Cooperative control method, system and equipment for provincial and regional power transmission network and medium

Abstract

The invention discloses a provincial and regional power transmission network cooperative control method, system and device and a medium, and relates to the technical field of power transmission, comprising the steps of monitoring a network in real time, identifying fault nodes and analyzing propagation, evaluating an influence range and resource availability, optimizing standby resource scheduling, and generating a calling priority in combination with load balancing. Cross-level permission allocation is adjusted through load difference analysis, a standby path is analyzed to determine a convergence acceleration sequence, real-time flow and sequence are combined, flow is reallocated to balance loads, resource scheduling parameters are dynamically adjusted, a resource calling and path switching sequence is iteratively optimized, and historical data feedback is utilized to realize resource scheduling. And generating a final resource allocation and path switching scheme. According to the method, a provincial and regional power transmission network cooperative control closed loop is constructed, the fault recovery speed, the resource utilization efficiency and the system toughness are improved, and the problems of response lag, resource waste and insufficient adaptability of a traditional method are effectively solved.

Description

Technical Field

[0001] This invention relates to the field of power transmission technology, specifically to a method, system, equipment, and medium for coordinated control of provincial and local power transmission networks. Background Technology

[0002] With the deepening of smart grid construction, power communication networks, as key infrastructure supporting the stable operation of the power grid, directly impact the safety and stability of the power system through transmission reliability and resource scheduling efficiency. In recent years, new network architectures such as software-defined networking and network function virtualization have been gradually applied to the field of power communication, providing a technical foundation for centralized control and dynamic scheduling of cross-level network resources. Existing research mainly focuses on resource optimization within a single level, which improves local resource utilization and fault response speed to a certain extent. In addition, traditional power communication networks typically employ static permission allocation and preset path switching strategies, which, while ensuring the transmission needs of basic services, are still insufficient in the face of sudden faults and cross-regional service collaboration scenarios.

[0003] However, existing collaborative control technologies for provincial and local power transmission networks still have significant limitations: First, in terms of fault propagation suppression, traditional methods rely heavily on preset threshold alarms and manual intervention, lacking accurate prediction of the fault's impact range and intelligent scheduling capabilities for cross-level backup resources, leading to prolonged fault recovery time and increased risk of service interruption. Second, in terms of load balancing, existing technologies struggle to dynamically perceive load differences across cross-level links, and rigid permission allocation mechanisms easily cause local network congestion and uneven resource utilization. Third, path switching strategies are mostly based on fixed rules, unable to adaptively adjust the activation order according to real-time network status and service priorities, resulting in path convergence delays and decreased service quality. Furthermore, existing systems lack the ability to deeply mine and optimize historical fault data, making it difficult to form a collaborative control mechanism for continuous improvement. Summary of the Invention

[0004] In view of the above-mentioned existing problems, the present invention provides a method, system, equipment and medium for coordinated control of provincial and local power transmission networks, in order to solve the problems of prolonged fault recovery time, increased risk of service interruption, local network congestion and uneven resource utilization, and decreased service quality in the prior art.

[0005] To address the aforementioned technical problems, a collaborative control method for provincial and local power transmission networks is proposed, including: Real-time network status monitoring is used to obtain information on device failures, identify faulty nodes, and perform fault propagation analysis to determine the scope of the fault's impact and resource availability. Based on this assessment, the order of backup resource calls is optimized through resource scheduling. Current load data is extracted and combined with load balancing to generate a resource call priority list. Permission level configuration data is obtained, and uneven load distribution across different levels of links is detected through load difference analysis. The permission allocation scheme is adjusted according to the resource call priority list to obtain optimized permission level allocation results. Based on these results, convergence requirements are determined, and the activation order and link latency data of backup paths are analyzed to determine... The system employs an adjustment sequence for path convergence acceleration. Real-time traffic data is extracted and combined with the path convergence acceleration adjustment sequence. Load balancing is used to recalculate the traffic distribution across provincial and regional links, obtaining the balanced link load distribution status. Based on the backup resource activation time requirements, resource scheduling perception parameters are adjusted using a real-time status awareness process to obtain adjusted resource response time data. The resource scheduling optimization scheme is obtained based on the difference between the resource response time and the expected activation time. Iterative calculations are performed on the resource call priority list and the path convergence acceleration adjustment sequence. Historical data is analyzed to extract feedback information on network performance and fault recovery effects from historical fault scenario data, generating a target resource allocation scheme and path switching mode.

[0006] As a preferred embodiment of the provincial and local power transmission network collaborative control method described in this invention, the real-time monitoring of network status includes: collecting real-time link data of the entire network, processing the real-time link data, calculating link saturation and node health scores, and evaluating the overall network operating status. Based on network topology and node characteristics, the potential propagation paths of faults are analyzed, an impact area matrix is ​​generated, the sequence of affected nodes and the distribution of backup resources are identified, real-time bandwidth utilization data is obtained, and combined with resource availability judgment, the scope of fault impact and resource allocation requirements are determined.

[0007] As a preferred embodiment of the provincial power transmission network collaborative control method of the present invention, the step of adjusting the order of backup resource calls includes obtaining real-time response time data of backup resource nodes and generating an initial resource scoring sequence in combination with load balancing requirements. The initial resource scoring sequence is iteratively optimized to minimize response time and balance load. Based on resource utilization data, resource capacity levels are divided, and the initial call order is determined according to business needs. The system obtains real-time load values ​​for each resource group, compares them with preset capacity allocation thresholds, optimizes load distribution, and generates a priority sequence for backup resource calls.

[0008] As a preferred embodiment of the provincial power transmission network collaborative control method described in this invention, the adjustment of the permission allocation scheme includes: acquiring network-wide permission distribution data, recording real-time load values ​​of links at each level, analyzing node load and bandwidth occupancy characteristics, and generating load difference indicators. The balance of cross-level load distribution is judged based on the indicators, and the backup resource data is extracted from the resource call priority sequence. The permission adjustment parameters are determined in combination with the link saturation. Calculate the adjusted permission allocation weights and generate the optimized permission scheme.

[0009] As a preferred embodiment of the provincial and local power transmission network collaborative control method of the present invention, the method of determining the adjustment sequence for accelerating path convergence includes: obtaining the convergence time of switching paths through path monitoring, comparing with a preset threshold to filter the set of paths to be optimized, and using a multilayer perceptron neural network to input path stability, link delay, hop count and historical success rate features, outputting path priority scores, and generating an initial score sequence. A preliminary adjustment plan is generated based on bandwidth usage and target convergence time. The random forest algorithm is then used to predict the total latency value of each activation sequence by taking into account path combination ID, activation sequence number, traffic pattern and network state characteristics. Calculate a comprehensive score by grouping according to business priority, select the highest score order to generate a path activation time sequence table, and determine the path execution sequence based on the time sequence table; The formula for calculating the overall score is expressed as follows: in, For the overall score, To predict delays, This is the business weight.

[0010] As a preferred embodiment of the provincial power transmission network collaborative control method described in this invention, the recalculation of cross-level provincial links includes: acquiring real-time traffic data of network nodes, calculating the bandwidth ratio of cross-level links, and using a convolutional neural network to input bandwidth utilization, number of data packets, number of connections, and error rate characteristics, outputting a balance score; when the balance score is greater than a preset threshold, calculating the traffic difference between levels, constructing a link weight matrix, analyzing the deviation between the current traffic and the preset threshold, generating a traffic adjustment scheme in combination with path convergence time parameters, and redistributing cross-level link traffic to verify the balance effect; when the load variance drops to within 0.2, confirming the generation of a stable load distribution state. The formulas for calculating the bandwidth ratio of cross-layer links and the traffic difference between layers are expressed as follows: in, This refers to the bandwidth capacity of the upper-level link. For the bandwidth capacity of the lower-level links, This represents the bandwidth ratio of cross-layer links. Due to differences in traffic between levels, The real-time traffic value for level i. The real-time traffic value for level j. and For hierarchical indexes; The adjustment of resource scheduling perception parameters includes: acquiring balanced link load data, extracting the real-time response time of backup resources, using a support vector machine classification algorithm, inputting load balancing degree, response time, CPU utilization and memory utilization features, classifying resources into three categories: high-quality resources, good resources and general resources, calculating state perception parameters, generating resource response benchmark values, comparing the actual response time with the expected activation time, generating a latency difference matrix, reordering backup resources based on node attributes and business priorities, verifying the adjusted response performance, and generating a resource scheduling optimization scheme. The formula for calculating state-aware parameters is expressed as follows: in, For load stability, In order to maintain consistency, For resource availability, , and These are the weighting coefficients. These are state-aware parameters.

[0011] As a preferred embodiment of the provincial and local power transmission network collaborative control method described in this invention, the method for obtaining resource scheduling optimization schemes includes: extracting fault scenarios and recovery data from a fault history database, calculating resource call priority scores, and using a long short-term memory network to extract scenario feature vectors by inputting fault type, network load, response time, and recovery time series, outputting scenario urgency classification probabilities through a fully connected layer, generating a path switching rule set by combining historical network performance curves, evaluating the success rate of rule set execution, combining historical execution feedback for combined optimization, and generating a set of alternative schemes through simulation testing. The formula for prioritizing resource allocation is expressed as follows: in, The weight of the k-th factor. Score the priority of resource access. This represents the k-th standardized factor value; The construction of scene feature vectors using a long short-term memory network includes inputting a time series of historical fault data, with the sequence length determined by the sampling interval. The input sequence is processed using a two-layer long short-term memory network, and the hidden state at the last moment is extracted as the scene feature vector. The feature vector is then mapped to the output dimension through a fully connected layer. The probability of three types of scenes is output using the softmax function, and the activation conditions of the path switching rule are adjusted according to the classification results. The generation of target resource allocation scheme and path switching mode includes: calculating the comprehensive score of each alternative scheme based on multi-objective optimization, calculating the cost-benefit ratio, selecting the scheme with the highest comprehensive score as the final scheme, updating the execution priority according to the real-time network status through a dynamic adjustment mechanism, and verifying the performance indicators of the final scheme under simulated fault scenarios. The formula for calculating the overall score of each alternative is expressed as follows: in, For comprehensive scoring, For recovery time, For resource utilization, For business continuity, Cost-effectiveness ratio.

[0012] The beneficial effects of this preferred technical solution are as follows: by combining the path convergence sequence to recalculate cross-level traffic allocation and adjusting resource scheduling parameters based on real-time state awareness, the refined management of network traffic distribution and the evolution of the resource scheduling model are realized. This can maintain the network load in a dynamically balanced and healthy state, avoid local congestion or link idleness, and enable the resource scheduling strategy to adapt to real-time changes in network performance, thus realizing closed-loop feedback optimization of the scheduling scheme.

[0013] As a preferred embodiment of the provincial power transmission network collaborative control system described in this invention, it is characterized by including a status monitoring and analysis module, a resource scheduling optimization module, a permission allocation and load balancing module, and a path optimization and rule generation module.

[0014] The status monitoring and analysis module is used to collect and evaluate the operating status of the power communication network in real time. It acquires real-time data of the entire network through link status monitors and node status collectors, calculates link saturation and node health scores using resource analysis algorithms, analyzes the fault propagation range through graph neural network models, and generates an impact area matrix.

[0015] The resource scheduling optimization module is used to dynamically adjust the resource allocation strategy according to the network status, obtain the real-time response time of the backup node through resource response detection, generate an initial resource score sequence using a genetic algorithm, divide the resource capacity level by combining a support vector machine classification model, optimize the load distribution through the least squares method, and generate the final resource call priority list.

[0016] The permission allocation and load balancing module is used to obtain network-wide permission distribution data, analyze node load characteristics using the random forest algorithm, generate load difference indicators, determine permission adjustment parameters based on support vector machine classification, calculate optimized permission allocation weights through a weight distribution analyzer, analyze traffic characteristics using a convolutional neural network, construct a link weight matrix, and redistribute cross-level link traffic.

[0017] The path optimization and rule generation module is used to obtain the convergence time of the switching path, generate path priority scores using a multilayer perceptron neural network, evaluate the latency performance of different activation orders using a random forest algorithm, and construct scene feature vectors using a long short-term memory network based on historical fault data to generate a path switching rule set, and select the final solution through multi-objective optimization.

[0018] A computer device includes a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement the steps of a method for coordinated control of a provincial power transmission network.

[0019] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of a method for coordinated control of a provincial power transmission network.

[0020] The beneficial effects of this invention are as follows: By real-time monitoring of network status and fault propagation analysis, this invention achieves a shift from passive processing to proactive early warning, significantly improving the accuracy and timeliness of fault identification; the resource scheduling mechanism based on multi-objective optimization and dynamic load balancing effectively overcomes the problem of uneven resource utilization caused by fixed priority allocation, improving resource utilization efficiency; and through intelligent permission allocation optimization and path convergence acceleration mechanisms, it achieves dynamic adjustment of permission levels and a significant improvement in path switching efficiency; the use of convolutional neural networks for traffic feature analysis and real-time status awareness optimization improves the accuracy of load balancing control; based on iterative learning of historical fault data and multi-objective optimization decision-making, it constructs a control strategy system with self-learning and evolutionary capabilities, and comprehensively considers performance indicators and economic factors in scheme evaluation, solving the problems of response lag, resource waste, and insufficient adaptability of traditional methods. Attached Figure Description

[0021] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 The above is a flowchart of a collaborative control method for provincial power transmission networks provided in one embodiment of the present invention.

[0023] Figure 2 The present invention provides a system scheme flowchart for a collaborative control system for provincial power transmission networks, which is an embodiment of the present invention. Detailed Implementation

[0024] To make the above-mentioned objects, features, and advantages of the present invention more apparent and understandable, specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the protection scope of the present invention.

[0025] Example 1, referring to Figure 1 As an embodiment of the present invention, a method for coordinated control of provincial power transmission networks is provided, comprising: S100: Monitors network status in real time, obtains information on device failures through status monitoring, identifies faulty nodes, performs fault propagation analysis, obtains the scope of fault impact and resource availability judgment, and optimizes the order of backup resource calls through resource scheduling based on the scope of fault impact and resource availability judgment, extracts current load data, and generates a resource call priority list in combination with load balancing processing.

[0026] S200: Obtain permission level configuration data, detect uneven load distribution across layers through load difference analysis, adjust the permission allocation scheme according to the resource call priority list, obtain the optimized permission level allocation result, and based on the optimized permission level allocation result, obtain convergence requirements, analyze the activation order of backup paths and link latency data, and determine the adjustment sequence for accelerating path convergence.

[0027] S300: Extract real-time traffic data, combine it with path convergence acceleration adjustment sequence, adopt load balancing processing, recalculate the traffic distribution of cross-level provincial links, obtain the balanced link load distribution status, and combine the backup resource activation time requirements to adjust the resource scheduling perception parameters using the real-time status perception process, obtain the adjusted resource response time data, and obtain the resource scheduling optimization scheme based on the difference between the resource response time and the expected activation time.

[0028] S400: Iteratively calculates the resource call priority list and path convergence acceleration adjustment sequence. By analyzing historical data, it extracts feedback information on network performance and fault recovery effect from historical fault scenario data, and generates target resource allocation scheme and path switching mode.

[0029] It should be noted that by generating dynamic resource call priorities through real-time fault propagation analysis, accurate and efficient fault response is achieved. Furthermore, by accelerating cross-level permission collaboration and intelligent path convergence, the problem of uneven load distribution is solved, ensuring the rapid recovery of core services. Combined with dynamic traffic balancing and adaptive adjustment of scheduling parameters, the continuous health of network load and closed-loop optimization of strategies are ensured. Through deep learning and iterative feedback of historical data, the overall resilience, intelligence level, and business continuity of the provincial and local power transmission networks under complex fault scenarios are significantly improved.

[0030] Example 2, refer to Figure 1 This is a second embodiment of the present invention, which provides a method for coordinated control of provincial power transmission networks, including: In this embodiment of the application, step S100, the real-time monitoring of network status includes steps S101 to S103: S101: Continuously collects real-time link data from the entire network, including link throughput, bandwidth utilization, and node operation indicators, processes the real-time link data, calculates link saturation and node health scores, and evaluates the overall network operation status. The formula for calculating link saturation is expressed as: in, For link saturation, and The weighting coefficients are used to balance bandwidth utilization and transmission quality. This represents the current throughput of the link. This represents the maximum theoretical bandwidth of the link. The bit error rate is the saturation value. The closer the saturation value is to 1, the closer the link is to saturation. The formula for calculating the health score of a computing node is expressed as follows: in, Assess node health. , and These are the normalized weighting coefficients. For node CPU utilization, For memory utilization, This represents the number of currently active alerts on the current node. This represents the total number of currently active alerts across all nodes.

[0031] S102: Based on network topology and node characteristics, analyze the potential propagation paths of faults using machine learning algorithms to generate an impact region matrix; Using a graph neural network (GNN) model, the input network topology is represented by a graph structure G=(V,E), where V is the set of nodes, E is the set of links, and the fault injection point (source node v) is the source node. fault The output influence region matrix M (with dimensions |V|) is given, where M[l] represents node v. l The probability value of being affected by the fault (between 0 and 1) is set, and a probability threshold (0.7) is set. Nodes with a score higher than 0.7 are marked as affected nodes.

[0032] S103: Identify the affected node sequence and the distribution of backup resources, obtain real-time bandwidth utilization data, and combine it with resource availability judgment to determine the scope of the fault impact and resource allocation requirements; For the affected node sequence, the K-shortest path algorithm is used to calculate multiple backup paths, and the switching latency and remaining bandwidth of each path are evaluated in combination with the performance model to generate path selection criteria. When link overload is detected, the SDN controller identifies traffic priorities and smoothly migrates low-priority traffic to backup paths that meet the conditions based on the evaluation results. After traffic scheduling, the system re-evaluates the network status, dynamically updates and determines an accurate fault impact range and resource availability judgment after mitigation measures by comparing the impact area matrix and resource health before and after optimization, providing a reliable basis for subsequent resource scheduling. The formula for calculating path switching delay is expressed as follows: in, For path switching delay, The path propagation delay is determined by the physical transmission distance. For node latency handling, empirical values ​​are set based on the node device model. The maximum queue delay is estimated based on the current link load. The formula for calculating the remaining bandwidth capacity of a path is expressed as: in, The available bandwidth capacity for path Pi. This refers to all links contained in path Pi. This represents the maximum theoretical bandwidth of a single link. This represents the current actual throughput of the link.

[0033] In an optional implementation, in step S100, the real-time monitoring of network status further includes deploying traffic probes at key nodes and links, periodically collecting traffic data and device status information, obtaining node resource utilization (CPU, memory) via SNMP protocol, and using a threshold detection algorithm (fixed threshold comparison) to determine link load and node health status; when the threshold exceeds the limit, triggering an alarm and generating an impact range list, and combining historical traffic patterns, using simple linear regression analysis to predict fault propagation trends.

[0034] In another optional implementation, in step S100, the real-time monitoring of network status may further include: centrally collecting traffic statistics and link status of switch ports through the SDN controller, obtaining real-time bandwidth utilization and packet loss rate using the OpenFlow protocol, calculating the link load index based on flow table rules, determining the scope of fault impact through the controller's built-in decision logic (rule-based engine), and generating a list of affected nodes in conjunction with the network topology database.

[0035] Furthermore, in this embodiment of the application, in step S102, the fault analysis includes steps A1 to A4: A1: Extract training data from the historical fault database, including network topology (set of nodes and links), fault injection points, and labels of affected nodes.

[0036] A2: Build a feature vector for each node, including node type, shortest path hop count to the fault point, link saturation, and health score.

[0037] A3: Use GraphSAGE or Graph Attention Network (GAT) for training to learn node representations and topological relationships.

[0038] A4: The model outputs an impact probability matrix. A threshold of 0.7 is set to mark affected nodes. When applied online, the current topology and node features are input in real time to generate a binarized impact region matrix.

[0039] In an optional implementation, in step S102, the fault analysis further includes defining a preset rule base, where the rules are based on the network topology hierarchy (core layer, aggregation layer, access layer) and link dependencies; when a fault occurs, the rule engine matches rules according to the fault type (node ​​failure, link interruption), infers the impact range layer by layer, and uses a graph traversal algorithm (breadth-first search) to start from the fault point, mark directly and indirectly connected nodes, and adjust the impact judgment in combination with real-time load data.

[0040] In another optional implementation, in step S102, the fault analysis may further include: constructing a Bayesian network structure, where nodes represent network components (links, nodes) and edges represent dependencies; learning a conditional probability table from historical data to reflect the causal relationship of component faults; inputting real-time fault events; calculating the posterior probability of affected nodes through probabilistic inference (Gibbs sampling); setting a probability threshold; generating a list of affected areas; and dynamically updating the probability model by combining node health data.

[0041] Furthermore, in this embodiment of the application, in step S100, the adjustment of the backup resource calling order includes steps S111~S114: S111: Obtain real-time response time data of backup resource nodes, and use a genetic algorithm combined with load balancing requirements to generate an initial resource scoring sequence; Initialize the population, with each individual representing a resource scoring scheme. Define a fitness function that combines the reciprocal of the response time and the load balancing score (weighted at 0.6 and 0.4). Perform selection operations (roulette wheel selection), crossover operations (single-point crossover with a probability of 0.8), and mutation operations (Gaussian mutation with a probability of 0.1). After 100 generations, output the optimal resource scoring sequence. The fitness function is expressed as: in, The weighting coefficient is 0.6. The weighting factor is 0.4. The average response time, For load balancing scoring, a higher fitness value indicates a better resource scoring scheme.

[0042] S112: Use support vector machines to classify capacity levels based on resource utilization and determine the initial call order; use the support vector machine classification model to classify resource levels based on CPU utilization and bandwidth capacity, including high-level resources (CPU utilization <60% and bandwidth capacity >10Gbps), medium-level resources (CPU utilization 60%~80% and bandwidth capacity 5-10Gbps), and low-level resources (CPU utilization >80% or bandwidth capacity <5Gbps).

[0043] S113: For the initial sequence, the resource balancer optimizes the real-time load distribution of each resource group using the least squares method, and calculates a comprehensive score based on response time, capacity level, reliability and cost factor. The least squares method is expressed as: in, This represents the real-time load value of the m-th resource group. For the target load, Given the total load, after optimization, verify whether the load variance has decreased to within the threshold of 0.2.

[0044] S114: The system groups resources according to business type and sorts them in descending order within each group based on comprehensive score, generating a final priority sequence for backup resource calls; The formula for comprehensive scoring of computing resources is expressed as follows: in, This is a comprehensive resource score; a higher score indicates a higher priority. For resource response time, Resource capacity levels (1~5). This represents a resource reliability index (0~1). Resource cost factor (0~1); key business groups include priority 1~100, important business groups include priority 101~200, and general business groups include priority 201~300.

[0045] In an optional implementation, step S100, adjusting the order of calling spare resources further includes initializing a particle swarm, where each particle represents a resource score vector, defining an objective function, minimizing the average response time and load variance, updating particle positions and velocities based on individual and global optima, iterating until convergence, outputting an optimized score sequence, and using K-means clustering to classify resource levels, and adjusting the calling order according to business priorities (such as medical, financial, and general).

[0046] In another optional implementation, in step S100, adjusting the order of backup resource calls may further include calculating priority scores based on resource response time, availability and cost factors, grouping by business type (critical, important, general), setting inter-group call intervals, and allocating resources using a weighted round-robin mechanism to ensure that high-priority businesses are given priority, and dynamically adjusting weights based on real-time load feedback.

[0047] In step S200, the adjustment of the permission allocation scheme includes steps S201 to S203: S201: Obtain network-wide data through permission monitoring and load collection, and use the random forest algorithm to analyze node load characteristics and generate load difference indicators to determine whether the cross-level load is balanced. The formula for calculating the load difference index is expressed as follows: in, This is a load difference indicator. The standard deviation of the load between levels, The average load across levels, when When this occurs, it is determined to be an uneven load distribution.

[0048] S202: When unevenness is detected, the system extracts backup resource data from the resource call priority sequence and, in combination with link saturation, determines specific permission adjustment parameters through a support vector machine classification algorithm; The support vector machine classification algorithm includes feature vectors such as link saturation, resource priority, and historical failure frequency. The classifier is trained to classify adjustment requirements into three categories: slight adjustment (parameter 0.1-0.3), moderate adjustment (0.3-0.6), and severe adjustment (0.6-0.9). The RBF kernel function is used with a regularization parameter of 1.0 and a gamma parameter of 0.1 to classify and predict the current state. The adjustment parameter values ​​are determined based on the classification results.

[0049] S203: The weight distribution analyzer calculates and adjusts the permission allocation weights of each level based on the current parameters, and generates an optimized permission scheme; The formula for adjusting permissions is expressed as follows: in, The adjustment factor is 0.5. This represents the load difference value. These are the permission values ​​before the adjustment. These are the adjusted permission values. Adjust the permissions value.

[0050] Furthermore, in step S200, the determination of the adjustment sequence for accelerating path convergence includes steps S211 to S214: S211: Obtain the convergence time of switching paths through path monitoring, and compare it with the preset threshold of 200 milliseconds to filter the set of paths to be optimized; S212: Using a multilayer perceptron neural network, input path stability, link delay, hop count and historical success rate features, output path priority score, and generate an initial score sequence; The multilayer perceptron neural network includes input features such as path stability (jitter rate), link latency, hop count, and historical success rate, and output path priority score (0-1); the network structure includes an input layer of 4 dimensions, an 8-layer hidden layer, a 4-layer hidden layer, and an output layer of 1, with ReLU as the activation function.

[0051] S213: Generate a preliminary adjustment plan based on bandwidth usage and target convergence time, and use the random forest algorithm to predict the total latency value of each activation sequence by inputting path combination ID, activation sequence number, traffic pattern and network state characteristics. The random forest algorithm is used, with input features including path combination ID, activation order number, traffic pattern, and network state, and outputting total latency. The model uses 200 decision trees, with a maximum depth of 15 and a minimum number of split samples of 5.

[0052] S214: Calculate the comprehensive score by grouping according to business priority, select the order of the highest score to generate the path activation time series table, and determine the path execution sequence based on the time series table. Among them, the weight of medical business is 0.4, the weight of financial business is 0.3, the weight of general business is 0.3, and the order of the highest score is selected to generate the time series table. The formula for calculating the overall score is expressed as follows: in, For the overall score, To predict delays, This is the business weight.

[0053] In step S300, the recalculation of traffic allocation for cross-level provincial links includes steps S301 to S304: S301: Obtain real-time traffic data of network nodes through traffic monitoring equipment, and calculate the bandwidth ratio of cross-level links using the bandwidth analysis module; The formula for calculating the bandwidth ratio of cross-layer links is expressed as: in, This refers to the bandwidth capacity of the upper-level link. For the bandwidth capacity of the lower-level links, This represents the bandwidth ratio of cross-level links.

[0054] S302: Use a convolutional neural network to analyze traffic characteristics and generate a balanced baseline value. Then, calculate the traffic difference between layers based on the current baseline value and construct a link weight matrix. The convolutional neural network includes input features such as bandwidth utilization, number of packets, number of connections, and error rate, and outputs a balance score (0-1). The network architecture includes Conv1D (filters=32, kernel size=3), MaxPool1D(2), Conv1D(64,3), GlobalAvgPool, and Dense(1, sigmoid). When the score is greater than 0.72, it indicates that adjustment is needed. The formula for calculating the flow difference between levels is expressed as: in, Due to differences in traffic between levels, The real-time traffic value for level i. The real-time traffic value for level j. and This is a hierarchical index.

[0055] S303: Analyze the deviation between the current traffic and the preset threshold based on the link weight matrix, and generate a traffic adjustment plan in combination with the path convergence time parameter; The link weight matrix is ​​represented as follows: in, For the link To link Traffic allocation weights, The attenuation coefficient is set to 2. Due to differences in traffic between levels, The scaling factor is set to 0.8. For the link To link reliability coefficient, It is an exponential function.

[0056] S304: Perform load balancing, redistribute cross-level link traffic, and verify the balancing effect. When the load variance drops to within 0.2, confirm the generation of a stable load distribution state. Furthermore, in this embodiment of the application, in step S304, the load balancing process includes steps B1 to B3: B1: Collect real-time load data for each resource group, calculate the load variance, and when the variance exceeds the threshold of 0.2, use the Lagrange multiplier method to solve for the optimal allocation and minimize the squared difference between each load and the target load.

[0057] B2: Adjust the traffic allocation weights, verify that the variance is reduced to within the threshold, and combine convolutional neural networks to analyze traffic characteristics and generate a balance benchmark value; the input of the convolutional neural network includes bandwidth utilization, number of packets, number of connections and error rate, and the output is a balance score.

[0058] B3: Construct a link weight matrix, and based on traffic differences and reliability, redistribute cross-level link traffic to generate a stable load distribution state.

[0059] In an optional implementation, in step S304, the load balancing process further includes setting weights based on link bandwidth and real-time load, allocating traffic to each link in a round-robin fashion according to the weights, periodically monitoring load differences, dynamically adjusting weights, and triggering redistribution using threshold detection.

[0060] In another optional implementation, in step S304, the load balancing process may further include using a time series model (ARIMA) to predict traffic trends, pre-allocating bandwidth resources based on the prediction results, dynamically adjusting the allocation strategy in conjunction with real-time monitoring data, and optimizing the prediction model through a feedback loop.

[0061] Furthermore, in step S300, adjusting the resource scheduling awareness parameters includes steps S311~S313: S311: Obtain balanced link load data through the load monitoring system and extract the real-time response time of backup resources from the resource response detection module; S312: A resource status table is constructed using a support vector machine classification algorithm, and status awareness parameters are calculated. Then, a resource response baseline value is generated based on the awareness parameters to guide subsequent scheduling. The feature vectors of the support vector machine classification algorithm include load balancing, response time, CPU utilization, and memory utilization. Kernel functions utilize RBF functions: in, For the sample and Similarity measurement in feature space and Let n and o be the feature vectors of the nth and oth resources. The kernel function parameter is set to 0.1. For Euclidean distance For exponential functions The formula for calculating state-aware parameters is expressed as follows: in, For load stability, In order to maintain consistency, For resource availability, , and These are the weighting coefficients. These are state-aware parameters; resources are categorized into high-quality resources, good resources, and general resources.

[0062] S313: Based on the resource response baseline, compare the actual response time with the expected activation time to generate a latency difference matrix, and reorder the standby resources in combination with node attributes and business priorities. By verifying the adjusted response performance, generate a resource scheduling optimization scheme.

[0063] In step S400, obtaining the resource scheduling optimization scheme includes steps S401 to S403: S401: Extract fault scenarios and recovery data from the fault history database, calculate resource call priority scores, and use a long short-term memory network to extract scenario feature vectors by inputting fault type, network load, response time, and recovery time series. The formula for prioritizing resource allocation is expressed as follows: in, The weight of the k-th factor. Score the priority of resource access. This is the k-th standardized factor value.

[0064] S402: The urgency level of the scene is classified by outputting the probability of the fully connected layer, and a path switching rule set is generated by combining the historical network performance curve.

[0065] S403: Evaluate the success rate of rule set execution, combine historical execution feedback for combined optimization, and generate a set of alternative solutions through simulation testing.

[0066] Furthermore, in step S401, constructing the scene feature vector using a long short-term memory network includes steps C1~C2: C1: Input historical fault data time series, the sequence length is determined according to the sampling interval, use two layers of long short-term memory network to process the input sequence, and extract the hidden state at the last moment as the scene feature vector; The input sequence includes fault type, network load, response time, and recovery time.

[0067] C2: The feature vector is mapped to the output dimension through a fully connected layer, and the probability of the three types of scenes is output using the softmax function. The activation conditions of the path switching rule are adjusted according to the classification results. The feature extraction formula is expressed as: in, For scene feature vectors, For the output of the Long Short-Term Memory network, Index for the last moment.

[0068] Furthermore, in step S400, the generation of the target resource allocation scheme and path switching mode includes steps S411-412: S411: Calculate the comprehensive score of each alternative based on multi-objective optimization, and calculate the cost-benefit ratio.

[0069] S412: Select the scheme with the highest comprehensive score as the final scheme, and use a dynamic adjustment mechanism to update the execution priority according to the real-time network status to verify the performance indicators of the final scheme under simulated fault scenarios. The formula for calculating the overall score of each alternative is expressed as follows: in, For comprehensive scoring, For recovery time, For resource utilization, For business continuity, For cost-effectiveness, the CBR should be greater than or equal to 1.5.

[0070] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

[0071] Example 3, referring to Figure 2 This is the third embodiment of the present invention. This embodiment provides a collaborative control system for provincial power transmission networks, including a status monitoring and analysis module, a resource scheduling optimization module, a permission allocation and load balancing module, and a path optimization and rule generation module.

[0072] The status monitoring and analysis module is used to collect and evaluate the operating status of the power communication network in real time. It acquires real-time data of the entire network through link status monitors and node status collectors, calculates link saturation and node health scores using resource analysis algorithms, analyzes the fault propagation range through graph neural network models, and generates an impact area matrix.

[0073] The resource scheduling optimization module is used to dynamically adjust the resource allocation strategy according to the network status, obtain the real-time response time of the backup node through resource response detection, generate an initial resource score sequence using a genetic algorithm, divide the resource capacity level by combining a support vector machine classification model, optimize the load distribution through the least squares method, and generate the final resource call priority list.

[0074] The permission allocation and load balancing module is used to obtain network-wide permission distribution data, analyze node load characteristics using the random forest algorithm, generate load difference indicators, determine permission adjustment parameters based on support vector machine classification, calculate optimized permission allocation weights through a weight distribution analyzer, analyze traffic characteristics using a convolutional neural network, construct a link weight matrix, and redistribute cross-level link traffic.

[0075] The path optimization and rule generation module is used to obtain the convergence time of the switching path, generate path priority scores using a multilayer perceptron neural network, evaluate the latency performance of different activation orders using a random forest algorithm, and construct scene feature vectors using a long short-term memory network based on historical fault data to generate a path switching rule set, and select the final solution through multi-objective optimization.

[0076] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

[0077] Example 4, the fourth embodiment of the present invention, differs from the previous three embodiments in that: If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0078] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-including system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device.

[0079] More specific examples of computer-readable media (a non-exhaustive list) include: electrical connections (electronic devices) having one or more wires, portable computer disk drives (magnetic devices), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Furthermore, computer-readable media can even be paper or other suitable media on which the program can be printed, because the program can be obtained electronically, for example, by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in computer memory.

[0080] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, multiple steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

Claims

1. A method for coordinated control of provincial and local power transmission networks, characterized in that: include, Real-time monitoring of network status; obtaining information on equipment failures through status monitoring; identifying fault nodes; conducting fault propagation analysis; obtaining the scope of fault impact and resource availability assessment; based on the scope of fault impact and resource availability assessment; optimizing and adjusting the order of backup resource calls through resource scheduling; extracting current load data; and generating a resource call priority list in conjunction with load balancing. Obtain permission level configuration data, detect uneven load distribution across layers through load difference analysis, adjust the permission allocation scheme according to the resource call priority list, obtain the optimized permission level allocation result, and based on the optimized permission level allocation result, obtain convergence requirements, analyze the activation order of backup paths and link latency data, and determine the adjustment sequence for accelerating path convergence. Extract real-time traffic data, combine it with path convergence acceleration adjustment sequence, adopt load balancing processing, recalculate the traffic distribution of cross-level provincial links, obtain the balanced link load distribution status, and combine the backup resource activation time requirements to adjust the resource scheduling perception parameters using the real-time status perception process, obtain the adjusted resource response time data, and obtain the resource scheduling optimization scheme based on the difference between the resource response time and the expected activation time. The resource call priority list and path convergence acceleration adjustment sequence are iteratively calculated. By analyzing historical data, feedback information on network performance and fault recovery effect is extracted from historical fault scenario data to generate target resource allocation scheme and path switching mode.

2. The provincial-level power transmission network collaborative control method as described in claim 1, characterized in that: The real-time network status monitoring includes collecting real-time link data from the entire network, processing the real-time link data, calculating link saturation and node health scores, and evaluating the overall network operating status. Based on network topology and node characteristics, the potential propagation paths of faults are analyzed, an impact area matrix is ​​generated, the sequence of affected nodes and the distribution of backup resources are identified, real-time bandwidth utilization data is obtained, and combined with resource availability judgment, the scope of fault impact and resource allocation requirements are determined.

3. The provincial-level power transmission network collaborative control method as described in claim 2, characterized in that: The adjustment of the backup resource call order includes obtaining real-time response time data of backup resource nodes and generating an initial resource scoring sequence based on load balancing requirements; The initial resource scoring sequence is iteratively optimized to minimize response time and balance load. Based on resource utilization data, resource capacity levels are divided, and the initial call order is determined according to business needs. The system obtains real-time load values ​​for each resource group, compares them with preset capacity allocation thresholds, optimizes load distribution, and generates a priority sequence for backup resource calls.

4. The provincial-level power transmission network collaborative control method as described in claim 3, characterized in that: The adjusted permission allocation scheme includes acquiring network-wide permission distribution data, recording real-time load values ​​of links at each level, analyzing node load and bandwidth usage characteristics, and generating load difference indicators. The balance of cross-level load distribution is judged based on the indicators, and the backup resource data is extracted from the resource call priority sequence. The permission adjustment parameters are determined in combination with the link saturation. Calculate the adjusted permission allocation weights and generate the optimized permission scheme.

5. The provincial-level power transmission network collaborative control method as described in claim 4, characterized in that: The adjustment sequence for determining path convergence acceleration includes: obtaining the convergence time of switching paths through path monitoring, comparing with a preset threshold to filter the set of paths to be optimized, and using a multilayer perceptron neural network to input path stability, link delay, hop count and historical success rate features, outputting path priority scores, and generating an initial score sequence. A preliminary adjustment plan is generated based on bandwidth usage and target convergence time. The random forest algorithm is then used to predict the total latency value of each activation sequence by taking into account path combination ID, activation sequence number, traffic pattern and network state characteristics. Calculate a comprehensive score by grouping according to business priority, select the highest score order to generate a path activation time sequence table, and determine the path execution sequence based on the time sequence table; The formula for calculating the overall score is expressed as follows: in, For the overall score, To predict delays, This is the business weight.

6. The provincial-level power transmission network collaborative control method as described in claim 5, characterized in that: The recalculation of cross-level provincial links includes: acquiring real-time traffic data of network nodes, calculating the bandwidth ratio of cross-level links, and using a convolutional neural network to input bandwidth utilization, number of data packets, number of connections, and error rate features, outputting a balance score. When the balance score is greater than a preset threshold, the traffic difference between levels is calculated, a link weight matrix is ​​constructed, the deviation between the current traffic and the preset threshold is analyzed, a traffic adjustment scheme is generated in combination with the path convergence time parameter, and cross-level link traffic is redistributed to verify the balance effect. When the load variance drops to within 0.2, a stable load distribution state is confirmed. The formulas for calculating the bandwidth ratio of cross-layer links and the traffic difference between layers are expressed as follows: in, This refers to the bandwidth capacity of the upper-level link. For the bandwidth capacity of the lower-level links, This represents the bandwidth ratio of cross-layer links. Due to differences in traffic between levels, The real-time traffic value for level i. The real-time traffic value for level j. and For hierarchical indexes; The adjustment of resource scheduling perception parameters includes: acquiring balanced link load data, extracting the real-time response time of backup resources, using a support vector machine classification algorithm, inputting load balancing degree, response time, CPU utilization and memory utilization features, classifying resources into three categories: high-quality resources, good resources and general resources, calculating state perception parameters, generating resource response benchmark values, comparing the actual response time with the expected activation time, generating a latency difference matrix, reordering backup resources based on node attributes and business priorities, verifying the adjusted response performance, and generating a resource scheduling optimization scheme. The formula for calculating state-aware parameters is expressed as follows: in, For load stability, In order to maintain consistency, For resource availability, , and These are the weighting coefficients. These are state-aware parameters.

7. The provincial-level power transmission network collaborative control method as described in claim 6, characterized in that: The obtained resource scheduling optimization scheme includes, The system extracts fault scenarios and recovery data from the fault history database, calculates resource call priority scores, and uses a long short-term memory network to extract scenario feature vectors by inputting fault type, network load, response time, and recovery time series. The system outputs the scenario urgency classification probability through a fully connected layer and generates a path switching rule set by combining it with historical network performance curves. The system evaluates the success rate of rule set execution, combines historical execution feedback for combined optimization, and generates a set of alternative solutions through simulation testing. The formula for prioritizing resource allocation is expressed as follows: in, The weight of the k-th factor. Score the priority of resource access. This represents the k-th standardized factor value; The construction of scene feature vectors using a long short-term memory network includes inputting a time series of historical fault data, with the sequence length determined by the sampling interval. The input sequence is processed using a two-layer long short-term memory network, and the hidden state at the last moment is extracted as the scene feature vector. The feature vector is then mapped to the output dimension through a fully connected layer. The probability of three types of scenes is output using the softmax function, and the activation conditions of the path switching rule are adjusted according to the classification results. The generation of target resource allocation scheme and path switching mode includes: calculating the comprehensive score of each alternative scheme based on multi-objective optimization, calculating the cost-benefit ratio, selecting the scheme with the highest comprehensive score as the final scheme, updating the execution priority according to the real-time network status through a dynamic adjustment mechanism, and verifying the performance indicators of the final scheme under simulated fault scenarios. The formula for calculating the overall score of each alternative is expressed as follows: in, For comprehensive scoring, For recovery time, For resource utilization, For business continuity, Cost-effectiveness ratio.

8. A provincial-level power transmission network collaborative control system, employing the provincial-level power transmission network collaborative control method as described in any one of claims 1 to 7, characterized in that, It includes a status monitoring and analysis module, a resource scheduling optimization module, a permission allocation and load balancing module, and a path optimization and rule generation module; The status monitoring and analysis module is used to collect and evaluate the operating status of the power communication network in real time. It acquires real-time data of the entire network through link status monitors and node status collectors, calculates link saturation and node health scores using resource analysis algorithms, analyzes the fault propagation range through graph neural network models, and generates an impact area matrix. The resource scheduling optimization module is used to dynamically adjust the resource allocation strategy according to the network status, obtain the real-time response time of the backup node through resource response detection, generate an initial resource score sequence using a genetic algorithm, divide the resource capacity level by combining a support vector machine classification model, optimize the load distribution through the least squares method, and generate the final resource call priority list. The permission allocation and load balancing module is used to obtain network-wide permission distribution data, analyze node load characteristics using the random forest algorithm, generate load difference indicators, determine permission adjustment parameters based on support vector machine classification, calculate optimized permission allocation weights through a weight distribution analyzer, analyze traffic characteristics using a convolutional neural network, construct a link weight matrix, and redistribute cross-level link traffic. The path optimization and rule generation module is used to obtain the convergence time of the switching path, generate path priority scores using a multilayer perceptron neural network, evaluate the latency performance of different activation orders using a random forest algorithm, and construct scene feature vectors using a long short-term memory network based on historical fault data to generate a path switching rule set, and select the final solution through multi-objective optimization.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the collaborative control method for provincial power transmission networks according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the collaborative control method for provincial power transmission networks according to any one of claims 1 to 7.

Images