Artificial intelligence-based power communication big data anomaly prediction and intelligent regulation system

Abstract

This invention discloses an artificial intelligence-based power communication big data anomaly prediction and intelligent control system, belonging to the field of intelligent power communication monitoring technology. It addresses the problem that traditional power communication monitoring systems often rely on communication alarms or manual monitoring to detect communication link anomalies or network edge node anomalies. The system includes an anomaly extraction module, an anomaly prediction module, an online detection module, and an intelligent control module. The anomaly extraction module constructs a set of anomaly values ​​for network edge nodes based on rated parameter data and real-time node power data. The anomaly prediction module trains an anomaly detection model based on historical anomaly network data. The online detection module detects the anomaly value set using the anomaly detection model. The intelligent control module intelligently controls anomaly network edge nodes based on anomaly types. This invention achieves trend prediction and intelligent control of power communication network anomalies.

Description

Technical Field

[0001] This invention belongs to the field of intelligent monitoring technology for power communication, specifically an artificial intelligence-based power communication big data anomaly prediction and intelligent control system. Background Technology

[0002] Power communication is a dedicated communication network used in power systems. It utilizes wired, radio, optical, or other electromagnetic systems to transmit information related to power generation, dispatch, and management. Along with relay protection systems and dispatch automation systems, it forms one of the three pillars of power security. As the foundation for grid dispatch automation, market-based network operation, and modernized management, this communication network is a crucial infrastructure of the power system. Built using various technologies such as fiber optics, satellite, microwave, and carrier wave, it forms a communication guarantee system covering the entire process of power generation, transmission, transformation, and distribution, possessing high reliability and rapid response characteristics.

[0003] Traditional power communication monitoring systems rely heavily on communication alarms or manual monitoring to detect communication link anomalies or network edge node anomalies. They cannot identify anomalies in advance, resulting in intervention only after a communication anomaly occurs. Furthermore, after detecting communication link anomalies or network edge node anomalies, they rely heavily on manual adjustments, which cannot achieve intelligent strategy generation and closed-loop control.

[0004] To this end, this invention proposes an artificial intelligence-based power communication big data anomaly prediction and intelligent control system. Summary of the Invention

[0005] The purpose of this invention is to propose an artificial intelligence-based power communication big data anomaly prediction and intelligent control system to solve the problems mentioned in the background art.

[0006] The technical problem to be solved by this invention is:

[0007] How to achieve trend prediction and intelligent control of anomalies in power communication networks.

[0008] The objective of this invention can be achieved through the following technical solutions:

[0009] An AI-based power communication big data anomaly prediction and intelligent control system includes a data acquisition module, an anomaly extraction module, an anomaly prediction module, an online detection module, an intelligent control module, and a database module. The data acquisition module collects real-time node power data from all network edge nodes in the power communication network at different sampling points and sends it to the anomaly extraction module. The database module stores the rated parameter data of all network edge nodes and sends it to the anomaly extraction module. The anomaly extraction module constructs an anomaly value set of network edge nodes based on the rated parameter data and real-time node power data, and sends the constructed anomaly value set to the online detection module.

[0010] The database module is also used to store historical abnormal network data of all network edge nodes and send it to the anomaly prediction module; the anomaly prediction module is used to train the anomaly detection model based on the historical abnormal network data, and the trained anomaly detection model is sent to the online detection module.

[0011] The online detection module is used to detect abnormal value sets through an anomaly detection model, and sends the detected anomaly warning signals and anomaly types of abnormal network edge nodes to the intelligent control module; the intelligent control module is used to intelligently control the abnormal network edge nodes according to the anomaly type.

[0012] Furthermore, the real-time node power data consists of the real-time signal delay, real-time signal packet loss rate, and real-time signal disconnection rate of all network edge nodes in the power communication network at different sampling points.

[0013] Furthermore, the rated parameter data are the rated signal delay, rated signal packet loss rate, and rated signal disconnection rate of all network edge nodes in the power communication network.

[0014] Furthermore, the process of constructing the set of abnormal values ​​is as follows:

[0015] Step B101: Traverse the real-time signal delay corresponding to the real-time node power data at all sampling points;

[0016] If no real-time signal delay is received at any sampling point or the real-time signal delay at any sampling point is empty, then the corresponding real-time signal delay is determined to be a missing value, and proceed to step B102.

[0017] If all sampling points receive real-time signal delay, and the real-time signal delay of all sampling points is not empty, then the corresponding real-time signal delay is determined to be normal, and step B104 is entered.

[0018] Step B102: Obtain the real-time signal delay V1 of the first sampling point and the real-time signal delay V3 of the third sampling point. If the real-time signal delay V2 is not collected at the second sampling point, subtract the time node corresponding to the first sampling point from the time node corresponding to the second sampling point to calculate the first time interval between the second sampling point and the first sampling point. At the same time, subtract the time node corresponding to the second sampling point from the time node corresponding to the third sampling point to calculate the second time interval between the third sampling point and the second sampling point. Divide the first time interval by the second time interval to calculate the time interval ratio JG.

[0019] Step B103: If the time interval ratio is greater than one, then the second sampling point is determined to be close to the third sampling point. The real-time signal delay corresponding to the second sampling point is V2 = [(V1 + V3) / 2] × JG.

[0020] If the time interval ratio is less than one, then the second sampling point is determined to be close to the first sampling point, and the real-time signal delay corresponding to the second sampling point is V2=[(V1+V3) / 2]×JG;

[0021] If the calculation result is equal to one, it is determined that the time interval between the second sampling point and the first and third sampling points is the same, and the real-time signal delay corresponding to the second sampling point is V2 = (V1 + V3) / 2;

[0022] Step B104: Obtain the real-time signal delay corresponding to any sampling point of the network edge node, divide the real-time signal delay by the rated signal delay, and calculate the standard value of the signal delay of the corresponding sampling point. Similarly, calculate the standard value of the signal packet loss rate and the standard value of the signal disconnection rate, and replace the values ​​in the real-time node power data with the calculated standard values.

[0023] Step B105: Subtract the standard value of the signal delay corresponding to the previous sampling point from the standard value of the signal delay corresponding to the next sampling point to calculate the signal delay difference between adjacent sampling points. Similarly, calculate the difference in signal packet loss rate and signal disconnection rate between adjacent sampling points.

[0024] Furthermore, the process of constructing the set of abnormal values ​​is as follows:

[0025] Step B106: If the signal delay difference between adjacent sampling points is greater than or equal to the first difference threshold, or the signal packet loss rate difference is greater than or equal to the second difference threshold, or the signal disconnection rate difference is greater than or equal to the third difference threshold, then the corresponding adjacent sampling points are determined to be abnormal sampling points, and proceed to step B108.

[0026] If the signal delay difference between adjacent sampling points is less than the first difference threshold, the signal packet loss rate difference is less than the second difference threshold, and the signal disconnection rate difference is less than the third difference threshold, then proceed to the next step.

[0027] Step B107: When the standard value of signal delay of any sampling point among the adjacent sampling points is greater than one, or the standard value of signal packet loss rate is greater than one, or the standard value of signal disconnection rate is greater than one, the corresponding sampling point is determined to be an abnormal sampling point, and the process proceeds to step B108.

[0028] When the standard value of signal delay of all adjacent sampling points is less than or equal to one, the standard value of signal packet loss rate is less than or equal to one, and the standard value of signal disconnection rate is less than or equal to one, the corresponding adjacent sampling points are determined to be normal sampling points.

[0029] Step B108: Record the standard value of signal delay at abnormal sampling points as the standard value of abnormal signal delay, the standard value of signal packet loss rate as the standard value of abnormal signal packet loss rate, and the standard value of signal disconnection rate as the standard value of abnormal signal disconnection rate. Then, summarize the standard values ​​of abnormal signal delay, abnormal signal packet loss rate, and abnormal signal disconnection rate into a set of abnormal values ​​for network edge nodes.

[0030] Furthermore, historical abnormal network data includes abnormal signal delays, abnormal signal packet loss rates, and abnormal signal disconnection rates for all network edge nodes in the power communication network.

[0031] Furthermore, the training process of the anomaly detection model is as follows:

[0032] Step C101: Obtain the abnormal signal delay of historical abnormal network data, and repeat step B104 to calculate the historical abnormal signal delay standard value.

[0033] Step C102: Divide the number of historical abnormal signal delay standard values ​​according to a fixed ratio to obtain a training set, a validation set, and a test set;

[0034] Step C103: Obtain the untrained anomaly detection model, input the training set into the anomaly detection model for training, and set the number of training rounds to a fixed number. Then, input the validation set into the anomaly detection model. The anomaly detection model performs anomaly detection on the historical anomaly signal delay standard values ​​in the validation set, detects the model validation anomaly type, and then compares the model validation anomaly type with the anomaly type of the values ​​in the historical anomaly network data.

[0035] Step C104: If the anomaly type is the same as the anomaly type verified by the model, the anomaly detection model is determined to be correct, and the anomaly detection model outputs one.

[0036] If the anomaly type is different from the anomaly type verified in the model, the anomaly detection model is judged to be wrong and the anomaly detection model outputs zero.

[0037] Step C105: Count the number of correct anomaly detection model outputs a value of 1, and then divide the number of correct outputs by the number of historical anomaly signal delay standard values ​​in the validation set to calculate the first accuracy of the anomaly detection model.

[0038] Step C106: When the first accuracy rate is greater than the first accuracy rate threshold, proceed to the next step;

[0039] When the first accuracy is less than or equal to the first accuracy threshold, historical abnormal network data is obtained again and steps C101 to C103 are repeated to train the anomaly detection model until the first accuracy is greater than the first accuracy threshold.

[0040] Step C107: Input the test set into the anomaly detection model. The anomaly detection model detects the test set and then obtains the corresponding output result based on the judgment result of the anomaly detection model. The number of output results that are one is divided by the number of historical anomaly signal delay standard values ​​in the test set to calculate the second accuracy of the anomaly detection model.

[0041] If the second accuracy rate is greater than the second accuracy rate threshold, the anomaly detection model is considered to have completed training.

[0042] If the second accuracy rate is less than or equal to the second accuracy rate threshold, then repeat step C106 until the second accuracy rate is greater than the second accuracy rate threshold; wherein the second accuracy rate threshold is greater than the first accuracy rate threshold, and the first accuracy rate threshold is greater than zero.

[0043] Furthermore, the detection process of the online detection module is as follows:

[0044] The set of abnormal values ​​is input into the anomaly detection model, which analyzes the standard value of the abnormal signal delay within the set of abnormal values ​​to obtain the anomaly type corresponding to the standard value of the abnormal signal delay.

[0045] When the number of identical anomalies is greater than or equal to a fixed number, the online detection module issues an anomaly warning signal and records the corresponding network edge node as an abnormal network edge node; when the number of identical anomalies is less than a fixed number, no action is taken.

[0046] Furthermore, the control process of the intelligent control module is as follows:

[0047] When the anomaly type of the abnormal network edge node is communication link anomaly, the power communication network is switched from the communication link to the abnormal network edge node to the backup communication link, and the abnormal network edge node is restarted; if the anomaly type of the abnormal network edge node is still communication link anomaly after restarting, a security check is performed on the abnormal network edge node; if the abnormal network edge node has no anomaly type after restarting, no operation is performed.

[0048] Furthermore, the control process of the intelligent control module is as follows:

[0049] When the anomaly type of the abnormal network edge node is node device anomaly, hardware detection is performed on the abnormal network edge node; if any unit in the abnormal network edge node is anomaly, the corresponding unit is restarted, and during the restart process, the number of communication tasks of the power communication network is limited to the minimum number of communication tasks; if the abnormal network edge node is a complete failure, the communication tasks required by the abnormal network edge node are taken over by the backup network edge node, and the abnormal network edge node is fault detected.

[0050] When the anomaly type of the abnormal network edge node is data transmission anomaly, the abnormal network edge node will sort the lost communication tasks according to their time sequence and then resend the lost communication tasks.

[0051] Compared with the prior art, the beneficial effects of the present invention are:

[0052] 1. On the one hand, the present invention constructs an abnormal value set of network edge nodes based on rated parameter data and real-time node power data through an anomaly extraction module, and sends the constructed abnormal value set to the online detection module. On the other hand, the anomaly prediction module is used to train the anomaly detection model in combination with historical abnormal network data, and the trained anomaly detection model is sent to the online detection module. Finally, the online detection module detects the abnormal value set in combination with the anomaly detection model, and sends the detected anomaly warning signal and the anomaly type of the abnormal network edge node to the intelligent control module, thereby realizing intelligent detection of the power communication network.

[0053] 2. This invention uses an intelligent control module to intelligently control abnormal network edge nodes based on the type of anomaly, thereby realizing trend prediction and intelligent control of power communication network anomalies. Attached Figure Description

[0054] To facilitate understanding by those skilled in the art, the present invention will be further described below with reference to the accompanying drawings.

[0055] Figure 1 This is an overall system block diagram of the present invention;

[0056] Figure 2 This is an example diagram illustrating the relationship between network edge nodes and the cloud system in this invention;

[0057] Figure 3 This is a flowchart of the method of the present invention. Detailed Implementation

[0058] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0059] Example 1, please refer to Figure 1 and Figure 2As shown, the technical solution provided by the present invention is: an artificial intelligence-based power communication big data anomaly prediction and intelligent control system. This system predicts anomaly trends and intelligently controls possible anomalies in the power communication network based on the data generated by the power communication network. The power communication network consists of a cloud system and multiple network edge nodes, including a data acquisition module, an anomaly extraction module, an anomaly prediction module, an online detection module, an intelligent control module, and a database module.

[0060] In this embodiment, the data acquisition module is used to collect real-time node power data of all network edge nodes in the power communication network at different sampling points, and send the real-time node power data to the anomaly extraction module;

[0061] It should be specifically noted that the real-time node power data specifically refers to the real-time signal delay, real-time signal packet loss rate, and real-time signal disconnection rate of all network edge nodes in the power communication network at different sampling points. The cloud system predicts abnormal conditions of the network edge nodes based on the real-time node power data collected by the network edge nodes. Real-time node power data at different sampling points can be collected by monitoring equipment deployed in the power communication network. In this embodiment, the sampling point is specifically a time node.

[0062] Furthermore, the database module is used to store the rated parameter data of all network edge nodes and send the rated parameter data to the anomaly extraction module; wherein, the rated parameter data specifically refers to the rated signal delay, rated signal packet loss rate and rated signal disconnection rate of all network edge nodes in the power communication network;

[0063] In this embodiment, the anomaly extraction module is used to construct a set of abnormal values ​​for network edge nodes based on rated parameter data and real-time node power data. The construction process is as follows:

[0064] Step B101: Traverse the real-time signal delay corresponding to the real-time node power data at all sampling points;

[0065] If no real-time signal delay is received at any sampling point or the real-time signal delay at any sampling point is empty, then the corresponding real-time signal delay is determined to be a missing value, and proceed to step B102.

[0066] If all sampling points receive real-time signal delay, and the real-time signal delay of all sampling points is not empty, then the corresponding real-time signal delay is determined to be normal, and step B104 is entered.

[0067] Among them, the real-time signal delay being null is manifested as the corresponding real-time signal delay in the real-time node power data being null, NaN, or None, etc.

[0068] It should be noted that the subsequent steps and modules of this method only analyze the real-time signal delay in the real-time node power data. The analysis process of real-time signal packet loss rate, real-time signal disconnection rate and node real-time load rate is the same as that of real-time signal delay, and will not be repeated here.

[0069] Step B102: Obtain the real-time signal delay V1 of the first sampling point and the real-time signal delay V3 of the third sampling point. If the real-time signal delay V2 is not collected at the second sampling point, subtract the time node corresponding to the first sampling point from the time node corresponding to the second sampling point to calculate the first time interval between the second sampling point and the first sampling point. At the same time, subtract the time node corresponding to the second sampling point from the time node corresponding to the third sampling point to calculate the second time interval between the third sampling point and the second sampling point. Divide the first time interval by the second time interval to calculate the time interval ratio JG.

[0070] Step B103: If the time interval ratio is greater than one, then the second sampling point is determined to be close to the third sampling point. The real-time signal delay corresponding to the second sampling point is V2 = [(V1 + V3) / 2] × JG.

[0071] If the time interval ratio is less than one, then the second sampling point is determined to be close to the first sampling point, and the real-time signal delay corresponding to the second sampling point is V2=[(V1+V3) / 2]×JG;

[0072] If the calculation result is equal to one, it is determined that the time interval between the second sampling point and the first and third sampling points is the same, and the real-time signal delay corresponding to the second sampling point is V2 = (V1 + V3) / 2;

[0073] Step B104: Obtain the real-time signal delay corresponding to any sampling point of the network edge node, divide the real-time signal delay by the rated signal delay, and calculate the standard value of the signal delay of the corresponding sampling point. Similarly, calculate the standard value of the signal packet loss rate and the standard value of the signal disconnection rate, and replace the values ​​in the real-time node power data with the standard values ​​calculated above.

[0074] Step B105: Subtract the standard value of the signal delay corresponding to the previous sampling point from the standard value of the signal delay corresponding to the next sampling point to calculate the signal delay difference between adjacent sampling points. Similarly, calculate the difference in signal packet loss rate and signal disconnection rate between adjacent sampling points.

[0075] Step B106: If the signal delay difference between adjacent sampling points is greater than or equal to the first difference threshold, or the signal packet loss rate difference is greater than or equal to the second difference threshold, or the signal disconnection rate difference is greater than or equal to the third difference threshold, then the corresponding adjacent sampling points are determined to be abnormal sampling points, and proceed to step B108.

[0076] If the signal delay difference between adjacent sampling points is less than the first difference threshold, the signal packet loss rate difference is less than the second difference threshold, and the signal disconnection rate difference is less than the third difference threshold, then proceed to the next step.

[0077] There is no comparative relationship between the first, second, and third difference thresholds; they are only used for comparison with their corresponding differences. The derivation process of the first difference threshold is as follows: The signal delay differences are summed and averaged to calculate the average signal delay difference between adjacent sampling points. Then, the standard deviation of the signal delay between adjacent sampling points is calculated using the standard deviation formula. At this point, the first difference threshold = average signal delay difference + k × standard deviation of signal delay. In the actual calculation, k = 3. Similarly, the second and third difference thresholds between adjacent sampling points are calculated. There is no comparative relationship between the first, second, and third difference thresholds; they are only used for comparison with their corresponding differences.

[0078] Step B107: When the standard value of signal delay of any sampling point among the adjacent sampling points is greater than one, or the standard value of signal packet loss rate is greater than one, or the standard value of signal disconnection rate is greater than one, the corresponding sampling point is determined to be an abnormal sampling point, and the process proceeds to step B108.

[0079] When the standard value of signal delay of all adjacent sampling points is less than or equal to one, the standard value of signal packet loss rate is less than or equal to one, and the standard value of signal disconnection rate is less than or equal to one, the corresponding adjacent sampling points are determined to be normal sampling points.

[0080] Step B108: Record the standard value of signal delay at abnormal sampling points as the standard value of abnormal signal delay, the standard value of signal packet loss rate as the standard value of abnormal signal packet loss rate, and the standard value of signal disconnection rate as the standard value of abnormal signal disconnection rate. Then, summarize the standard values ​​of abnormal signal delay, abnormal signal packet loss rate, and abnormal signal disconnection rate into a set of abnormal values ​​for network edge nodes.

[0081] The anomaly extraction module sends the set of abnormal values ​​to the online detection module.

[0082] As a further embodiment, the database module is also used to store historical abnormal network data of all network edge nodes and send it to the anomaly prediction module;

[0083] Specifically, the historical abnormal network data includes abnormal signal delay, abnormal signal packet loss rate, and abnormal signal disconnection rate of all network edge nodes in the power communication network. The database stores multiple sets of historical abnormal network data, but this embodiment selects only one set for training the anomaly detection model. All values ​​in the historical abnormal network data are labeled with their corresponding anomaly types, including communication link anomalies, node device anomalies, and data transmission anomalies. It should be noted that the anomaly prediction module is deployed in a cloud system.

[0084] In this embodiment, the anomaly prediction module is used to train the anomaly detection model based on historical anomaly network data. The training process is as follows:

[0085] Step C101: Obtain the abnormal signal delay of historical abnormal network data, and repeat step B104 to calculate the historical abnormal signal delay standard value.

[0086] Step C102: Divide the number of historical abnormal signal delay standard values ​​according to a fixed ratio to obtain a training set, a validation set, and a test set;

[0087] In practice, this embodiment only inputs the historical abnormal signal delay standard values ​​into the anomaly detection model for training. The training process for other data in the historical abnormal network data is the same as the training process for training the anomaly detection model using the historical abnormal signal delay standard values, and will not be elaborated here. When dividing according to a fixed ratio, the ratio can be 7:2:1. For example, if there are 100 historical abnormal signal delay standard values, 70 of them are used as the training set, 20 as the validation set, and 10 as the test set. The training set is used as training samples to train the anomaly detection model. The validation set is used to verify the training status of the anomaly detection model. The detection results of the validation set are used to determine whether the parameters of the anomaly detection model need to be adjusted and whether training can be stopped. The test set is not used in the training process of the anomaly detection model, but only to test the training results of the anomaly detection model.

[0088] It should be specifically noted that the anomaly detection model framework used in this embodiment is based on a spatiotemporal deep learning framework, which can adaptively adjust the model structure and combine the representation learning capability of deep learning with the inference capability of statistical learning.

[0089] Step C103: Obtain the untrained anomaly detection model, input the training set into the anomaly detection model for training, and set the number of training rounds to a fixed number. Then, input the validation set into the anomaly detection model. The anomaly detection model performs anomaly detection on the historical anomaly signal delay standard values ​​in the validation set, detects the model validation anomaly type, and then compares the model validation anomaly type with the anomaly type of the values ​​in the historical anomaly network data.

[0090] Specifically, the process of training the anomaly detection model using training, validation, and test sets is existing technology and will not be elaborated here; in practice, the training rounds can be up to 50.

[0091] Step C104: If the anomaly type is the same as the anomaly type verified by the model, the anomaly detection model is determined to be correct, and the anomaly detection model outputs one.

[0092] If the anomaly type is different from the anomaly type verified in the model, the anomaly detection model is judged to be wrong and the anomaly detection model outputs zero.

[0093] Step C105: Count the number of correct anomaly detection model outputs a value of 1, and then divide the number of correct outputs by the number of historical anomaly signal delay standard values ​​in the validation set to calculate the first accuracy of the anomaly detection model.

[0094] Step C106: When the first accuracy rate is greater than the first accuracy rate threshold, proceed to the next step;

[0095] When the first accuracy is less than or equal to the first accuracy threshold, historical abnormal network data is obtained again and steps C101 to C103 are repeated to train the anomaly detection model until the first accuracy is greater than the first accuracy threshold.

[0096] Step C107: Input the test set into the anomaly detection model. The anomaly detection model detects the test set and then obtains the corresponding output result based on the judgment result of the anomaly detection model. The number of output results that are one is divided by the number of historical anomaly signal delay standard values ​​in the test set to calculate the second accuracy of the anomaly detection model.

[0097] If the second accuracy rate is greater than the second accuracy rate threshold, the anomaly detection model is considered to have completed training.

[0098] If the second accuracy rate is less than or equal to the second accuracy rate threshold, repeat step C106 until the second accuracy rate is greater than the second accuracy rate threshold.

[0099] Among them, the second accuracy threshold is greater than the first accuracy threshold, and the first accuracy threshold is greater than zero;

[0100] The anomaly prediction module sends the anomaly detection model to the online detection module.

[0101] Specifically, the online detection module is used to detect abnormal value sets through an anomaly detection model. The detection process is as follows:

[0102] Step D101: Input the set of abnormal values ​​into the anomaly detection model. The anomaly detection model analyzes the standard value of the abnormal signal delay within the set of abnormal values ​​to obtain the anomaly type corresponding to the standard value of the abnormal signal delay.

[0103] Among them, existing technologies use anomaly detection models to detect existing data and determine the corresponding anomaly types.

[0104] Step D102: When the number of the same anomaly type is greater than or equal to a fixed number, the online detection module issues an anomaly warning signal and records the corresponding network edge node as an abnormal network edge node.

[0105] If the number of the same exception type is less than a fixed number, no action is taken.

[0106] The online detection module sends abnormal early warning signals and abnormal network edge node anomaly types to the intelligent control module.

[0107] In this embodiment, the intelligent control module is used to intelligently control abnormal network edge nodes according to the anomaly type. The control process is as follows:

[0108] Step E101: When the anomaly type of the abnormal network edge node is communication link anomaly, the power communication network is switched from the communication link to which the abnormal network edge node belongs to the backup communication link, and the abnormal network edge node is restarted.

[0109] If the abnormality type of the abnormal network edge node is still communication link abnormal after the restart is completed, the staff will conduct a security test on the abnormal network edge node.

[0110] If the abnormal network edge node has no abnormal type after the restart is completed, no operation will be performed;

[0111] Step E102: When the anomaly type of the abnormal network edge node is node device anomaly, perform hardware detection on the abnormal network edge node.

[0112] If any unit in the abnormal network edge node is abnormal, the corresponding unit will be restarted, and during the restart process, the number of communication tasks in the power communication network will be limited to the minimum number of communication tasks.

[0113] If the entire abnormal network edge node fails, the communication tasks required by the abnormal network edge node will be taken over by the backup network edge node, and the staff will conduct fault detection on the abnormal network edge node.

[0114] The power-saving equipment at the network edge node includes a main control processing unit, a memory storage unit, a signal acquisition unit, a communication network unit, and a power protection unit.

[0115] Step E103: When the anomaly type of the abnormal network edge node is data transmission anomaly, the abnormal network edge node sorts the lost communication tasks according to their time sequence and then resends the lost communication tasks.

[0116] Example 2, based on another concept of the same invention, proposes a method for anomaly prediction and intelligent control of power communication big data based on artificial intelligence, including the following steps:

[0117] Step S100: Collect real-time node power data of all network edge nodes in the power communication network;

[0118] Step S200: Construct a set of abnormal values ​​for network edge nodes based on rated parameter data and real-time node power data;

[0119] Step S300: Train the anomaly detection model based on historical anomaly network data;

[0120] Step S400: The abnormal value set is detected by the anomaly detection model to obtain the abnormal network edge nodes and the corresponding anomaly types.

[0121] Step S500: Intelligent control is performed on abnormal network edge nodes according to the anomaly type.

[0122] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. An artificial intelligence-based power communication big data anomaly prediction and intelligent control system, characterized in that, The system includes a data acquisition module, an anomaly extraction module, an anomaly prediction module, an online detection module, an intelligent control module, and a database module. The data acquisition module is used to collect real-time node power data of all network edge nodes in the power communication network at different sampling points and send it to the anomaly extraction module. The database module is used to store the rated parameter data of all network edge nodes and send it to the anomaly extraction module. The anomaly extraction module is used to construct an anomaly value set of network edge nodes based on the rated parameter data and real-time node power data, and send the constructed anomaly value set to the online detection module. The specific process for constructing the set of abnormal values ​​is as follows: Step B101: Traverse the real-time signal delay corresponding to the real-time node power data at all sampling points; If no real-time signal delay is received at any sampling point or the real-time signal delay at any sampling point is empty, then the corresponding real-time signal delay is determined to be a missing value, and proceed to step B102. If all sampling points receive real-time signal delay, and the real-time signal delay of all sampling points is not empty, then the corresponding real-time signal delay is determined to be normal, and step B104 is entered. Step B102: Obtain the real-time signal delay V1 of the first sampling point and the real-time signal delay V3 of the third sampling point. If the real-time signal delay V2 is not collected at the second sampling point, subtract the time node corresponding to the first sampling point from the time node corresponding to the second sampling point to calculate the first time interval between the second sampling point and the first sampling point. At the same time, subtract the time node corresponding to the second sampling point from the time node corresponding to the third sampling point to calculate the second time interval between the third sampling point and the second sampling point. Divide the first time interval by the second time interval to calculate the time interval ratio JG. Step B103: If the time interval ratio is greater than one, then the second sampling point is determined to be close to the third sampling point. The real-time signal delay corresponding to the second sampling point is V2 = [(V1 + V3) / 2] × JG. If the time interval ratio is less than one, then the second sampling point is determined to be close to the first sampling point, and the real-time signal delay corresponding to the second sampling point is V2=[(V1+V3) / 2]×JG; If the calculation result is equal to one, it is determined that the time interval between the second sampling point and the first and third sampling points is the same, and the real-time signal delay corresponding to the second sampling point is V2 = (V1 + V3) / 2; Step B104: Obtain the real-time signal delay corresponding to any sampling point of the network edge node, divide the real-time signal delay by the rated signal delay, and calculate the standard value of the signal delay of the corresponding sampling point. Similarly, calculate the standard value of the signal packet loss rate and the standard value of the signal disconnection rate, and replace the values ​​in the real-time node power data with the calculated standard values. Step B105: Subtract the standard value of the signal delay corresponding to the previous sampling point from the standard value of the signal delay corresponding to the next sampling point to calculate the signal delay difference between adjacent sampling points. Similarly, calculate the difference in signal packet loss rate and signal disconnection rate between adjacent sampling points. Step B106: If the signal delay difference between adjacent sampling points is greater than or equal to the first difference threshold, or the signal packet loss rate difference is greater than or equal to the second difference threshold, or the signal disconnection rate difference is greater than or equal to the third difference threshold, then the corresponding adjacent sampling points are determined to be abnormal sampling points, and proceed to step B108. If the signal delay difference between adjacent sampling points is less than the first difference threshold, the signal packet loss rate difference is less than the second difference threshold, and the signal disconnection rate difference is less than the third difference threshold, then proceed to the next step. Step B107: When the standard value of signal delay of any sampling point among the adjacent sampling points is greater than one, or the standard value of signal packet loss rate is greater than one, or the standard value of signal disconnection rate is greater than one, the corresponding sampling point is determined to be an abnormal sampling point, and the process proceeds to step B108. When the standard value of signal delay of all adjacent sampling points is less than or equal to one, the standard value of signal packet loss rate is less than or equal to one, and the standard value of signal disconnection rate is less than or equal to one, the corresponding adjacent sampling points are determined to be normal sampling points. Step B108: Record the standard value of signal delay at abnormal sampling points as the standard value of abnormal signal delay, the standard value of signal packet loss rate as the standard value of abnormal signal packet loss rate, and the standard value of signal disconnection rate as the standard value of abnormal signal disconnection rate. Then, summarize the standard values ​​of abnormal signal delay, abnormal signal packet loss rate, and abnormal signal disconnection rate into a set of abnormal values ​​for network edge nodes. The database module is also used to store historical abnormal network data of all network edge nodes and send it to the anomaly prediction module; the anomaly prediction module is used to train the anomaly detection model based on the historical abnormal network data, and the trained anomaly detection model is sent to the online detection module. The online detection module is used to detect abnormal value sets through an anomaly detection model, and sends the detected anomaly warning signals and anomaly types of abnormal network edge nodes to the intelligent control module; the intelligent control module is used to intelligently control the abnormal network edge nodes according to the anomaly type.

2. The artificial intelligence-based power communication big data anomaly prediction and intelligent control system according to claim 1, characterized in that, Real-time node power data refers to the real-time signal delay, real-time signal packet loss rate, and real-time signal disconnection rate of all network edge nodes in the power communication network at different sampling points.

3. The artificial intelligence-based power communication big data anomaly prediction and intelligent control system according to claim 2, characterized in that, The rated parameter data are the rated signal delay, rated signal packet loss rate, and rated signal disconnection rate of all network edge nodes in the power communication network.

4. The artificial intelligence-based power communication big data anomaly prediction and intelligent control system according to claim 1, characterized in that, Historical abnormal network data includes abnormal signal delay, abnormal signal packet loss rate, and abnormal signal disconnection rate of all network edge nodes in the power communication network.

5. The artificial intelligence-based power communication big data anomaly prediction and intelligent control system according to claim 4, characterized in that, The training process of the anomaly detection model is as follows: Step C101: Obtain the abnormal signal delay of historical abnormal network data, and repeat step B104 to calculate the historical abnormal signal delay standard value. Step C102: Divide the number of historical abnormal signal delay standard values ​​according to a fixed ratio to obtain a training set, a validation set, and a test set; Step C103: Obtain the untrained anomaly detection model, input the training set into the anomaly detection model for training, and set the number of training rounds to a fixed number. Then, input the validation set into the anomaly detection model. The anomaly detection model performs anomaly detection on the historical anomaly signal delay standard values ​​in the validation set, detects the model validation anomaly type, and then compares the model validation anomaly type with the anomaly type of the values ​​in the historical anomaly network data. Step C104: If the anomaly type is the same as the anomaly type verified by the model, the anomaly detection model is determined to be correct, and the anomaly detection model outputs one. If the anomaly type is different from the anomaly type verified in the model, the anomaly detection model is judged to be wrong and the anomaly detection model outputs zero. Step C105: Count the number of correct anomaly detection model outputs a value of 1, and then divide the number of correct outputs by the number of historical anomaly signal delay standard values ​​in the validation set to calculate the first accuracy of the anomaly detection model. Step C106: When the first accuracy rate is greater than the first accuracy rate threshold, proceed to the next step; When the first accuracy is less than or equal to the first accuracy threshold, historical abnormal network data is obtained again and steps C101 to C103 are repeated to train the anomaly detection model until the first accuracy is greater than the first accuracy threshold. Step C107: Input the test set into the anomaly detection model. The anomaly detection model detects the test set and then obtains the corresponding output result based on the judgment result of the anomaly detection model. The number of output results that are one is divided by the number of historical anomaly signal delay standard values ​​in the test set to calculate the second accuracy of the anomaly detection model. If the second accuracy rate is greater than the second accuracy rate threshold, the anomaly detection model is considered to have completed training. If the second accuracy rate is less than or equal to the second accuracy rate threshold, then repeat step C106 until the second accuracy rate is greater than the second accuracy rate threshold; wherein the second accuracy rate threshold is greater than the first accuracy rate threshold, and the first accuracy rate threshold is greater than zero.

6. The artificial intelligence-based power communication big data anomaly prediction and intelligent control system according to claim 5, characterized in that, The detection process of the online detection module is as follows: The set of abnormal values ​​is input into the anomaly detection model, which analyzes the standard value of the abnormal signal delay within the set of abnormal values ​​to obtain the anomaly type corresponding to the standard value of the abnormal signal delay. When the number of identical anomalies is greater than or equal to a fixed number, the online detection module issues an anomaly warning signal and records the corresponding network edge node as an abnormal network edge node; when the number of identical anomalies is less than a fixed number, no action is taken.

7. The artificial intelligence-based power communication big data anomaly prediction and intelligent control system according to claim 1, characterized in that, The control process of the intelligent control module is as follows: When the anomaly type of the abnormal network edge node is communication link anomaly, the power communication network is switched from the communication link to the abnormal network edge node to the backup communication link, and the abnormal network edge node is restarted; if the anomaly type of the abnormal network edge node is still communication link anomaly after restarting, a security check is performed on the abnormal network edge node; if the abnormal network edge node has no anomaly type after restarting, no operation is performed.

8. The artificial intelligence-based power communication big data anomaly prediction and intelligent control system according to claim 7, characterized in that, The control process of the intelligent control module is as follows: When the anomaly type of the abnormal network edge node is node device anomaly, hardware detection is performed on the abnormal network edge node; if any unit in the abnormal network edge node is anomaly, the corresponding unit is restarted, and during the restart process, the number of communication tasks of the power communication network is limited to the minimum number of communication tasks. If the entire abnormal network edge node fails, the backup network edge node will take over the communication tasks required by the abnormal network edge node and perform fault detection on the abnormal network edge node. When the anomaly type of the abnormal network edge node is data transmission anomaly, the abnormal network edge node will sort the lost communication tasks according to their time sequence and then resend the lost communication tasks.

Images