Optical cable information transmission monitoring system based on intelligent control
Through the intelligent control system, the use of deep learning and reinforcement learning technology to automatically extract optical cable characteristics and optimize control strategies, the problem of insufficient adaptability of traditional optical cable monitoring is solved, the network reliability and transmission efficiency are improved, and the operation cost and energy consumption are reduced.
Patent Information
- Application Number
- CN202411355386.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-27
- Publication Date
- 2025-01-03
- Estimated Expiration
- 2044-09-27
AI Technical Summary
Traditional optical cable monitoring and control methods cannot effectively adapt to complex and changeable dynamic environments, resulting in distortion and attenuation of optical signals and degradation of transmission quality. The existing machine learning-based methods have problems such as high cost of manual labeling data, poor real-time and insufficient adaptability.
A fiber cable information transmission monitoring system based on intelligent control is proposed, including sample collection module, abnormal model training module, real-time data collection module, reinforcement model building module and intelligent control module. Through deep learning and reinforcement learning technology, features and optimization control strategies are automatically extracted. Generate real-time action decisions.
It improves the reliability and transmission efficiency of the optical cable network, reduces operating costs and energy consumption, and provides guarantees for the efficient and stable operation of the network.
Smart Images
Figure CN118890089B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to the technical field of intelligent control, in particular to an optical cable information transmission monitoring system based on intelligent control. Background Art
[0002] With the rapid development of communication networks, large-capacity optical cable lines have assumed increasingly important information transmission tasks. However, in the process of long-distance, high-speed optical transmission, optical signals are easily distorted and attenuated by various nonlinear effects and environmental interference, resulting in increased bit error rates and reduced transmission quality. Traditional optical cable monitoring and control methods mostly use artificial experience rules, which cannot fully adapt to complex and changing dynamic environments, and urgently need to introduce intelligent technology.
[0003] At present, although some progress has been made in the control methods based on machine learning, there are obvious limitations. Traditional supervised learning methods require a large amount of manually annotated historical data for model training. The cost of obtaining annotated data is high and the real-time performance is poor. Other machine learning models, such as decision trees, are often based on static feature spaces and cannot learn better control strategies in dynamic environments. They lack adaptability and robustness.
[0004] A Chinese patent with authorization announcement number CN111369769B discloses an optical cable monitoring and early warning system based on data analysis, including an optical cable sensing module, a data acquisition module, a data analysis module, a controller, a comprehensive judgment module and a feedback storage module; the optical cable sensing module is used to collect peripheral coverage information of each optical cable segment of the optical cable line and transmit it to the data acquisition module, and the peripheral coverage information of the optical cable segment is composed of infrared ranging data of the optical cable segment and light intensity data of the optical cable segment. However, this method fails to solve the problem of intelligent control of optical cable signal transmission power.
[0005] To this end, the present invention proposes an optical cable information transmission monitoring system based on intelligent control. Summary of the invention
[0006] The present invention aims to solve at least one of the technical problems existing in the prior art. To this end, the present invention proposes an optical cable information transmission monitoring system based on intelligent control, which improves the reliability and transmission efficiency of the network, reduces operating costs and energy consumption, and provides a strong guarantee for the efficient and stable operation of the network.
[0007] To achieve the above purpose, an optical cable information transmission monitoring system based on intelligent control is proposed, which includes a sample collection module, an abnormal model training module, a real-time data collection module, an enhanced model construction module and an intelligent control module; wherein each module is electrically connected;
[0008] The sample collection module collects historical monitoring data and historical annotation data of each collection cycle in advance in the test environment, converts the historical monitoring data into an input feature set, and sends the input feature set and historical annotation data to the anomaly model training module;
[0009] The method of pre-collecting the historical monitoring data and historical annotation data of each collection cycle is:
[0010] Collect historical monitoring data consisting of parameter time series of various optical cable monitoring parameters regularly collected by sensor monitoring equipment, and historical annotation data consisting of abnormal label value series;
[0011] The parameter time series is a parameter time series composed of the parameter values of each optical cable monitoring parameter at each moment in each acquisition cycle in chronological order;
[0012] The abnormal label value sequence is a time sequence of labels composed of abnormal label values of the optical cable at each moment in each collection cycle in time order; the abnormal label value is 0 or 1;
[0013] The method of converting historical monitoring data into an input feature set is:
[0014] In each acquisition cycle, the parameter time series of each optical cable monitoring parameter in the acquisition cycle is converted into a set of monitoring feature vectors, and all monitoring feature vectors in the acquisition cycle constitute a set of input feature sets;
[0015] The anomaly model training module takes the input feature set of the current collection cycle as input, the historical annotation data of the next collection cycle as the prediction target, trains the anomaly assessment model, and sends the anomaly assessment model to the real-time data collection module;
[0016] The method of training the anomaly assessment model is:
[0017] The input feature set of each acquisition cycle is used as the input of the anomaly assessment model. The anomaly assessment model uses the predicted value of the anomaly probability sequence in the next acquisition cycle as the output. The anomaly probability sequence is the probability of the optical cable anomaly at each moment in the next acquisition cycle. The anomaly label value sequence corresponding to the next acquisition cycle is used as the prediction target. The Euclidean distance between the anomaly probability sequence and the anomaly label value sequence is used as the prediction error. Minimizing the sum of the prediction errors is used as the training target. The anomaly assessment model is trained until the sum of the prediction errors reaches convergence and the training is stopped. The anomaly assessment model is a time series prediction model.
[0018] The real-time data collection module collects the real-time monitoring data of the optical cable to be monitored; inputs the real-time monitoring data into the abnormality assessment model to obtain the abnormal probability sequence, and converts the abnormal probability sequence into an abnormal risk scalar through the sequence conversion model; and sends the real-time monitoring data and the abnormal risk scalar to the intelligent control module;
[0019] The method of collecting the real-time monitoring data of the optical cable to be monitored is:
[0020] For the optical cable to be monitored, at every collection cycle, the parameter time series of each optical cable monitoring parameter within each collection cycle is collected as real-time monitoring data in a manner consistent with the historical monitoring data collected in the test environment;
[0021] The converting of the abnormal probability sequence into an abnormal risk scalar by the sequence conversion model comprises the following steps:
[0022] Step 11: Mark each abnormal probability sequence as P, and use moving average or Gaussian kernel filtering method to smooth P to suppress high-frequency disturbances and obtain a smoothed abnormal probability sequence P';
[0023] Step 12: Count the maximum value pmax and the minimum value pmin of P', and linearly scale P' by the maximum value pmax and the minimum value pmin to obtain the standard sequence P'';
[0024] Step 13: construct an LSTM encoder for the sequence conversion model;
[0025] The input layer of the LSTM encoder receives the standard sequence P'';
[0026] The LSTM layer of the LSTM encoder returns the hidden state h(t) at each time step, where t is the time step of the sequence conversion model;
[0027] The attention layer of the LSTM encoder: performs self-attention calculation on the hidden state h(t), outputs the attention weight sequence α(t), and calculates the context vector c;
[0028] Step 14: Encode the anomaly risk scalar:
[0029] The attention layer is connected to the fully connected layer, and the fully connected layer is connected to the output layer;
[0030] Input the context vector c into the fully connected layer to obtain the abnormal risk scalar output by the output layer;
[0031] Step 15: construct a risk value sample for the sequence conversion model, the input of the risk value sample is M% abnormal label value sequences in the historical annotation data, the output of the risk value sample is a risk scalar set manually according to the risk degree of the abnormal probability sequence, and the sequence conversion model is trained through the risk value sample; M is the proportion of the preset selected abnormal label value sequence;
[0032] Strengthen the model building module, build the state space, action space and reward function of the power decision model, so as to complete the design of the intelligent agent of the power decision model, and send the designed power decision model to the intelligent control module;
[0033] The state space, action space and reward function of the power decision model are constructed as follows:
[0034] The power decision model is set to an Actor-Critic model;
[0035] The state space is marked as S, and the state space S reflects the real-time operating state of the entire optical cable. The state space includes real-time monitoring data and abnormal risk scalars in each monitoring cycle; the state of each acquisition cycle in the state space S is marked as s;
[0036] The action space is marked as A, wherein the action space A includes the adjustment amount of the optical signal output power that can be provided by the transmitting end of the optical cable output by the Actor model, and the optical signal output power taken by the optical cable in the next acquisition cycle is marked as a;
[0037] The reward function is labeled as Q, and the reward function Q=r(s, a) measures the effect of performing action a in the current state s;
[0038] The agent is designed as follows:
[0039] Design the intelligent agent as an Actor-Critic structure based on RNN;
[0040] The input of the agent is state s;
[0041] The Critic network outputs the state-behavior value V(s);
[0042] The Actor network outputs the action probability π(a|s);
[0043] The intelligent control module generates the action decision for the next acquisition cycle for the monitored optical cable based on real-time monitoring data, abnormal risk scalar and power decision model;
[0044] The method of generating the action decision of the next acquisition cycle for the optical cable to be monitored is:
[0045] The real-time monitoring data and the abnormal risk scalar are used as the current state s; the current state s is input to the Actor model of the power decision model, and the Actor model outputs an action in the action space A so as to maximize the state-behavior value output by the Critic model and output an action decision on the adjustment amount of the output power of the optical signal;
[0046] Execute action decision to obtain the next state s' and actual reward value;
[0047] The critic network outputs the state-behavior value after executing the action decision in the current state, calculates the TD error δ of the critic network, updates the critic network parameters according to δ, and minimizes the critic loss function;
[0048] The Actor model takes the state-behavior value output by the Critic model as the output target and uses the policy gradient update method to update the parameters of the Actor network model.
[0049] A method for monitoring optical cable information transmission based on intelligent control is proposed, comprising the following steps:
[0050] Step 1: Collect historical monitoring data and historical annotation data of each collection cycle in advance in the test environment, and convert the historical monitoring data into an input feature set;
[0051] Step 2: Take the input feature set of the current collection cycle as input, and the historical annotated data of the next collection cycle as the prediction target to train the anomaly assessment model;
[0052] Step 3: Collect real-time monitoring data of the optical cable to be monitored; input the real-time monitoring data into the anomaly assessment model to obtain an anomaly probability sequence, and convert the anomaly probability sequence into an anomaly risk scalar through a sequence conversion model;
[0053] Step 4: Construct the state space, action space and reward function of the power decision model to complete the design of the intelligent agent of the power decision model;
[0054] Step 5: Based on the real-time monitoring data, abnormal risk scalar and power decision model, generate an action decision for the next acquisition cycle for the optical cable to be monitored.
[0055] An electronic device is proposed, comprising: a processor and a memory, wherein the memory stores a computer program that can be called by the processor;
[0056] The processor executes the above-mentioned optical cable information transmission monitoring method based on intelligent control by calling the computer program stored in the memory.
[0057] A computer-readable storage medium is provided, on which a rewritable computer program is stored;
[0058] When the computer program is executed on a computer device, the computer device is enabled to execute the above-mentioned optical cable information transmission monitoring method based on intelligent control.
[0059] Compared with the prior art, the present invention has the following beneficial effects:
[0060] This application first collects the original real-time monitoring data of the optical cable line and constructs an anomaly detection data set. Then a deep learning model is used to automatically extract features from the original data and encode a score sequence that reflects the risk of line anomalies without relying on manual labeling. The anomaly score is then input into the reinforcement learning agent together with other state features. The agent obtains the control strategy of how to optimize the control of the transmission power under various working conditions through autonomous learning to achieve the goal of balancing the quality of optical signals and transmission efficiency. By integrating deep learning and reinforcement learning technology, the deep learning model automatically extracts high-quality features from the original data without relying on manual labeling; and the reinforcement agent continuously optimizes the control strategy through trial and error interaction, and has excellent decision-making ability and environmental adaptability, thereby improving the reliability and transmission efficiency of the network, reducing operating costs and energy consumption, and providing strong guarantees for the efficient and stable operation of the network. BRIEF DESCRIPTION OF THE DRAWINGS
[0061] Figure 1 This is a module connection relationship diagram of the optical cable information transmission monitoring system based on intelligent control in Example 1 of the present invention;
[0062] Figure 2 This is a flow chart of an optical cable information transmission monitoring method based on intelligent control in Embodiment 2 of the present invention;
[0063] Figure 3 This is a schematic diagram of the structure of an electronic device in Embodiment 3 of the present invention;
[0064] Figure 4 This is a schematic diagram of the computer-readable storage medium structure in Example 4 of the present invention. DETAILED DESCRIPTION
[0065] The technical solution of the present invention will be clearly and completely described below in conjunction with the embodiments. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0066] Example 1
[0067] like Figure 1As shown, the optical cable information transmission monitoring system based on intelligent control includes a sample collection module, an abnormal model training module, a real-time data collection module, an enhanced model building module and an intelligent control module; wherein each module is electrically connected;
[0068] The sample collection module collects historical monitoring data and historical annotation data of each collection cycle in advance in the test environment, converts the historical monitoring data into an input feature set, and sends the input feature set and historical annotation data to the anomaly model training module;
[0069] Specifically, the test environment refers to:
[0070] The testers select multiple optical cable lines in different environments in advance, and deploy optical power, loss, BER, temperature and humidity and other sensor monitoring equipment in the lines. At preset collection cycle times, the real-time operation data of the lines are regularly collected, including normal and abnormal conditions. The abnormal conditions include but are not limited to power loss, noise interference, etc. It is understandable that power loss will cause a significant decrease in optical power value, while noise interference will cause a significant increase in bit error rate value and greater power fluctuation.
[0071] Specifically, the method of pre-collecting the historical monitoring data and historical annotation data of each collection period is:
[0072] Collect historical monitoring data consisting of parameter time series of various optical cable monitoring parameters regularly collected by sensor monitoring equipment, and historical annotation data consisting of abnormal label value series;
[0073] Specifically, the optical cable monitoring parameters are transmission parameters and environmental parameters that can affect or show whether the optical cable transmission is normal; the transmission parameters include but are not limited to optical power, optical loss, bit error rate, radio frequency spectrum, signal transmission power, modulation parameters, etc.; the environmental parameters include but are not limited to environmental parameters such as temperature, humidity, and pressure; among them, the optical power is collected by optical power monitoring equipment, the optical loss is obtained by calculating the difference between the transmission power and the reception power, the bit error rate is obtained by deploying a bit error rate test device at the receiving end, the radio frequency spectrum is obtained by connecting the optoelectronic conversion and radio frequency signal analysis equipment at the receiving end, and the modulation parameters are parameters pre-set at the transmitting end;
[0074] The parameter time series is a parameter time series composed of the parameter values of each optical cable monitoring parameter at each moment in each acquisition cycle in chronological order;
[0075] The abnormal label value sequence is a time sequence of labels composed of abnormal label values of the optical cable at each moment in each collection cycle in time order; the abnormal label value is 0 or 1. When the abnormal label value is 1, it indicates that there is an abnormality in the information transmission of the optical cable line (power loss, noise interference, etc.); when the abnormal label value is 0, it indicates that there is no abnormality in the information transmission of the optical cable line;
[0076] Furthermore, the method of converting the historical monitoring data into an input feature set is:
[0077] In each acquisition cycle, the parameter time series of each optical cable monitoring parameter in the acquisition cycle is converted into a set of monitoring feature vectors, and all monitoring feature vectors in the acquisition cycle constitute a set of input feature sets;
[0078] It is understandable that when controlling the transmission power of a long intercity optical cable line, it is not enough to rely solely on the original monitoring data such as optical power, loss, BER, etc. as model input, because there is a complex nonlinear mapping relationship between these data, and it is difficult to directly reflect the abnormal risk level of the line. However, if a deep learning model is used to process these raw data first and output a scalar real-time abnormality score, the health status of the line can be well summarized; therefore, it is necessary to evaluate the abnormal probability of the optical cable;
[0079] The anomaly model training module takes the input feature set of the current collection cycle as input, the historical annotation data of the next collection cycle as the prediction target, trains the anomaly assessment model, and sends the anomaly assessment model to the real-time data collection module;
[0080] Specifically, the method of training the anomaly assessment model is:
[0081] The input feature set of each acquisition cycle is used as the input of the anomaly assessment model, and the anomaly assessment model takes the predicted value of the anomaly probability sequence in the next acquisition cycle as the output, the anomaly probability sequence is the probability of the optical cable anomaly at each moment in the next acquisition cycle, the anomaly label value sequence corresponding to the next acquisition cycle is used as the prediction target, the Euclidean distance between the anomaly probability sequence and the anomaly label value sequence is used as the prediction error, and minimizing the sum of the prediction errors is used as the training target; the anomaly assessment model is trained until the sum of the prediction errors reaches convergence and the training is stopped; the anomaly assessment model is a time series prediction model; the time series prediction model is any one of an RNN model or an LSTM model;
[0082] The real-time data collection module collects the real-time monitoring data of the optical cable to be monitored; inputs the real-time monitoring data into the abnormality assessment model to obtain the abnormal probability sequence, and converts the abnormal probability sequence into an abnormal risk scalar through the sequence conversion model; and sends the real-time monitoring data and the abnormal risk scalar to the intelligent control module;
[0083] Specifically, the method of collecting the real-time monitoring data of the optical cable to be monitored is:
[0084] For the optical cable to be monitored, at every collection cycle, the parameter time series of each optical cable monitoring parameter within each collection cycle is collected as real-time monitoring data in a manner consistent with the historical monitoring data collected in the test environment;
[0085] It is understandable that inputting the real-time monitoring data into the anomaly assessment model to obtain the anomaly probability sequence is a conventional method in the art, and this application will not elaborate on it here;
[0086] The converting of the abnormal probability sequence into an abnormal risk scalar by the sequence conversion model comprises the following steps:
[0087] Step 11: Mark each abnormal probability sequence as P, and use moving average or Gaussian kernel filtering method to smooth P to suppress high-frequency disturbances and obtain a smoothed abnormal probability sequence P';
[0088] Step 12: Count the maximum value pmax and the minimum value pmin of P', and linearly scale P' by the maximum value pmax and the minimum value pmin to obtain the standard sequence P'', that is, P'' = (P' - pmin) / (pmax - pmin), to ensure that all values of P'' are in the interval [0, 1];
[0089] Step 13: construct an LSTM encoder for the sequence conversion model;
[0090] The input layer of the LSTM encoder receives the standard sequence P'';
[0091] The LSTM layer of the LSTM encoder returns the hidden state h(t) at each time step, where t is the time step of the sequence conversion model;
[0092] The attention layer of the LSTM encoder: performs self-attention calculation on the hidden state h(t), outputs the attention weight sequence α(t), and calculates the context vector c. The context vector ;
[0093] Step 14: Encode the anomaly risk scalar:
[0094] The attention layer is connected to the fully connected layer, and the fully connected layer is connected to the output layer;
[0095] Input the context vector c into the fully connected layer to obtain the abnormal risk scalar output by the output layer;
[0096] Step 15: construct a risk value sample for the sequence conversion model, the input of the risk value sample is M% abnormal label value sequences in the historical annotation data, the output of the risk value sample is a risk scalar set manually according to the risk degree of the abnormal probability sequence, and the sequence conversion model is trained through the risk value sample; M is the proportion of the preset selected abnormal label value sequence; preferably, M can be set to 70;
[0097] Preferably, the setting of the risk scalar may refer to the ratio of the duration of the optical cable communication anomaly to the duration of the collection period in each collection period. It can be understood that the greater the probability of anomaly at each moment, the closer the duration of the communication anomaly is to the actual abnormal situation.
[0098] Strengthen the model building module, build the state space, action space and reward function of the power decision model, so as to complete the design of the intelligent agent of the power decision model, and send the designed power decision model to the intelligent control module;
[0099] The state space, action space and reward function of the power decision model are constructed as follows:
[0100] The power decision model is set to an Actor-Critic model;
[0101] The state space is marked as S, and the state space S reflects the real-time operating state of the entire optical cable. The state space includes real-time monitoring data and abnormal risk scalars in each monitoring cycle; the state of each acquisition cycle in the state space S is marked as s;
[0102] The action space is marked as A, and the action space A includes the adjustment amount of the optical signal output power that can be provided by the transmitting end of the optical cable output by the Actor model, and the adjustment amount of the optical signal output power taken by the optical cable in the next acquisition cycle is marked as a; it can be understood that the adjustment amount is for the previous acquisition cycle; in addition, when an abnormality is detected in the optical cable, it means that the entire line is on the edge of an abnormality. The intelligent network should decisively increase the transmission power based on this risk signal to improve the line quality.
[0103] The reward function is labeled as Q, and the reward function Q=r(s, a) measures the effect of performing action a in the current state s;
[0104] Preferably, the reward function Q can be specifically set as:
[0105] ;
[0106] Among them, Pt represents the optical power received by the receiving end, and Pm represents the preset target optical power; It is used to measure the effect of attenuating the optical power at the receiving end after executing action a. The better the execution effect, the The larger the value, the larger the reward function Q; BERt is the actual attenuation rate of information in the optical cable. The smaller the BERt is, the better the execution effect of reducing the bit error rate of information in the optical cable is, and the larger the reward function Q is. a is the adjustment amount of the optical signal output power. Obviously, the smaller the adjustment amount is, the smaller the disturbance to the optical cable is, and the better the execution effect is, and the larger the reward function Q is; a1, a2 and a3 are all preset proportional coefficients;
[0107] It can be understood that the setting of the above reward function Q is only a simple example provided in this embodiment. In practical applications, the reward function Q can be adaptively modified, deleted or added;
[0108] The agent is designed as follows:
[0109] Design the intelligent agent as an Actor-Critic structure based on RNN;
[0110] The input of the agent is state s;
[0111] The Critic network outputs the state-behavior value V(s);
[0112] The Actor network outputs the action probability π(a|s);
[0113] The intelligent control module generates the action decision for the next acquisition cycle for the monitored optical cable based on real-time monitoring data, abnormal risk scalar and power decision model;
[0114] Specifically, the method of generating an action decision for the next acquisition cycle for the optical cable to be monitored based on the real-time monitoring data, the abnormal risk scalar and the power decision model is:
[0115] The real-time monitoring data and the abnormal risk scalar are used as the current state s; the current state s is input to the Actor model of the power decision model, and the Actor model outputs an action in the action space A so as to maximize the state-behavior value output by the Critic model and output an action decision on the adjustment amount of the output power of the optical signal;
[0116] Execute action decision to obtain the next state s' and actual reward value;
[0117] The critic network outputs the state-behavior value after executing the action decision in the current state, calculates the TD error δ of the critic network, updates the critic network parameters according to δ, and minimizes the critic loss function; the state-behavior value output by the critic model is calculated by the state-behavior value function generated by the temporal difference optimization method;
[0118] Specifically, the state-behavior value function generated by the temporal difference optimization method can be:
[0119] , γ is the discount factor, V(s') is the estimated value in state s'; generally, the estimated value can be replaced by Q(s', a'), that is, V(s') represents the estimated value of the state-behavior value function corresponding to the next state s' and action a', so as to ensure that Q(s, a) can consider the long term instead of only the current state;
[0120] Then the TD error δ is ;
[0121] The Actor model takes the state-behavior value output by the Critic model as the output target and uses the policy gradient update method to update the parameters of the Actor network model.
[0122] Example 2
[0123] like Figure 2 As shown, the optical cable information transmission monitoring method based on intelligent control includes the following steps:
[0124] Step 1: Collect historical monitoring data and historical annotation data of each collection cycle in advance in the test environment, and convert the historical monitoring data into an input feature set;
[0125] Step 2: Take the input feature set of the current collection cycle as input, and the historical annotated data of the next collection cycle as the prediction target to train the anomaly assessment model;
[0126] Step 3: Collect real-time monitoring data of the optical cable to be monitored; input the real-time monitoring data into the anomaly assessment model to obtain an anomaly probability sequence, and convert the anomaly probability sequence into an anomaly risk scalar through a sequence conversion model;
[0127] Step 4: Construct the state space, action space and reward function of the power decision model to complete the design of the intelligent agent of the power decision model;
[0128] Step 5: Based on the real-time monitoring data, abnormal risk scalar and power decision model, generate an action decision for the next acquisition cycle for the optical cable to be monitored.
[0129] Example 3
[0130] Figure 3 1 is a schematic diagram of the structure of an electronic device provided by an embodiment of the present application. Figure 3As shown, according to another aspect of the present application, an electronic device 100 is also provided. The electronic device 100 may include one or more processors and one or more memories. The memories store computer readable codes, and when the computer readable codes are run by one or more processors, the optical cable information transmission monitoring method based on intelligent control as described above may be executed.
[0131] The method or device according to the embodiment of the present application can also be used by Figure 3 The electronic device architecture shown in FIG. Figure 3 As shown, the electronic device 100 may include a bus 101, one or more CPUs 102, a ROM 103, a RAM 104, a communication port 105 connected to a network, an input / output component 106, a hard disk 107, etc. A storage device in the electronic device 100, such as the ROM 103 or the hard disk 107, may store the optical cable information transmission monitoring method based on intelligent control provided in the present application.
[0132] Furthermore, the electronic device 100 may further include a user interface 108. Figure 3 The architecture shown is only exemplary and can be omitted according to actual needs when implementing different devices. Figure 3 One or more components of an electronic device are shown.
[0133] Example 4
[0134] Figure 4 Schematic diagram of the computer-readable storage medium structure provided by an embodiment of the present application. Figure 4 As shown, it is a computer-readable storage medium 200 according to an embodiment of the present application. Computer-readable instructions are stored on the computer-readable storage medium 200. When the computer-readable instructions are executed by the processor, the optical cable information transmission monitoring method based on intelligent control according to the embodiment of the present application described with reference to the above figures can be executed. The computer-readable storage medium 200 includes, but is not limited to, for example, volatile memory and / or non-volatile memory. Volatile memory may include, for example, random access memory (RAM) and cache memory (cache), etc. Non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc.
[0135] In addition, according to the implementation of the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, the present application provides a non-transitory machine-readable storage medium, the non-transitory machine-readable storage medium stores machine-readable instructions, the machine-readable instructions can be run by a processor to execute instructions corresponding to the method steps provided by the present application, and when the computer program is executed by a central processing unit (CPU), the above functions defined in the method of the present application are executed.
[0136] The methods, apparatuses, and devices of the present application may be implemented in many ways. For example, the methods, apparatuses, and devices of the present application may be implemented by software, hardware, firmware, or any combination of software, hardware, and firmware. The above order of steps for the method is for illustration only, and the steps of the method of the present application are not limited to the order specifically described above unless otherwise specified. In addition, in some embodiments, the present application may also be implemented as programs recorded in a recording medium, which include machine-readable instructions for implementing the method according to the present application. Therefore, the present application also covers recording media storing programs for executing the method according to the present application.
[0137] In addition, the parts of the above-mentioned technical solutions provided in the embodiments of the present application that are consistent with the implementation principles of the corresponding technical solutions in the prior art are not described in detail to avoid excessive redundancy.
[0138] The specific implementation modes as described above further describe the purpose, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific implementation mode of the present invention and is not intended to limit the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.
[0139] The above preset parameters or preset thresholds are all set by technicians in this field according to actual conditions or obtained through large-scale data simulation.
[0140] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.
Claims
1. The optical cable information transmission monitoring system based on intelligent control is characterized by: It includes a sample collection module, an abnormal model training module, a real-time data collection module, an enhanced model building module and an intelligent control module; wherein each module is electrically connected; The sample collection module collects historical monitoring data and historical annotation data of each collection cycle in advance in the test environment, converts the historical monitoring data into an input feature set, and sends the input feature set and historical annotation data to the anomaly model training module; The anomaly model training module takes the input feature set of the current collection cycle as input, the historical annotation data of the next collection cycle as the prediction target, trains the anomaly assessment model, and sends the anomaly assessment model to the real-time data collection module; The real-time data collection module collects the real-time monitoring data of the optical cable to be monitored; inputs the real-time monitoring data into the abnormality assessment model to obtain the abnormal probability sequence, and converts the abnormal probability sequence into an abnormal risk scalar through the sequence conversion model; and sends the real-time monitoring data and the abnormal risk scalar to the intelligent control module; Strengthen the model building module, build the state space, action space and reward function of the power decision model, so as to complete the design of the intelligent agent of the power decision model, and send the designed power decision model to the intelligent control module; The intelligent control module generates the action decision for the next acquisition cycle for the monitored optical cable based on real-time monitoring data, abnormal risk scalar and power decision model; The method of pre-collecting the historical monitoring data and historical annotation data of each collection cycle is: Collect historical monitoring data consisting of parameter time series of various optical cable monitoring parameters regularly collected by sensor monitoring equipment, and historical annotation data consisting of abnormal label value series; The parameter time series is a parameter time series composed of the parameter values of each optical cable monitoring parameter at each moment in each acquisition cycle in chronological order; The abnormal label value sequence is a time sequence of labels composed of abnormal label values of the optical cable at each moment in each collection cycle in time order; the abnormal label value is 0 or 1; The method of converting historical monitoring data into an input feature set is: In each acquisition cycle, the parameter time series of each optical cable monitoring parameter in the acquisition cycle is converted into a set of monitoring feature vectors, and all monitoring feature vectors in the acquisition cycle constitute a set of input feature sets; The method of training the anomaly assessment model is: The input feature set of each acquisition cycle is used as the input of the anomaly assessment model. The anomaly assessment model uses the predicted value of the anomaly probability sequence in the next acquisition cycle as the output. The anomaly probability sequence is the probability of the optical cable anomaly at each moment in the next acquisition cycle. The anomaly label value sequence corresponding to the next acquisition cycle is used as the prediction target. The Euclidean distance between the anomaly probability sequence and the anomaly label value sequence is used as the prediction error. Minimizing the sum of the prediction errors is used as the training target. The anomaly assessment model is trained until the sum of the prediction errors reaches convergence and the training is stopped. The anomaly assessment model is a time series prediction model. The method of collecting the real-time monitoring data of the optical cable to be monitored is: For the optical cable to be monitored, at every collection cycle, the parameter time series of various optical cable monitoring parameters within each collection cycle is collected as real-time monitoring data in a manner consistent with the historical monitoring data collected in the test environment.
2. The optical cable information transmission monitoring system based on intelligent control according to claim 1 is characterized in that: The converting of the abnormal probability sequence into an abnormal risk scalar by the sequence conversion model comprises the following steps: Step 11: Mark each abnormal probability sequence as P, and use moving average or Gaussian kernel filtering method to smooth P to suppress high-frequency disturbances and obtain a smoothed abnormal probability sequence P'; Step 12: Count the maximum value pmax and the minimum value pmin of P', and linearly scale P' by the maximum value pmax and the minimum value pmin to obtain the standard sequence P''; Step 13: construct an LSTM encoder for the sequence conversion model; The input layer of the LSTM encoder receives the standard sequence P''; The LSTM layer of the LSTM encoder returns the hidden state h(t) at each time step, where t is the time step of the sequence conversion model; The attention layer of the LSTM encoder: performs self-attention calculation on the hidden state h(t), outputs the attention weight sequence α(t), and calculates the context vector c; Step 14: Encode the anomaly risk scalar: The attention layer is connected to the fully connected layer, and the fully connected layer is connected to the output layer; Input the context vector c into the fully connected layer to obtain the abnormal risk scalar output by the output layer; Step 15: Construct a risk value sample for the sequence conversion model. The input of the risk value sample is M% abnormal label value sequences in the historical annotation data. The output of the risk value sample is a risk scalar set manually according to the risk level of the abnormal probability sequence. The sequence conversion model is trained through the risk value sample; M is the proportion of the preset selected abnormal label value sequence.
3. The optical cable information transmission monitoring system based on intelligent control according to claim 2 is characterized in that: The state space, action space and reward function of the power decision model are constructed as follows: The power decision model is set to an Actor-Critic model; The state space is marked as S, and the state space S reflects the real-time operation state of the entire optical cable, and the state space includes real-time monitoring data and abnormal risk scalars in each monitoring cycle; The state of each acquisition cycle in the state space S is marked as s; The action space is marked as A, wherein the action space A includes the adjustment amount of the optical signal output power that can be provided by the transmitting end of the optical cable output by the Actor model, and the optical signal output power taken by the optical cable in the next acquisition cycle is marked as a; The reward function is marked as Q, and the reward function Q=r(s, a) measures the effect of performing action a in the current state s.
4. The optical cable information transmission monitoring system based on intelligent control according to claim 3 is characterized in that: The agent is designed as follows: Design the intelligent agent as an Actor-Critic structure based on RNN; The input of the agent is state s; The state-behavior value is output by the Critic network; The action probability is output by the Actor network.
5. The optical cable information transmission monitoring system based on intelligent control according to claim 4 is characterized in that: The method of generating the action decision of the next acquisition cycle for the optical cable to be monitored is: The real-time monitoring data and abnormal risk scalar are taken as the current state s; Inputting the current state s to the Actor model of the power decision model, the Actor model outputting an action in the action space A so as to maximize the state-behavior value output by the Critic model, and outputting an action decision of the adjustment amount of the output power of the optical signal; Execute action decision to obtain the next state s' and actual reward value; The critic network outputs the state-behavior value after executing the action decision in the current state, calculates the TD error δ of the critic network, updates the critic network parameters according to δ, and minimizes the critic loss function; The Actor model takes the state-behavior value output by the Critic model as the output target and uses the policy gradient update method to update the parameters of the Actor network model.
6. An optical cable information transmission monitoring method based on intelligent control, which is implemented based on an optical cable information transmission monitoring system based on intelligent control according to any one of claims 1 to 5, characterized in that: The following steps are involved: Step 1: Collect historical monitoring data and historical annotation data of each collection cycle in advance in the test environment, and convert the historical monitoring data into an input feature set; Step 2: Take the input feature set of the current collection cycle as input, and the historical annotated data of the next collection cycle as the prediction target to train the anomaly assessment model; Step 3: Collect real-time monitoring data of the optical cable to be monitored; input the real-time monitoring data into the anomaly assessment model to obtain an anomaly probability sequence, and convert the anomaly probability sequence into an anomaly risk scalar through a sequence conversion model; Step 4: Construct the state space, action space and reward function of the power decision model to complete the design of the intelligent agent of the power decision model; Step 5: Based on the real-time monitoring data, abnormal risk scalar and power decision model, generate an action decision for the next acquisition cycle for the optical cable to be monitored.
7. An electronic device, characterized in that: include: A processor and a memory, wherein: The memory stores a computer program that can be called by the processor; The processor executes the optical cable information transmission monitoring method based on intelligent control as claimed in claim 6 in the background by calling the computer program stored in the memory.
8. A computer-readable storage medium, characterized in that: A rewritable computer program is stored thereon; When the computer program is executed on a computer device, the computer device executes the optical cable information transmission monitoring method based on intelligent control as claimed in claim 6 in the background.