Identification method, device, and medium

Abstract

The present application relates to the technical field of communications, and discloses an identification method, a device, and a medium. The identification method comprises: acquiring abnormal signal data, and identifying the abnormal signal data to obtain rogue optical network unit information. The dependence on manual identification experience can be reduced, and the identification efficiency and reliability of rogue optical network units are effectively improved.

Description

Identification method, device and medium

[0001] Related applications

[0002] This application claims priority to Chinese Patent Application No. 202410700812.0, filed on May 31, 2024, the contents of which are incorporated herein by reference in its entirety. TECHNICAL FIELD

[0003] The present application relates to the field of communication, in particular to an identification method, a network device and a storage medium. BACKGROUND

[0004] Passive optical network (PON) technology, as a mainstream technology for fiber to the home, has been widely deployed to provide users with stable and reliable high access bandwidth.

[0005] A PON system is a point-to-multipoint network topology, in which an optical line terminal (OLT) is connected to multiple optical network units (ONUs) through an optical distribution network (ODN). In the uplink direction, the ONUs transmit signals under the authorization of the OLT. However, in actual operation of the PON system, due to problems in the hardware and software of the ONUs, malicious user behavior, etc., the ONUs may transmit signals in the uplink direction without authorization from the OLT. At this time, the signals transmitted by the ONUs without authorization from the OLT will conflict with the signals transmitted by other ONUs in the uplink direction, and will affect the signals transmitted by the normal ONUs in the uplink direction. Such an ONU that transmits signals in the uplink direction without authorization from the OLT is referred to as a rogue ONU.

[0006] Currently, the identification of rogue ONUs generally relies on manual operation, which results in low reliability and low efficiency of the identification of rogue ONUs. SUMMARY

[0007] The main purpose of the present application is to provide an identification method, an identification method, a network device and a storage medium.

[0008] To achieve the above purpose, the present application provides an identification method, which is applied to an optical line terminal, and the method comprises: acquiring abnormal signal data; identifying the abnormal signal data to obtain rogue optical network unit information.

[0009] The embodiment of the present application further provides a recognition method, which is applied to a base station central unit and a computing center, and comprises the following steps: obtaining training data and related optical network unit information from an optical line terminal; and training based on the training data and the related optical network unit information to obtain a recognition model, wherein the recognition model is used to recognize rogue optical network unit information according to abnormal signal data.

[0010] The embodiment of the present application further provides a network device, which comprises a memory, a processor and a computer program stored in the memory and executable on the processor, and the computer program is configured to implement the steps of the recognition method.

[0011] The embodiment of the present application further provides a storage medium, which is a computer readable storage medium, and the storage medium stores a computer program, and the computer program is executed by a processor to implement the steps of the recognition method. BRIEF DESCRIPTION OF DRAWINGS

[0012] Fig. 1 is a structural schematic diagram of a running device of a hardware running environment related to the embodiment of the present application;

[0013] Fig. 2 is a flow schematic diagram of a recognition method according to a first embodiment;

[0014] Fig. 3 is a schematic diagram of an overall architecture according to the first embodiment;

[0015] Fig. 4 is a flow schematic diagram of a recognition method according to a second embodiment;

[0016] Fig. 5 is a schematic diagram of an architecture according to the second embodiment;

[0017] Fig. 6 is a flow schematic diagram of a recognition method according to a third embodiment;

[0018] Fig. 7 is a flow schematic diagram of a recognition method according to a fourth embodiment;

[0019] Fig. 8 is a structural schematic diagram one of a recognition device provided by the embodiment of the present application;

[0020] Fig. 9 is a structural schematic diagram two of a recognition device provided by the embodiment of the present application.

[0021] The implementation, functional features and advantages of the present application will be further described with reference to the accompanying drawings. DETAILED DESCRIPTION

[0022] It should be understood that the specific embodiments described herein are merely intended to explain the present application, and are not intended to limit the present application.

[0023] At present, as the mainstream technology of FTTH, the passive optical network (PON) technology has been deployed on a large scale to provide users with stable and reliable high access bandwidth. The PON system is a point-to-multipoint network topology, and one OLT connects multiple ONUs through an ODN. In the uplink direction, the ONUs send signals under the authorization of the OLT. However, in the actual operation of the PON system, due to the software and hardware problems of the ONUs, malicious user behaviors, and the like, the ONUs may send signals in the uplink direction without the authorization of the OLT. At this time, the uplink signals of the ONUs may conflict with each other and affect the uplink signals of the normal uplink ONUs. Such an ONU that sends signals in the uplink direction without the authorization of the OLT is a rogue ONU. At present, there is no reliable method for identifying and locating the rogue ONU. In actual engineering, the identification of the rogue ONU generally depends on manual operation, which is a relatively long and difficult process. Generally, only the rogue ONU that sends signals for a long time can be identified, and it is difficult to locate the rogue ONU that sends signals for a short time or at random.

[0024] To solve the above technical problems, referring to FIG. 1, FIG. 1 is a schematic diagram of the running device structure of the hardware running environment related to the embodiments of the present application.

[0025] As shown in FIG. 1, the running device can include a processor 1001, for example, a central processing unit (CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. The communication bus 1002 is used to realize the connection and communication between the components. The user interface 1003 can include a display screen, an input unit such as a keyboard, and can also include a standard wired interface and a wireless interface. The network interface 1004 can include a standard wired interface and a wireless interface (such as a wireless fidelity (WI-FI) interface). The memory 1005 can be a high-speed random access memory (RAM) or a stable non-volatile memory (NVM) such as a disk memory. The memory 1005 can also be a storage device independent of the aforementioned processor 1001.

[0026] The structure shown in FIG. 1 does not constitute a limitation on the running device, and can include more or fewer components than those shown, or combine certain components, or different component arrangements.

[0027] As shown in FIG. 1, the memory 1005 as a storage medium can include an operating system, a data storage module, a network communication module, a user interface module, and a computer program.

[0028] In the running device shown in FIG. 1, the network interface 1004 is mainly used for data communication with other devices; the user interface 1003 is mainly used for data interaction with the user; the processor 1001 and the memory 1005 in the running device of the application can be arranged in the running device, and the running device calls the computer program stored in the memory 1005 through the processor 1001 and performs the following operations: acquiring abnormal signal data; identifying the abnormal signal data to obtain rogue optical network unit information.

[0029] For example, the step of acquiring abnormal signal data further includes: continuously collecting and analyzing the uplink signal, and determining the collected data as the abnormal signal data in the case of detecting mutation of the uplink signal in the collected data; and / or, performing error correction processing on the uplink signal, and determining the information block corresponding to the error code as the abnormal signal data in the case of detecting the error code.

[0030] For example, the step of continuously collecting and analyzing the uplink signal and / or error correction processing further includes: allocating uplink bandwidth to the target optical network unit to form a bandwidth allocation result; and sending the bandwidth allocation result to the target optical network unit, so that the target optical network unit sends the uplink signal according to the bandwidth allocation result.

[0031] For example, the step of acquiring abnormal signal data includes: acquiring the abnormal signal data; and acquiring the optical network unit information to which the uplink bandwidth corresponding to the abnormal signal data belongs according to the bandwidth allocation result.

[0032] For example, the step of identifying the abnormal signal data to obtain rogue optical network unit information includes: identifying the abnormal signal data and the optical network unit information to which the corresponding uplink bandwidth belongs through a pre-trained identification model to obtain the rogue optical network unit information.

[0033] For example, the step of identifying the abnormal signal data and the optical network unit information to which the corresponding uplink bandwidth belongs through a pre-trained identification model to obtain the rogue optical network unit information further includes: acquiring training data and related optical network unit information; training based on the training data and related optical network unit information to obtain the identification model; and / or, sending the training data and related optical network unit information to a computing center, so that the computing center trains based on the training data and related optical network unit information to obtain the identification model, and obtains the identification model from the computing center.

[0034] The step of obtaining the training data comprises: collecting sample signals of the optical network units, and obtaining synthetic collision data according to the sample signals of the optical network units, and / or sending the sample signals of the optical network units to the computing center, so that the computing center obtains the synthetic collision data according to the sample signals of the optical network units, and obtaining the synthetic collision data from the computing center, and / or collecting actual collision data; and taking the synthetic collision data and / or the actual collision data as the training data.

[0035] The step of collecting sample signals of the optical network units and obtaining synthetic collision data according to the sample signals of the optical network units comprises: calculating a sampling starting point according to a local sampling clock and a transmission clock of the optical network unit; collecting sample signals of two optical network units according to the sampling starting point; and superimposing the sample signals of the two optical network units to obtain the synthetic collision data.

[0036] The step of collecting actual collision data comprises at least one of the following: allocating uplink bandwidths with ascending starting times and overlapping each other to two optical network units, and collecting signal data corresponding to the overlapping part of the uplink bandwidths as the actual collision data; allocating completely overlapping uplink bandwidths to two optical network units when the optical network units are set not to detect the starting time ascending relationship of uplink bandwidth mapping entries allocated by an OLT to an ONU, and collecting signal data corresponding to the overlapping uplink bandwidths as the actual collision data, the uplink bandwidth mapping entry at least indicating a starting time of the uplink bandwidth and a length or an ending time of the uplink bandwidth; allocating the same transmission container to two optical network units, allocating uplink bandwidths to the transmission container, and collecting uplink signals sent by the two optical network units in the uplink bandwidths as the actual collision data.

[0037] The step of training based on the training data and related optical network unit information to obtain the identification model comprises: screening training data according to one optical network unit information in the related optical network information, training based on the screened training data and another optical network unit information to obtain the identification model; and / or training based on the training data and related optical network units to obtain the identification model.

[0038] The step of training the identification model based on the training data and the related optical network unit information includes: inputting the training data and the related optical network unit information into the neural network for forward propagation to obtain a prediction result; calculating a loss function according to the prediction result and the related optical network unit information; performing backward propagation according to the loss function to update network parameters of the neural network, and returning to perform the step of inputting the training data and the related optical network unit information into the neural network for forward propagation and subsequent steps until the neural network converges, the training is terminated, and the identification model is obtained.

[0039] The identification model includes a single decision model. Based on the single decision model, the step of identifying the abnormal signal data to obtain the rogue optical network unit information corresponding to the abnormal signal data includes: determining the single decision model according to the optical network unit information to which the uplink bandwidth belongs, and identifying the abnormal signal data to output at least one number of a rogue optical network unit corresponding to the abnormal signal data; and / or identifying the abnormal signal data through the single decision model to output two optical network unit information that generates the abnormal signal data, determining the optical network unit information that is affected in the two optical network unit information according to the optical network unit information to which the uplink bandwidth belongs, and determining the other optical network unit information as the rogue optical network unit information.

[0040] The identification model includes at least one classification decision model. Based on the at least one classification decision model, the step of identifying the abnormal signal data to obtain the rogue optical network unit information corresponding to the abnormal signal data includes: identifying a data type corresponding to the abnormal signal data; calling a corresponding at least one classification decision model according to the data type, inputting the abnormal signal data into the at least one classification decision model to obtain at least one number of a rogue optical network unit.

[0041] The processor 1001 can call a computer program stored in the memory 1005, and further perform the following operations: obtaining training data and related optical network unit information from an optical line terminal; training based on the training data and the related optical network unit information to obtain an identification model; wherein the identification model is used to identify rogue optical network unit information according to abnormal signal data.

[0042] Exemplarily, the step of obtaining the training data comprises: receiving the synthetic conflict data and / or the actual conflict data sent by the optical line terminal, and / or receiving the optical network unit sample signal sent by the optical line terminal, and obtaining the synthetic conflict data according to the optical network unit sample signal; taking the synthetic conflict data and / or the actual conflict data as the training data, and / or sending the synthetic conflict data to the optical line terminal, so that the optical line terminal takes the synthetic conflict data as the training data.

[0043] Exemplarily, the step of training based on the training data and the related optical network unit information to obtain the identification model comprises: inputting the training data and the related optical network unit information into the neural network to perform forward propagation, to obtain a prediction result; calculating a loss function according to the prediction result and the related optical network unit information; performing backward propagation according to the loss function, updating network parameters of the neural network, and returning to perform the step of inputting the training data and the related optical network unit information into the neural network to perform forward propagation and subsequent steps, until the neural network converges, terminating the training, and obtaining the identification model.

[0044] Exemplarily, the step of training based on the training data and the related optical network unit information to obtain the identification model further comprises: sending the identification model to the optical line terminal, so that the optical line terminal identifies the obtained abnormal signal data based on the identification model, to obtain the rogue optical network unit information corresponding to the abnormal signal data.

[0045] Exemplarily, the step of training based on the training data and the related optical network unit information to obtain the identification model further comprises: receiving the abnormal signal data sent by the optical line terminal; identifying the abnormal signal data based on the identification model, to obtain the rogue optical network unit information corresponding to the abnormal signal data.

[0046] Exemplarily, the identification model comprises a single decision model and / or at least one classification decision model.

[0047] Exemplarily, the step of identifying the abnormal signal data based on the identification model to obtain the rogue optical network unit information corresponding to the abnormal signal data comprises at least one of the following: determining the single decision model according to the optical network unit information to which the uplink bandwidth belongs, identifying the abnormal signal data, and outputting the number of at least one rogue optical network unit corresponding to the abnormal signal data; identifying the abnormal signal data through the single decision model, outputting the information of two optical network units that generate the abnormal signal data, determining the optical network unit information that is affected from the two optical network unit information according to the optical network unit information to which the uplink bandwidth belongs, and determining the other optical network unit information as the rogue optical network unit information; identifying the data type corresponding to the abnormal signal data, and calling at least one classification decision model according to the data type, inputting the abnormal signal data into the at least one classification decision model, and obtaining the number of at least one rogue optical network unit.

[0048] First embodiment

[0049] Referring to FIG. 2, FIG. 2 is a flowchart of an identification method according to the first embodiment, and the identification method of the embodiment of the application can be applied to an optical line terminal. The method comprises the following steps:

[0050] In step S10, abnormal signal data is obtained.

[0051] A passive optical network (PON) system is a point-to-multipoint network topology. An optical line terminal (OLT) is connected to multiple optical network units (ONUs) through an optical distribution network (ODN). In the uplink direction, the ONUs send signals under the authorization of the OLT. However, in actual operation of the PON system, due to problems in the hardware and software of the ONUs, malicious user behavior, and the like, the ONUs may send signals in the uplink direction without obtaining the authorization of the OLT. At this time, the signals may conflict with the signals of other ONUs that normally send signals in the uplink direction, and affect the signals of the normally sending ONUs in the uplink direction. Such an ONU that sends signals in the uplink direction without obtaining the authorization of the OLT is a rogue ONU. Currently, there is no reliable method for identifying and locating rogue ONUs. In actual engineering, manual work is generally relied on, and only rogue ONUs that send signals for a long time can be identified. It is difficult to identify rogue ONUs that send signals for a short time or randomly.

[0052] The embodiment of the application provides a PON system architecture based on artificial intelligence, in an ONU normal working phase, an OLT obtains signal data sent by an ONU in an uplink direction, the OLT sends the data to a training center for training, completes training and forms an identification model, then the OLT monitors and obtains to-be-identified data, inputs the to-be-identified data into the identification model and outputs a decision result, for example, outputs a rogue ONU ID, outputs an identity abnormal ONU, etc.

[0053] Referring to FIG. 3, FIG. 3 is a schematic diagram of an overall architecture according to the first embodiment, as shown in FIG. 3, the architecture mainly includes three functional modules, including data acquisition (sampling), data training (machine learning, ML) and data decision (identification model).

[0054] For example, data acquisition is divided into acquisition of training data and acquisition of decision data, the acquisition of training data is generally performed on a specific signal, the OLT acquires a signal normally sent by the ONU, or instructs the ONU to send a specific signal and then the OLT acquires the signal, for data training and forming a decision model, the acquisition of decision data is generally performed on a target signal and input into a decision model for data decision. The data obtained by acquiring the signal is generally a series of waveform amplitude values at corresponding time points according to a certain sampling rate, based on the series of data, other format data can also be obtained by conversion.

[0055] In the embodiment of the application, data acquisition is performed by the OLT, and data training and data decision can be performed by the OLT or in a computing center (for example, a network management system, a computing single board in the same machine frame, etc.). For example, the data acquisition functional module is in the OLT, the data training and data decision modules are in the computing center outside the OLT, and the OLT sends training data and decision data and related information to the computing center; for example, the data acquisition and data decision are in the OLT, and the data training is in the computing center outside the OLT, and the OLT sends the training data to the computing center, and the computing center returns an identification model trained and related parameters to the OLT.

[0056] For example, the step of obtaining abnormal signal data further includes: continuously acquiring and analyzing the uplink signal, and determining the acquired data as the abnormal signal data in a case where a mutation of the uplink signal in the acquired data is detected; and / or performing error correction processing on the uplink signal, and determining an information block corresponding to an error code as the abnormal signal data in a case where the error code is detected.

[0057] The step of continuously collecting and analyzing and / or error correction processing the uplink signal further comprises: allocating uplink bandwidth to the target optical network unit to form a bandwidth allocation result; and sending the bandwidth allocation result to the target optical network unit, so that the target optical network unit sends the uplink signal according to the bandwidth allocation result.

[0058] The step of obtaining the abnormal signal data comprises: obtaining the abnormal signal data; and obtaining optical network unit information corresponding to uplink bandwidth of the abnormal signal data according to the bandwidth allocation result.

[0059] In the embodiment, the OLT built-in DSP (Digital Signal Processor) module samples the uplink signal from the ONU and sends the uplink signal to the computing center for training and decision-making. In order to capture the data to be decided, the OLT DSP also buffers and analyzes the uplink signal data, and if an abnormality is found, the data is snapshoted to form the data to be decided and sent to the network management. The OLT DSP can also snapshot the buffered data on demand, periodically or triggered by a service data error code, and send the data to the computing center for training or as data to be decided.

[0060] In the embodiment, there are two main ways to capture the data to be decided and obtain related information: one is that the OLT continuously analyzes the uplink signal, and if an abnormality is found, such as a sudden change in power, i.e., a significant increase or decrease in signal power, the data is captured. Since the ONU uplink transmission is based on the bandwidth allocation of the OLT, the OLT knows exactly which ONU the data belongs to, and the OLT sends the abnormal signal, ONU-ID, and signal location (which can be superframe number and location in the superframe, length) to the OLT. The other is that the OLT performs FEC (Forward Error Correction) correction, and if an error code appears, the signal corresponding to the FEC code word is captured. Since the ONU uplink transmission is based on the bandwidth allocation of the OLT, the OLT knows exactly which ONU the data belongs to, and the OLT sends the abnormal signal, ONU-ID, and signal location (which can be superframe number and location in the superframe, length) to the OLT.

[0061] In step S20, the abnormal signal data is identified to obtain rogue optical network unit information.

[0062] Further, after the OLT collects the abnormal signal data, the abnormal signal data can be identified to obtain the rogue optical network unit information.

[0063] Exemplarily, the step of identifying the abnormal signal data to obtain rogue optical network unit information comprises: identifying the abnormal signal data and corresponding uplink bandwidth belonging optical network unit information by a pre-trained identification model to obtain the rogue optical network unit information.

[0064] Exemplarily, the identification model can be trained by the OLT according to collected training data, or can be trained by the OLT sending the collected training data to a computing center, the computing center can send the trained identification model to the OLT, or the computing center can receive the abnormal signal data sent by the OLT and identify the abnormal signal data based on the trained identification model to obtain the rogue optical network unit corresponding to the abnormal signal data.

[0065] Exemplarily, in the embodiment, the data collection function module is in the OLT, and the data training and data decision module are in the computing center outside the OLT. In this way, the computing center contains a training module and an identification model. The computing center can directly train the OLT DSP output signal data, or can train the output signal data after being synthesized, or can train according to external input data. The OLT sends training data and decision data to the computing center, including data types, such as training data or decision data, data types corresponding to training data and decision data (such as ONU identification, link quality monitoring data, or abnormal data), and data related information (such as time when the data is obtained, ONU SN and other related information).

[0066] Exemplarily, the identification model includes a single decision model, and the step of identifying the abnormal signal data to obtain the rogue optical network unit information corresponding to the abnormal signal data based on the single decision model comprises: determining the single decision model according to the uplink bandwidth belonging optical network unit information, identifying the abnormal signal data, and outputting at least one number of the rogue optical network unit corresponding to the abnormal signal data; and / or identifying the abnormal signal data by the single decision model to output two optical network unit information generating the abnormal signal data, determining the affected optical network unit information in the two optical network unit information according to the uplink bandwidth belonging optical network unit information, and determining the other optical network unit information as the rogue optical network unit information.

[0067] Exemplarily, the identification model includes at least one classification decision model, and the step of identifying the abnormal signal data to obtain the rogue optical network unit corresponding to the abnormal signal data based on the at least one classification decision model comprises:

[0068] identifying a data type corresponding to the abnormal signal data; and calling at least one classification decision model corresponding to the data type according to the data type, and inputting the abnormal signal data into the at least one classification decision model to obtain the number of at least one rogue optical network unit.

[0069] For example, the decision mode in the embodiment includes two modes. One is to identify a model (i.e., a decision model) calling an ONU-ID corresponding decision model, inputting a signal into the decision model for inference decision, and outputting which ONU is a rogue ONU or which ONUs are suspicious rogue ONUs. Correspondingly, data training is based on signals of conflict between a certain affected ONU and other ONUs as a training set for training. The other mode is to input decision data into a decision model for inference decision, and output conflict signals of which two ONUs. Then, according to an ONU-ID of an affected ONU, an ONU-ID of another ONU in the conflict is output as a rogue ONU. Correspondingly, data training is based on signals of all two-by-two conflicts as a training set for training.

[0070] For example, in the embodiment of the application, in addition to obtaining abnormal signal data, information of an optical network unit to which an uplink bandwidth corresponding to the abnormal signal data belongs also needs to be obtained.

[0071] For example, the OLT can allocate an idle uplink bandwidth, which does not belong to any ONU. Therefore, the information of the optical network unit to which the uplink bandwidth corresponding to the abnormal signal data belongs is not information of any ONU in the network, but a special information. The abnormal signal data obtained in such an uplink bandwidth is a signal sent by a rogue ONU. Therefore, training data can be sample signals (which can be understood as a special case of conflict data) and corresponding ONU information (which can be understood as a special case of related ONU information). An identification model identifies sample signals (which can be understood as a special case of abnormal signal data) and special ONU information (which can be understood as a special case of information of an ONU to which an uplink bandwidth belongs), and outputs rogue ONU information.

[0072] The embodiment can reduce dependence on manual identification experience, and effectively improve identification efficiency and reliability of a rogue optical network unit by obtaining abnormal signal data, identifying the abnormal signal data, and obtaining rogue optical network unit information.

[0073] Second Embodiment

[0074] Referring to FIG. 4, which is a flowchart of an identification method according to the second embodiment, the identification method further includes the following steps before step S20 based on the first embodiment.

[0075] Step S01: Obtain training data and related optical network unit information.

[0076] Exemplarily, the training data can be collected by the OLT, the recognition model can be obtained by the OLT based on the training data and the related optical network unit information, and the training data and the related optical network unit information can be sent to the computing center by the OLT, and the recognition model can be obtained by training in the computing center.

[0077] Referring to FIG. 5, which is a schematic diagram of an architecture according to a second embodiment, as shown in FIG. 5, in this embodiment, the OLT samples the signal output by the signal amplifier, samples the uplink signal at a certain sampling rate (for example, 2 times the sampling rate), calculates the required buffer size according to the required time for service processing or signal analysis, and takes the generation of error codes by FEC decoding as an example. In a 50G-PON, the size of one FEC code block is 17280 bits, and the buffer size is at least 17280* sampling rate* sampling bit number. The sampling bit number is the number of bits used to represent one sampling point. The larger the number of bits, the more accurate the value represented by the sampling point. The OLT continuously analyzes the data in the buffer, captures signal mutations such as power, amplitude, frequency, phase, SNR (Signal-to-Noise Ratio), and outputs a snapshot of the buffer when a mutation is captured. Exemplarily, the snapshot triggers the output of data in the buffer, including data type (training data or decision data), abnormal position (error bit position, signal mutation position), and snapshot time, etc. The snapshot trigger conditions include: (1) triggered by external demand, such as periodicity, FEC decoding error code; (2) signal analysis captures signal mutation.

[0078] Exemplarily, the length of the training data and the decision data can be fixed in this embodiment, for example, 200ns sampling data, and the training data can be collected and updated regularly, or the length of the training data can be determined according to the length of the obtained decision data, and retraining can be performed.

[0079] Step S02, training based on the training data and the related optical network unit information to obtain the recognition model; and / or, sending the training data and the related optical network unit information to the computing center, so that the computing center trains based on the training data and the related optical network unit information to obtain the recognition model, and obtaining the recognition model from the computing center.

[0080] Illustratively, the step of training based on the training data and the related optical network unit information to obtain the identification model comprises: inputting the training data and the related optical network unit information into the neural network for forward propagation to obtain a prediction result; calculating a loss function according to the prediction result and the related optical network unit information; performing back propagation according to the loss function to update network parameters of the neural network, and returning to perform the step of inputting the training data and the related optical network unit information into the neural network for forward propagation and subsequent steps until the neural network converges, terminating the training, and obtaining the identification model.

[0081] Illustratively, in the embodiment of the present application, the CNN convolutional neural network is used as the training model, including an input layer, a convolutional layer, a pooling layer, a ReLU layer, and an output layer. The OLT DSP obtains a sampling rate 8 times the upstream rate after actual sampling or down sampling or interpolation sampling, and inputs the sampled data sequence into the CNN for training and decision. The data is shown in Table 1, and the test accuracy is 90.0%.

[0082] Table 1: Example of identification model test results

[0083] Illustratively, in the embodiment of the present application, the CNN convolutional neural network is used as the training model, including an input layer, a convolutional layer, a pooling layer, a ReLU layer, and an output layer, a total of 12 layers. The OLT DSP obtains a sampling rate 2 times the upstream rate after actual sampling or down sampling or interpolation sampling, and converts the sampled data into an eye diagram, and inputs the CNN for training and decision. The data is shown in Table 1, and the test accuracy is 98.0%.

[0084] In the above scheme, the training data and the related optical network unit information are obtained, the training data and the related optical network unit information are trained based on the training data and the related optical network unit information to obtain the identification model, and / or the training data and the related optical network unit information are sent to the computing center for training based on the training data and the related optical network unit information to obtain the identification model, and the identification model is obtained from the computing center. The OLT or the computing center can train based on the obtained training data and the related optical network unit information to obtain the identification model, which is used for identifying abnormal signal data and obtaining rogue optical network units corresponding to the abnormal signal data. The dependence on manual identification experience can be reduced, and the identification efficiency and reliability of the rogue optical network unit can be effectively improved.

[0085] Third embodiment

[0086] Referring to FIG. 6, FIG. 6 is a flowchart of an identification method according to the third embodiment. In the embodiment, the step of obtaining the training data comprises:

[0087] In step S011, the sample signal of the optical network unit is collected, and the synthetic collision data is obtained according to the sample signal of the optical network unit, and / or the sample signal of the optical network unit is sent to the computing center, so that the computing center obtains the synthetic collision data according to the sample signal of the optical network unit, and the synthetic collision data is obtained from the computing center, and / or the actual collision data is collected.

[0088] In step S012, the synthetic collision data and / or the actual collision data is taken as the training data.

[0089] For example, the step of collecting the sample signal of the optical network unit and obtaining the synthetic collision data according to the sample signal of the optical network unit includes: calculating a sampling starting point according to a local sampling clock and a transmission clock of the optical network unit; collecting the sample signal of the optical network unit of two optical network units according to the sampling starting point; and superimposing the sample signal of the optical network unit of the two optical network units to obtain the synthetic collision data.

[0090] For example, the training is performed based on the synthetic collision signal of the single ONU signal. Each ONU uses the downlink recovery clock as the uplink transmission clock, and the frequency of the local clock of the OLT and the transmission clock of each ONU is basically consistent, but the phases of each other are different. The phase difference between the local clock of the OLT and the transmission clock of each ONU is determined. The OLT calculates the sampling starting point according to the local sampling clock and the transmission clock of the ONU, and directly superimposes the two sampling signals. The sampling starting points of different ONUs are not at the same time, but considering the relationship between the sampling clock and the ONU transmission clock, after a sampling point, after an integer multiple of the sampling clock period and the ONU transmission clock period, the sampling point is determined again. In this way, the sampling point phase of the ONU signal is the same.

[0091] For example, the step of collecting the actual collision data includes at least one of the following: allocating uplink bandwidths with ascending starting times and overlapping each other to two optical network units, and collecting signal data corresponding to the overlapping part of the uplink bandwidth as the actual collision data; setting the starting time of the uplink bandwidth mapping entry not to be detected by the optical network unit, allocating completely overlapping uplink bandwidths to two optical network units, and collecting signal data corresponding to the overlapping uplink bandwidth as the actual collision data. The uplink bandwidth mapping entry is allocated by the OLT to the ONU, and at least indicates the starting time of the uplink bandwidth and the length or end time of the uplink bandwidth; allocating the same transmission container to two optical network units, allocating uplink bandwidths to the transmission container, and collecting uplink signals sent by the two optical network units in the uplink bandwidth as the actual collision data.

[0092] Exemplarily, the step of training the identification model based on the training data and the related optical network unit information comprises: screening training data according to one optical network unit information in the related optical network information, training the identification model based on the screened training data and another optical network unit information; and / or training the identification model based on the training data and the related optical network unit.

[0093] Exemplarily, in the embodiment, the conflict signals between each two ONUs are collected for training, and the actual conflict signals are closer to the training signals.

[0094] Exemplarily, the OLT sends the training data and the related optical network unit information to the computing center, and the related optical network unit information further comprises two ONU-IDs in conflict. During the training process, the training can be performed based on all the conflict data, i.e., the label of the conflict data is the two ONU-IDs in conflict, or the training data can be classified, i.e., one ONU-ID is determined, and the conflict data containing the specified ONU-ID in the two ONU-IDs in conflict is the training set.

[0095] Exemplarily, as one of the implementation manners, according to the standard requirement, the uplink bandwidths allocated to the two ONUs have ascending order of StartTime, but there is overlapping uplink bandwidth between them, and the OLT collects the signal data corresponding to the overlapping uplink bandwidth as the training data; as another of the implementation manners, the ascending order of StartTime of the BWmap entry is set on the ONU side, the OLT allocates completely overlapping bandwidths to the two ONUs, and the OLT collects the signal data corresponding to the overlapping uplink bandwidth as the training data. In the OMC IONU2-G ME, the attribute BWmapCheck attribute is added, which is 1 byte, a write attribute (W), when the value is 0, it means not to check the BWmap entry, and when the value is 1, it means to check the validity of the BWmap entry, in the embodiment, the attribute is set to 0; as still another of the implementation manners, the T-CONT transmission container attribute is added, the T-CONT is associated with the two ONUs, the OLT allocates uplink bandwidth to the T-CONT, and the two ONUs respond respectively. Referring to Table 2, the Assign_Alloc-ID PLOAM message is modified, when byte 7 = 0x02, the Alloc-ID specified by bytes 5-6 can be allocated to multiple ONUs, through the message, one Alloc-ID can be allocated to two ONUs, i.e., one T-CONT is associated with multiple ONUs, the OLT allocates bandwidth to the T-CONT, and the two ONUs will send uplink signals in the bandwidth.

[0096] Table 2, bandwidth allocation example table

[0097] The embodiment provides a plurality of ways of collecting sample signals of the optical network unit and generating training data, so as to make the training data more abundant, thereby improving the reliability of the obtained identification model.

[0098] Fourth embodiment

[0099] Referring to FIG. 7, FIG. 7 is a flowchart of an identification method according to the fourth embodiment, and the identification method of the embodiment can be applied to the computing center. The method comprises the following steps.

[0100] Step A10, obtaining training data and related optical network unit information from the optical line terminal.

[0101] Step A20, training based on the training data and the related optical network unit information to obtain an identification model; wherein the identification model is used to identify rogue optical network unit information according to abnormal signal data.

[0102] For example, the step of obtaining training data by the computing center comprises the following steps: receiving the synthetic conflict data and / or the actual conflict data sent by the optical line terminal, and / or receiving the sample signal of the optical network unit sent by the optical line terminal, and obtaining the synthetic conflict data according to the sample signal of the optical network unit; taking the synthetic conflict data and / or the actual conflict data as the training data, and / or sending the synthetic conflict data to the optical line terminal, so that the optical line terminal takes the synthetic conflict data as the training data.

[0103] For example, the OLT can collect the training data, the OLT can directly train the identification model based on the training data and the related optical network unit information, and the OLT can also send the training data and the related optical network unit information to the computing center, and the computing center trains the identification model.

[0104] Exemplarily, in the embodiment of the present application, the OLT built-in DSP module samples the uplink signal from the ONU and sends it to the computing center for training and decision-making. In order to capture the data to be decided, the OLT DSP also buffers and analyzes the uplink signal data, and if an anomaly is found, the data is snapshoted to form the data to be decided and sent to the network management. The OLT DSP can also snapshot the buffered data as needed, periodically or triggered by a service data error code, and send it to the computing center for training or as data to be decided.

[0105] Exemplarily, in the embodiment of the present application, there are mainly two ways to capture the data to be decided and obtain its related information: one is that the OLT continuously analyzes the uplink signal, and if an anomaly is found, such as a sudden change in power, i.e., a significant increase or decrease in signal power, the data is captured. Since the ONU uplink transmission is based on the bandwidth allocation of the OLT, the OLT knows exactly which ONU the data belongs to, and the OLT sends the abnormal signal, ONU-ID, and signal location (which can be superframe number and location in the superframe, length) to the OLT. The other is that the OLT performs FEC (Forward Error Correction) error correction, and if an error code is found, the signal corresponding to the FEC code word is captured. Since the ONU uplink transmission is based on the bandwidth allocation of the OLT, the OLT knows exactly which ONU the data belongs to, and the OLT sends the abnormal signal, ONU-ID, and signal location (which can be superframe number and location in the superframe, length) to the OLT.

[0106] Exemplarily, in the embodiment, the OLT samples the signal output by the signal amplifier, samples the uplink signal at a certain sampling rate (such as 2 times the sampling rate), calculates the required buffer size according to the required time for service processing or signal analysis, and takes the FEC decoding error as an example. In a 50G-PON, the size of a FEC code block is 17280 bits, and the buffer size is at least 17280* sampling rate* sampling bit number. The sampling bit number is the number of bits used to represent a sampling point. The larger the bit number, the more accurate the value represented by the sampling point. The OLT continuously analyzes the data in the buffer, captures signal mutations such as power, amplitude, frequency, phase, SNR (Signal-to-Noise Ratio), etc., and outputs the snapshot of the buffer if a mutation is captured. Exemplarily, the snapshot triggers the output of the data in the buffer, including data type (training data or decision data), abnormal position (error bit position, signal mutation position), and snapshot time, etc. The snapshot trigger conditions include: (1) capturing by external trigger as needed, such as periodically, FEC decoding error code; (2) signal analysis captures signal mutation.

[0107] Exemplarily, the training data and the decision data length can be fixed in the embodiment, for example, 200ns sampling data, the training data is periodically collected and updated, or the training data length can be determined according to the obtained decision data length, and retraining is performed.

[0108] Exemplarily, the step of training based on the training data and the related optical network unit information to obtain the identification model comprises: inputting the training data and the related optical network unit information into the neural network for forward propagation to obtain a prediction result; calculating a loss function according to the prediction result and the related optical network unit information; performing backward propagation according to the loss function to update the network parameters of the neural network, and returning to execute the step of inputting the training data and the related optical network unit information into the neural network for forward propagation and subsequent steps until the neural network converges, the training is terminated, and the identification model is obtained.

[0109] Exemplarily, the step of training based on the training data and the related optical network unit information to obtain the identification model further comprises: sending the identification model to the optical line terminal, so that the optical line terminal identifies the obtained abnormal signal data based on the identification model to obtain the rogue optical network unit corresponding to the abnormal signal data.

[0110] Exemplarily, the step of training based on the training data and the related optical network unit information to obtain the identification model further comprises: receiving the abnormal signal data sent by the optical line terminal; identifying the abnormal signal data based on the identification model to obtain the rogue optical network unit corresponding to the abnormal signal data.

[0111] Exemplarily, the identification model comprises a single decision model and / or at least one classification decision model.

[0112] Exemplarily, the step of identifying the abnormal signal data based on the identification model to obtain the rogue optical network unit corresponding to the abnormal signal data comprises at least one of the following: determining the single decision model according to the optical network unit information to which the uplink bandwidth belongs, and identifying the abnormal signal data to output the number of at least one rogue optical network unit corresponding to the abnormal signal data; identifying the abnormal signal data through the single decision model to output the information of two optical network units that produce the abnormal signal data, determining the optical network unit information that is affected in the two optical network unit information according to the optical network unit information to which the uplink bandwidth belongs, and determining the other optical network unit information as the rogue optical network unit information; identifying the data type corresponding to the abnormal signal data, and calling at least one classification decision model corresponding to the data type, inputting the abnormal signal data into the at least one classification decision model to obtain the number of at least one rogue optical network unit.

[0113] Exemplarily, the decision mode in the embodiment includes two kinds. One is that a recognition model (i.e., a decision model) calls the decision model corresponding to the ONU-ID, inputs a signal into the decision model for inference decision, and outputs which ONU is a rogue ONU or which ONUs are possible rogue ONUs. Correspondingly, data training is performed based on the signal of conflict between a certain affected ONU and other ONUs as a training set. The other way is that the OLT inputs decision data into the decision model for inference decision, and outputs the conflict signal of which two ONUs. Then, according to the ONU-ID of the affected ONU, the ONU-ID of the other ONU in the conflict is output as a rogue ONU. Correspondingly, data training is performed based on all two-by-two conflict signals as a training set.

[0114] The embodiment obtains training data and related optical network unit information from an optical line terminal through the above scheme; and trains based on the training data and the related optical network unit information to obtain a recognition model. The recognition model is used to identify rogue optical network unit information according to abnormal signal data. Through the computing center, data training and / or data decision are performed to share the computing pressure of the OLT. Through the recognition model, the dependence on manual identification experience can be reduced, and the identification efficiency and reliability of the rogue optical network unit can be effectively improved.

[0115] Fifth embodiment

[0116] Referring to FIG. 8, FIG. 8 is a structural schematic diagram of a recognition device provided by an embodiment of the present application. The device can be carried on or is the terminal device in the method embodiment described above. The recognition device shown in FIG. 8 can be used to execute part or all of the functions in the method embodiment described in the above embodiments. As shown in FIG. 8, the recognition device includes: a first acquisition module configured to acquire abnormal signal data; and an identification module configured to identify the abnormal signal data to obtain rogue optical network unit information. Exemplarily, before the step of acquiring the abnormal signal data, the method further includes: continuously collecting and analyzing the uplink signal, and determining the collected data as the abnormal signal data in the case of detecting that the uplink signal in the collected data has a mutation; and / or performing error correction processing on the uplink signal, and determining the information block corresponding to the error code as the abnormal signal data in the case of detecting the error code.

[0117] Exemplarily, before the steps of continuously collecting and analyzing the uplink signal and / or performing error correction processing, the method further includes: allocating uplink bandwidth to a target optical network unit to form a bandwidth allocation result; and sending the bandwidth allocation result to the target optical network unit, so that the target optical network unit sends the uplink signal according to the bandwidth allocation result.

[0118] Exemplarily, the step of acquiring the abnormal signal data comprises: acquiring the abnormal signal data; and acquiring optical network unit information to which an uplink bandwidth corresponding to the abnormal signal data belongs according to the bandwidth allocation result.

[0119] Exemplarily, the step of identifying the abnormal signal data to obtain rogue optical network unit information comprises: identifying the abnormal signal data and the optical network unit information to which the uplink bandwidth corresponding to the abnormal signal data belongs by using a pre-trained identification model to obtain the rogue optical network unit information.

[0120] Exemplarily, before the step of identifying the abnormal signal data by using the pre-trained identification model, the method further comprises: acquiring training data and related optical network unit information; training based on the training data and the related optical network unit information to obtain the identification model; and / or sending the training data and the related optical network unit information to a computing center, so that the computing center trains based on the training data and the related optical network unit information to obtain the identification model, and the identification model is acquired from the computing center.

[0121] Exemplarily, the step of acquiring the training data comprises: collecting optical network unit sample signals, and obtaining synthetic collision data according to the optical network unit sample signals, and / or sending the optical network unit sample signals to the computing center, so that the computing center obtains synthetic collision data according to the optical network unit sample signals, and the synthetic collision data is acquired from the computing center, and / or collecting actual collision data; and taking the synthetic collision data and / or the actual collision data as the training data.

[0122] Exemplarily, the step of collecting optical network unit sample signals and obtaining synthetic collision data according to the optical network unit sample signals comprises: calculating a sampling starting point according to a local sampling clock and a transmission clock of an optical network unit; collecting optical network unit sample signals of two optical network units according to the sampling starting point; and superimposing the optical network unit sample signals of the two optical network units to obtain the synthetic collision data.

[0123] The step of collecting actual conflict data includes at least one of the following: allocating uplink bandwidths with ascending start times and overlaps for two optical network units, and collecting signal data corresponding to the overlapping part of the uplink bandwidths as the actual conflict data; setting the start time of the uplink bandwidth mapping entries in the optical network units in ascending order, allocating completely overlapping uplink bandwidths for two optical network units, and collecting signal data corresponding to the overlapping uplink bandwidths as the actual conflict data, the uplink bandwidth mapping entries being allocated by the OLT to the ONUs and indicating at least the start time of the uplink bandwidth and the length or end time of the uplink bandwidth; allocating the same transmission container for two optical network units, allocating uplink bandwidths for the transmission container, and collecting uplink signals sent by the two optical network units in the uplink bandwidths as the actual conflict data.

[0124] The step of training the identification model based on the training data and the related optical network unit information includes: filtering training data according to one optical network unit information in the related optical network information, and training the identification model based on the filtered training data and another optical network unit information; and / or training the identification model based on the training data and the related optical network unit information.

[0125] The step of training the identification model based on the training data and the related optical network unit information includes: inputting the training data and the related optical network unit information into the neural network for forward propagation to obtain a prediction result; calculating a loss function based on the prediction result and the related optical network unit information; performing backward propagation based on the loss function to update network parameters of the neural network, and returning to the step of inputting the training data and the related optical network unit information into the neural network for forward propagation and subsequent steps until the neural network converges, terminating the training, and obtaining the identification model.

[0126] The identification model includes a single decision model, and the step of identifying the abnormal signal data based on the single decision model to obtain rogue optical network unit information corresponding to the abnormal signal data includes: determining the single decision model according to optical network unit information to which the uplink bandwidth belongs, identifying the abnormal signal data, and outputting at least one number of a rogue optical network unit corresponding to the abnormal signal data; and / or identifying the abnormal signal data through the single decision model to output information of two optical network units generating the abnormal signal data, determining the optical network unit information affected in the information of the two optical network units according to the optical network unit information to which the uplink bandwidth belongs, and determining another optical network unit information as the rogue optical network unit information.

[0127] Illustratively, the identification model comprises at least one classification decision model, and based on the at least one classification decision model, the step of identifying the rogue optical network unit corresponding to the abnormal signal data comprises: identifying a data type corresponding to the abnormal signal data; calling a corresponding at least one classification decision model according to the data type, inputting the abnormal signal data into the at least one classification decision model, and obtaining a number of at least one rogue optical network unit.

[0128] The identification device provided by the embodiments of the present application has similar implementation principles and beneficial effects to the technical solutions shown in the corresponding method embodiments described above, and thus will not be described in detail here.

[0129] Sixth embodiment

[0130] Referring to FIG. 9, FIG. 9 is a structural schematic diagram of an identification device provided by an embodiment of the present application. The device can be mounted on or is the terminal device in the method embodiments described above. The identification device shown in FIG. 9 can be used to perform part or all of the functions in the method embodiments described in the above embodiments. As shown in FIG. 9, the identification device comprises: a second acquisition module configured to acquire training data and related optical network unit information from an optical line terminal; and a training module configured to train based on the training data and the related optical network unit information to obtain an identification model, wherein the identification model is configured to identify rogue optical network unit information according to abnormal signal data.

[0131] Illustratively, the step of acquiring the training data comprises: receiving synthetic conflict data and / or actual conflict data sent by the optical line terminal, and / or receiving optical network unit sample signals sent by the optical line terminal, and obtaining the synthetic conflict data according to the optical network unit sample signals; taking the synthetic conflict data and / or the actual conflict data as the training data, and / or sending the synthetic conflict data to the optical line terminal so that the optical line terminal takes the synthetic conflict data as the training data.

[0132] Illustratively, the step of training based on the training data and the related optical network unit information to obtain the identification model comprises: inputting the training data and the related optical network unit information into the neural network for forward propagation to obtain a prediction result; calculating a loss function according to the prediction result and the related optical network unit information; performing back propagation according to the loss function to update network parameters of the neural network, and returning to perform the step of inputting the training data and the related optical network unit information into the neural network for forward propagation and subsequent steps until the neural network converges, the training is terminated, and the identification model is obtained.

[0133] An exemplary embodiment of the method comprises the following steps: receiving the training data and the information of the optical network units; training the identification model based on the training data and the information of the optical network units; and sending the identification model to the optical line terminal, so that the optical line terminal identifies the abnormal signal data based on the identification model to obtain the rogue optical network unit corresponding to the abnormal signal data.

[0134] An exemplary embodiment of the method comprises the following steps: receiving the training data and the information of the optical network units; training the identification model based on the training data and the information of the optical network units; and sending the identification model to the optical line terminal, so that the optical line terminal identifies the abnormal signal data based on the identification model to obtain the rogue optical network unit corresponding to the abnormal signal data.

[0135] An exemplary embodiment of the method comprises the following steps: receiving the training data and the information of the optical network units; training the identification model based on the training data and the information of the optical network units; and sending the identification model to the optical line terminal, so that the optical line terminal identifies the abnormal signal data based on the identification model to obtain the rogue optical network unit corresponding to the abnormal signal data.

[0136] An exemplary embodiment of the method comprises the following steps: receiving the training data and the information of the optical network units; training the identification model based on the training data and the information of the optical network units; and sending the identification model to the optical line terminal, so that the optical line terminal identifies the abnormal signal data based on the identification model to obtain the rogue optical network unit corresponding to the abnormal signal data.

[0137] An exemplary embodiment of the method comprises the following steps: receiving the training data and the information of the optical network units; training the identification model based on the training data and the information of the optical network units; and sending the identification model to the optical line terminal, so that the optical line terminal identifies the abnormal signal data based on the identification model to obtain the rogue optical network unit corresponding to the abnormal signal data.

[0138] In addition, the present application further provides a network device, which comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, and the computer program is configured to implement the steps of the above-mentioned identification method.

[0139] In addition, the present application further provides a storage medium, which is a computer readable storage medium, and the storage medium stores a computer program, and the computer program is executed by a processor to implement the steps of the above-mentioned identification method.

[0140] It should be noted that, in the present document, the terms "comprises", "comprising", or any other variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can also include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element preceded by "comprises a" does not, without more constraints, foreclose the existence of additional identical elements in the process, method, article, or apparatus that comprises the element.

[0141] Those skilled in the art can clearly understand, through the description of the above embodiments, that the above-mentioned embodiment methods can be realized by means of software and the necessary general hardware platform, of course, they can also be realized by hardware, but in many cases the former is a better embodiment. Based on such understanding, the technical solutions of the present application can be embodied in the form of a software product, which is stored in a storage medium (such as a ROM / RAM, a magnetic disk, or an optical disk) as described above, and includes a number of instructions for causing an end device (which can be a mobile phone, a computer, a server, or a network device, etc.) to execute the methods described in the various embodiments of the present application.

[0142] The above is only an optional embodiment of the present application, and does not limit the patent scope of the present application, and any equivalent structure or equivalent process transformation made by using the content of the present application specification and drawings, or directly or indirectly applied to other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. An identification method applied to an optical line terminal, wherein, The method includes: Acquire abnormal signal data; The abnormal signal data is identified to obtain rogue optical network unit information.

2. The identification method as described in claim 1, wherein, The step of acquiring abnormal signal data is preceded by: The uplink signal is continuously acquired and analyzed. If a sudden change in the uplink signal is detected in the acquired data, the acquired data is identified as the abnormal signal data; and / or, Error correction processing is performed on the uplink signal. If a bit error is detected, the information block corresponding to the bit error is identified as the abnormal signal data.

3. The identification method as described in claim 2, wherein, Prior to the step of continuously acquiring and analyzing and / or correcting the uplink signal, the method further includes: Uplink bandwidth is allocated to the target optical network unit to form a bandwidth allocation result; The bandwidth allocation result is sent to the target optical network unit so that the target optical network unit sends the uplink signal according to the bandwidth allocation result.

4. The identification method as described in claim 3, wherein, The steps for acquiring abnormal signal data include: Based on the bandwidth allocation result, obtain the optical network unit information to which the uplink bandwidth corresponding to the abnormal signal data belongs.

5. The identification method as described in claim 1, wherein, The step of identifying the abnormal signal data to obtain information about the rogue optical network unit includes: The abnormal signal data and the corresponding uplink bandwidth optical network unit information are identified by a pre-trained identification model to obtain the rogue optical network unit information.

6. The identification method as described in claim 5, wherein, Before the step of identifying the abnormal signal data using a pre-trained identification model, the method further includes: Acquire training data and related optical network unit information; The recognition model is obtained by training based on the training data and related optical network unit information; and / or, the training data and related optical network unit information are sent to the computing center so that the computing center can train based on the training data and related optical network unit information to obtain the recognition model, and obtain the recognition model from the computing center.

7. The identification method as described in claim 6, wherein, The steps for obtaining training data include: Acquire optical network unit sample signals, and obtain synthetic collision data based on the optical network unit sample signals, and / or send the optical network unit sample signals to the computing center, so that the computing center obtains synthetic collision data based on the optical network unit sample signals, and obtains the synthetic collision data from the computing center, and / or collects actual collision data; The synthetic conflict data and / or the actual conflict data are used as the training data.

8. The identification method as described in claim 7, wherein, The steps of acquiring sample signals from optical network units and obtaining synthetic collision data based on the sample signals from optical network units include: The sampling start point is calculated based on the local sampling clock and the transmission clock of the optical network unit; The optical network unit sample signal of the optical network unit is acquired according to the sampling start point; The composite conflict data is obtained by superimposing the optical network unit sample signals of the two optical network units.

9. The identification method as described in claim 7, wherein, The steps for collecting actual conflict data include at least one of the following: Uplink bandwidth with an ascending start time and overlapping characteristics is allocated to two optical network units, and signal data corresponding to the overlapping portion of the uplink bandwidth is collected as the actual collision data. In the optical network unit, the starting time ascending order of the uplink bandwidth mapping entries is not detected. The two optical network units are allocated completely overlapping uplink bandwidths, and the signal data corresponding to the overlapping uplink bandwidths are collected as the actual conflict data. Two optical network units are assigned the same transmission container, uplink bandwidth is allocated to the transmission container, and the uplink signals sent by the two optical network units in the uplink bandwidth are collected as the actual collision data.

10. The identification method as described in claim 6, wherein, The step of training the recognition model based on the training data and related optical network unit information includes: Training data is selected based on one optical network unit (ONU) information from the relevant ONU information, and the recognition model is trained based on the selected training data and the other ONU information; and / or, The recognition model is obtained by training based on all the training data and related optical network unit information.

11. The identification method as described in claim 10, wherein, The step of training the recognition model based on the training data and related optical network unit information includes: The training data and related optical network unit information are input into the neural network for forward propagation to obtain the prediction result; The loss function is calculated based on the prediction results and the relevant optical network unit information. Backpropagation is performed based on the loss function to update the network parameters of the neural network. Then, the process of inputting the training data and related optical network unit information into the neural network for forward propagation and subsequent steps is repeated until the neural network converges, at which point training is terminated, and the recognition model is obtained.

12. The identification method as described in claim 5, wherein, The identification model includes a single decision model. The steps of identifying the abnormal signal data based on the single decision model to obtain the rogue optical network unit information corresponding to the abnormal signal data include: The single decision model is determined based on the optical network unit information of the uplink bandwidth, and the abnormal signal data is identified, outputting the number of at least one rogue optical network unit corresponding to the abnormal signal data; and / or, The abnormal signal data is identified by the single decision model, and the information of the two optical network units that generated the abnormal signal data is output. The optical network unit information affected by the two optical network unit information is determined according to the optical network unit information to which the uplink bandwidth belongs, and the other optical network unit information is determined to be the rogue optical network unit information.

13. The identification method as described in claim 5, wherein, The identification model includes at least one classification decision model. The step of identifying the abnormal signal data based on the at least one classification decision model to obtain the rogue optical network unit information corresponding to the abnormal signal data includes: Identify the data type corresponding to the abnormal signal data; Based on the data type, call at least one corresponding classification decision model, input the abnormal signal data into the at least one decision model, and obtain the number of at least one rogue optical network unit.

14. An identification method applied to a computing center, wherein, The method includes: Acquire training data and related optical network unit information from the optical line terminal; A recognition model is obtained by training based on the training data and related optical network unit information; The identification model is used to identify rogue optical network unit information based on abnormal signal data.

15. The identification method as described in claim 14, wherein, The steps for obtaining training data include: Receive synthetic collision data and / or actual collision data sent by the optical line terminal, and / or receive optical network unit sample signals sent by the optical line terminal, and obtain the synthetic collision data based on the optical network unit sample signals; The synthetic conflict data and / or the actual conflict data are used as the training data, and / or the synthetic conflict data is sent to the optical line terminal so that the optical line terminal uses the synthetic conflict data as the training data.

16. The identification method as described in claim 14, wherein, The step of training the recognition model based on the training data and related optical network unit information includes: The training data and related optical network unit information are input into the neural network for forward propagation to obtain the prediction result; The loss function is calculated based on the prediction results and the relevant optical network unit information. Backpropagation is performed based on the loss function to update the network parameters of the neural network. Then, the process of inputting the training data and related optical network unit information into the neural network for forward propagation and subsequent steps is repeated until the neural network converges, training is terminated, and the recognition model is obtained.

17. The identification method as described in claim 15, wherein, After the step of training the neural network based on the training data and related optical network unit information to obtain the recognition model, the method further includes: The identification model is sent to the optical line terminal so that the optical line terminal can identify the acquired abnormal signal data based on the identification model and obtain information about rogue optical network units.

18. The identification method as described in claim 15, wherein, After the step of training the recognition model based on the training data and related optical network unit information to obtain the recognition model, the method further includes: Receive abnormal signal data sent by the optical line terminal; Based on the identification model, the abnormal signal data and the corresponding uplink optical network unit information are identified to obtain rogue optical network unit information.

19. The identification method as described in claim 18, wherein, The identification model includes a single decision model and / or at least one classification decision model.

20. The identification method as described in claim 19, wherein, The step of identifying the abnormal signal data based on the identification model to obtain the rogue optical network unit information corresponding to the abnormal signal data includes at least one of the following: The single decision model is determined based on the optical network unit information of the uplink bandwidth, and the abnormal signal data is identified, and the number of at least one rogue optical network unit corresponding to the abnormal signal data is output. And / or, The abnormal signal data is identified by the single decision model, and the information of the two optical network units that generated the abnormal signal data is output. The optical network unit information affected by the two optical network unit information is determined according to the optical network unit information to which the uplink bandwidth belongs, and the other optical network unit information is determined to be the rogue optical network unit information. Identify the data type corresponding to the abnormal signal data, call at least one corresponding classification decision model according to the data type, input the abnormal signal data into the at least one decision model, and obtain the number of at least one rogue optical network unit.

21. A network device, wherein, The network device includes: a memory, a processor, and a computer program stored in the memory and executable on the processor, the computer program being configured to implement the steps of the identification method as described in any one of claims 1 to 13 or the identification method as described in any one of claims 14 to 20.

22. A storage medium, wherein, The storage medium is a computer-readable storage medium, and a computer program is stored on the storage medium. When the computer program is executed by a processor, it implements the steps of the identification method as described in any one of claims 1 to 13 or the identification method as described in any one of claims 14 to 20.

Images