector cannot simultaneously consider spatiotemporal characteristics. Fault detection and judgment are limited to either a single spatial or temporal feature vector, resulting in low fault detection sensitivity. The extraction of spatial and temporal feature vectors is also unrelated to the decision model, further reducing fault detection accuracy during distribution shifts. SDN control strategies require manual configuration, and the application of intelligent algorithms such as deep reinforcement learning suffers from slow training convergence and high online fine-tuning costs, leading to… Figure 1 The configuration method shown requires a lot of time to configure in the early stages of use and cannot quickly adapt to different network failure scenarios.

[0025] Furthermore, the use of a purely electronic processing architecture for feature extraction and decision calculation results in significant latency throughout the entire process of data acquisition and configuration command issuance in optical communication networks, which cannot meet the sub-millisecond hardware reconfiguration requirements.

[0026] In view of this, embodiments of this application provide a configuration method for an optical communication network, see below. Figure 2 This is a flowchart illustrating another configuration method for an optical communication network according to some embodiments, the configuration method being applied, for example, to an optical network management server. Exemplarily, the configuration method is applied to an optical network management server equipped with a Field Programmable Gate Array (FPGA).

[0027] like Figure 2 As shown, the configuration method includes the following steps: S201, in the first step of the optical communication network In this observation, raw data is obtained.

[0028] The raw data includes physical layer data and service layer data of the optical communication network. It is a positive integer greater than 0.

[0029] The raw data includes multimodal fundamental data generated during the operation of the optical communication network, such as physical layer parameters reflecting the physical state of the fiber optic link, such as optical power, signal-to-noise ratio, and bit error rate, as well as service layer indicators reflecting network load and service demands, such as wavelength utilization, link traffic load, and spectrum usage. These data collectively constitute the information source for subsequent feature extraction and fault diagnosis. For example, the signal-to-noise ratio is, for instance, the optical signal-to-noise ratio (OSNR).

[0030] S202. Determine the link state matrix and time-series traffic matrix based on the raw data.

[0031] The link state matrix has a dimension of N. M, the dimension of the time-series flow matrix is ​​N W. For example, N is the number of nodes in the optical communication network, M is the number of parameters in the raw data, and W is the sliding window size.

[0032] S203. Based on the link-state matrix and the time-series traffic matrix, determine the fused feature vector. The fused feature vector includes the spatial features of the link-state matrix and the temporal features of the time-series traffic matrix; S204. Determine the cumulative log-likelihood ratio based on the fused feature vector.

[0033] Step S204 may include steps 1 and 2.

[0034] Step 1: Determine the fault hypothesis probability and the normal hypothesis probability based on the fused feature vector. The fault hypothesis probability represents the probability that the optical communication network observes the fused feature vector under fault conditions, and the normal hypothesis probability represents the probability that the optical communication network observes the fused feature vector under normal conditions.

[0035] Step 2: Determine the cumulative log-likelihood ratio based on the probability of failure and the probability of normality.

[0036] S205. Based on the initialization threshold group, determine the decision threshold group. The initialization threshold group includes a first initialization threshold and a second initialization threshold, and the decision threshold group includes a first decision threshold and a second decision threshold.

[0037] S206. Determine the state of the optical communication network based on at least one of the first decision threshold and the second decision threshold, and the cumulative log-likelihood ratio.

[0038] The state determination rule for optical communication networks is based on the Modified Sequential Probability Ratio Test (MSPRT), determined for example by the following formula: (1) in, To improve the MSPRT neural probability model, the input fused feature vector is used. Under the given conditions, the probability distribution of the observed variable z; These are trainable parameters; It follows a Gaussian normal distribution; , The three-layer MLP is based on the input features The mean and covariance of a Gaussian distribution are dynamically fitted.

[0039] In formula (1), the mean and covariance are driven by the fused feature vector. When the optical communication network links and traffic change, the distribution pattern changes synchronously, so that the probability distribution changes in real time with the optical communication network operating conditions, distinguishes the statistical differences between normal links and faulty links, and improves the detection accuracy under new fault and noise environments.

[0040] In some embodiments, the optical communication network is determined to be in a normal state if the cumulative log-likelihood ratio is less than or equal to a second decision threshold.

[0041] If the optical communication network is in normal working order, maintain the current configuration of the optical communication network.

[0042] The configuration method provided in this application incorporates multimodal raw data from both the physical and business layers, resulting in a more comprehensive state awareness dimension and avoiding missed detections when a single data source is insufficient for fault characterization. The link state matrix and time-series traffic matrix determined based on the raw data enable subsequent processing to focus on structured spatial snapshots and time series, reducing the complexity of data fusion.

[0043] Furthermore, by fusing physical layer data and service layer data of optical communication networks, a fused feature vector containing spatial and temporal features is extracted. This vector combines the correlation of physical states between optical communication network nodes and / or links with the evolution pattern of traffic over time, making subtle anomalies caused by faults more separable in the feature space.

[0044] Furthermore, by continuously accumulating subtle anomalies from multiple observations to determine the cumulative log-likelihood ratio, a significant statistical bias can be observed in the early stages of a fault. This ensures detection sensitivity while naturally suppressing false triggers caused by random noise using a probabilistic framework. The configuration method can also adjust the decision threshold set according to the real-time conditions of the optical communication network, allowing the decision boundary to dynamically adapt to different service and / or channel environments. This overcomes the shortcomings of fixed thresholds, which are prone to false alarms during peak hours and missed alarms during off-peak hours, achieving highly reliable and low-latency fault detection.

[0045] In some embodiments, Figure 2 Step S206 shown in the figure also includes: If the cumulative log-likelihood ratio is greater than the second decision threshold but less than the first decision threshold, the optical communication network is determined to be in the observation state.

[0046] When the optical communication network is in observation mode, the configuration method for the optical communication network also includes performing the following steps: +1 observation.

[0047] In some embodiments, Figure 2 Step S206 shown in the figure also includes: If the cumulative log-likelihood ratio is greater than or equal to the first decision threshold, the optical communication network is determined to be in a fault state.

[0048] In some embodiments, when the optical communication network is in a faulty state, the configuration method for the optical communication network provided in this application embodiment further includes, as follows: Figure 3 The step S301 shown is as follows: Figure 3 This is a flowchart illustrating another method for configuring an optical communication network according to some embodiments.

[0049] S301. Based on the physical layer constraints of the optical communication network, determine the configuration instructions. The physical layer constraints include at least one of wavelength continuity, spectral fragmentation, and optical switch response time. The configuration instructions include at least one of adjusting the modulation format, calculating the dispersion compensation map, and switching links.

[0050] In some embodiments, continue reading Figure 3 The configuration method also includes steps S302 and S303.

[0051] S302. Generate at least one optical switch control signal based on the configuration command.

[0052] S303. Based on at least one optical switch control signal, perform at least one of link switching, power adjustment, or modulation format change on the optical communication network.

[0053] In some embodiments, when the optical switch control signal is in a first state, the link corresponding to the optical switch control signal is activated; when the optical switch control signal is in a second state, the current state of the link corresponding to the optical switch control signal is maintained.

[0054] In some embodiments, when the optical switch control signal is in a first state, the state of the link corresponding to the optical switch control signal is switched; when the optical switch control signal is in a second state, the current state of the link corresponding to the optical switch control signal is maintained.

[0055] For example, the optical switch control signal is determined based on the following formula: (2) in, for The optical switch control signal matrix at any given time, where L is the number of links in the optical communication network; For the first The control signal for the link is 1 for enabling / switching and 0 for maintaining.

[0056] In the configuration method provided in this application embodiment, physical layer constraints such as wavelength continuity, spectral fragmentation, and optical switch response time are explicitly integrated into the generation process of configuration instructions. This ensures that the generated configuration instructions, such as adjusting modulation format, calculating dispersion compensation diagram, and switching links, follow the basic laws of optical transmission, thus avoiding configuration failures, signal interruptions, or increased blocking rates caused by physical layer infeasibility from the source.

[0057] Furthermore, based on the optical switch information, the configuration execution unit performs link switching, power adjustment, or modulation format change on the optical communication network in a one-time or sequential manner, realizing the atomicity of reconfiguration actions. Because the physical layer constraints are embedded when the control signal is generated, the execution process does not require real-time physical feasibility verification. Therefore, it can significantly shorten the execution time of configuration instructions, meet the optical switch response time limit, and effectively suppress bandwidth fragmentation and connection establishment failures caused by spectrum fragmentation and wavelength continuity conflicts. It can quickly and reliably complete the adaptive reconfiguration of the optical communication network in dynamic service scenarios, ensuring transmission quality and improving the resource utilization of the optical communication network.

[0058] In the configuration method provided in this application embodiment, configuration commands are input into an optoelectronic collaborative processing pipeline composed of a photonic computing unit and an electronic control unit. An optical switch control signal is generated via a dynamic reconfiguration interface, and then sent to optical hardware to complete operations such as link switching, power adjustment, and modulation format change. This enables sub-millisecond fault recovery and bandwidth configuration, as well as coordinated detection and resource scheduling. Afterwards, the original data can be reacquired, and the process can proceed to the next step of the optical communication network... +1 observation and fault detection.

[0059] The photoelectric co-processing pipeline is determined based on the following formula: (3) In some embodiments, the configuration method for an optical communication network provided in this application further includes, for example, Figure 4 Step S401 shown, Figure 4 This is a flowchart illustrating another method for configuring an optical communication network according to some embodiments.

[0060] S401, Perform the following steps for the optical communication network: +1 observation.

[0061] In some embodiments, when the optical communication network is in a faulty state, the configuration method for the optical communication network provided in this application embodiment further includes, as follows: Figure 5 The step S501 shown is... Figure 5 This is a flowchart illustrating another method for configuring an optical communication network according to some embodiments.

[0062] S501. Based on the network path optimization model and the physical layer constraints of the optical communication network, determine the routing path and dynamic bandwidth allocation scheme of the optical communication network.

[0063] When the optical communication network is in a fault state, physical layer constraints such as wavelength continuity, spectrum fragmentation, and optical switch response time are input into the Network Path Optimization (NPO) algorithm to generate routing paths, dynamic bandwidth allocation schemes, and fault recovery operation instructions, which are then integrated into SDN configuration instructions.

[0064] For example, step S501 can be executed after step S301, before step S301, or in parallel with step S301. This application does not limit this.

[0065] In some embodiments, steps S301 and S501 can be the same step, namely, determining a configuration strategy based on the physical layer constraints of the optical communication network, the network path optimization model, and the physical layer constraints of the optical communication network. The configuration strategy includes configuration instructions, routing paths of the optical communication network, and dynamic bandwidth allocation schemes.

[0066] The configuration method provided in this application takes into account physical layer constraints such as wavelength continuity, spectrum fragmentation, and optical switch response time, ensuring that the obtained routing path naturally meets wavelength consistency requirements. This effectively utilizes fragmented spectrum resources and avoids path establishment failures caused by spectrum conflicts. The optical switch response time constraint guarantees its feasibility in terms of physical execution time and prevents secondary failures caused by configuration command and hardware conflicts.

[0067] In some embodiments, Figure 2 Step S201 shown may also include at least one of the following preprocessing operations: nonlinear compensation calibration of optical power value, quadrature phase shift keying (QPSK) signal BER estimation, and maximum-minimum normalization of the flux matrix.

[0068] In some embodiments, Figure 2 Step S201 shown may also include mapping physical parameters with different dimensions and vastly different numerical ranges, such as optical power, OSNR, BER, and wavelength utilization, to the same numerical range in order to eliminate dimensional interference.

[0069] In some embodiments, Figure 2 Step S201 shown may include, for example, at least one of the following: 1) Obtain the raw optical power value of the optical communication network, and determine the calibrated optical power value based on the raw optical power value and the nonlinear compensation coefficient. The calibrated optical power value is determined based on the following formula: (4) in, The calibrated optical power value (unit: dBm); Raw optical power values ​​(unit: dBm) collected by a coherent receiver; The coefficients are dimensionless nonlinear compensation polynomials. The calibration method is as follows: A set of optical power values ​​with known true values ​​is output from a standard optical power source as the calibration dataset. The raw readings acquired by the receiver are then used for calibration. As an independent variable, the standard source outputs the truth value. As the fitting objective, the least squares method is used to fit the third-order polynomial, and the coefficients are obtained by solving the problem. This polynomial is used to directly perform numerical fitting on the dBm reading space, without requiring dimensional matching for each term of the polynomial, and is used to compensate for the reading nonlinearity deviation introduced by coherent receiver devices.

[0070] 2) Obtain the signal-to-noise ratio (SNR) of the optical communication network, and determine the bit error rate (BER) of the optical communication network based on the SNR and the complementary error function. The BER is determined based on the following formula: (5) in, Bit error rate is a core performance indicator for nodes / links in optical communication networks. It is a complementary error function; for It is the ratio of energy per bit to noise power spectral density; its value is compared with the receiver's power signal-to-noise ratio. Satisfy conversion relationship In the formula Here, is the bit rate, and is the channel bandwidth; in this embodiment, the received signal-to-noise ratio is first obtained through the constellation diagram deviation of the pilot symbols. Combined with known bit rates Channel bandwidth Converted .

[0071] In some embodiments, the normalized time-series flow matrix is ​​determined based on the following formula: (6) in, These are the normalized flow matrix element values, ranging from [0,1]. The first of the original time-series flow matrix Line 1 Column elements; This represents the maximum value of the flow data within the sliding window; This represents the minimum value of the flow data within the sliding window.

[0072] This embodiment corrects the original optical power value by introducing a nonlinear compensation coefficient for calibration fitting, effectively eliminating measurement deviations caused by hardware distortion. This ensures that the optical power values ​​recorded in the link state matrix more accurately reflect the actual physical state of each node / link, thereby guaranteeing the input quality of the entire fault detection process from the data source. Furthermore, the theoretical bit error rate under the QPSK modulation format is calculated based on the signal-to-noise ratio of the optical communication network and the complementary error function, thus enabling real-time and sensitive acquisition of data reflecting the transmission quality of the optical link. This provides a high-quality and highly reliable data foundation for subsequent link state matrix construction, spatial feature vector extraction, and final fault determination.

[0073] In some embodiments, Figure 2 Step S202 shown in the figure may specifically include determining the link state matrix based on the normalized monitoring parameters, with the number of nodes included in the optical communication network as the number of rows and the monitoring parameters corresponding to a single node or a single link as the number of columns.

[0074] In some embodiments, Figure 2 Step S202 shown in the figure may specifically include: The time-series flow matrix is ​​determined based on the original time-series flow matrix, the maximum value of the flow data within the sliding window, and the minimum value of the flow data within the sliding window.

[0075] In some embodiments, Figure 2 Step S203 shown can be specifically implemented as follows: Figure 6 Steps S601 to S603 are shown. Figure 6 This is a flowchart illustrating another configuration method for an optical communication network according to some embodiments. Step S203 may specifically include: S601. The spatial feature vector of the link state matrix is ​​extracted using the first feature extraction module. The first feature extraction module includes a convolutional network structure.

[0076] For example, the first feature extraction module is a convolutional neural network (CNN). The link state matrix is ​​input into the first feature extraction module, and the spatial feature vector is obtained by parallel convolution with 3×3 convolution kernels and processing with the Swish activation function.

[0077] Spatial feature vectors are extracted based on the following formula: (7) in, Spatial feature vectors extracted for CNN branches; This is the Swish activation function, used to improve nonlinear modeling capabilities and ensure gradient stability; This represents the number of convolutional kernels; For the first The weight matrix of each convolutional kernel has a size of 3×3; " represents convolution operation; For the first Bias terms of each convolution kernel; A standardized node / link state matrix.

[0078] S602. The second feature extraction module is used to extract the temporal feature vector of the temporal flow matrix. The second feature extraction module includes a recurrent network structure with memory function.

[0079] For example, the second feature extraction module is a Long Short-Term Memory (LSTM) network. The standardized temporal flow matrix is ​​input into the second feature extraction module, and the temporal feature vector is obtained by updating the hidden state at each time step.

[0080] Temporal feature vectors are extracted based on the following formula: (8) in, This represents the hidden state of the LSTM at the current time step, i.e., the extracted temporal feature vector. The cell state at the current moment of LSTM; A standardized time-series flow matrix; These represent the hidden state and cell state of the LSTM at the previous time step. This is the parameter set for the LSTM, which includes the parameters for the input gate, forget gate, and output gate.

[0081] S603. Based on the spatial feature vector and the temporal feature vector, determine the fused feature vector. The dimension of the fused feature vector is the sum of the dimensions of the spatial feature vector and the temporal feature vector.

[0082] The fused feature vector is determined based on the following formula: (9) in, The fused spatiotemporal feature vector; For the dimension of fusion features, and The sum of dimensions; " is a feature splicing operation.

[0083] In the configuration method provided in this application embodiment, spatial feature vectors and temporal feature vectors are input into the feature splicing layer to complete the fusion, and a high-dimensional fused feature vector is output to realize the joint modeling of spatial correlation and temporal evolution mode of optical communication network.

[0084] Specifically, the first feature extraction module uses a convolutional neural network to extract spatial features from the link state matrix. By utilizing the local perception capability of the 3×3 convolutional kernel and the smooth nonlinearity of the Swish activation function, it can effectively identify abnormal spatial distribution patterns of physical layer parameters in optical communication networks. The second feature extraction module uses a long short-term memory network to extract temporal features from the temporal traffic matrix. With the gating mechanism of LSTM and the long-term memory capability of cell states, it can capture dynamic behaviors such as gradual trends, periodic fluctuations, or sudden spikes in service traffic over time. The combination of the two complements the instantaneous physical state anomalies reflected by spatial features with the service behaviors reflected by temporal features, realizing the spatiotemporal alignment and joint representation of multimodal data of the physical layer and service layer. This overcomes the limitation of being unable to identify physical layer faults without traffic changes when relying solely on temporal features.

[0085] In the configuration method provided in this application embodiment, a Mach-Zehnder interferometer (MZI) array is used to implement simulated optical convolution of CNN, and an optical delay line is used to process the timing operations of LSTM; an FPGA is used to implement numerical calculations of MSPRT, and nanosecond-level data interaction between photons and electronic units is achieved through the SerDes interface. The pipeline clock cycle is designed to be 10 ns, supporting 10 ms per second. 8 This decision update can significantly reduce the time required for complex calculations, compressing the entire process latency from raw data acquisition to configuration command generation to within 500μs, thereby improving the configuration speed of optical communication networks.

[0086] In some embodiments, Figure 2 In step S204 shown, determining the fault hypothesis probability and the normal hypothesis probability based on the fused feature vector may specifically include: Based on the multilayer perceptron corresponding to the dimension of the fused feature vector, the probability of fault hypothesis and the probability of normal hypothesis are determined.

[0087] For example, the feature vectors will be fused. The input is a three-layer multi-layer perceptron (MLP) classifier consisting of an input layer, a hidden layer, and an output layer. The posterior probability is obtained after Softmax normalization. (Fusion features observed) The network is assumed to be faulty. posterior probability) (Fusion features observed) The network is assumed to be normal. (posterior probability); based on Bayes' theorem, combined with the prior probability of failure. Normal prior probability The likelihood is calculated as follows:

[0088] Assumption of failure Observed features The likelihood; Normal assumption Observed features The likelihood of the hypothesis is calculated. In this three-layer MLP, the input layer dimension is the same as the fused feature vector dimension, the hidden layer contains 256 neurons, and the output layer generates the failure hypothesis probability and the normal hypothesis probability through Softmax normalization.

[0089] In some embodiments, Figure 2 In step S204 shown, determining the cumulative log-likelihood ratio based on the failure hypothesis probability and the normal hypothesis probability may specifically include: exist When the cumulative log-likelihood ratio is equal to 1, it is the th The single log-likelihood ratio determined by the nth observation is based on the nth observation. The probability of the fault hypothesis and the probability of the normal hypothesis are determined by the second observation.

[0090] exist When the cumulative log-likelihood ratio is greater than 1, it is the sum of the single log-likelihood ratio determined by the first observation up to the second observation. The sum of single log-likelihood ratios determined by each observation.

[0091] After determining the probability of the failure hypothesis and the probability of the normal hypothesis based on MLP, their log-likelihood ratio is calculated. The cumulative log-likelihood ratio is determined by summing the log-likelihood ratios obtained from multiple observations.

[0092] The converted likelihood , Substitute the values ​​into the following formula to calculate the log-likelihood ratio for a single observation. The cumulative log-likelihood ratio is determined based on the following formula: (10) in, For the first The cumulative log-likelihood ratio of all observations is the core decision statistic for fault detection; For the first The fusion feature vector of the two observations; This is a fault assumption, meaning the optical communication network is in a faulty state; This assumes normal operation, meaning the optical communication network is in a normal state. This is a probability density estimator implemented using a three-layer MLP. For the parameter set of the MLP; Assumption of failure Observed features The likelihood, Normal assumption Observed features The likelihood; For the first The log-likelihood ratio determined by the first observation.

[0093] In the configuration method provided in this application, the single log-likelihood ratio is determined based on the probability of fault assumption and the probability of normal assumption. On this basis, the single log-likelihood ratio is accumulated one after another, and the weak fault features that may be masked by noise in a single observation are continuously superimposed, gradually amplifying the statistical difference between the fault state and the normal state. This allows the configuration method provided in this application to obtain significant cumulative evidence through the accumulation of multiple observations in the early stage of the fault and when a single observation is insufficient for reliable judgment, thus overcoming the defect that single threshold decision is easily affected by random fluctuations.

[0094] In some embodiments, Figure 2 Step S205 shown in the figure can be specifically implemented as follows: Figure 7 Steps S701 and S702 are shown. Figure 7 This is a flowchart illustrating another configuration method for an optical communication network according to some embodiments. Step S205 may specifically include: S701. Perform a gradient descent step on at least one meta-task in the meta-task group to determine intermediate threshold parameters. The meta-tasks include failure modes of the optical communication network.

[0095] Step S701 can be specifically implemented as follows: performing a gradient descent step on at least one meta-task in the meta-task group based on the initial threshold group and the first cyclic learning rate to determine the intermediate threshold parameter; the meta-tasks include the failure modes of the optical communication network.

[0096] The intermediate threshold parameter is determined based on the following formula: (11) in, This is the inner loop learning rate, i.e., the first loop learning rate; For the first Individual Tasks Corresponding loss function For global threshold parameters The gradient is obtained, where , The fault determination threshold is... This is the threshold for normal maintenance.

[0097] S702. Determine the aggregated gradient of the meta-task group based on the intermediate threshold parameter, and determine the decision threshold group based on the aggregated gradient.

[0098] Step S702 can be specifically implemented as follows: determining the aggregated gradient of the meta-task group based on the intermediate threshold parameter, and determining the decision threshold group based on the aggregated gradient and the second cyclic learning rate.

[0099] The decision threshold set is determined based on the following formula: (12) in, This is the outer loop learning rate, i.e., the second loop learning rate; For all meta-task losses The aggregation gradient; The gradient operator is used to characterize how fast the loss function changes with parameters and the direction of its increase or decrease. The loss function includes at least one of the following: fault determination error loss, optical switch, and L2 regularization terms for hardware constraints such as spectral range; The kth optical network fault meta-task corresponds to a link fault scenario. The hardware-constrained loss function for the k-th fault meta-task, for the decision threshold group The gradient is calculated. The decision threshold set θ includes the first decision threshold A and the second decision threshold B.

[0100] In the configuration method provided in this application embodiment, based on multiple meta-tasks for different fault modes, the threshold learning process can perceive the differences in the statistical characteristics of fused features and cumulative log-likelihood ratio under different fault scenarios, avoiding insufficient generalization ability caused by the decision threshold group being optimized only for a single fault mode. First, the inner loop learning quickly extracts the first-order optimization signals for different fault scenarios, while preventing over-adaptation to a specific fault mode and forgetting information from other modes. Then, the outer loop aggregation update unit aggregates the loss gradients at the intermediate threshold parameters of all meta-tasks, and uniformly updates the initial threshold group, so that the final generated decision threshold group can be balanced under different fault modes, reducing the probability of false positives and false negatives.

[0101] Furthermore, a hardware constraint regularization term with the optical switch response threshold as a prior is introduced into the MAML framework. The decision threshold is dynamically optimized through inner and outer double-loop gradient updates, which solves the problems of conflict between meta-learning and hardware constraints and slow algorithm convergence, and realizes fast adaptive threshold under hardware constraints.

[0102] In some embodiments, Figure 7 Step S702 shown in the figure may specifically include: The aggregate gradient is determined based on the intermediate threshold parameter and hardware constraints.

[0103] The hardware constraints are determined based on the expectations of the fused feature vector and the true fault label, the true fault label, the cumulative log-likelihood ratio, the decision threshold set, and the optical switch response threshold.

[0104] Hardware constraints, for example, are cross-entropy loss functions with hardware constraints, determined based on the following formula: (13) in, For the first Loss function for each meta-task (specific failure mode scenario); To the fusion features and real labels Expectations; The labels are for actual faults, with 1 indicating a fault and 0 indicating normal. The sigmoid function calculates the cumulative log-likelihood ratio. Mapped to probability values; These are the regularization coefficients, used to balance loss and hardware constraints; For the decision threshold group, A is the fault determination threshold, and B is the normal maintenance threshold; To meet the optical switch response threshold; This is an L2 regularization term that limits the threshold adjustment range to adapt to hardware constraints.

[0105] The configuration method provided in this application embodiment designs an improved model-agnostic meta-learning (MAML) framework oriented towards the hardware constraints of optical communication networks, and optimizes the decision threshold for fault detection through inner and outer double-loop gradient updates. Furthermore, hardware constraints such as optical switch response time and spectrum adjustment range are incorporated into the loss function to achieve dynamic adaptation of the decision threshold group.

[0106] Specifically, the decision threshold parameters are first initialized to determine the initial threshold group. Then, the meta-tasks for various fault modes are input into the inner loop to complete single-task gradient calculation and one-step gradient descent, obtaining intermediate threshold parameters. Subsequently, based on the results of the inner loop, multi-task loss aggregation and global gradient calculation are performed by combining hardware constraint regularization terms. Finally, a dynamic adaptive decision threshold adapted to hardware constraints is output through global parameter updates, thus adapting to the physical layer constraints of the optical communication network.

[0107] In some embodiments, the configuration method for optical communication networks can also be applied to industrial internet networks in smart manufacturing, industrial automation scenarios, and other similar applications. This involves adapting the multimodal data acquisition layer to sensor data (such as temperature, speed, and voltage) from industrial equipment and transmission data from industrial links; matching the feature extraction dimensions of CNN and LSTM models to the spatiotemporal characteristics of industrial data; and replacing the optoelectronic co-execution layer with hardware control units of industrial switches and gateways to achieve rapid reconfiguration of industrial links. This enables intelligent management and control of industrial internet networks, improves the real-time performance and reliability of industrial data transmission, and ensures the continuity of industrial automated production.

[0108] In summary, the configuration method provided in this application, through the end-to-end closed-loop process of collecting and preprocessing multimodal raw data, fusing spatial and temporal feature vectors using CNN-LSTM, applying MSPRT neural likelihood ratio, MAML threshold adaptation, SDN intelligent decision-making, and optoelectronic collaborative execution, effectively improves the resource utilization of optical communication networks, the robustness of fault detection, and the adaptive speed under hardware constraints, and significantly reduces the latency from monitoring to executing configuration instructions.

[0109] In this application, the spatiotemporal state of the optical communication network is perceived in real time based on CNN-LSTM fusion feature vectors, and routing paths and bandwidth allocation are dynamically adjusted in conjunction with SDN decision-making. Under burst traffic scenarios, the spectrum fragmentation rate can be reduced to 8.7%, improving bandwidth efficiency by 37% compared to traditional static allocation schemes, effectively meeting the dynamic bandwidth requirements under multiple service priorities. Furthermore, this application introduces a three-layer MLP into the SPRT framework to construct a neural likelihood ratio estimator, learning the probability density from the CNN-LSTM spatiotemporal fusion features. At the same false alarm rate (1e-5), the configuration method provided in this application can achieve a detection probability of 92.3% for unknown fault types, significantly enhancing the robustness and sensitivity of fault detection.

[0110] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods according to the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method.

[0111] This application also provides a configuration device for an optical communication network. The configuration device is, for example, an optical network management server. Exemplarily, the configuration device is an optical network management server equipped with an FPGA.

[0112] The configuration device includes an acquisition module, a preprocessing module, a fusion feature extraction module, a log-likelihood ratio calculation module, a decision threshold generation module, and a judgment module.

[0113] The acquisition module is used for the acquisition of data from the optical communication network. In this observation, raw data was acquired; the raw data includes physical layer data and service layer data of the optical communication network. It is a positive integer greater than 0.

[0114] The preprocessing module is used to determine the link state matrix and time-series traffic matrix based on the raw data.

[0115] The fusion feature extraction module is used to determine the fusion feature vector based on the link state matrix and the time-series traffic matrix. The fusion feature vector includes the spatial features of the link state matrix and the temporal features of the time-series traffic matrix. The log-likelihood ratio calculation module is used to determine the cumulative log-likelihood ratio based on the fused feature vector.

[0116] The decision threshold generation module is used to determine a decision threshold group based on the initial threshold group. The decision threshold group includes a first decision threshold and a second decision threshold.

[0117] The judgment module is used to determine the state of the optical communication network based on at least one of a first decision threshold and a second decision threshold, and the cumulative log-likelihood ratio; if the cumulative log-likelihood ratio is less than or equal to the second decision threshold, the optical communication network is determined to be in a normal state.

[0118] In some embodiments, obtaining raw data includes at least one of the following: obtaining the raw optical power value of the optical communication network, and determining the calibration optical power value based on the raw optical power value and the nonlinear compensation coefficient; obtaining the signal-to-noise ratio of the optical communication network, and determining the bit error rate of the optical communication network based on the signal-to-noise ratio and the complementary error function.

[0119] In some embodiments, determining the link state matrix and the time-series traffic matrix based on the original data includes: determining the time-series traffic matrix based on the original time-series traffic matrix, the maximum value of traffic data within the sliding window, and the minimum value of traffic data within the sliding window.

[0120] In some embodiments, determining the fused feature vector based on the link state matrix and the temporal traffic matrix includes: extracting the spatial feature vector of the link state matrix using a first feature extraction module; the second feature extraction module includes a convolutional network structure; extracting the temporal feature vector of the temporal traffic matrix using the second feature extraction module; the second feature extraction module includes a recurrent network structure with memory function; determining the fused feature vector based on the spatial feature vector and the temporal feature vector; the dimension of the fused feature vector is the sum of the dimension of the spatial feature vector and the dimension of the temporal feature vector.

[0121] In some embodiments, determining the cumulative log-likelihood ratio based on the fused feature vector includes: The probability of the fault hypothesis and the probability of the normal hypothesis are determined based on the fused feature vector. The probability of the fault hypothesis represents the probability that the optical communication network observes the fused feature vector under fault conditions, and the probability of the normal hypothesis represents the probability that the optical communication network observes the fused feature vector under normal conditions.

[0122] The cumulative log-likelihood ratio is determined based on the probability of failure and the probability of normality.

[0123] In some embodiments, determining the probability of a fault hypothesis and the probability of a normal hypothesis based on the fused feature vector includes: determining the probability of a fault hypothesis and the probability of a normal hypothesis based on a multilayer perceptron corresponding to the dimension of the fused feature vector.

[0124] In some embodiments, determining the cumulative log-likelihood ratio based on the failure hypothesis probability and the normal hypothesis probability includes: When the cumulative log-likelihood ratio is equal to 1, it is the th The single log-likelihood ratio determined by the nth observation is based on the nth observation. The probability of the failure hypothesis and the probability of the normal hypothesis are determined by the second observation; in When the cumulative log-likelihood ratio is greater than 1, it is the sum of the single log-likelihood ratio determined by the first observation up to the second observation. The sum of single log-likelihood ratios determined by each observation.

[0125] In some embodiments, determining a decision threshold group based on an initialization threshold group includes: performing a gradient descent step on at least one meta-task in a meta-task group based on the initialization threshold group to determine an intermediate threshold parameter; the meta-task includes a failure mode of an optical communication network; determining an aggregated gradient of the meta-task group based on the intermediate threshold parameter; and determining a decision threshold group based on the aggregated gradient.

[0126] In some embodiments, determining the aggregation gradient of the meta-task group based on the intermediate threshold parameter includes: determining the aggregation gradient based on the intermediate threshold parameter and hardware constraints; the hardware constraints are determined based on the expectation of the fused feature vector and the fault true label, the fault true label, the cumulative log-likelihood ratio, the decision threshold group, and the optical switch response threshold.

[0127] In some embodiments, determining a decision threshold set based on an initial threshold set includes: performing a gradient descent step on at least one meta-task in a meta-task group based on the initial threshold set and a first cyclic learning rate to determine an intermediate threshold parameter; the meta-task includes a failure mode of an optical communication network; determining an aggregated gradient of the meta-task group based on the intermediate threshold parameter; and determining a decision threshold based on the aggregated gradient and a second cyclic learning rate.

[0128] In some embodiments, determining the state of the optical communication network based on at least one of a first decision threshold and a second decision threshold, and the cumulative log-likelihood ratio, further includes: determining the optical communication network as a fault state if the cumulative log-likelihood ratio is greater than or equal to the first decision threshold.

[0129] In some embodiments, when the optical communication network is in a fault state, the configuration method further includes: determining configuration instructions based on the physical layer constraints of the optical communication network; the physical layer constraints include at least one of wavelength continuity, spectral fragmentation, and optical switch response time; the configuration instructions include at least one of adjusting the modulation format, calculating the dispersion compensation map, and switching links.

[0130] In some embodiments, the method further includes: determining the routing path and dynamic bandwidth allocation scheme of the optical communication network based on the network path optimization model and the physical layer constraints of the optical communication network.

[0131] In some embodiments, the optical communication network includes at least one optical switch, and the configuration method further includes: generating at least one optical switch control signal based on configuration instructions; and performing at least one of link switching, power adjustment, or modulation format change on the optical communication network based on the at least one optical switch control signal.

[0132] In some embodiments, after performing at least one of link switching, power adjustment, or modulation format change on the optical communication network based on at least one optical switch control signal, the configuration method further includes: performing a step on the optical communication network... +1 observation.

[0133] In some embodiments, when the optical switch control signal is in a first state, the link corresponding to the optical switch control signal is activated; when the optical switch control signal is in a second state, the current state of the link corresponding to the optical switch control signal is maintained; or, when the optical switch control signal is in a first state, the state of the link corresponding to the optical switch control signal is switched; when the optical switch control signal is in a second state, the current state of the link corresponding to the optical switch control signal is maintained.

[0134] In some embodiments, determining the state of the optical communication network based on at least one of a first decision threshold and a second decision threshold, and the cumulative log-likelihood ratio, further includes: determining the optical communication network as an observation state if the cumulative log-likelihood ratio is greater than the second decision threshold and less than the first decision threshold; and, if the optical communication network is in an observation state, performing a step on the optical communication network... +1 observation.

[0135] For a description of the features in the embodiment corresponding to the configuration device of the optical communication network, please refer to the relevant description in the embodiment corresponding to the configuration method of the optical communication network, which will not be repeated here.

[0136] Embodiments of this application also provide an electronic device, including a memory and a processor, wherein the memory stores a computer program and the processor is configured to run the computer program to perform the steps in any of the above-described embodiments of the configuration method for an optical communication network.

[0137] Embodiments of this application also provide a computer-readable storage medium storing a computer program, wherein the computer program is configured to execute the steps in any of the above-described embodiments of the configuration method for an optical communication network when it is run.

[0138] In one exemplary embodiment, the aforementioned computer-readable storage medium may include, but is not limited to, various media capable of storing computer programs, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard disk, magnetic disk, or optical disk.

[0139] Embodiments of this application also provide a computer program product, which includes a computer program that, when executed by a processor, implements the steps in any of the above-described configuration method embodiments of an optical communication network.

[0140] Embodiments of this application also provide another computer program product, including a non-volatile computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps in any of the above configuration method embodiments.

[0141] Any of the components, modules, units, parts, methods, and operations described herein can be implemented using software, firmware, hardware (e.g., fixed logic circuitry), manual processing, or any combination thereof. Alternatively or additionally, any functionality described herein can be executed at least in part by one or more hardware logic components, such as, but not limited to, a central processing unit (CPU), a field-programmable gate array (FPGA), an application-specific integrated circuit (ASIC), an application-specific standard product (ASSP), a system-on-a-chip (SoC), a complex programmable logic device (CPLD), a microprocessor (MCU), etc. The terms "system," "computing device," or "apparatus" as used herein encompass various means, devices, and machines for processing data, including, for example, one or more programmable processors, computers, SoCs, or combinations thereof. The apparatus may also include code that creates an execution environment for the computer program in question, such as code constituting processor firmware, a protocol stack, a database management system, an operating system, a cross-platform runtime environment, a virtual machine, or one or more combinations thereof. The aforementioned computer program (also known as a program, software, software application, app, script, or code) can be written in any form of programming language, including compiled or interpreted languages, declarative or procedural languages, and can be deployed in any form, including as a standalone program or as a module, component, subroutine, object, or other unit suitable for a computing environment.

[0142] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0143] The configuration method and apparatus for an optical communication network provided in this application have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the embodiments above are only for the purpose of helping to understand the method and core ideas of this application. It should be noted that those skilled in the art can make several improvements and modifications to this application without departing from the principles of this application, and these improvements and modifications also fall within the protection scope of the claims of this application.

Claims

1. A configuration method for an optical communication network, characterized in that, include: In the first aspect of the optical communication network In this observation, raw data is acquired; the raw data includes the physical layer data and the service layer data of the optical communication network. It is a positive integer greater than 0; The link state matrix and time-series traffic matrix are determined based on the raw data; Based on the link state matrix and the time-series traffic matrix, a fusion feature vector is determined; The fused feature vector includes the spatial features of the link state matrix and the temporal features of the temporal traffic matrix; The cumulative log-likelihood ratio is determined based on the fused feature vector; Based on the initialization threshold group, a decision threshold group is determined, wherein the decision threshold group includes a first decision threshold and a second decision threshold. The state of the optical communication network is determined based on at least one of the first decision threshold and the second decision threshold, and the cumulative log-likelihood ratio. If the cumulative log-likelihood ratio is less than or equal to the second decision threshold, the optical communication network is determined to be in a normal state.

2. The configuration method according to claim 1, characterized in that, The acquisition of raw data includes at least one of the following: Obtain the original optical power value of the optical communication network, and determine the calibration optical power value based on the original optical power value and the nonlinear compensation coefficient; The signal-to-noise ratio (SNR) of the optical communication network is obtained, and the bit error rate (BER) of the optical communication network is determined based on the SNR and the complementary error function.

3. The configuration method according to claim 1, characterized in that, The process of determining the link state matrix and time-series traffic matrix based on the raw data includes: The time-series flow matrix is ​​determined based on the original time-series flow matrix, the maximum value of the flow data within the sliding window, and the minimum value of the flow data within the sliding window.

4. The configuration method according to claim 1, characterized in that, The determination of the fused feature vector based on the link state matrix and the time-series traffic matrix includes: The spatial feature vector of the link state matrix is ​​extracted using a first feature extraction module; the first feature extraction module includes a convolutional network structure. The temporal feature vector of the temporal flow matrix is ​​extracted using a second feature extraction module; the second feature extraction module includes a recurrent network structure with memory function; Based on the spatial feature vector and the temporal feature vector, the fused feature vector is determined; the dimension of the fused feature vector is the sum of the dimension of the spatial feature vector and the dimension of the temporal feature vector.

5. The configuration method according to claim 1, characterized in that, The step of determining the cumulative log-likelihood ratio based on the fused feature vector includes: The probability of a fault hypothesis and the probability of a normal hypothesis are determined based on the fused feature vector; the probability of a fault hypothesis represents the probability that the optical communication network observes the fused feature vector in a fault state, and the probability of a normal hypothesis represents the probability that the optical communication network observes the fused feature vector in a normal state. The cumulative log-likelihood ratio is determined based on the failure hypothesis probability and the normal hypothesis probability.

6. The configuration method according to claim 5, characterized in that, The determination of the fault hypothesis probability and the normal hypothesis probability based on the fused feature vector includes: Based on the multilayer perceptron corresponding to the dimension of the fused feature vector, the probability of the fault hypothesis and the probability of the normal hypothesis are determined.

7. The configuration method according to claim 5, characterized in that, Determining the cumulative log-likelihood ratio based on the failure hypothesis probability and the normal hypothesis probability includes: exist When the cumulative log-likelihood ratio is equal to 1, it is the th The single log-likelihood ratio determined by the nth observation, the single log-likelihood ratio being based on the nth observation. The probability of the fault hypothesis and the probability of the normal hypothesis generated by the second observation are determined; exist When the cumulative log-likelihood ratio is greater than 1, it is the sum of the single log-likelihood ratio determined by the first observation up to the second observation. The sum of single log-likelihood ratios determined by each observation.

8. The configuration method according to claim 1, characterized in that, The process of determining the decision threshold set based on the initialization threshold set includes: A gradient descent step is performed on at least one meta-task in the meta-task group based on an initialization threshold group to determine an intermediate threshold parameter; the meta-tasks include the fault modes of the optical communication network. The aggregated gradient of the meta-task group is determined based on the intermediate threshold parameter, and the decision threshold group is determined based on the aggregated gradient.

9. The configuration method according to claim 8, characterized in that, Determining the aggregated gradient of the meta-task group based on the intermediate threshold parameter includes: The aggregation gradient is determined based on the intermediate threshold parameter and hardware constraints. The hardware constraints are determined based on the expectation of the fused feature vector and the fault true label, the fault true label, the cumulative log-likelihood ratio, the decision threshold set, and the optical switch response threshold.

10. The configuration method according to claim 8, characterized in that, The process of determining the decision threshold set based on the initialization threshold set includes: Based on the initialization threshold group and the first cyclic learning rate, perform a gradient descent step on at least one meta-task in the meta-task group to determine the intermediate threshold parameter; the meta-task includes the fault modes of the optical communication network. The aggregated gradient of the meta-task group is determined based on the intermediate threshold parameter, and the decision threshold group is determined based on the aggregated gradient and the second cyclic learning rate.

11. The configuration method according to any one of claims 1-10, characterized in that, The step of determining the state of the optical communication network based on at least one of the first decision threshold and the second decision threshold, and the cumulative log-likelihood ratio, further includes: If the cumulative log-likelihood ratio is greater than or equal to the first decision threshold, the optical communication network is determined to be in a fault state.

12. The configuration method according to claim 11, characterized in that, In the event that the optical communication network is in a faulty state, the configuration method further includes: Based on the physical layer constraints of the optical communication network, configuration instructions are determined; the physical layer constraints include at least one of wavelength continuity, spectral fragmentation, and optical switch response time; the configuration instructions include at least one of adjusting the modulation format, calculating the dispersion compensation map, and switching links.

13. The configuration method according to claim 12, characterized in that, Also includes: Based on the network path optimization model and the physical layer constraints of the optical communication network, the routing path and dynamic bandwidth allocation scheme of the optical communication network are determined.

14. The configuration method according to claim 12, characterized in that, The optical communication network includes at least one optical switch, and the configuration method further includes: At least one optical switch control signal is generated based on the configuration instructions; Based on the at least one optical switch control signal, at least one of link switching, power adjustment, or modulation format change is performed on the optical communication network.

15. The configuration method according to claim 14, characterized in that, After performing at least one of link switching, power adjustment, or modulation format change on the optical communication network based on the at least one optical switch control signal, the configuration method further includes: Execute the first step on the optical communication network +1 observation.

16. The configuration method according to claim 14, characterized in that, When the optical switch control signal is in the first state, the link corresponding to the optical switch control signal is activated; when the optical switch control signal is in the second state, the current state of the link corresponding to the optical switch control signal is maintained. Alternatively, when the optical switch control signal is in the first state, the state of the link corresponding to the optical switch control signal is switched; when the optical switch control signal is in the second state, the current state of the link corresponding to the optical switch control signal is maintained.

17. The configuration method according to any one of claims 1-10, characterized in that, The step of determining the state of the optical communication network based on at least one of the first decision threshold and the second decision threshold, and the cumulative log-likelihood ratio, further includes: If the cumulative log-likelihood ratio is greater than the second decision threshold and less than the first decision threshold, the optical communication network is determined to be in an observation state. When the optical communication network is in observation mode, perform the first step on the optical communication network. +1 observation.

18. A configuration device for an optical communication network, characterized in that, include: The acquisition module is used to acquire data from the optical communication network at the first... In this observation, raw data is acquired; the raw data includes the physical layer data and the service layer data of the optical communication network. It is a positive integer greater than 0; The preprocessing module is used to determine the link state matrix and the time-series traffic matrix based on the raw data; The fusion feature extraction module is used to determine the fusion feature vector based on the link state matrix and the time-series traffic matrix; The fused feature vector includes the spatial features of the link state matrix and the temporal features of the temporal traffic matrix; The log-likelihood ratio calculation module is used to determine the cumulative log-likelihood ratio based on the fused feature vector; A decision threshold generation module is used to determine a decision threshold group based on an initial threshold group, wherein the decision threshold group includes a first decision threshold and a second decision threshold. The judgment module is used to determine the state of the optical communication network based on at least one of the first decision threshold and the second decision threshold, and the cumulative log-likelihood ratio. If the cumulative log-likelihood ratio is less than or equal to the second decision threshold, the optical communication network is determined to be in a normal state.

19. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor, configured to implement the steps of the configuration method as described in any one of claims 1-17 when executing the computer program.

20. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program, wherein the computer program, when executed by a processor, implements the steps of the configuration method as described in any one of claims 1 to 17.

21. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the configuration method as described in any one of claims 1 to 17.