Monitoring and control system of 5G communication network based on artificial intelligence

Abstract

The invention discloses a 5G communication network monitoring and control system based on artificial intelligence, which belongs to the technical field of communication and comprises an influence coefficient construction module, a service and performance evaluation module and a performance index monitoring and control module. The system collects a three-dimensional data system to construct a service-network incidence matrix, and calculates resource consumption through a reference-calibration-dynamic influence coefficient; fusing a service level agreement, a user level, a scene value and a resource scarcity adaptation degree dimension, and dynamically evaluating service importance and calculating a performance index weight by adopting an entropy evaluation method; constructing a performance index change curve, introducing a threshold super / threshold drop area and a fluctuation coefficient to calculate a risk value, and generating an exception handling priority sequence to guide targeted management and control; the problems of extensive resource accounting, one-sided evaluation, weight solidification and management and control lagging are solved, 5G network resource scheduling precision, service guarantee dynamic and abnormal response targeting are realized, and the network stability and service quality guarantee capability in a multi-service concurrent scene are remarkably improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of communication, and particularly relates to a monitoring and control system for a 5G communication network based on artificial intelligence. BACKGROUND

[0002] The 5G network carries diversified services such as high-definition video, industrial control and artificial intelligence interaction, and the competition for network resources is intensified, and the difficulty of quality of service guarantee is increasing. The traditional monitoring and control system relies on static threshold alarm and manual experience decision, and has low precision and poor adaptability in service resource consumption quantification, importance evaluation and abnormal response, and cannot realize intelligent and accurate operation and maintenance. The following technical problems are proposed in detail: It is difficult to dynamically establish a precise mapping relationship between the operation parameters of the service and the occupation of the network resources, the resource consumption accounting granularity is coarse, the different influence degrees of different service characteristics on bandwidth, calculation, storage and other resources cannot be reflected in real time, and the network scheduling lacks accurate quantitative basis; The importance evaluation of the service excessively depends on the static indicators such as preset service level agreement priority, and does not dynamically integrate multi-dimensional factors such as user level, scene value and resource scarcity state, and lacks adaptive weight adjustment capability for service distribution changes, and it is difficult to comprehensively identify high-value services; The performance index weight is fixed and unchanged, the risk identification only depends on the single-point threshold over-limit judgment, the index fluctuation trend and cumulative risk are ignored, and when multiple abnormalities occur, there is no scientific priority sorting mechanism, the targeting of the control measures is poor, the response is lagging, and the network stable operation is affected. Therefore, we propose a monitoring and control system for a 5G communication network based on artificial intelligence. SUMMARY

[0003] The purpose of the present application is to provide a monitoring and control system for a 5G communication network based on artificial intelligence to solve the problems proposed in the background.

[0004] To achieve the above purpose, the present application provides the following technical scheme: a monitoring and control system for a 5G communication network based on artificial intelligence, comprising: An influence coefficient construction module: a three-dimensional data system of the running service of the 5G communication network is collected, a service-network associated data matrix is constructed, the base, calibration and dynamic influence coefficients of each fine-grained parameter of the running service and the corresponding network resources are calculated, an influence coefficient library is constructed, and the sub-consumption value and the comprehensive resource consumption value of the running service to each network resource are calculated; A service and performance evaluation module: based on the evaluation dimension of the running service, the importance evaluation value of the running service is calculated, the urgency coefficient of the performance index in the 5G network is analyzed, and the importance evaluation value of the running service and the sub-consumption value of the running service to the network resources corresponding to the performance index are combined to calculate the importance evaluation value and the weight coefficient of the performance index; The performance index monitoring and control module: constructs a performance index change curve, calculates a performance index risk value, marks an abnormal performance index and calculates a corresponding risk overrun value, calculates an abnormal evaluation value in combination with a weight coefficient corresponding to the abnormal index, constructs an abnormal index processing priority sequence, and accordingly sequentially targets the network resources and associated operation services corresponding to each abnormal index for control.

[0005] Preferably, the specific process of constructing the business-network association data matrix is as follows: The three-dimensional data system of all operation services supported by the 5G communication network includes: basic attribute layer parameters, fine-grained running layer parameters, and network association layer parameters. The business-network association data matrix is constructed by aligning the channel data, classifying the parameters, and structuring the integration, with each operation service as the core, and with the business unique identifier, the collection timestamp, the basic attribute layer parameters, the fine-grained running layer parameters, and the network association layer parameters as the columns.

[0006] Preferably, the specific process of calculating the reference, calibration, and dynamic influence coefficients of each fine-grained parameter of the operation service and the corresponding network resources is as follows: For each current operation service and for each fine-grained parameter of the operation service, based on the historical fine-grained parameter values of each sampling period, the historical network resource occupation values corresponding to the fine-grained parameters, the average values of all historical sampling period fine-grained parameter values, the average values of all historical sampling period corresponding network resource occupation values, and the total number of historical sampling periods, a comprehensive analysis is performed to obtain the reference influence coefficient between the fine-grained parameter and the corresponding network resource under the operation service. The real-time values of the fine-grained parameters of the operation service and the real-time occupation values of the corresponding network resources in the current sampling period are obtained, and a comprehensive analysis is performed to obtain the calibration coefficient between the fine-grained parameter and the corresponding network resource under the operation service. The reference influence coefficient and the calibration coefficient between the fine-grained parameter and the corresponding network resource under the operation service are combined, and a comprehensive analysis is performed in combination with the preset reference coefficient weight to obtain the dynamic influence coefficient between the fine-grained parameter and the corresponding network resource under the operation service.

[0007] Preferably, the specific process of constructing the influence coefficient library is as follows: All services supported by the communication network are integrated into the influence coefficient library according to the unique mapping relationship of business-fine-grained parameter-network resource and the corresponding coefficients. Each service currently running in the influence coefficient library has a real-time calibrated dynamic influence coefficient corresponding to each fine-grained parameter-network resource combination thereof, and each service that has appeared historically but is not currently running has a baseline influence coefficient fitted based on historical data reserved for each fine-grained parameter-network resource combination thereof; when the service is accessed for running again, the dynamic influence coefficient is automatically calculated using the newly collected real-time fine-grained parameter value and the real-time occupation value of the corresponding network resource, and the baseline influence coefficient is updated to the dynamic influence coefficient.

[0008] Preferably, the specific process of accounting for the itemized consumption value and the comprehensive resource consumption value of the running service to each network resource is as follows: For each currently running service, the uplink and downlink real-time traffic is first obtained and the uplink and downlink traffic proportion is calculated, and then the uplink and downlink traffic correction factor is obtained by combining the baseline value of the conventional uplink and downlink traffic proportion of the service and the preset influence coefficient of the uplink and downlink traffic proportion; For each network resource of the current service, the real-time value of each fine-grained parameter corresponding to the network resource is combined with the dynamic influence coefficient of each fine-grained parameter matching the network resource in the influence coefficient library to calculate the itemized consumption value of the current service to the network resource; A preset weight coefficient is assigned to each network resource of the current service, each network resource itemized consumption value is multiplied by the corresponding weight coefficient, and the sum of all the products is summed to obtain the comprehensive resource consumption value of the currently running service.

[0009] Preferably, the specific process of calculating the importance evaluation value of the running service is as follows: The four evaluation dimensions and core parameters of the currently running service are extracted, which are: service level agreement priority, user level coefficient, scene value coefficient, and resource scarcity adaptation degree; The calculation process of the resource scarcity adaptation degree is as follows: Based on the network hardware configuration file and the network comprehensive resource maximum carrying value preset according to the 5G protocol, the sum of the comprehensive resource consumption values of all currently running services is subtracted to obtain the network comprehensive resource remaining amount; The comprehensive resource consumption value of the current running service and the network remaining resource amount are then comprehensively analyzed to obtain the resource scarcity adaptation degree; After the core parameters of each dimension are normalized and de-dimensioned, the dynamic weight of each dimension is calculated based on the core parameter entropy value corresponding to each dimension; and the normalized core parameter actual value is comprehensively analyzed based on the dynamic weight to obtain the importance evaluation value of the current running service.

[0010] Preferably, the specific process of calculating the importance evaluation value and the weight coefficient of the performance indicator is as follows: Based on the network association layer, the performance indicators corresponding to all network resources of the 5G network are obtained; For each performance indicator, the upper limit of the agreed delay of its associated service level agreement, the current network real-time delay, and a preset timeout penalty coefficient are combined to obtain an urgency coefficient of the performance indicator; For each performance indicator, the importance evaluation value of the performance indicator is obtained by combining the itemized consumption values of all current operating services on the corresponding network resources, the importance evaluation values of the operating services, and the urgency coefficient of the performance indicator. The importance evaluation values of all performance indicators are linearly normalized to obtain the weight coefficient of each indicator. An operating service update scenario library is established, and when any update scenario in the library occurs, the weight coefficient update process is automatically triggered, and the calculation is restarted when the update condition is met, to ensure that it matches the current service distribution.

[0011] Preferably, the specific process of constructing the performance indicator change curve and calculating the performance indicator risk value is as follows: For each performance indicator, the value of the performance indicator is taken as the ordinate, and the time is taken as the abscissa to construct a two-dimensional rectangular coordinate system. The values of the performance indicators corresponding to each sampling period are obtained in real time, and are marked in the two-dimensional rectangular coordinate system to obtain a plurality of data points. According to the time sequence, the adjacent data points are connected by a smooth curve to obtain a performance indicator change curve. Two performance indicator threshold lines are preset, namely a threshold overshoot threshold line and a threshold drop threshold line. The area surrounded by the performance indicator change curve and the threshold overshoot threshold line is defined as the threshold overshoot area, and the area surrounded by the performance indicator change curve and the threshold drop threshold line is defined as the threshold drop area. For each sampling period, the sampling period at which the performance indicator weight was last updated is taken as the starting point, and the current sampling period is taken as the ending point to define the current network state monitoring window. The sum of all threshold overshoot areas in the current network state monitoring window is calculated and defined as the threshold overshoot total, and the sum of all threshold drop areas in the current network state monitoring window is calculated and defined as the threshold drop total. The standard deviation value of the performance indicator change curve in the current network state monitoring window is calculated to obtain a performance indicator fluctuation coefficient. The threshold overshoot total, the threshold drop total, and the performance indicator fluctuation coefficient in the current network state monitoring window are comprehensively analyzed to obtain the performance indicator risk value of the current network state monitoring window.

[0012] Preferably, an abnormal indicator processing priority sequence is constructed, and the specific process of sequentially targeting and controlling the network resources and associated operating services corresponding to each abnormal indicator is as follows: The performance index risk of the current network state monitoring window is compared with the corresponding threshold value, and if greater than or equal to the corresponding threshold value, the performance index is marked as an abnormal index; For each abnormal index, the risk value of the abnormal index is subtracted from the corresponding preset threshold value to obtain an abnormal index risk overrun value; The abnormal evaluation value is obtained by multiplying the risk overrun value corresponding to the abnormal index by the weight coefficient of the corresponding performance index; All abnormal indexes are integrated and sorted in descending order based on the corresponding abnormal evaluation values to obtain an abnormal index processing priority sequence; The abnormal index processing priority sequence is sent to the operation and maintenance terminal, and the operation and maintenance terminal carries out targeted management and control on the network resources and associated operation and maintenance services corresponding to each abnormal index based on the abnormal index processing priority sequence.

[0013] Compared with the prior art, the beneficial effects of the present application are: (1) The monitoring and control system for the 5G communication network based on artificial intelligence upgrades the mapping relationship between the business fine-grained operation parameters and the network resource occupation from static empirical estimation to real-time adaptive calculation by constructing a three-dimensional data matrix and a dynamic influence coefficient library, so that the consumption accounting of bandwidth, calculation, storage and other resources is more accurate and timely, providing a reliable quantitative basis for network scheduling, and effectively solving the fundamental problem of the traditional scheme of coarse granularity and poor timeliness.

[0014] (2) The monitoring and control system for the 5G communication network based on artificial intelligence, by fusing a four-dimensional evaluation system of service level agreement priority, user level coefficient, scene value coefficient and resource scarcity adaptation degree, and combining with the entropy method to dynamically allocate weights, makes the operation service importance evaluation value and the network load state real-time coupled; the resource scarcity adaptation degree quantifies the correlation between business resource occupation and network residual capacity, and the operation service updates the scene library to automatically trigger weight update, ensuring that the evaluation system evolves dynamically with business distribution. The performance index importance evaluation value generated thereby provides a comprehensive and dynamic decision basis for high-value business identification.

[0015] (3) The monitoring and control system for the 5G communication network based on artificial intelligence, by constructing a performance index change curve, introducing threshold overrun area, threshold reduction area and performance index fluctuation coefficient to establish a three-dimensional risk quantification model, and delimiting the periodical network state monitoring window to realize the time-sequential and accurate identification of risks; based on the abnormal index risk overrun value and the weight coefficient, the abnormal evaluation value is calculated, and the abnormal index processing priority sequence is generated to guide the operation and maintenance terminal to implement targeted management and control; this mechanism changes the disordered disposal of multiple abnormal concurrent into ordered and accurate intervention, realizes the closed-loop management from risk identification, evaluation sorting to resource regulation, and improves the stability of 5G network and the service quality guarantee capability. BRIEF DESCRIPTION OF DRAWINGS

[0016] Figure 1 Flow chart of the present application; Figure 2 Performance index change curve of the present application. DETAILED DESCRIPTION

[0017] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work belong to the scope of protection of the present application.

[0018] Embodiment one; Please refer to Figure 1 Figure 2 The present application provides a monitoring and control system for 5G communication network based on artificial intelligence, comprising: An influence coefficient construction module: collects the three-dimensional data system of all running services of the 5G communication network, constructs a service-network correlation data matrix, calculates the benchmark, calibration and dynamic influence coefficients of each fine-grained parameter of the running service and the corresponding network resources, constructs an influence coefficient library, calculates the uplink and downlink traffic correction factors, calculates the itemized consumption value and comprehensive resource consumption value of the service to each network resource, and the specific process is as follows: Collect the three-dimensional data system of all running services of the 5G communication network, including: Basic attribute layer: service type (video, industrial control, GenAI interaction, etc.), user QoS level; SLA protocol agreement parameters (upper limit of delay, packet loss rate threshold, etc.); Fine-grained running layer: code rate, definition, frame rate, instruction period, transmission frequency, uplink material size, generation complexity, etc.; Network correlation layer: bandwidth occupancy rate, delay fluctuation value, packet loss rate, uplink and downlink real-time traffic, number of access terminals, link occupancy state, etc.; Taking each running service as the core, taking the service unique identifier, the collection timestamp, the basic attribute layer parameters, the fine-grained running layer parameters and the network correlation layer parameters as columns, and through the collection channel data alignment-parameter classification mapping-structured integration method, a service-network correlation data matrix is constructed; Further, the construction method of the service-network correlation data matrix is as follows: Assign a unique identifier to each running service as the core index of each row of the matrix; Align the timestamps of the three-dimensional data system collected through the above-mentioned cooperative collection mechanism; ​The three-dimensional data system after timestamp alignment is classified according to the basic attribute layer, the fine-grained running layer and the network association layer, and is respectively mapped to the corresponding column field; Based on the parameter value range agreed in the 5G network communication protocol, the mapped data is subjected to missing value completion (using the mean value of the same type of business at the same period) and abnormal value elimination (data exceeding the protocol value range ± 3σ), to form a structured business-network association data matrix: {operating business identifier, collection timestamp, business type, user QoS level, SLA protocol parameter, fine-grained running parameter, business occupied base station port, uplink and downlink real-time traffic, access terminal quantity, link occupation state, etc.}.

[0019] For each current operating business and for each fine-grained parameter of the operating business, the same historical data as the current operating business in the past several sampling periods are retrieved, including: the historical value of the fine-grained parameter and the historical occupation value of the network resource corresponding to the fine-grained parameter; wherein the network resource includes: bandwidth resource, computing resource, storage resource, etc. The fine-grained parameter historical value and the corresponding network resource historical occupation value of each sampling period in the historical data are substituted into the formula: to obtain the reference influence coefficient between the fine-grained parameter and the corresponding network resource under the operating business ; wherein i is the label of the historical sampling period, is the fine-grained parameter value of the i-th historical sampling period; is the average value of the fine-grained parameter values of all historical sampling periods; is the occupation value of the network resource corresponding to the fine-grained parameter of the i-th historical sampling period; is the average value of the network resource occupation values corresponding to the fine-grained parameter of all historical sampling periods; and n is the total number of historical sampling periods. Based on the business-network association data matrix, the real-time value of the fine-grained parameter of the operating business in the current sampling period and the real-time occupation value of the corresponding network resource and the corresponding reference influence coefficient are obtained, and substituted into the formula: to obtain the calibration coefficient between the fine-grained parameter and the corresponding network resource under the operating business; By substituting the reference influence coefficient and the calibration coefficient between the fine-grained parameter and the corresponding network resource under the operating business into the formula: , the dynamic influence coefficient between the fine-grained parameter and the corresponding network resource under the operating business is obtained; wherein a is a preset reference coefficient weight. All services supported by the communication network are integrated according to the unique mapping relationship of service-fine-grained parameter-network resource to form an influence coefficient library, specifically: For each service currently running in the 5G communication network in the influence coefficient library: for each fine-grained parameter-network resource combination contained in the service, there is a dynamically calibrated dynamic influence coefficient corresponding to it; For each service that has appeared in the past but is not currently running in the influence coefficient library: for each fine-grained parameter-network resource combination contained in the service, the baseline influence coefficient fitted based on the historical data of the service is temporarily retained; when the service is accessed and run again, the dynamic influence coefficient is automatically calculated based on the newly collected corresponding real-time fine-grained parameter value and the corresponding real-time network resource occupation value, and the baseline influence coefficient is updated to the dynamic influence coefficient; For each currently running service, obtain the uplink and downlink real-time traffic of the running service, and calculate the uplink and downlink traffic ratio γ (γ = uplink real-time traffic ÷ (uplink real-time traffic + downlink real-time traffic)); Using the formula: , the uplink and downlink traffic correction factor β of the running service is obtained; Wherein, is the conventional uplink and downlink traffic ratio baseline value (which can be preset based on the 3GPP protocol), is the preset influence coefficient of the uplink and downlink traffic ratio; For each network resource of the current running service, the real-time value of each fine-grained parameter corresponding to the network resource , the dynamic influence coefficient in the influence coefficient library corresponding to each fine-grained parameter and the network resource type , are substituted into the formula: , to obtain the sub-consumption value of the current running service for the network resource ; wherein j is the label of the fine-grained parameter; is the real-time value of the jth fine-grained parameter of the current service, is the dynamic influence coefficient of the jth fine-grained parameter of the current service and the network resource, and m is the total number of fine-grained parameters corresponding to the network resource; For each network resource of the current running service, a preset weight coefficient is assigned, and then the sub-consumption values of the network resources of the current running service are multiplied by the corresponding preset weight coefficients, and the sum of all the products is obtained. The comprehensive resource consumption value R of the current running service is obtained.

[0020] It should be noted that by constructing a three-dimensional data system and a service-network associated data matrix, the standardized integration of multi-source heterogeneous data is realized, and the problem of unstable quantification caused by the disconnection of service parameters and network resource data and the asynchronous collection channel in traditional solutions is solved. Through the benchmark-calibration-dynamic three-level influence coefficient progressive computing mechanism, the combination of static historical fitting and real-time sampling calibration enables the mapping of business fine-grained running parameters to bandwidth, computing, and storage resource consumption to upgrade from a fixed empirical model to a dynamic adaptive model, significantly improving the timeliness and accuracy of resource consumption accounting. Through the historical business retention and dynamic updating strategy of the influence coefficient library, the rapid response capability of newly accessed businesses is guaranteed, and the continuous optimization and calibration of normal operation businesses are realized, breaking the limitations of traditional methods that rely only on real-time data or only on historical data. The introduction of uplink and downlink traffic correction factors differentiates the traffic direction characteristics of different business types, avoiding resource evaluation distortion caused by uniform conversion. The final output of the sub-consumption value and the comprehensive resource consumption value provides quantifiable decision-making basis for subsequent business importance evaluation and performance indicator weight calculation, fundamentally solving the technical problem of lack of accurate quantitative basis for network scheduling, and realizing the leap from extensive estimation to precise quantification of resource consumption evaluation.

[0021] Business and performance evaluation module: based on the evaluation dimensions of the operating business and the corresponding core parameters, calculate the operating business importance evaluation value, analyze the urgency coefficient of performance indicators in the 5G network, and combine the importance evaluation value of the operating business and the sub-consumption value of the network resources corresponding to the performance indicators to calculate the performance indicator importance evaluation value and weight coefficient, build the operating business update scene library, and trigger the performance indicator weight update, the specific process is: Extract the evaluation dimensions of the current operating business and the corresponding core parameters, specifically: First dimension: service level agreement priority, core parameter is the priority level extracted from the service level agreement configuration file generated based on the business opening; Second dimension: user level coefficient, core parameter is the preset coefficient extracted based on the user management log of the user to which the business belongs; the user management log records the preset coefficient configured by the network system according to the user type (ordinary user, VIP user, government and enterprise user) when the user registers; Third dimension: scene value coefficient, core parameter is the scene value weight extracted based on the business attribute registration table preset by the network system; the registration table pre-records all supported business types and corresponding scene values (such as emergency communication 2.0, industrial control 1.8, generative artificial intelligence interaction 1.5, and ordinary entertainment 0.9) according to industry general standards; Dimension four: resource scarcity adaptation degree, core parameter is the resource scarcity adaptation degree corresponding to the current operating business, the specific analysis process is: Using the formula: , the network comprehensive resource remaining amount is obtained , wherein a network comprehensive resource maximum bearing value preset based on a network hardware configuration file and a 5G protocol, a total of all current all operating service comprehensive resource consumption values; combining the current operating service comprehensive resource consumption value R and the network residual resource amount , using the formula: , the resource scarcity adaptation degree corresponding to the current operating service is obtained , which is used to quantify the influence degree of the operating service on the network resource scarcity; wherein, is a preset minimum value; is a preset resource scarcity adaptation coefficient; After obtaining the core parameters corresponding to each evaluation dimension and performing normalization and dimensionless processing, using the formula: , the dynamic weight of each evaluation dimension is obtained ; wherein g is the label of the evaluation dimension, is the entropy value of the core parameter of the gth evaluation dimension; After obtaining the actual value of the core parameter of each evaluation dimension and performing normalization and dimensionless processing, using the formula: , the importance evaluation value of the current operating service is obtained ; wherein, is the actual value of the core parameter of the gth evaluation dimension, and E is a preset scene adaptation factor; Obtain the importance evaluation value of each operating service in the current collection period and the dynamic weight corresponding to each evaluation dimension, and record each item in the field order of business unique identifier-importance evaluation value-service level agreement priority weight-user level coefficient weight-scene value coefficient weight-resource scarcity adaptation degree weight, and arrange in descending order of importance evaluation value. Arrange to form the current collection period operating service importance evaluation table; Based on the network association layer, obtain the performance indicators corresponding to all network resources of the 5G network, including: bandwidth occupancy rate, delay fluctuation value, packet loss rate, uplink capacity utilization rate, etc. For each performance indicator, the service level agreement upper limit of the performance indicator and the current network real-time delay corresponding to the performance indicator are called, and the formula: , the urgency coefficient of the performance indicator is obtained , wherein is the upper limit of the delay agreed by the service level agreement associated with the performance indicator, is the current network real-time delay of the performance indicator, is a preset timeout penalty coefficient; For each performance indicator, by substituting the item consumption value of all current operating services on the network resource corresponding to the performance indicator, the importance evaluation value of each operating service, and the urgency coefficient of the performance indicator into the formula: , to obtain the importance evaluation value of the performance index ; wherein d is the label of the running service, D is the total number of the current running services, s is the label of the performance index, is the sub-consumption value of the network resource corresponding to the performance index of the dth current running service; is the importance evaluation value of the dth current running service, is the urgency coefficient of the st performance index; By linearly normalizing the importance evaluation values of all performance indexes (mapping to the interval [0, 1]), the weight coefficients of the performance indexes are obtained; A running service update scene library is established, and when any update scene in the running service update scene library actually occurs, the weight coefficient updating process is automatically triggered; The running service update scene library includes: the number of newly added / online running services ≥ the preset number, the service QoS level change ratio is greater than the preset value, etc. When the weight coefficient updating condition is met, the weight coefficients of the performance indexes are recalculated to ensure matching with the current service distribution.

[0022] It should be noted that by constructing a four-dimensional evaluation system of service level agreement priority, user level coefficient, scene value coefficient and resource scarcity adaptation degree, the problem of one-sidedness caused by the traditional method of relying only on static SLA priority is fundamentally solved; The entropy value method is introduced to calculate the dynamic weight, so that the business importance evaluation can be adjusted in real time according to the network resource consumption state, and especially the quantitative calculation of the resource scarcity adaptation degree, which analyzes the correlation between business resource occupation and network remaining capacity, realizes the dynamic coupling of the evaluation result and the real load state of the network; By establishing a performance index urgency coefficient and importance evaluation value calculation model, the business sub-consumption value, business importance and service level agreement delay constraint are fused, and the disadvantages of fixed and unchanged performance index weight are solved; The running service update scene library is constructed and the weight updating is automatically triggered, which ensures that the evaluation system evolves synchronously when the business increases or decreases or the service quality changes, provides accurate, dynamic and evolvable decision basis for subsequent abnormal control, and realizes the leap from static classification to dynamic comprehensive evaluation of business value identification.

[0023] The performance index monitoring and control module: construct the performance index change curve, calculate the performance index risk value, mark the abnormal performance index and calculate the corresponding risk overrun value, calculate the abnormal evaluation value in combination with the weight coefficient corresponding to the abnormal index, construct the abnormal index processing priority sequence, and accordingly sequentially carry out targeted control on the network resources and associated running services corresponding to each abnormal index, and the specific process is: For each performance indicator, a two-dimensional rectangular coordinate system is constructed with the numerical value of the performance indicator as the ordinate and time as the abscissa; The values corresponding to each sampling period are obtained in real time, and the values corresponding to each sampling period of the performance indicator are substituted into the two-dimensional rectangular coordinate system to obtain a plurality of data points. According to the time sequence, the data points are sequentially connected by a smooth curve to obtain a performance indicator change curve, such as Figure 2 ; Two performance indicator threshold lines are preset in the performance indicator change curve, respectively a threshold increase threshold line and a threshold decrease threshold line; The area surrounded by the performance indicator change curve exceeding the threshold increase threshold line and the threshold increase threshold value is denoted as a threshold increase area, and the area surrounded by the performance indicator change curve below the threshold decrease threshold line and the threshold decrease threshold line is denoted as a threshold decrease area; For each sampling period, the performance indicator weight is updated in the last sampling period as the starting point and the current sampling period as the end point, and the current network state monitoring window is determined; The sum of all threshold increase areas in the current network state monitoring window is denoted as a threshold increase total integral CS, and the sum of all threshold decrease areas in the current network state monitoring window is denoted as a threshold decrease total integral JS; The standard deviation value of the performance indicator change curve in the current network state monitoring window is calculated to obtain a performance indicator fluctuation coefficient BD; After normalizing and de-dimensioning the corresponding threshold increase total integral CS, threshold decrease total integral JS and performance indicator fluctuation coefficient BD in the current network state monitoring window, the performance indicator risk value FXZ of the current network state monitoring window is obtained by using the formula: FXZ=CS×z1+JS×z2+BD×z3; wherein z1, z2, z3 are preset weight coefficients; A performance indicator risk threshold is preset, and the performance indicator risk of the current network state monitoring window is compared with the corresponding threshold value. If it is greater than or equal to the corresponding threshold value, the performance indicator is marked as an abnormal indicator; For each abnormal indicator, the risk value of the abnormal indicator is subtracted from the corresponding preset threshold to obtain an abnormal indicator risk overrun value; The abnormal indicator risk overrun value is multiplied by the weight coefficient of the corresponding performance indicator to obtain an abnormal evaluation value; All abnormal indicators are integrated and sorted in descending order based on the corresponding abnormal evaluation values to obtain an abnormal indicator processing priority sequence; By sending the abnormal index processing priority sequence to the operation and maintenance terminal, the operation and maintenance terminal carries out targeted management and control on the network resources and associated operation and maintenance services corresponding to each abnormal index in sequence based on the abnormal index processing priority sequence - preferentially taking measures such as resource allocation optimization, link parameter adjustment, temporary reduction of non-core services for high-priority abnormal indexes, and gradually processing low-priority abnormal indexes, thereby achieving the goal of quickly relieving network performance abnormal risk, guaranteeing high-importance service SLA compliance, and maintaining stable operation of the 5G communication network.

[0024] It should be noted that by constructing the performance index change curve and introducing the three-dimensional risk quantification model of threshold excess area, threshold drop area and fluctuation coefficient, the defects of traditional single-point threshold alarm ignoring the cumulative effect and fluctuation trend of the index are solved, and the timing and refinement of risk identification are realized. By delineating a dynamic monitoring window and synchronizing it with the business weight update period, it is ensured that risk calculation is always based on the latest business distribution state, avoiding the problem of lagging evaluation benchmarks. Based on the abnormal risk over-limit value and the performance index weight coefficient, an abnormal evaluation value is constructed, and a processing priority sequence is generated accordingly, which changes the disordered disposal in the case of multiple abnormal concurrent into ordered targeted management and control, significantly improving the efficiency and fairness of resource optimization configuration. Finally, the operation and maintenance terminal implements differentiated management and control strategies to preferentially guarantee high-value business service level agreements, realizing closed-loop management from risk identification, evaluation sorting to resource regulation, and making network management and control change from passive response to active and accurate intervention, thereby systematically improving the stability and service quality guarantee capability of the 5G network in the multi-service concurrent scenario.

[0025] Although embodiments of the present application have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and alterations can be made without departing from the principles and spirit of the present application, and the scope of the present application is defined by the appended claims and their equivalents.

Claims

1. A monitoring and control system for 5G communication networks based on artificial intelligence, characterized in that: include: Influence coefficient construction module: Collects a three-dimensional data system of 5G communication network operation services and constructs a service-network correlation data matrix; Calculate the baseline, calibration and dynamic impact coefficients of each fine-grained parameter of the running service and the corresponding network resources, construct an impact coefficient library, and calculate the individual consumption value and comprehensive resource consumption value of the running service on each network resource. Service and performance evaluation module: Based on the operational service evaluation dimension, calculate the importance evaluation value of operational services, analyze the urgency coefficient of performance indicators in 5G network, and combine the importance evaluation value of operational services and the sub-item consumption value of network resources corresponding to the performance indicators by operational services to calculate the importance evaluation value and weight coefficient of performance indicators. Performance indicator monitoring and control module: Constructs performance indicator change curves, calculates performance indicator risk values, marks abnormal performance indicators and calculates corresponding risk exceedance values, calculates abnormal assessment values ​​by combining the weight coefficients corresponding to abnormal indicators, constructs an abnormal indicator processing priority sequence, and conducts targeted control on network resources and related operating services corresponding to each abnormal indicator in sequence.

2. The monitoring and control system for 5G communication networks based on artificial intelligence according to claim 1, characterized in that: The specific process of constructing the business-network related data matrix is ​​as follows: A three-dimensional data system for collecting all currently operating services of the 5G communication network is established, including: basic attribute layer parameters, fine-grained operation layer parameters, and network association layer parameters. With each operational service as the core, and using unique service identifiers, collection timestamps, basic attribute layer parameters, fine-grained operational layer parameters, and network association layer parameters as columns, a service-network association data matrix is ​​constructed through data alignment of collection channels, parameter classification mapping, and structured integration.

3. The monitoring and control system for 5G communication networks based on artificial intelligence according to claim 2, characterized in that: The specific process for calculating the baseline, calibration, and dynamic impact coefficients of each fine-grained parameter of the running service and the corresponding network resources is as follows: For each currently running service, and for each fine-grained parameter of the running service, a comprehensive analysis is conducted based on the historical values ​​of the fine-grained parameters in each historical sampling period, the historical network resource occupancy values ​​corresponding to the fine-grained parameters, the average value of the fine-grained parameter values ​​in all historical sampling periods, the average value of the network resource occupancy values ​​corresponding to all historical sampling periods, and the total number of historical sampling periods, to obtain the benchmark influence coefficient between the fine-grained parameters and the corresponding network resources under the running service. The real-time values ​​of fine-grained parameters of the running services in the current sampling period, the real-time occupancy values ​​of the corresponding network resources, and the corresponding baseline influence coefficients are obtained. A comprehensive analysis is then performed to obtain the calibration coefficients between the fine-grained parameters and the corresponding network resources under the running services. By combining the baseline influence coefficient and calibration coefficient between fine-grained parameters and corresponding network resources under the operational service, as well as the preset baseline coefficient weights, a comprehensive analysis is conducted to obtain the dynamic influence coefficient between fine-grained parameters and corresponding network resources under the operational service.

4. The monitoring and control system for 5G communication networks based on artificial intelligence according to claim 3, characterized in that: The specific process of constructing the influence coefficient database is as follows: All services supported by the communication network are integrated into an influence coefficient library based on the unique mapping relationship between service, fine-grained parameter, and network resource. For each currently running service in the influence coefficient database, each fine-grained parameter-network resource combination corresponds to a dynamically calibrated influence coefficient in real time; for each service that has appeared in the past but is not currently running, each fine-grained parameter-network resource combination temporarily retains a baseline influence coefficient based on historical data fitting; when the service is accessed and running again, the dynamic influence coefficient is automatically calculated using the newly collected real-time fine-grained parameter values ​​and the corresponding real-time network resource occupancy values, and the baseline influence coefficient is updated to the dynamic influence coefficient.

5. The monitoring and control system for 5G communication networks based on artificial intelligence according to claim 4, characterized in that: The specific process for calculating the individual and total resource consumption values ​​of operational services on each network resource is as follows: For each currently running service, first obtain its real-time uplink and downlink traffic and calculate the uplink and downlink traffic ratio. Then, combine the service's regular uplink and downlink traffic ratio benchmark value with the preset influence coefficient of the uplink and downlink traffic ratio to obtain the uplink and downlink traffic correction factor. For each network resource of the current service, the real-time values ​​of each fine-grained parameter corresponding to the network resource are combined with the dynamic influence coefficients of each fine-grained parameter in the influence coefficient library that match the network resource to calculate the itemized consumption value of the current service for the network resource. Assign a preset weight coefficient to each network resource for the current service, multiply the consumption value of each network resource item by the corresponding weight coefficient, and sum all the product results to obtain the comprehensive resource consumption value of the current running service.

6. The monitoring and control system for 5G communication networks based on artificial intelligence according to claim 5, characterized in that: The specific process for calculating the operational business importance assessment value is as follows: The four evaluation dimensions and core parameters of the current operating business are extracted as follows: Service Level Agreement priority, User Level coefficient, Scenario Value coefficient, and Resource Scarcity Adaptability. The calculation process for resource scarcity fit is as follows: The remaining amount of network resources is obtained by subtracting the sum of the comprehensive resource consumption of all currently running services from the maximum capacity value of network comprehensive resources preset by the network hardware configuration file and 5G protocol. By combining the overall resource consumption of the current operating services with the remaining resources of the network, a comprehensive analysis is conducted to determine the resource scarcity suitability. After normalizing and removing the dimensions of the core parameters of each dimension, the dynamic weights of each dimension are calculated based on the entropy values ​​of the core parameters corresponding to each dimension. The normalized core parameter values ​​are then combined with dynamic weights for comprehensive analysis to obtain the current operational business importance assessment value.

7. The monitoring and control system for 5G communication networks based on artificial intelligence according to claim 6, characterized in that: The specific process for calculating the importance evaluation value and weight coefficient of performance indicators is as follows: Based on the network association layer, obtain the performance indicators corresponding to all network resources of the 5G network; For each performance metric, the agreed latency limit of its associated service level agreement and the current real-time network latency are retrieved, and combined with the preset timeout penalty coefficient, to obtain the urgency coefficient of that performance metric. For each performance metric, its importance evaluation value is obtained by combining the component consumption values ​​of all currently running services on its corresponding network resources, the importance assessment value of each running service, and the urgency coefficient of the metric. The importance evaluation values ​​of all performance indicators are linearly normalized to obtain the weight coefficients of each indicator; Establish a business update scenario library, and automatically trigger the weight coefficient update process when any update scenario in the library occurs in real time. If the update conditions are met, the weight coefficient will be recalculated to ensure that it matches the current business distribution.

8. The monitoring and control system for 5G communication networks based on artificial intelligence according to claim 7, characterized in that: The specific process for constructing performance indicator change curves and calculating performance indicator risk values ​​is as follows: For each performance metric, a two-dimensional Cartesian coordinate system is constructed with the numerical value of the performance metric as the vertical axis and time as the horizontal axis. The values ​​of performance indicators for each sampling period are acquired in real time and marked in a two-dimensional rectangular coordinate system to obtain several data points. According to the time order, adjacent data points are connected sequentially through a smooth curve to obtain the performance indicator change curve. Two performance indicator threshold lines are preset: the threshold exceeding the threshold line and the threshold falling below the threshold line. The area enclosed by the performance index change curve exceeding the threshold threshold line is called the threshold exceedance area; the area enclosed by the performance index change curve below the threshold descent area is called the threshold descent area. For each sampling period, the network status monitoring window for the current period is defined with the sampling period in which the performance index weights were last updated as the starting point and the current sampling period as the ending point. The sum of all threshold outflow areas within the current network status monitoring window is denoted as the threshold outflow product; the sum of all threshold descent areas within the current network status monitoring window is denoted as the threshold descent product. Calculate the standard deviation of the performance index change curve within the current network status monitoring window to obtain the performance index fluctuation coefficient; By comprehensively analyzing the total product of threshold overshoot, total product of threshold descent, and performance index fluctuation coefficient within the current network status monitoring window, the performance index risk value of the current network status monitoring window is obtained.

9. The monitoring and control system for 5G communication networks based on artificial intelligence according to claim 8, characterized in that: The specific process of constructing a priority sequence for handling abnormal indicators, and then carrying out targeted control of the network resources and related operational services corresponding to each abnormal indicator in sequence, is as follows: The performance indicator risk value of the current network status monitoring window is compared with the corresponding threshold. If it is greater than or equal to the corresponding threshold, the performance indicator is marked as an abnormal indicator. For each abnormal indicator, the risk exceeding limit value of the abnormal indicator is obtained by subtracting the corresponding preset threshold from the risk value of the abnormal indicator. An anomaly assessment value is obtained by multiplying the risk exceeding the limit corresponding to the abnormal indicator by the weighting coefficient of the corresponding performance indicator. All abnormal indicators are integrated and sorted from largest to smallest based on their corresponding abnormal assessment values ​​to obtain an abnormal indicator processing priority sequence. The priority sequence for handling abnormal indicators is sent to the operation and maintenance terminal. The operation and maintenance terminal then performs targeted control on the network resources and related operational services corresponding to each abnormal indicator based on the priority sequence.

Images