A delay optimization method for power communication network based on fgOTN

Abstract

The present invention relates to the technical field of network delay optimization, and specifically to a method for optimizing delay of an electric power communication network based on fgOTN. The method comprises the following steps: setting a dynamic delay data acquisition model at each node of the electric power communication network to acquire bidirectional transmission delay data of the node; setting a delay and jitter threshold of the node; fusing and analyzing the acquired bidirectional transmission delay data with electric power business data if the threshold is exceeded to obtain a node to be optimized; obtaining a node evaluation matrix of the node to be optimized and extracting features to obtain a first delay feature; obtaining a node evaluation matrix from historical data and performing feature extraction to obtain a second delay feature; matching the first delay feature and the second delay feature according to a matching algorithm to output a list of delay optimization methods; obtaining a delay optimization method from the list of delay optimization methods and performing delay optimization on the node to be optimized; and issuing an early warning based on the number of times the delay optimization method is skipped.

Description

Technical Field

[0001] The present invention relates to the technical field of network delay optimization, and in particular to a method for optimizing the delay of an electric power communication network based on fgOTN. Background Art

[0002] The power communication network is a specialized communications network designed specifically for power systems. Its primary function is to provide stable, secure, and real-time data transmission services for the power grid, ensuring the normal operation of key services such as relay protection, dispatch automation, monitoring, and energy collection. Typically, power communication networks primarily utilize fiber-optic communication technology (sometimes supplemented by wireless technology) to achieve high-speed, low-latency, and interference-resistant data transmission across the entire network, thereby ensuring real-time response and coordinated operation across all aspects of the power grid, including production, dispatch, and protection.

[0003] Power communication network latency primarily refers to the time required for data packets to travel from source to destination within a dedicated communication network designed to support critical services such as real-time control, dispatching, and relay protection. This latency typically includes transmission delay, queuing delay, and processing delay. The fgOTN-based power communication network is a specialized communication network specifically tailored for power systems, utilizing fgOTN technology to achieve high-precision, low-latency data transmission. This ensures highly reliable operation of the power grid in production, dispatching, and monitoring, providing strict hard isolation and stable latency. This makes data transmission across all links of the power system more reliable and timely, which is particularly important for control and protection services within the power grid that require millisecond-level response times.

[0004] Therefore, a delay optimization method for power communication network based on fgOTN is proposed. Summary of the Invention

[0005] The present invention aims to provide a method for optimizing the delay of a power communication network based on fgOTN, so as to achieve real-time delay optimization of local nodes in the power communication network. First, a dynamic delay data acquisition model is set at each node in the power communication network to collect the node's two-way transmission delay data; a delay and jitter threshold is set for the node. If the threshold is exceeded, the collected two-way transmission delay data is integrated and analyzed with the power business data to obtain the node to be optimized; a node evaluation matrix of the node to be optimized is obtained and features are extracted to obtain a first delay feature; a node evaluation matrix from historical data is obtained and features are extracted to obtain a second delay feature; the first delay feature and the second delay feature are matched according to a matching algorithm, and a list of delay optimization methods is output; a delay optimization method in the delay optimization method list is obtained and delay optimization is performed on the node to be optimized; and an early warning is issued based on the number of delay optimization methods skipped.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] A method for optimizing power communication network latency based on fgOTN, comprising:

[0008] A dynamic delay data collection model is established using historical power data and inter-node data flow distribution data. This model is set up at each node in the power communication network to independently collect bidirectional transmission delay data between the node itself and adjacent nodes.

[0009] Furthermore, the time delay data acquisition model includes:

[0010] Data preprocessing and feature extraction are performed on the node's historical power data and inter-node data flow distribution data to obtain a delay correlation feature matrix. The historical power data includes the power equipment served by the node, timestamp, service type, and power load. The inter-node data flow distribution data includes the inter-node communication flow matrix, protocol type, packet size distribution, and transmission path topology.

[0011] Obtaining a delay correlation feature matrix and bidirectional transmission delay data corresponding to the node, and obtaining a feature set in the delay correlation feature matrix that has a strong correlation with the bidirectional transmission delay data through a correlation extraction model, wherein the correlation extraction model includes a principal component analysis method, a machine learning model, and a reinforcement learning model;

[0012] The data of the corresponding features in the feature set are collected in real time and a delay collection index is established. If the threshold is exceeded, the two-way transmission delay data between the node itself and the adjacent nodes is collected.

[0013] Delay and jitter thresholds are set based on the node's fgOTN characteristic data. If the thresholds are exceeded, the collected bidirectional transmission delay data and power business data are integrated and analyzed to identify the node to be optimized.

[0014] Furthermore, the specific steps for setting the delay and jitter thresholds include:

[0015] Obtain historical data of corresponding features in the feature set of the node to build a historical baseline, obtain the delay and delay jitter corresponding to the historical baseline, and obtain the basic delay threshold and basic delay jitter threshold through weighted processing;

[0016] Obtain data on corresponding features from the feature set collected in real time to build a real-time baseline. Calculate the deviation between the real-time baseline and the historical baseline, and adjust the basic delay threshold and basic delay jitter threshold based on the deviation to obtain the real-time delay threshold and delay jitter threshold corresponding to the node.

[0017] Furthermore, the step of obtaining the node to be optimized includes:

[0018] If the real-time delay of the node exceeds the delay threshold and the real-time delay jitter of the node exceeds the delay jitter threshold, the collected two-way transmission delay data is obtained to construct a delay difference vector;

[0019] A node evaluation matrix is ​​constructed according to the delay difference vector and the power business data of the node, and the node evaluation matrix is ​​prioritized according to the decision tree to obtain the nodes to be optimized, which include slightly abnormal nodes, moderately abnormal nodes and severely abnormal nodes.

[0020] Obtain the node evaluation matrix of the node to be optimized and extract features to obtain the first delay feature; obtain the node evaluation matrix in the historical data and extract features to obtain the second delay feature; match the first delay feature and the second delay feature according to the matching algorithm, and output a list of delay optimization methods;

[0021] Furthermore, the specific steps of constructing a list of latency optimization methods include:

[0022] Perform feature extraction on the node evaluation matrix to obtain the first delay feature;

[0023] Obtaining the node evaluation matrix from the node's historical data and performing feature extraction to obtain the second delay feature;

[0024] The first delay feature and the second delay feature are matched according to a matching algorithm, which includes similarity calculation, cluster analysis, and machine learning classification methods. The corresponding delay optimization method in the historical data is obtained according to the matching degree to obtain a delay optimization method list.

[0025] Obtain a delay optimization method from the delay optimization method list and perform delay optimization on the node to be optimized. If the delay optimization time exceeds the preset time window, skip and execute the next delay optimization method in the delay optimization method list.

[0026] Furthermore, the delay optimization methods in the list are obtained in turn, and delay optimization is performed on the node to be optimized; if the delay optimization method takes more than the preset time window during execution, the method is automatically skipped and the next optimization method in the list is executed instead; if all candidate methods in the list cannot successfully optimize the delay within the preset time, the node early warning mechanism is triggered, prompting the administrator to intervene; the execution results and time consumption of each candidate method are recorded in real time, and the ranking of the candidate methods is adjusted according to the actual effect.

[0027] Issue an early warning based on the number of times the latency optimization method is skipped.

[0028] Furthermore, for each node to be optimized, the number of candidate optimization methods that are skipped due to execution timeout and failure to improve latency within a fixed time period is counted; an early warning threshold is set based on the number of consecutive skips and the cumulative timeout ratio. If the early warning threshold is exceeded, an early warning is automatically triggered. The early warning information includes: the node to be optimized, the list of skipped methods and the corresponding frequency, and analysis of the reasons for optimization failure; the early warning information is fed back to the network operation and maintenance personnel through the system management platform.

[0029] Compared with the prior art, the present invention has the following beneficial effects:

[0030] 1. The dynamic delay data acquisition model cleans, normalizes, and extracts features from historical power data and inter-node data flow distribution data to establish a delay-related feature matrix. It then uses principal component analysis, machine learning, and reinforcement learning models to screen out key features highly correlated with two-way delay, thereby constructing an accurate delay acquisition index for each node. When a node's delay acquisition index exceeds a preset threshold, the system automatically triggers the collection of two-way transmission delay data, enabling real-time monitoring of delay fluctuations and jitter in the power communication network.

[0031] 2. By acquiring and processing the historical data corresponding to the feature set, a historical baseline is constructed and the basic delay and jitter thresholds are determined. The real-time baseline is then constructed using the real-time data collected, and the deviation from the historical baseline is calculated. The basic threshold is dynamically adjusted based on the deviation. This process can adapt the corresponding delay threshold and delay jitter threshold for the node, ensuring the smooth operation of the power communication network under the high real-time and high reliability requirements.

[0032] 3. First, through data processing, key delay features are extracted, effectively removing noise interference and improving data quality. Secondly, combined with improved cosine similarity and cluster analysis, efficient matching of the first delay feature with the second delay feature is achieved, ensuring the accuracy and applicability of the matching results. Subsequently, a machine learning classification model is used to further optimize the matching results, predict the best delay optimization method, and sort it according to historical optimization results, making the optimization plan more intelligent and improving optimization efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0033] Figure 1 A schematic diagram of a method for optimizing power communication network latency based on fgOTN provided by an embodiment of the present invention;

[0034] Figure 2 A flowchart of obtaining a node to be optimized provided by an embodiment of the present invention;

[0035] Figure 3 A flowchart of a list of delay optimization methods provided in an embodiment of the present invention. DETAILED DESCRIPTION

[0036] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0037] Example 1:

[0038] The fgOTN-based power communication network is a dedicated communication network customized for power systems that uses fgOTN technology to achieve high-precision, low-latency data transmission, ensuring high-reliability operation of the power grid in production, scheduling, monitoring, etc. In order to optimize the communication delay of nodes in the power communication network, a company introduced a power communication network delay optimization method based on fgOTN provided by the present invention. The method process is as follows: Figure 1 As shown, the specific implementation is as follows:

[0039] First, a dynamic delay data acquisition model is established using historical power data and inter-node data flow distribution data. A dynamic delay data acquisition model is set up at each node in the power communication network to independently collect bidirectional transmission delay data between the node itself and its neighboring nodes.

[0040] Furthermore, the time delay data acquisition model includes:

[0041] Data preprocessing and feature extraction are performed on the node's historical power data and inter-node data flow distribution data to obtain a delay correlation feature matrix. The historical power data includes the power equipment served by the node, timestamp, service type, and power load. The inter-node data flow distribution data includes the inter-node communication flow matrix, protocol type, packet size distribution, and transmission path topology.

[0042] Obtaining a delay correlation feature matrix and bidirectional transmission delay data corresponding to the node, and obtaining a feature set in the delay correlation feature matrix that has a strong correlation with the bidirectional transmission delay data through a correlation extraction model, wherein the correlation extraction model includes a principal component analysis method, a machine learning model, and a reinforcement learning model;

[0043] The data of the corresponding features in the feature set are collected in real time and a delay collection index is established. If the threshold is exceeded, the two-way transmission delay data between the node itself and the adjacent nodes is collected.

[0044] Furthermore, by cleaning, normalizing and extracting features from historical power data and inter-node data flow distribution data, a comprehensive delay correlation feature matrix is ​​established. This matrix not only reflects the impact of power equipment and business load on delay, but also takes into account factors such as network traffic, protocols and transmission paths, providing basic data for subsequent correlation analysis.

[0045] Furthermore, after obtaining the delay correlation feature matrix, a correlation extraction model is used to filter out feature sets that are highly correlated with the two-way delay data. Various methods can be used, including but not limited to principal component analysis, machine learning models, and reinforcement learning models. In this embodiment, principal component analysis is used to reduce the dimensionality of the multidimensional feature matrix, extracting principal components that can explain the main fluctuations in the data and filtering out noise and redundant information. Taking node i as an example, the above processing yields a feature set that is strongly correlated with the two-way transmission delay data for node i, including power equipment characteristics, load characteristics, traffic characteristics, packet characteristics, and data peak characteristics. After determining the strong correlation between node i and these features, data from the feature set of node i is collected within a time window T, and a delay collection index is constructed based on the collected data. In this embodiment, the data can be pre-processed and then weighted, or the delay collection index can be constructed based on the degree of fluctuation in the collected data. If the delay collection index exceeds a threshold, the delay data collection model automatically collects two-way transmission delay data between node i and adjacent nodes. Table 1 shows some of the monitored delay data.

[0046] Table 1. Partial delay data

[0047]

[0048] The dynamic delay data collection model can deeply integrate historical power data and inter-node data flow distribution data, establish a delay-related feature matrix through cleaning, normalization and feature extraction, and then use principal component analysis, machine learning and reinforcement learning models to screen out key features that are highly correlated with two-way delay, thereby building accurate delay collection indicators for each node; when the node's delay collection indicator exceeds the preset threshold, the system can automatically trigger the collection of two-way transmission delay data. This not only realizes real-time monitoring of delay fluctuations and jitter in the power communication network, but also provides a scientific basis for subsequent delay optimization, resource scheduling and fault warning through data-driven dynamic adjustment, effectively improving the stability of network transmission and the real-time guarantee of power services.

[0049] Delay and jitter thresholds are set based on the node's fgOTN characteristic data. If the thresholds are exceeded, the collected bidirectional transmission delay data and power business data are integrated and analyzed to identify the node to be optimized.

[0050] Furthermore, the specific steps for setting the delay and jitter thresholds include:

[0051] Obtain historical data of corresponding features in the feature set of the node to build a historical baseline, obtain the delay and delay jitter corresponding to the historical baseline, and obtain the basic delay threshold and basic delay jitter threshold through weighted processing;

[0052] Obtain data on corresponding features from the feature set collected in real time to build a real-time baseline. Calculate the deviation between the real-time baseline and the historical baseline, and adjust the basic delay threshold and basic delay jitter threshold based on the deviation to obtain the real-time delay threshold and delay jitter threshold corresponding to the node.

[0053] Furthermore, after obtaining the feature set of the node above, the data corresponding to the feature set after triggering the collection is obtained to build a historical baseline. Taking node i as an example, the feature set includes power equipment features, load features, flow features, data packet features and data peak features. The data corresponding to several groups of feature sets collected in the historical data are obtained, and the data are cleaned, normalized, noise and outlier processing are performed, and then weighted to obtain the historical baseline; correspondingly, several groups of delays and delay jitters are obtained and cleaned, normalized, noise and outlier processing are performed, and then weighted to obtain the basic delay threshold and basic delay jitter. Delay and jitter thresholds; then, data corresponding to features in the real-time collected feature set are obtained to construct a real-time baseline, and the deviation between the real-time baseline and the historical baseline is calculated. The deviation can be constructed using the difference between the real-time baseline and the historical baseline, or a vector can be constructed using the real-time baseline and the historical baseline. The deviation is measured by the angle between the vectors. This deviation can reveal the changes in delay and jitter in the current network environment due to load changes, link abnormalities, or device performance fluctuations. After obtaining the deviation, the basic delay threshold and basic delay jitter threshold are adjusted according to the deviation to obtain the node's delay threshold and delay jitter threshold at the current moment.

[0054] By obtaining historical data corresponding to the feature set and performing cleaning, normalization and weighting processing, a historical baseline is constructed and the basic delay and jitter thresholds are determined; then, the real-time baseline is constructed using real-time collected data, and the deviation from the historical baseline is calculated. The degree of change is measured by vector or difference method, thereby dynamically adjusting the basic threshold; this process can adapt the corresponding delay threshold and delay jitter threshold for the node, ensuring the smooth operation of the power communication network under high real-time and high reliability requirements, and improving the overall service quality and network stability.

[0055] Furthermore, the steps to obtain the node to be optimized are as follows: Figure 2 Shown, including:

[0056] If the real-time delay of the node exceeds the delay threshold and the real-time delay jitter of the node exceeds the delay jitter threshold, the collected two-way transmission delay data is obtained to construct a delay difference vector;

[0057] A node evaluation matrix is ​​constructed according to the delay difference vector and the power business data of the node, and the node evaluation matrix is ​​prioritized according to the decision tree to obtain the nodes to be optimized, which include slightly abnormal nodes, moderately abnormal nodes and severely abnormal nodes.

[0058] Furthermore, if the real-time delay of the node does not exceed the threshold, and the real-time delay jitter of the node does not exceed the delay jitter threshold, it indicates that the node is operating normally; if the real-time delay of the node exceeds the threshold, or the real-time delay jitter of the node exceeds the threshold, it indicates that the node is relatively stable and can still operate normally; if the real-time delay of the node exceeds the delay threshold, and the real-time delay jitter of the node also exceeds the delay jitter threshold, it indicates that the data transmission of the node is unstable and delay optimization is required.

[0059] Furthermore, the two-way transmission delay data includes the real-time delay and real-time delay jitter between the node and the adjacent node. The two-way real-time delay and real-time delay jitter are obtained and subtracted to obtain a delay difference vector, which reflects the asymmetry of the two-way delay of the node.

[0060] Furthermore, a node evaluation matrix is ​​constructed based on the delay difference vector and the power business data of the node. The power business data includes dispatch instructions, business load, and network dispatch parameters. This matrix maps multidimensional data into a node evaluation matrix, which not only contains quantitative data on network transmission delay and jitter, but also incorporates the real-time operation status of the power business, comprehensively reflecting the extent of the impact of node performance on the power business.

[0061] Furthermore, based on the constructed node evaluation matrix, a decision tree algorithm is used to prioritize each node. The decision tree model automatically generates classification rules based on various evaluation metrics (such as latency deviation, jitter amplitude, and load conditions), categorizing nodes into three levels: mild, moderate, and severe. Mildly abnormal nodes may only require fine-tuning parameters or local optimization; moderately abnormal nodes require optimized scheduling for specific links or devices; and severely abnormal nodes may require global focused scheduling or manual intervention for in-depth investigation. Latency optimization is prioritized for each abnormal node, with severe abnormal nodes being prioritized.

[0062] By integrating the real-time collected two-way transmission delay data with the power business data and automatically dividing the node evaluation matrix with the help of a decision tree model, it is possible to efficiently and accurately identify abnormal nodes that have a greater impact on business operations in critical periods and areas, thereby ensuring that the power communication network can trigger optimization measures in a timely manner, prioritize the optimization of severely abnormal nodes, and ensure the high reliability and real-time performance of the entire system.

[0063] Obtain the node evaluation matrix of the node to be optimized and extract features to obtain the first delay feature; obtain the node evaluation matrix in the historical data and extract features to obtain the second delay feature; match the first delay feature and the second delay feature according to the matching algorithm, and output a list of delay optimization methods;

[0064] Furthermore, the specific steps for constructing a list of delay optimization methods are as follows: Figure 3 Shown, including:

[0065] Perform feature extraction on the node evaluation matrix to obtain the first delay feature;

[0066] Obtaining the node evaluation matrix from the node's historical data and performing feature extraction to obtain the second delay feature;

[0067] The first delay feature and the second delay feature are matched according to a matching algorithm, which includes similarity calculation, cluster analysis, and machine learning classification methods. The corresponding delay optimization method in the historical data is obtained according to the matching degree to obtain a delay optimization method list.

[0068] Furthermore, the node evaluation matrix of the node is first normalized and denoised. Then, it is processed using adaptive filters such as Kalman filtering or particle filtering to further reduce noise interference. Statistical analysis and dimensionality reduction techniques (such as principal component analysis) are then used to extract key features that can accurately reflect the current node delay status. These key features constitute the so-called first delay features.

[0069] Furthermore, a similar method is used to process the node evaluation matrix in the historical data of all nodes to obtain the second delay feature;

[0070] Furthermore, the matching algorithm in this embodiment is a multi-level matching algorithm. First, preliminary matching is performed using improved cosine similarity, and the second delay features with similarity greater than the similarity threshold are retained to obtain a first matching set. The formula for improved cosine similarity is:

[0071] ;

[0072] in, represents the improved cosine similarity, The delay characteristics The weight coefficient of a component can be determined by calculating the variance of the component and other historical components. The first delay characteristic A quantity, The second delay characteristic A quantity, represents the first delay characteristic, represents the second delay characteristic, represents the two-norm.

[0073] Furthermore, after obtaining the first matching set, the distance between the first feature vector and the cluster center of the first matching set is calculated through cluster analysis, and the second delay feature smaller than the distance is retained to obtain the second matching set;

[0074] Furthermore, the classification model is trained using the labeled second matching set. When real-time data is input, the model predicts several latency optimization methods that best match the first latency characteristic. Finally, the matched latency optimization methods are ranked based on their improvement effect in historical data and their real-time performance, resulting in a latency optimization method list. Table 2 shows the data from the latency optimization method list for node i.

[0075] Table 2. Some delay optimization methods

[0076]

[0077] First, key delay features are extracted through normalization, adaptive filtering, and principal component analysis, effectively removing noise interference and improving data quality. Secondly, combined with improved cosine similarity and cluster analysis, efficient matching of the first and second delay features is achieved, ensuring the accuracy and applicability of the matching results. Subsequently, a machine learning classification model is used to further optimize the matching results, predict the optimal delay optimization method, and sort it according to historical optimization results, making the optimization plan more intelligent and personalized.

[0078] Obtain a delay optimization method from the delay optimization method list and perform delay optimization on the node to be optimized. If the delay optimization time exceeds the preset time window, skip and execute the next delay optimization method in the delay optimization method list.

[0079] Furthermore, the delay optimization methods in the list are tried in turn, and delay optimization is performed on the node to be optimized. If a delay optimization method takes longer than the preset time window during execution, the method is automatically skipped and the next optimization method in the list is executed instead. If all candidate methods in the list cannot successfully optimize the delay within the preset time, the node early warning mechanism is triggered, prompting the administrator to intervene. The system records the execution results and time consumption of each candidate method in real time, and adjusts the ranking of the candidate methods according to the actual results.

[0080] Furthermore, during the execution process, a timing monitoring mechanism is designed to record the execution time and real-time optimization effect of the latency optimization method. If the execution time reaches the expected improvement goal within the preset time window, the method is considered successful and its execution result is recorded; if the execution time of a candidate method exceeds the preset time window, the system will automatically terminate the execution of the method and record the candidate method as "timed out". After terminating the current method that takes too long, the next optimization method is immediately selected from the list for trial until a method that can effectively reduce node latency within the specified time is found.

[0081] Through a timed monitoring mechanism, the system records the execution time and actual optimization results of each candidate method in real time. If the preset time window is exceeded, the current method will be automatically terminated and the next optimization solution will be quickly switched to, ensuring that node latency can be effectively reduced within the specified time. If all candidate methods fail to meet the optimization requirements, an early warning mechanism will promptly alert the administrator to intervene, thus ensuring the stable operation of the power communication network under the high real-time requirements.

[0082] Issue an early warning based on the number of times the latency optimization method is skipped.

[0083] Furthermore, for each node to be optimized, the number of candidate optimization methods that are skipped due to execution timeout and failure to improve latency within a fixed period of time is counted; an early warning threshold is set based on the number of consecutive skips or the cumulative timeout ratio. Once the early warning threshold is exceeded, an early warning is automatically triggered. The early warning information includes: the node to be optimized, the list of skipped methods and the corresponding frequency, and an analysis of possible reasons for optimization failure. The early warning information is fed back to the network operation and maintenance personnel through the system management platform.

[0084] Furthermore, fixed time refers to a long period, such as a day, week, and month. The number of times the candidate delay optimization method is skipped due to execution timeout or failure to achieve the expected improvement effect is counted. The warning threshold can be set based on the number of skips, or based on the total time proportion of the optimized delay.

[0085] This early warning mechanism based on the number of skips and timeout ratio can not only effectively monitor the operating status of each node to be optimized, but also automatically issue an alarm when a problem occurs, providing data support and decision-making basis for timely correction of network anomalies, thereby ensuring the high reliability and low latency of the power communication network in critical business operations.

[0086] The fgOTN-based power communication network delay optimization method provided by the present invention realizes the full process of "monitoring-collection-analysis-optimization" of the power communication network delay; first, by constructing a dynamic delay data collection model, the system can accurately capture the two-way transmission delay data between each node and its adjacent nodes; then, a baseline is constructed based on historical data, and the delay and jitter thresholds are adjusted in real time; when the node delay data exceeds the set threshold, the system further integrates the power business data and delay data, and uses decision trees and other methods to prioritize the nodes, identifying mild, moderate and severe abnormal nodes; then, a list of candidate optimization methods is generated according to a multi-level matching algorithm, and each optimization strategy is tried in turn, automatically skipping the candidate method that has timed out. If all candidate methods cannot improve the delay within the preset time, an early warning mechanism is triggered, ensuring that the power communication network always maintains low latency, low jitter and high stability in critical business scenarios, effectively reducing business risks caused by network anomalies.

[0087] Example 2:

[0088] When optimizing the latency of a power communication network, a company introduced the fgOTN-based power communication network latency optimization method provided by this invention. The specific implementation method is as follows:

[0089] A dynamic delay data collection model is established using historical power data and inter-node data flow distribution data. This model is set up at each node in the power communication network to independently collect bidirectional transmission delay data between the node itself and adjacent nodes.

[0090] Furthermore, the time delay data acquisition model includes:

[0091] Data preprocessing and feature extraction are performed on the node's historical power data and inter-node data flow distribution data to obtain a delay correlation feature matrix. The historical power data includes the power equipment served by the node, timestamp, service type, and power load. The inter-node data flow distribution data includes the inter-node communication flow matrix, protocol type, packet size distribution, and transmission path topology.

[0092] Obtaining a delay correlation feature matrix and bidirectional transmission delay data corresponding to the node, and obtaining a feature set in the delay correlation feature matrix that has a strong correlation with the bidirectional transmission delay data through a correlation extraction model, wherein the correlation extraction model includes a principal component analysis method, a machine learning model, and a reinforcement learning model;

[0093] The data of the corresponding features in the feature set are collected in real time and a delay collection index is established. If the threshold is exceeded, the two-way transmission delay data between the node itself and the adjacent nodes is collected.

[0094] Delay and jitter thresholds are set based on the node's fgOTN characteristic data. If the thresholds are exceeded, the collected bidirectional transmission delay data and power business data are integrated and analyzed to identify the node to be optimized.

[0095] Furthermore, the specific steps for setting the delay and jitter thresholds include:

[0096] Obtain historical data of corresponding features in the feature set of the node to build a historical baseline, obtain the delay and delay jitter corresponding to the historical baseline, and obtain the basic delay threshold and basic delay jitter threshold through weighted processing;

[0097] Obtain data on corresponding features from the feature set collected in real time to build a real-time baseline. Calculate the deviation between the real-time baseline and the historical baseline, and adjust the basic delay threshold and basic delay jitter threshold based on the deviation to obtain the real-time delay threshold and delay jitter threshold corresponding to the node.

[0098] Furthermore, the step of obtaining the node to be optimized includes:

[0099] If the real-time delay of the node exceeds the delay threshold and the real-time delay jitter of the node exceeds the delay jitter threshold, the collected two-way transmission delay data is obtained to construct a delay difference vector;

[0100] A node evaluation matrix is ​​constructed according to the delay difference vector and the power business data of the node, and the node evaluation matrix is ​​prioritized according to the decision tree to obtain the nodes to be optimized, which include slightly abnormal nodes, moderately abnormal nodes and severely abnormal nodes.

[0101] Obtain the node evaluation matrix of the node to be optimized and extract features to obtain the first delay feature; obtain the node evaluation matrix in the historical data and extract features to obtain the second delay feature; match the first delay feature and the second delay feature according to the matching algorithm, and output a list of delay optimization methods;

[0102] Furthermore, the specific steps of constructing a list of latency optimization methods include:

[0103] Perform feature extraction on the node evaluation matrix to obtain the first delay feature;

[0104] Obtaining the node evaluation matrix from the node's historical data and performing feature extraction to obtain the second delay feature;

[0105] The first delay feature and the second delay feature are matched according to a matching algorithm, which includes similarity calculation, cluster analysis, and machine learning classification methods. The corresponding delay optimization method in the historical data is obtained according to the matching degree to obtain a delay optimization method list.

[0106] Obtain a delay optimization method from the delay optimization method list and perform delay optimization on the node to be optimized. If the delay optimization time exceeds the preset time window, skip and execute the next delay optimization method in the delay optimization method list.

[0107] Furthermore, the delay optimization methods in the list are obtained in turn, and delay optimization is performed on the node to be optimized; if the delay optimization method takes more than the preset time window during execution, the method is automatically skipped and the next optimization method in the list is executed instead; if all candidate methods in the list cannot successfully optimize the delay within the preset time, the node early warning mechanism is triggered, prompting the administrator to intervene; the execution results and time consumption of each candidate method are recorded in real time, and the ranking of the candidate methods is adjusted according to the actual effect.

[0108] Issue an early warning based on the number of times the latency optimization method is skipped.

[0109] Furthermore, for each node to be optimized, the number of candidate optimization methods that were skipped due to execution timeouts and failure to improve latency within a fixed period of time is counted. A warning threshold is set based on the number of consecutive skips and the cumulative timeout ratio. If the threshold is exceeded, an alert is automatically triggered. The alert information includes: the node to be optimized, a list of skipped methods and their corresponding frequencies, and an analysis of the reasons for the optimization failure. This alert information is fed back to network operations personnel via the system management platform. Table 3 shows the alert information for some nodes.

[0110] Table 3. Warning information of some nodes

[0111]

[0112] The fgOTN-based power communication network delay optimization method provided by the present invention can realize real-time delay monitoring, analysis and optimization of local nodes of the power communication network, improve the delay optimization efficiency, and ensure the stability and reliability of the power communication network.

[0113] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A method for optimizing the delay of a power communication network based on fgOTN, characterized in that: include: A dynamic delay data collection model is established using historical power data and inter-node data flow distribution data. This model is set up at each node in the power communication network to independently collect bidirectional transmission delay data between the node itself and adjacent nodes. Set corresponding delay and jitter thresholds based on the node's fgOTN characteristic data; If the threshold is exceeded, the collected two-way transmission delay data and power business data are integrated and analyzed to obtain the node to be optimized; Obtain the node evaluation matrix of the node to be optimized and extract features to obtain the first delay feature; obtain the node evaluation matrix in the historical data and perform feature extraction to obtain the second delay feature; Matching the first delay feature and the second delay feature according to a matching algorithm, and outputting a list of delay optimization methods; Obtain a delay optimization method from the delay optimization method list and perform delay optimization on the node to be optimized. If the delay optimization time exceeds the preset time window, skip and execute the next delay optimization method in the delay optimization method list. Issue an early warning based on the number of times the latency optimization method is skipped.

2. The method for optimizing the delay of an electric power communication network based on fgOTN according to claim 1, characterized in that: The time delay data acquisition model includes: Data preprocessing and feature extraction are performed on the node's historical power data and inter-node data flow distribution data to obtain a delay correlation feature matrix. The historical power data includes the power equipment served by the node, timestamp, service type, and power load. The inter-node data flow distribution data includes the inter-node communication flow matrix, protocol type, packet size distribution, and transmission path topology. Obtaining a delay correlation feature matrix and bidirectional transmission delay data corresponding to the node, and obtaining a feature set in the delay correlation feature matrix that has a strong correlation with the bidirectional transmission delay data through a correlation extraction model, wherein the correlation extraction model includes a principal component analysis method, a machine learning model, and a reinforcement learning model; The data of the corresponding features in the feature set are collected in real time and a delay collection index is established. If the threshold is exceeded, the two-way transmission delay data between the node itself and the adjacent nodes is collected.

3. The method for optimizing the delay of an electric power communication network based on fgOTN according to claim 1, characterized in that: The specific steps for setting latency and jitter thresholds include: Obtain historical data of corresponding features in the feature set of the node to build a historical baseline, obtain the delay and delay jitter corresponding to the historical baseline, and obtain the basic delay threshold and basic delay jitter threshold through weighted processing; Obtain data on corresponding features from the feature set collected in real time to build a real-time baseline. Calculate the deviation between the real-time baseline and the historical baseline, and adjust the basic delay threshold and basic delay jitter threshold based on the deviation to obtain the real-time delay threshold and delay jitter threshold corresponding to the node.

4. The method for optimizing the delay of an electric power communication network based on fgOTN according to claim 1, characterized in that: The steps of obtaining the node to be optimized include: If the real-time delay of the node exceeds the delay threshold and the real-time delay jitter of the node exceeds the delay jitter threshold, the collected two-way transmission delay data is obtained to construct a delay difference vector; A node evaluation matrix is ​​constructed according to the delay difference vector and the power business data of the node, and the node evaluation matrix is ​​prioritized according to the decision tree to obtain the nodes to be optimized, which include slightly abnormal nodes, moderately abnormal nodes and severely abnormal nodes.

5. The method for optimizing the delay of an electric power communication network based on fgOTN according to claim 1, characterized in that: The specific steps for building a list of latency optimization methods include: Perform feature extraction on the node evaluation matrix to obtain the first delay feature; Obtaining the node evaluation matrix from the node's historical data and performing feature extraction to obtain the second delay feature; The first delay feature and the second delay feature are matched according to a matching algorithm, which includes similarity calculation, cluster analysis, and machine learning classification methods. The corresponding delay optimization method in the historical data is obtained according to the matching degree to obtain a delay optimization method list.

6. The method for optimizing the delay of an electric power communication network based on fgOTN according to claim 1, characterized in that: Also includes: Obtain the delay optimization methods in the list in turn, and perform delay optimization on the node to be optimized; If a latency optimization method takes longer than the preset time window during execution, the method will be automatically skipped and the next optimization method in the list will be executed instead; If all candidate methods in the list fail to successfully optimize latency within the preset time, the node warning mechanism will be triggered, prompting the administrator to intervene; the execution results and time consumption of each candidate method will be recorded in real time to adjust the ranking of the candidate methods.

7. The method for optimizing the delay of an electric power communication network based on fgOTN according to claim 1, characterized in that: Also includes: For each node to be optimized, count the number of candidate optimization methods that are skipped due to execution timeout and failure to improve latency within a fixed time. Set warning thresholds based on the number of consecutive skips and the cumulative timeout ratio. If the warning threshold is exceeded, an alert is automatically triggered. The alert information includes: nodes to be optimized, a list of skip methods and their corresponding frequencies, and an analysis of the reasons for optimization failure. The early warning information is fed back to the network operation and maintenance personnel through the system management platform.

Images