Method and system for evaluating reliability of power communication network

Abstract

The invention relates to the technical field of electric power system communication, and provides an electric power communication network reliability evaluation method and system, and the method comprises the steps: dividing the network state space of the operation state of an electric power communication network into a plurality of subspaces; generating an evaluation space point set by the space points extracted from each subspace based on a preset state evaluation space mapping relation; according to the key evaluation space points in the evaluation space point set, obtaining original network nodes in the network state space; and running a simulation evaluation process based on an evolved network state space generated by injecting each original network node into the corresponding preset control logic, performing reverse calibration on each preset control logic, and injecting each obtained calibration control logic into the evolved network state space to execute reliability evaluation so as to generate a dynamic reliability evaluation result. According to the method, the state can be actively and accurately sensed from the inside of the network, dynamic adjustment and optimization are carried out according to self feedback of the network, and the accuracy and adaptability of deep reliability state evaluation of the network are effectively improved.

Description

Technical Field

[0001] This invention relates to the field of power system communication technology, and in particular to a method and system for evaluating the reliability of power communication networks. Background Technology

[0002] Reliability assessment of power communication networks typically relies on deploying monitoring equipment at critical network nodes for external data collection, or on offline simulation based on historical operating data and fixed rule models. However, external monitoring technologies can only passively record the network's existing performance parameters and cannot actively explore the network's deep response mechanisms under potential faults or extreme conditions. Meanwhile, the evaluation logic based on fixed rule models is statically preset and difficult to adapt to dynamic changes in network topology and service flows. These two mainstream approaches result in a disconnect between the evaluation process and the network's real-time, endogenous state evolution. The obtained evaluation results have limitations in accuracy and scenario adaptability, lack depth and flexibility, and fail to reveal the propagation paths of complex faults within the system and the network's true resilience. Summary of the Invention

[0003] To address the aforementioned technical problems, this invention provides a method and system for evaluating the reliability of power communication networks. By proactively and accurately sensing the network's internal state and dynamically optimizing the reliability evaluation mechanism based on the network's own feedback, the accuracy and adaptability of the network's deep reliability status evaluation are effectively improved.

[0004] In a first aspect, embodiments of the present invention provide a method for evaluating the reliability of a power communication network, the method comprising: A network state space representing the operating state of the power communication network is constructed, and the network state space is spatially partitioned to form multiple subspaces with related relationships. Extract spatial points with reliability characterization capabilities from each of the subspaces, and map each of the spatial points to a preset evaluation space based on a preset state evaluation space mapping relationship to generate an evaluation space point set. Obtain the key evaluation space points in the set of evaluation space points, and obtain the corresponding original network nodes in the network state space based on each key evaluation space point; Each of the original network nodes is injected with corresponding preset control logic to generate an evolutionary network state space, and a simulation evaluation process is run based on the evolutionary network state space to obtain multiple sets of logic execution records. Based on all the logic execution records, each of the preset control logics is reverse-calibrated, and the resulting calibration control logics are injected into the state space of the evolution network to perform reliability evaluation, generating corresponding dynamic reliability evaluation results.

[0005] Furthermore, the step of performing spatial partitioning on the network state space to form multiple related subspaces includes: Based on the historical state change data of the power communication network, state change analysis is performed to obtain multiple state change patterns. Based on each of the state transition modes, a core state region corresponding to the state transition mode is defined in the network state space. Centered on each of the core state regions, the network extends outward according to a preset correlation rule to form multiple subspaces covering the network state space, and establishes correlation relationships between adjacent subspaces.

[0006] Furthermore, the steps for constructing the preset state evaluation space mapping relationship include: Based on preset evaluation requirements, the evaluation dimensions of the preset evaluation space are determined; the preset evaluation requirements include that the number of evaluation dimensions of the preset evaluation space is greater than the number of state dimensions of the network state space. Construct a mapping rule from each state dimension in the network state space to at least one evaluation dimension in the preset evaluation space; Based on all the mapping rules, the preset state evaluation space mapping relationship is formed.

[0007] Furthermore, the step of obtaining the key evaluation space points in the evaluation space point set includes: Calculate the evaluation distance between every two evaluation space points in the set of evaluation space points; Based on all the evaluation distances, construct a distance relationship network for the set of evaluation space points; The degree centrality is obtained based on the number of direct edges connecting each evaluation space point in the distance relationship network. Based on the proportion of the shortest path traversed by each evaluation space point in the distance relationship network, the corresponding betweenness centrality is obtained; The corresponding proximity centrality is obtained by taking the inverse of the sum of the shortest path distances between each evaluation space point and the other evaluation space points in the distance relationship network. Based on preset key point search conditions, the degree centrality, betweenness centrality, and proximity centrality of each evaluation space point are analyzed, and the evaluation space points that satisfy the preset key point search conditions are marked as the key evaluation space points; the preset key point search conditions are that the degree centrality is greater than a preset connectivity threshold, and the betweenness centrality or proximity centrality is greater than a preset centrality threshold.

[0008] Furthermore, the step of injecting each of the original network nodes with corresponding preset control logic to generate the evolutionary network state space includes: The multi-dimensional state data of each original network node within a preset historical time period is obtained, and the multi-dimensional state data is preprocessed to obtain the corresponding multi-dimensional state data to be analyzed; the multi-dimensional state data includes a state value sequence, a state change timestamp sequence, and a related event sequence. The state value sequences in each of the multi-dimensional state data to be analyzed are subjected to time-domain and frequency-domain feature extraction to obtain the corresponding state time-domain features and state frequency-domain features; the state time-domain features include mean, variance, skewness, and kurtosis; the state frequency-domain features include dominant frequency components; Event interval analysis is performed on the state change timestamp sequences in each of the multi-dimensional state data to be analyzed to obtain the corresponding event interval features; the event interval features include the average event interval time and the interval time variance. The state time-domain features, state frequency-domain features, and event interval features corresponding to each of the multi-dimensional state data to be analyzed are combined to obtain the node behavior features of the corresponding original network nodes. The node behavior characteristics of each of the original network nodes are used to fill in the preset control logic template to generate the corresponding preset control logic; the preset control logic template includes condition judgment rules and corresponding logic execution paths; Each of the preset control logics is injected into the corresponding original network node to generate a corresponding logic injection node; Based on the positions of each logically injected node in the network state space, the spatial topology of the network state space is recalculated to generate the evolved network state space.

[0009] Further, the step of recalculating the spatial topology of the network state space based on the positions of each of the logically injected nodes in the network state space, and generating the evolved network state space, includes: Each of the aforementioned logic injection nodes is marked as a special node in the network state space; Based on preset weight update conditions, the connection weights between each of the special nodes and all other nodes in the network state space are updated to obtain the corresponding logical injection connection weights; the preset weight update conditions include the new data flow interaction mode or preset node coupling rules generated after the introduction of logical injection nodes. Based on the logically injected connection weights of each of the special nodes, the spatial connection matrix of the network state space is updated, and the evolved network state space is generated based on the updated spatial connection matrix.

[0010] Furthermore, the step of obtaining multiple sets of logic execution records based on the state space operation simulation evaluation process of the evolutionary network includes: Multiple simulated operating scenarios are pre-set for the state space of the evolutionary network; In each of the simulated operating scenarios, the evolution network state space is driven to run from the initial state to the termination state, and the control logic operation information of each logic injection node is continuously collected to generate corresponding logic execution records; the control logic operation information includes the time when the control logic is triggered, the control logic triggering conditions, and the control logic execution results.

[0011] Furthermore, the step of reverse calibrating each of the preset control logics based on all the logic execution records includes: Obtain execution exception records from all the aforementioned logic execution records, generate an exception execution record group, and parse each exception execution record in the exception execution record group to obtain the corresponding control exception information; the control exception information includes the exception trigger time, the exception trigger input condition set, and the exception output result; Based on the logical execution path in the preset control logic corresponding to each of the aforementioned abnormal execution records, a complete execution path diagram is constructed from the logical execution starting point to the abnormal output node; In each of the complete execution path diagrams, starting from the abnormal output node, traverse backwards to check whether the actual output of each logical judgment node is consistent with the expected output, and mark the first logical judgment node with deviation as the target adjustment logical unit. The internal parameters of the target adjustment logic unit are adjusted, and the preset control logic after parameter adjustment is verified by simulation until the corresponding verification execution record matches the preset expected execution record, thus obtaining the corresponding calibration control logic.

[0012] Furthermore, the step of injecting the obtained calibration control logic into the state space of the evolution network to perform reliability evaluation and generate corresponding dynamic reliability evaluation results includes: Multiple observation points are set in the state space of the evolutionary network, and a preset dynamic evolution process of the state space of the evolutionary network is initiated. In the preset dynamic evolution process, the state evolution flow containing the running state of each logic injection node is synchronously collected from all the observation points; The dynamic reliability evaluation results are generated in real time by analyzing all the state evolution flows.

[0013] Secondly, embodiments of the present invention provide a power communication network reliability evaluation system, the system comprising: The state space partitioning module is used to construct a network state space that represents the operating state of the power communication network, and to perform space partitioning on the network state space to form multiple subspaces with related relationships. The evaluation space generation module is used to extract space points with reliability characterization capabilities from each of the subspaces, and map each of the space points to a preset evaluation space based on a preset state evaluation space mapping relationship to generate an evaluation space point set. The key node extraction module is used to obtain key evaluation space points in the evaluation space point set, and to obtain the corresponding original network nodes in the network state space based on each key evaluation space point. The network state simulation module is used to inject corresponding preset control logic into each of the original network nodes, generate an evolved network state space, and run a simulation evaluation process based on the evolved network state space to obtain multiple sets of logic execution records. The network dynamic evaluation module is used to reverse-calibrate each of the preset control logics based on all the logic execution records, and inject the obtained calibration control logics into the evolution network state space to perform reliability evaluation and generate corresponding dynamic reliability evaluation results.

[0014] This invention provides a method and system for reliability evaluation of power communication networks. The method involves constructing a network state space characterizing the operating state of the power communication network, dividing the network state space into multiple related subspaces, extracting at least one space point with reliability characterization capability from each subspace, mapping each space point to a preset evaluation space based on a preset state evaluation space mapping relationship to generate an evaluation space point set, obtaining key evaluation space points from the evaluation space point set, obtaining corresponding original network nodes in the network state space based on each key evaluation space point, injecting each original network node into corresponding preset control logic to generate an evolved network state space, running a simulation evaluation process based on the evolved network state space to obtain multiple sets of logic execution records, performing reverse calibration on each preset control logic based on all logic execution records, injecting the obtained calibration control logic into the evolved network state space to perform reliability evaluation, and generating corresponding dynamic reliability evaluation results. Compared with existing technologies, this power communication network reliability evaluation method, based on the proactive and accurate perception of the network's internal state and dynamic optimization based on the network's own feedback, can effectively overcome the passivity of existing external observations and the rigidity of static models. This makes the evaluation behavior itself closer to the real dynamic operation of the network, and improves the accuracy and adaptability of the deep reliability status evaluation of the network. Attached Figure Description

[0015] Figure 1 This is a flowchart illustrating the reliability evaluation method for power communication networks in an embodiment of the present invention; Figure 2 This is a schematic diagram comparing the number of logic execution records under different simulated running scenarios in this embodiment of the invention; Figure 3 This is a schematic diagram illustrating the relationship between the delay threshold and the abnormal execution rate during the reverse calibration iteration in this embodiment of the invention. Figure 4 This is a schematic diagram of the structure of the power communication network reliability evaluation system in an embodiment of the present invention; The attached figures are labeled as follows: 1. State space partitioning module; 2. Evaluation space generation module; 3. Key node extraction module; 4. Network state simulation module; 5. Network dynamic evaluation module. Detailed Implementation

[0016] To make the objectives, technical solutions, and beneficial effects of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. Obviously, the embodiments described below are only part of the embodiments of this invention and are used to illustrate the invention, but are not intended to limit the scope of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0017] In one embodiment, such as Figure 1 As shown, a reliability evaluation method for power communication networks is provided, including: S11. Construct a network state space that characterizes the operating status of the power communication network, and perform spatial partitioning on the network state space to form multiple subspaces with related relationships; wherein, the network state space can be understood as a space composed of multiple state dimensions for evaluating the operating status of the power communication network, each state dimension corresponding to a network operating parameter, such as node CPU utilization, link bandwidth utilization, packet loss rate and latency.

[0018] In this embodiment, performing spatial partitioning of the network state space can be understood as dividing a continuous space into multiple discrete regions based on the correlation and historical patterns between state variables of the power communication network, forming multiple subspaces with correlations, and the subspaces are connected by boundary conditions; specifically, the step of performing spatial partitioning of the network state space to form multiple subspaces with correlations includes: Based on the historical state change data of the power communication network, state change analysis was performed to obtain multiple state change patterns. The historical state change data comes from time-series state information (snapshot data) collected every minute by the power communication network monitoring system over the past year. This time-series state information includes operating parameters of various devices in the power communication network, link transmission status, service flow information, and related event records. Device operating parameters include the processor operating status, memory usage, and port working status of different communication devices. Link transmission status includes the link transmission rate, data transmission delay, data loss, and link connectivity. Service flow information includes the transmission paths, bandwidth usage, service request volume, and processing results of various power services. Related event records include the occurrence time, event type, and associated object information of events such as equipment fault alarms, link interruptions and recovery, network configuration adjustments, and service switching. Correspondingly, the state transition pattern can be understood as the transition law of state variables under specific conditions. The specific conditions refer to typical operating conditions in the operation of the power communication network, such as peak business, equipment start-up and shutdown, link switching, etc. The transition law is the orderly change trend of state variables from the initial value range to the target value range under the operating condition. The change has time continuity and value correlation. For example, from 9:00 to 11:00 every morning, the node CPU utilization and link bandwidth utilization increase synchronously, while the data packet loss rate shows an exponential growth trend after the CPU utilization exceeds 80%.

[0019] Based on each of the aforementioned state transition modes, a core state region corresponding to each state transition mode is defined in the network state space. The core state region is defined by the value range of state variables. In practical applications, when defining the core state region, the concentration trend of the values ​​of each relevant state variable in the targeted state transition mode is first analyzed. For example, for the "daytime peak traffic" mode, the numerical distribution of key state variables such as node CPU utilization and link bandwidth utilization in its historical data is analyzed, and the range with the highest frequency of occurrence is taken as the boundary of the core state region. That is, its core state region can be defined by the range of node CPU utilization between 70% and 90% and link bandwidth utilization between 75% and 95%. The final core state region should be able to typically represent the network operating state under this state transition mode.

[0020] Centered on each of the core state regions, multiple subspaces are formed by extending outwards according to a preset correlation rule, covering the network state space, and correlation relationships are established between adjacent subspaces; wherein, the preset correlation rule is based on the correlation or distance metric between state variables, and the calculation formula is: in, For variables x With variablesy The Pearson correlation coefficient; Representing variables x In the i The value at each historical moment; Representing variables x Average over a historical period; Representing variables y In the i The value at each historical moment, Representing variables y Average over a historical period.

[0021] In practical applications, when the absolute value of the correlation coefficient between two state variables is greater than the threshold (0.6), these two variables are considered highly correlated. During the spatial extension process, their respective dimensions are treated as a single correlated unit and expanded to form multiple subspaces covering the network state space (subspaces may overlap). During the expansion process, the boundaries of the subspaces are indeed determined by the physical limits of the state variables or the historically observed maximum and minimum values. Correlation relationships between adjacent subspaces are established. Specifically, historical state transition data is analyzed to calculate the frequency of transitions from the centroid state vector of one subspace to the centroid state vector of its adjacent subspace. This frequency is normalized and used as the state transition probability, thereby constructing a state transition probability matrix describing the correlation relationships between all subspaces. In other words, the correlation relationships between subspaces are described by shared boundaries or state transition probabilities. It should be noted that the threshold for judging the correlation strength between state variables is determined based on the correlation statistics of state variables in the historical operating data of the power communication network, referring to the conventional values ​​for correlation judgment in similar network reliability evaluations, to ensure accurate differentiation between highly correlated and weakly correlated variables.

[0022] S12. Extract spatial points with reliability characterization capabilities from each of the subspaces, and map each spatial point to a preset evaluation space based on a preset state evaluation space mapping relationship to generate an evaluation space point set; wherein, the selection of spatial points is based on their typicality or criticality within the subspace, and in practical applications, at least one spatial point is extracted from each subspace. Correspondingly, the preset evaluation space is a mathematical space composed of multiple evaluation dimensions that carry reliability evaluation indicators, and it is assumed that the evaluation space needs to accommodate more dimensions of reliability evaluation indicators, that is, the number of dimensions of the preset evaluation space is greater than the number of dimensions of the network state space.

[0023] In this embodiment, the preset state evaluation space mapping relationship is a mapping rule that maps the state dimension of the network state space to the evaluation dimension of the preset evaluation space. Specifically, the steps for constructing the preset state evaluation space mapping relationship include: Based on preset evaluation requirements, the evaluation dimensions of the preset evaluation space are determined; the preset evaluation requirements include that the number of evaluation dimensions in the preset evaluation space is greater than the number of state dimensions in the network state space; wherein, the evaluation dimensions of the preset evaluation space can be determined based on the preset evaluation requirements. For example, if the network state space contains 4 dimensions, the preset evaluation space can be designed to have 7 dimensions. The additional dimensions compared to the network state space may include node reliability score, link reliability score, path redundancy, and network vulnerability index.

[0024] Construct mapping rules from each state dimension in the network state space to at least one evaluation dimension in the preset evaluation space; that is, it is necessary to formulate mapping rules that assign each state dimension in the network state space to at least one evaluation dimension in the preset evaluation space. For example, the node CPU utilization dimension in the network state space can be mapped to two dimensions in the preset evaluation space: node reliability score and network vulnerability index. This mapping rule can be a linear transformation, a nonlinear function, or a conditional mapping (piecewise linear function), etc. For example, when the CPU utilization is below 50%, the node reliability score is 1, and when the CPU utilization is between 50% and 100%, the node reliability score linearly decreases to 0. It should be noted that the construction of this mapping rule depends on the correspondence between the network state in historical data and the reliability indicators of the post-evaluation, and the parameters of the mapping function can be determined through regression analysis.

[0025] Based on all the mapping rules, the preset state evaluation space mapping relationship is formed by combining them; that is, all the mapping rules obtained above can be combined to form a complete spatial mapping relationship from the network state space to the preset evaluation space, that is, the desired preset state evaluation space mapping relationship is obtained; the preset state evaluation space mapping relationship is a multi-dimensional vector function, for example, it can be a function that takes a 4-dimensional network state vector as input and outputs a 7-dimensional evaluation space vector as output.

[0026] S13. Obtain the key evaluation space points in the set of evaluation space points, and obtain the corresponding original network nodes in the network state space based on each key evaluation space point; wherein, the key evaluation space points represent the hub positions in the evaluation space, and are selected based on the connectivity requirements and centrality requirements of the evaluation space points. The connectivity requirements ensure the degree of connection of the points in the network, which can be measured by degree centrality, and the centrality requirements measure the importance of the points in the network, which can be measured by betweenness centrality or proximity centrality.

[0027] Specifically, the step of obtaining the key evaluation space points in the evaluation space point set includes: Calculate the evaluation distance between every two evaluation space points in the evaluation space point set; wherein, the evaluation distance uses Euclidean distance or other metrics (for example, Mahalanobis distance can be used to assess the correlation between evaluation dimensions); assuming the evaluation space point set contains 100 evaluation space points, and each evaluation space point is a 7-dimensional vector, then the evaluation distance between two evaluation space points is: in, Representing evaluation space points A With evaluation space points B The evaluation distance between them; Representing evaluation space points A In the k Coordinate values ​​on each evaluation dimension Representing evaluation space points B In the k Coordinate values ​​on each evaluation dimension.

[0028] Based on all the evaluation distances, a distance relationship network is constructed for the set of evaluation space points. Nodes in the distance relationship network are evaluation space points, and edges are connected based on the evaluation distance. In practical applications, a distance relationship network is constructed based on all evaluation distances. A distance threshold is set (e.g., 1.2 times the average of all evaluation distances). When the evaluation distance between two evaluation space points is less than or equal to the distance threshold, a connection edge is established between these two evaluation space points in the distance relationship network. It should be noted that a distance relationship network can also be constructed based on the K-nearest neighbor method, connecting each evaluation space point to the K other points with the smallest evaluation distance.

[0029] Based on the number of direct edges connecting each evaluation space point in the distance relationship network, the corresponding degree centrality is obtained; that is, the degree centrality is the number of edges directly connecting each evaluation space point.

[0030] Based on the proportion of the shortest paths passing through each evaluation space point in the distance relationship network, the corresponding betweenness centrality is obtained. In practical applications, for an evaluation space point P, the betweenness centrality calculation process is to find the shortest path between all other pairs of evaluation space points in the distance relationship network, count the number of shortest paths passing through evaluation space point P, and divide this number by the total number of all shortest paths in the network.

[0031] The proximity centrality is obtained by taking the reciprocal of the sum of the shortest path distances between each evaluation space point and the other evaluation space points in the distance relationship network. In practical applications, the proximity centrality of each evaluation space point is calculated. For evaluation space point Q, the shortest path distances from evaluation space point Q to all other evaluation space points in the distance relationship network are first calculated, and these distances are summed. The proximity centrality is the reciprocal of the sum of these distances.

[0032] Based on preset key point search conditions, the degree centrality, betweenness centrality, and proximity centrality of each evaluation space point are analyzed, and evaluation space points that meet the preset key point search conditions are marked as key evaluation space points. The preset key point search conditions are that the degree centrality is greater than a preset connectivity threshold, and the betweenness centrality or proximity centrality is greater than a preset centrality threshold. The preset connectivity threshold and preset centrality threshold can be obtained by analyzing and calibrating historical network reliability events. Preferably, the connectivity threshold can be set to 5 to ensure that the evaluation space points have sufficient direct connections, and the centrality threshold can be set to 0.1 to ensure that the evaluation space points are in critical hub positions in the network. Based on the preset key point search conditions, evaluation space points whose degree centrality is greater than the connectivity threshold and whose betweenness centrality or proximity centrality is greater than the centrality threshold are selected from the set of evaluation space points as key evaluation space points that meet the connectivity and centrality requirements. For example, if the degree centrality of evaluation space point P is 8 and the betweenness centrality is 0.15, then evaluation space point P is marked as a key evaluation space point.

[0033] After obtaining each key evaluation space point through the above methods and steps, the original network node (physical or logical component in the power communication network) corresponding to the key evaluation space point can be located in the network state space. By injecting preset control logic containing a series of condition triggering instructions into the original network node, a logic injection node with active evaluation function is generated, and an evolutionary network state space is generated. Based on the data collected and reported by the logic injection node during operation, the reliability is dynamically evaluated.

[0034] S14. Inject corresponding preset control logic into each of the original network nodes to generate an evolved network state space, and run a simulation evaluation process based on the evolved network state space to obtain multiple sets of logic execution records; wherein, the evolved network state space can be understood as a network state space generated by recalculating the spatial topology (describing the connection relationship between nodes) of the network state space based on the position of the logic-injected nodes in the network state space, reflecting the changes in network structure after the introduction of logic-injected nodes. Specifically, the step of injecting corresponding preset control logic into each of the original network nodes to generate the evolved network state space includes: The process involves acquiring multi-dimensional state data of each original network node within a preset historical time period, and preprocessing the multi-dimensional state data to obtain corresponding multi-dimensional state data to be analyzed. The multi-dimensional state data includes a state value sequence, a state change timestamp sequence, and a sequence of associated events. Assuming the original network node is a core router located in a regional dispatch center, the state value sequence of the acquired multi-dimensional state data within the preset historical time period (e.g., the last 30 days) includes CPU utilization percentage, memory utilization percentage, and inbound traffic rate of major ports collected every minute. The state change timestamp sequence records the moments when various parameters in the state value sequence reach preset conditions or undergo significant changes, including but not limited to each moment when CPU utilization exceeds a set threshold, memory utilization fluctuates significantly, and inbound traffic rate of ports exceeds a critical value. The associated event sequence records event identifiers and occurrence times related to routing oscillations and link failure alarms. The preprocessing in this embodiment includes cleaning multi-dimensional state data, removing noise data and outliers, and imputing missing values. For example, for a state value sequence, a sliding window average filter is used to remove instantaneous spike noise, data points that fluctuate by more than 50% within three consecutive minutes are considered outliers and removed, and missing values ​​are imputed by using linear interpolation of two valid data points before and after to fill in the missing minutes of data.

[0035] The state value sequences in each of the multi-dimensional state data to be analyzed are subjected to time-domain and frequency-domain feature extraction to obtain the corresponding state time-domain features and state frequency-domain features. The state time-domain features include mean, variance, skewness, and kurtosis. For example, taking the CPU utilization sequence as an example, its mean, variance, skewness, and kurtosis are calculated. The mean reflects the average load level, the variance reflects the degree of load fluctuation, the skewness describes the asymmetry of the distribution, and the kurtosis describes the sharpness of the distribution. The state frequency-domain features include the dominant frequency component, which is extracted by calculating the power spectral density through Fourier transform, as shown in the formula: in, In frequency The power spectral density value at that location; For Fourier transform operators; For time Extracting the dominant frequency component from the changing state value sequence, i.e. The frequency values ​​exhibiting distinct peaks may correspond to periodic patterns in network load. It should be noted that the aforementioned state time-domain and state frequency-domain characteristics can also be combined using wavelet transform to obtain joint time-frequency characteristics.

[0036] Event interval analysis is performed on the state change timestamp sequences in each of the multi-dimensional state data to be analyzed to obtain the corresponding event interval features. The event interval features include the average event interval time and the interval time variance. The average event interval time describes the frequency of CPU utilization exceeding the limit events, and the interval time variance describes the regularity of event occurrence.

[0037] The temporal, frequency, and event interval features of each of the multi-dimensional state data to be analyzed are combined to obtain the node behavior features of the corresponding original network node. For example, the vector corresponding to the node behavior features of a certain original network node obtained by combination can be represented as [CPU mean, CPU variance, CPU skewness, CPU kurtosis, first clock frequency, second clock frequency, average event interval, interval time variance]. It should be noted that in practical applications, the co-occurrence pattern of events can be further extracted as part of the node behavior features from the analysis of associated event sequences, and the dimensionality of the node behavior features can be reduced by principal component analysis to simplify subsequent logic.

[0038] The node behavior characteristics of each original network node are used to fill in a preset control logic template to generate corresponding preset control logic. The preset control logic template includes conditional judgment rules and corresponding logic execution paths. For example, a typical conditional judgment rule template is in the form of "If [feature A] exceeds [threshold A] and [feature B] is lower than [threshold B], then execute [action X]". In practical applications, the node behavior characteristics of the original network node are filled into the conditional judgment rules in the preset control logic template to generate customized control logic for the original network node. For example, based on the node behavior characteristics of an average CPU of 65% and an average event interval of 120 minutes, [feature A] in the preset control logic template is specified as "CPU utilization", [threshold A] is set to 70%, [feature B] is specified as "time since the last over-limit event", and [threshold B] is set to 60 minutes to obtain the corresponding customized control logic. This makes the injected control logic more in line with the actual operating mode of the original network node, so as to ensure the relevance and effectiveness of the injected logic.

[0039] Each of the preset control logics is injected into the corresponding original network node to generate a corresponding logic injection node; customized control logic is injected into the original network node to transform it into a logic injection node. The specific injection process can be implemented by loading a customized monitoring daemon at the level of the original network node, so that the logic injection node has the ability to perceive its own state and execute preset evaluation actions.

[0040] Based on the positions of each logically injected node in the network state space, the spatial topology of the network state space is recalculated to generate the evolved network state space. This recalculation of the spatial topology can be understood as adjusting the connection weights between the logically injected nodes and all other nodes in the network state space based on the state correlations or physical distances between nodes. Specifically, the step of recalculating the spatial topology of the network state space based on the positions of each logically injected node in the network state space to generate the evolved network state space includes: Each of the logic injection nodes is marked as a special node in the network state space. In practical applications, if a customized control logic is injected into an SDH (Synchronous Digital Hierarchy) transmission device located in a hub substation, it is marked as a special node identifier N105, and its coordinates in the network state space are determined by the normalized values ​​of its multiple state variables.

[0041] Based on preset weight update conditions, the connection weights between each of the special nodes and all other nodes in the network state space are updated to obtain the corresponding logical injection connection weights. The preset weight update conditions include the new data flow interaction modes or preset node coupling rules generated after the introduction of the logical injection node. The preset weight update conditions are pre-analyzed and determined before the preset control logic of the logical injection node is deployed to each target original node. Specifically, they include: 1) parsing the condition judgment rules and logic execution paths defined in the preset control logic template one by one to identify the data packet types, data transmission directions, data interaction frequencies and their triggering conditions that may be newly initiated to execute the active evaluation function; 2) combining the target original network node's business role in the power communication network, data processing capabilities and the conventional data processing load and main communication objects reflected in its historical operation data, inferring the possible new data flow interaction scenarios between the logical injection node and other nodes (communication objects) when activating the evaluation function; judging the real-time and reliability requirements of the data processed by the node according to its business type, thereby inferring the priority and service quality requirements that the new data flow should have, and based on the node's data processing capabilities. The evaluation assesses whether the node can stably handle the data flow generated by the newly added proactive evaluation task while completing the original business processing. Simultaneously, by referring to the node's historical data interaction objects, the set of target nodes most likely to which the new data flow points is determined. Considering all these factors, the potential data flow interaction scenarios between the logical injection node and other nodes in the network are derived, and their correlation strength is qualitatively assessed. 3) Based on the physical connection relationships between nodes in the network topology, data transmission protocol requirements, business priority configurations, and inherent data dependencies between nodes, coupling rules are formulated to clarify the correlation methods and impact levels between the logical injection node and other nodes during data interaction, providing a basis for recalculating connection weights. Specifically, the primary basis is the physical connection relationships between nodes in the network topology; the more direct the physical connection, the higher the basic connection weight usually indicated by the coupling rules. The protocol requirements followed by data transmission between nodes are considered, as this affects the complexity and latency of data flow interaction, thus affecting the coupling strength. The priority configurations of various power services operating in the network are analyzed; the evaluation data flow generated by the logical injection node must follow this priority system to avoid impacting critical businesses. Coupling between business nodes with similar priorities may be stronger. The inherent data dependencies between nodes are also an important reference for formulating coupling rules. Node pairs with strong data dependencies should be given a high weight in terms of logical coupling, even if they are not physically directly connected. For example, periodically sending probe messages to specific network management nodes or adjacent nodes; and formulating coupling rules between nodes based on the physical connection topology of the power communication network, the communication protocols followed by data transmission between nodes, the priority settings of different power services, and the inherent data dependencies between nodes.

[0042] Specifically, when recalculating connection weights based on the new data flow interaction patterns or preset inter-node coupling rules generated after the introduction of logically injected nodes, it is necessary to first obtain the correlation index of historical state data between nodes. Combined with the topological distance of the physical deployment of nodes, a weighted fusion method is used to adjust the original connection weights. The higher the correlation and the closer the physical distance, the more closely the connection weight adjustment reflects the correlation strength. This can be expressed as: In the formula, Assign logical injection connection weights between logically injected node N105 and another node j in the network; Inject the original connection weights between node N105 and node j into the logic; To logically inject the set of state variables associated with node N105 in the network state space; Let j be the set of state variables associated with node j in the network state space; The preset coupling gain coefficient is used to adjust the magnitude of the impact. It can be set differently according to the type or importance of the logical injection node. For logical injection nodes of the core hub node type, a larger value is taken to strengthen its coupling with other nodes. For logical injection nodes of the edge access node type, a smaller value is taken to match its association range and degree of influence in the network.

[0043] Based on the logically injected connection weights of each of the special nodes, the spatial connection matrix of the network state space is updated, and the evolved network state space is generated based on the updated spatial connection matrix. In practical applications, the spatial connection matrix of the network state space is updated based on the recalculated logically injected connection weights. Assuming that the spatial connection matrix of the network state space is an n x n symmetric matrix, where n is the total number of nodes in the network state space, and the element in the i-th row and j-th column of the matrix records the connection weights between node i and node j, the updated spatial connection matrix obtained based on the above weight recalculation formula generates the evolved network state space, and the evolved network state space uses the new spatial connection matrix as the mathematical description of its topology, as shown in Table 1.

[0044] Table 1. Logical Injection Connection Weights of Some Node Pairs Calculated Again After generating the evolutionary network state space based on the above method steps, the network behavior under various operating conditions can be simulated within the state space. Multiple sets of logic execution records generated by the logic injection nodes during the simulation are recorded to calibrate relevant control logic parameters to adapt to real-world evaluation scenarios. Specifically, the step of running the simulation evaluation process based on the evolutionary network state space to obtain multiple sets of logic execution records includes: Multiple simulated operating scenarios are pre-set for the state space of the evolved network; among them, the simulated operating scenarios can be constructed based on historical fault data or preset stress test models, such as "normal load scenario", "single router failure scenario", "regional link congestion scenario" and "distributed denial-of-service attack simulation scenario", etc.

[0045] In each of the aforementioned simulation scenarios, the evolutionary network state space is driven from an initial state to a terminated state, and the control logic operation information of each logic injection node is continuously collected to generate corresponding logic execution records. The control logic operation information includes the time when the control logic is triggered, the triggering condition, and the execution result. In practical applications, in each simulation scenario, the evolutionary network state space is driven from an initial state to a terminated state. The initial state is a snapshot of the evolutionary network state space loaded with specific scenario parameter configurations. The driving process simulates network dynamics by iteratively updating the state values ​​of nodes in the spatial connection matrix. The terminated state is when a preset simulation time length is reached or the system enters a stable or faulty state. During operation, the time when the logic injection node is triggered, the triggering condition, and the execution result are continuously collected to form a logic execution record. For example, in one simulation, it is recorded that "at simulation time T=150 seconds, the CPU utilization state value of logic injection node N105 reaches 72%, triggering the control logic 'record current network flow characteristics,' and the execution result is the generation and storage of a snapshot file containing ten flow records," etc.

[0046] It should be noted that driving the state space operation of the evolving network can be based on a differential equation model. This model establishes a rate-of-change equation for each state variable relative to time to describe the network dynamics. The rate-of-change equation captures the coupling relationships between state variables and the influence of external simulated operating scenarios, such as the driving force of normal load or fault events on state evolution. The driving process starts from a set initial state, using numerical integration to solve a system of differential equations, gradually updating the values ​​of all state variables in the network state space, simulating the continuous evolution trajectory of the network state. Evolution continues until a termination condition is met, such as reaching a preset simulation time or the system entering stable equilibrium. During this process, the behavior of the logic injection nodes is monitored in real time and recorded as logic execution records. Simultaneously, random disturbance factors can be introduced into the simulated operating scenarios to simulate uncertain network environments. It can be understood that by setting multiple simulation scenarios for the evolving network state space, the behavior of logic injection nodes under different stress or fault conditions can be comprehensively tested. The structured storage of the final logic execution records serves as the data foundation for subsequent reverse calibration. Each record includes a timestamp, input state vector, triggered logic rule identifier, and output action result.

[0047] The visualization effect of various simulated scenarios running in this embodiment is as follows: Figure 2 As shown, the horizontal axis represents the simulated running scenario, and the vertical axis represents the number of execution records, distinguishing between the expected number of execution records and the number of abnormal execution records; for example... Figure 2 As shown, scenario complexity is positively correlated with abnormal execution records: under normal load scenarios, the proportion of expected records is extremely high, reflecting the consistency of logic injection node execution when the network state is stable; as the scenario is upgraded, expected records gradually decrease while abnormal records continue to increase; through multi-scenario simulation, the behavior of logic injection nodes under different working conditions is comprehensively tested, and the collected execution records are the core data source for subsequent reverse calibration control logic, demonstrating the role of front-end data support from simulation to collection to optimization process.

[0048] S15. Perform reverse calibration on each of the preset control logics based on all the logic execution records, and inject the obtained calibrated control logics into the state space of the evolved network to perform reliability evaluation, generating corresponding dynamic reliability evaluation results; wherein, performing reverse calibration on each preset control logic can be understood as identifying the gap between the initially injected control logic parameters and the actual network feedback exposed during simulation by comparing the abnormal execution records and expected execution records (control logic execution records that meet the design goals) in the logic execution records, automatically correcting the deviation of the evaluation logic parameters, and making its output match the actual dynamic response characteristics of the network after calibration process; specifically, the step of performing reverse calibration on each of the preset control logics based on all the logic execution records includes: The system retrieves all execution anomaly records from the logic execution records, generates anomaly execution record groups, and parses each anomaly execution record in these groups to obtain corresponding control anomaly information. This control anomaly information includes the anomaly trigger time, the set of anomaly trigger input conditions, and the anomaly output result. For example, in 100 groups of logic execution records, the expected number of execution records is 85. In these 85 groups, the output actions of the logic injection nodes are consistent with the preset reliability assessment target. The number of anomaly execution records is 15. In these 15 groups, the logic injection nodes performed unexpected output actions under specific conditions, such as triggering a "network degradation warning" action that should only be triggered during severe congestion when the link latency slightly increases. The corresponding parsed control anomaly information includes the anomaly trigger time as simulated time T = 320 seconds, the set of input conditions at the time of the anomaly trigger as {CPU utilization = 65%, memory utilization = 50%, average latency = 45ms}, and the anomaly output result as "send network degradation warning," etc.

[0049] Based on the logical execution path in the preset control logic corresponding to each of the aforementioned abnormal execution records, a complete execution path diagram from the logical execution start point to the abnormal output node is constructed. In practical applications, by decomposing the hierarchical structure of the logical execution path in the preset control logic, the association order of the logical start point, each condition judgment node, execution branch, and abnormal output node is clarified. The complete execution path diagram is drawn in the form of nodes and directed edges according to the logical link from the start point to the judgment node, then to the branch, and finally to the output.

[0050] In each of the complete execution path diagrams, starting from the abnormal output node, a reverse traversal is performed, sequentially checking whether the actual output of each logical judgment node matches the expected output. The first logical judgment node with a deviation is marked as the target adjustment logic unit. It should be noted that the actual control logic is represented in the form of a decision tree. The logical starting point is "reading the current state input," and the logical path passes through a series of condition judgment nodes, ultimately branching to different output action nodes. In the complete execution path diagram, starting from the abnormal output node "send network degradation warning," a reverse traversal is performed, sequentially checking each logical judgment node. During the backtracking process, the actual output of each logical judgment node is compared with the expected output. For example, the logical judgment node "whether the delay is greater than 50ms" should actually output "no" to guide the logic towards "maintaining steady-state monitoring" since the average input delay is 45ms. However, the record shows that its output is "yes," leading to the abnormal output. The first logical judgment node with a deviation is identified and marked as the target adjustment logic unit; that is, the judgment node "whether the delay is greater than 50ms" is positioned as the target adjustment logic unit.

[0051] The internal parameters of the target adjustment logic unit are adjusted, and the preset control logic after parameter adjustment is verified through simulation until the corresponding verification execution record matches the preset expected execution record, thus obtaining the corresponding calibration control logic. In practical applications, based on the deviation type of the target adjustment logic unit, the internal parameters such as the condition threshold and judgment logic priority of the target adjustment logic unit are adjusted accordingly, and the control logic after parameter adjustment is verified through simulation, gradually fine-tuning until its execution record matches the expected execution record, thus obtaining the required calibration control logic. For example, the threshold parameter in the "whether the latency is greater than 50ms" of the target adjustment logic unit is adjusted from 50ms to 40ms, and then the simulation is restarted until its execution record matches the expected execution record. After modifying the threshold and simulating again, under the same input conditions {CPU utilization = 65%, memory utilization = 50%, average latency = 45ms}, the control logic no longer triggers the "network degradation warning" but instead executes the "maintain steady-state monitoring" action, achieving consistency with the expected behavior. The specific iterative calibration process is as follows: Figure 3 As shown. Figure 3The horizontal axis represents the calibration iteration steps, the left vertical axis represents the time delay threshold, and the right vertical axis corresponds to the abnormal execution rate. This visually presents the iterative optimization process of the threshold for this unit. Lowering the threshold makes the logic judgment more in line with the actual time delay characteristics of the node. The significant decrease in the abnormal execution rate verifies the closed-loop calibration mechanism, which includes the positioning deviation unit, parameter adjustment, and verification effect. It demonstrates the role of reverse calibration in improving the accuracy of control logic and is a visual representation of the simulation, feedback, and optimization data-driven process.

[0052] After loading the obtained calibration control logic into the corresponding logic injection nodes, dynamic reliability evaluation can be performed in the evolutionary network state space. Dynamic reliability evaluation assesses network reliability by monitoring node state changes in real time. Specifically, the steps of injecting the obtained calibration control logic into the evolutionary network state space to perform reliability evaluation and generate corresponding dynamic reliability evaluation results include: Multiple observation points are set in the evolved network state space, and a preset dynamic evolution process for the evolved network state space is initiated. The observation points are a pre-selected set of nodes in the network state space used to collect state evolution flows. The selection of observation points can be based on the criticality of the network topology, such as setting observation points on network state space nodes corresponding to core routers, aggregation switches, and important service servers. In this embodiment, the preset dynamic evolution process can be understood as driving the continuous change of state values ​​in the evolved network state space over time by simulating tidal changes in network traffic, random device fault injection, and link state fluctuations. After setting up multiple observation points, the preset dynamic evolution process can be initiated, and reliability analysis can be performed using data collected from the observation points.

[0053] In the pre-defined dynamic evolution process, a state evolution stream containing the operational status of each logical injection node is synchronously collected from all observation points. This state evolution stream is time-series data, containing multi-dimensional state variable values ​​for each observation point at various times. For example, the CPU utilization, link throughput, and current control logic state of the logical injection node marker N105 are collected once per second for the core router node. It should be noted that the logical injection node continuously operates as an active sensing and response unit during the dynamic evolution process, and its behavior directly affects the evolution path of the network state space, thus being captured by the observation points and reflected in the state evolution stream. The synchronous collection of observation points is coordinated through a global simulation clock to ensure strict alignment of the timestamps in the state evolution stream.

[0054] The dynamic reliability evaluation results are generated in real time by analyzing all the aforementioned state evolution flows; that is, the collected state evolution flows are analyzed in real time, and the comprehensive reliability index of the power communication network at the current moment is calculated based on the following formula, so that network operation and maintenance personnel can visually perceive the real-time changing trend of network reliability: in, In order to be in t Dynamic reliability evaluation index of time-based networks; The number of key state parameters monitored in the state evolution flow; For time t, the first m Normalized values ​​of key state parameters; For the first m Preset reliability thresholds for key state parameters; function Used for quantization parameters Relative to threshold The degree of deviation, when Superior When the time is 0, the function value is 0; otherwise, the function value increases between 0 and 1. It should be noted that the function... The specific form can be defined according to the characteristics of different state parameters, and can be a linear piecewise function or an exponential function.

[0055] The process of generating dynamic reliability evaluation results in real time can be combined with the evaluation output of the logic injection node itself for weighted fusion. The dynamic reliability evaluation results can be displayed in real time on the monitoring interface in the form of time series curves. The lower the evaluation result value, the worse the instantaneous reliability of the network.

[0056] This invention provides a reliability evaluation mechanism based on proactive and precise state perception from within the network and dynamic optimization based on network feedback, generating corresponding dynamic reliability evaluation results. This mechanism not only transforms traditional external observation into proactive internal exploration by locating key original network nodes in the network state space and injecting preset control logic into them to create logic injection nodes, but also allows the injected logic nodes to dynamically generate specific data streams or trigger preset test states on the network according to evaluation requirements. This software-defined probe implantation mechanism directly and proactively acquires deep state information about path connectivity, fault propagation, and resource capacity from within the network, making the evaluation process itself part of the network's programmable functionality. This approach flexibly adapts to different evaluation scenarios and objectives, ensuring that the acquired state information is more direct and closer to the actual operating scenario of the network. Moreover, by establishing an evolutionary network state space and running a simulation process, it can reverse-calibrate the control logic based on multiple sets of logic execution records of the recorded logic injection nodes, generating a data-driven closed-loop design of calibrated control logic parameters. This optimizes the logic control behavior pattern, improves the accuracy of the evaluation action itself and the credibility of the evaluation results, and enables the reliability evaluation to have the ability to self-correct and continuously improve. In this way, it effectively overcomes the passivity of existing external observations and the rigidity of static models, making the evaluation behavior itself closer to the real dynamic operating essence of the network, and improving the accuracy and adaptability of the deep reliability state evaluation of the network.

[0057] It should be noted that although the steps in the flowchart above are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise explicitly stated in this document, there is no strict order requirement for the execution of these steps, and they can be executed in other orders.

[0058] In one embodiment, such as Figure 4 As shown, a reliability evaluation system for power communication networks is provided, the system comprising: State space partitioning module 1 is used to construct a network state space that represents the operating state of the power communication network, and to perform space partitioning on the network state space to form multiple subspaces with related relationships. Evaluation space generation module 2 is used to extract at least one space point with reliability characterization capability from each of the subspaces, and map each of the space points to a preset evaluation space based on a preset state evaluation space mapping relationship to generate an evaluation space point set. The key node extraction module 3 is used to obtain key evaluation space points in the evaluation space point set, and to obtain the corresponding original network nodes in the network state space based on each key evaluation space point. The network state simulation module 4 is used to inject each of the original network nodes into the corresponding preset control logic, generate the evolved network state space, and run the simulation evaluation process based on the evolved network state space to obtain multiple sets of logic execution records. The network dynamic evaluation module 5 is used to perform reverse calibration on each of the preset control logics based on all the logic execution records, and inject the obtained calibration control logics into the evolution network state space to perform reliability evaluation and generate corresponding dynamic reliability evaluation results.

[0059] Specific limitations regarding the power communication network reliability evaluation system can be found in the limitations of the power communication network reliability evaluation method described above; the corresponding technical effects are equivalent and will not be repeated here. Each module in the aforementioned power communication network reliability evaluation system can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0060] In summary, the reliability evaluation method and system for power communication networks provided by this invention can effectively overcome the passivity of existing external observations and the rigidity of static models by actively and accurately sensing the state from within the network and dynamically optimizing the reliability evaluation mechanism based on the network's own feedback. This makes the evaluation behavior itself closer to the real dynamic operation of the network and improves the accuracy and adaptability of the deep reliability status evaluation of the network.

[0061] The various embodiments in this specification are described in a progressive manner. For directly identical or similar parts of the embodiments, refer to each other. Each embodiment focuses on describing the differences from other embodiments. In particular, the system embodiments are basically similar to the method embodiments, so the description is relatively simple; relevant parts can be referred to the descriptions in the method embodiments. It should be noted that the technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as the combination of these technical features does not contradict each other, it should be considered within the scope of this specification.

[0062] The above-described embodiments are merely preferred embodiments of the present invention, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of the invention. It should be noted that those skilled in the art can make various improvements and substitutions without departing from the principles of the present invention, and these improvements and substitutions should also be considered within the scope of protection of the present invention. Therefore, the scope of protection of this invention should be determined by the scope of the claims.

Claims

1. A reliability evaluation method for power communication networks, characterized in that, The method includes: A network state space representing the operating state of the power communication network is constructed, and the network state space is spatially partitioned to form multiple subspaces with related relationships. Extract spatial points with reliability characterization capabilities from each of the subspaces, and map each of the spatial points to a preset evaluation space based on a preset state evaluation space mapping relationship to generate an evaluation space point set. Obtain the key evaluation space points in the set of evaluation space points, and obtain the corresponding original network nodes in the network state space based on each key evaluation space point; Each of the original network nodes is injected with corresponding preset control logic to generate an evolutionary network state space, and a simulation evaluation process is run based on the evolutionary network state space to obtain multiple sets of logic execution records. Based on all the logic execution records, each of the preset control logics is reverse-calibrated, and the resulting calibration control logics are injected into the state space of the evolution network to perform reliability evaluation, generating corresponding dynamic reliability evaluation results.

2. The power communication network reliability evaluation method as described in claim 1, characterized in that, The step of performing spatial partitioning on the network state space to form multiple related subspaces includes: Based on the historical state change data of the power communication network, state change analysis is performed to obtain multiple state change patterns. Based on each of the state transition modes, a core state region corresponding to the state transition mode is defined in the network state space. Centered on each of the core state regions, the network extends outward according to a preset correlation rule to form multiple subspaces covering the network state space, and establishes correlation relationships between adjacent subspaces.

3. The power communication network reliability evaluation method as described in claim 1, characterized in that, The steps for constructing the preset state evaluation space mapping relationship include: Based on preset evaluation requirements, the evaluation dimensions of the preset evaluation space are determined; the preset evaluation requirements include that the number of evaluation dimensions of the preset evaluation space is greater than the number of state dimensions of the network state space. Construct a mapping rule from each state dimension in the network state space to at least one evaluation dimension in the preset evaluation space; Based on all the mapping rules, the preset state evaluation space mapping relationship is formed.

4. The power communication network reliability evaluation method as described in claim 1, characterized in that, The step of obtaining the key evaluation space points in the set of evaluation space points includes: Calculate the evaluation distance between every two evaluation space points in the set of evaluation space points; Based on all the evaluation distances, construct a distance relationship network for the set of evaluation space points; The degree centrality is obtained based on the number of direct edges connecting each evaluation space point in the distance relationship network. Based on the proportion of the shortest path traversed by each evaluation space point in the distance relationship network, the corresponding betweenness centrality is obtained; The corresponding proximity centrality is obtained by taking the inverse of the sum of the shortest path distances between each evaluation space point and the other evaluation space points in the distance relationship network. Based on preset key point search conditions, the degree centrality, betweenness centrality, and proximity centrality of each evaluation space point are analyzed, and the evaluation space points that satisfy the preset key point search conditions are marked as the key evaluation space points; the preset key point search conditions are that the degree centrality is greater than a preset connectivity threshold, and the betweenness centrality or proximity centrality is greater than a preset centrality threshold.

5. The power communication network reliability evaluation method as described in claim 1, characterized in that, The step of injecting each of the original network nodes with corresponding preset control logic to generate the evolutionary network state space includes: The multi-dimensional state data of each original network node within a preset historical time period is obtained, and the multi-dimensional state data is preprocessed to obtain the corresponding multi-dimensional state data to be analyzed; the multi-dimensional state data includes a state value sequence, a state change timestamp sequence, and a related event sequence. The state value sequences in each of the multi-dimensional state data to be analyzed are subjected to time-domain and frequency-domain feature extraction to obtain the corresponding state time-domain features and state frequency-domain features; the state time-domain features include mean, variance, skewness, and kurtosis; the state frequency-domain features include dominant frequency components; Event interval analysis is performed on the state change timestamp sequences in each of the multi-dimensional state data to be analyzed to obtain the corresponding event interval features; the event interval features include the average event interval time and the interval time variance. The state time-domain features, state frequency-domain features, and event interval features corresponding to each of the multi-dimensional state data to be analyzed are combined to obtain the node behavior features of the corresponding original network nodes. The node behavior characteristics of each of the original network nodes are used to fill in the preset control logic template to generate the corresponding preset control logic; the preset control logic template includes condition judgment rules and corresponding logic execution paths; Each of the preset control logics is injected into the corresponding original network node to generate a corresponding logic injection node; Based on the positions of each logically injected node in the network state space, the spatial topology of the network state space is recalculated to generate the evolved network state space.

6. The power communication network reliability evaluation method as described in claim 5, characterized in that, The step of recalculating the spatial topology of the network state space based on the positions of each logically injected node in the network state space, and generating the evolved network state space, includes: Each of the aforementioned logic injection nodes is marked as a special node in the network state space; Based on preset weight update conditions, the connection weights between each of the special nodes and all other nodes in the network state space are updated to obtain the corresponding logical injection connection weights; the preset weight update conditions include the new data flow interaction mode or preset node coupling rules generated after the introduction of logical injection nodes. Based on the logically injected connection weights of each of the special nodes, the spatial connection matrix of the network state space is updated, and the evolved network state space is generated based on the updated spatial connection matrix.

7. The power communication network reliability evaluation method as described in claim 5, characterized in that, The steps of obtaining multiple sets of logic execution records based on the state space operation simulation evaluation process of the evolutionary network include: Multiple simulated operating scenarios are pre-set for the state space of the evolutionary network; In each of the simulated operating scenarios, the evolution network state space is driven to run from the initial state to the termination state, and the control logic operation information of each logic injection node is continuously collected to generate corresponding logic execution records; the control logic operation information includes the time when the control logic is triggered, the control logic triggering conditions, and the control logic execution results.

8. The power communication network reliability evaluation method as described in claim 1, characterized in that, The step of reverse calibrating each of the preset control logics based on all the logic execution records includes: Obtain execution exception records from all the aforementioned logic execution records, generate an exception execution record group, and parse each exception execution record in the exception execution record group to obtain the corresponding control exception information; the control exception information includes the exception trigger time, the exception trigger input condition set, and the exception output result; Based on the logical execution path in the preset control logic corresponding to each of the aforementioned abnormal execution records, a complete execution path diagram is constructed from the logical execution starting point to the abnormal output node; In each of the complete execution path diagrams, starting from the abnormal output node, traverse backwards to check whether the actual output of each logical judgment node is consistent with the expected output, and mark the first logical judgment node with deviation as the target adjustment logical unit. The internal parameters of the target adjustment logic unit are adjusted, and the preset control logic after parameter adjustment is verified by simulation until the corresponding verification execution record matches the preset expected execution record, thus obtaining the corresponding calibration control logic.

9. The power communication network reliability evaluation method as described in claim 1, characterized in that, The step of injecting the obtained calibration control logic into the state space of the evolutionary network to perform reliability evaluation and generate corresponding dynamic reliability evaluation results includes: Multiple observation points are set in the state space of the evolutionary network, and a preset dynamic evolution process of the state space of the evolutionary network is initiated. In the preset dynamic evolution process, the state evolution flow containing the running state of each logic injection node is synchronously collected from all the observation points; The dynamic reliability evaluation results are generated in real time by analyzing all the state evolution flows.

10. A reliability evaluation system for power communication networks, characterized in that, The system includes: The state space partitioning module is used to construct a network state space that represents the operating state of the power communication network, and to perform space partitioning on the network state space to form multiple subspaces with related relationships. The evaluation space generation module is used to extract space points with reliability characterization capabilities from each of the subspaces, and map each of the space points to a preset evaluation space based on a preset state evaluation space mapping relationship to generate an evaluation space point set. The key node extraction module is used to obtain key evaluation space points in the evaluation space point set, and to obtain the corresponding original network nodes in the network state space based on each key evaluation space point. The network state simulation module is used to inject corresponding preset control logic into each of the original network nodes, generate an evolved network state space, and run a simulation evaluation process based on the evolved network state space to obtain multiple sets of logic execution records. The network dynamic evaluation module is used to reverse-calibrate each of the preset control logics based on all the logic execution records, and inject the obtained calibration control logics into the evolution network state space to perform reliability evaluation and generate corresponding dynamic reliability evaluation results.

Images