Power dispatching data network resource intelligent matching method, system, equipment and medium

Abstract

The invention relates to the technical field of resource matching, in particular to an intelligent power dispatching data network resource matching method, system, equipment and medium, comprising the following steps: extracting network node state measurement consumption and collecting service requirements, splicing feature vectors for service attributes and calculating distances to screen a matching resource pool; the states are recombined into a node matrix, depth features are extracted through a graph network, a resource pool feature mapping flow table is combined to deduce a path and a distribution strategy, and a logic is constructed for device configuration according to the strategy to generate a scheduling result. According to the method, distance evaluation and neighborhood aggregation are performed on a multi-source service request through fusion of dynamic topology, global state constraints are extracted, interference terms are eliminated, optimal distribution is explored in an enhanced deduction space in combination with a multi-dimensional observation sequence, and a mapping relation between a service boundary and equipment configuration is directly established based on decision evolution. Dependence of passive fixed threshold and artificial log research and judgment is abandoned, and the risk of flow cutoff and congestion caused by intervention delay is weakened so as to guarantee communication of a communication link.

Description

Technical Field

[0001] This invention relates to the field of resource matching technology, and in particular to a method, system, device and medium for intelligent resource matching in power dispatching data networks. Background Technology

[0002] The field of resource matching technology involves a technical system for establishing a correspondence between a limited set of resources and a set of demands. It covers core aspects such as attribute modeling of resource elements, structured expression of demand characteristics, formulation of matching rules, execution control of the matching process, and adjustment and updating of matching results. The whole revolves around the collection, organization, comparison, and association of multi-source data. By uniformly encoding and storing information such as resource capacity, type, location, time window, and usage constraints, and combining the quantity requirements, performance indicators, priority order, and constraints on the demand side, an executable matching process is constructed. In the information system, operations such as resource allocation, path determination, relationship mapping, and result output are completed. It is widely used in scenarios such as power system dispatching, communication network configuration, logistics allocation, and computing resource allocation.

[0003] The traditional intelligent resource matching method for power dispatch data networks refers to a processing approach that allocates bandwidth resources, link channels, node ports, transmission time slots, and service priorities in a power dispatch data network environment. The technical issue it addresses is determining specific link combinations and bandwidth usage patterns based on service type, real-time level, data traffic volume, and network topology during power dispatch service access, data transmission path establishment, and capacity allocation. Traditional methods typically involve pre-establishing network topology tables, link capacity tables, and service demand tables. Based on manually set priority orders and fixed threshold rules, each service request is searched for available links level by level according to node order. Under the premise of meeting bandwidth, port status, and transmission delay conditions, a specific link path is selected, and the occupied port number, link number, and corresponding bandwidth value are recorded in the resource allocation record table. Simultaneously, the allocated resources are periodically manually checked and adjusted based on the operation log to complete the matching process for dispatch data network resources.

[0004] Traditional matching methods typically rely on manually preset fixed threshold rules and static priority order to perform node retrieval. When faced with complex and ever-changing network operating environments and massive concurrent requests, they are difficult to dynamically adapt and adjust. Mechanical step-by-step comparisons can easily lead to link congestion or resource idleness. At the same time, the verification and maintenance in the later stages of allocation rely heavily on manual periodic analysis of operation logs. This lagging and passive management mode not only consumes huge manpower costs, but also causes communication delays or even interruptions due to untimely intervention, which cannot meet the stringent transmission requirements of modern power business. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of existing technologies and to propose a method, system, equipment and medium for intelligent matching of power dispatch data network resources.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a method for intelligent matching of power dispatch data network resources, comprising the following steps:

[0007] S1: Obtain the topology of the power dispatch data network, total bandwidth of physical links, remaining available bandwidth of the network, and node memory occupancy rate through network probes; calculate the bandwidth consumed by the links; collect service voltage level parameters, terminal equipment type parameters, service application scenario parameters, service guarantee bandwidth threshold, and maximum allowable communication latency to obtain service processing data.

[0008] S2: Based on the business data, splice together the business voltage level parameters, terminal device type parameters and business application scenario parameters to construct a feature vector, calculate the distance between the feature vector and the node attribute values, and filter the matching resource pool in ascending order according to the attribute distance values.

[0009] S3: Based on the business data, perform data reorganization, construct a network topology node matrix, input the network topology node matrix into a graph convolutional neural network model to aggregate node feature information, and obtain a resource status feature matrix;

[0010] S4: Perform flow table rule mapping on the matching resource pool and resource status feature matrix to construct a network forwarding flow table. Input the network forwarding flow table, service guarantee bandwidth threshold and maximum allowable communication latency into the deep enhanced network model to perform Markov decision process deduction to obtain data transmission routing path and bandwidth resource allocation strategy.

[0011] S5: Use the data transmission routing path to filter nodes and obtain the transmission node device identifier. Match the device identifier to the configuration list according to the bandwidth resource allocation strategy and generate a resource scheduling matching result.

[0012] The improvements of this invention include: the service processing data includes service load throughput, protocol handshake overhead, and reserved margin; the matching resource pool includes host network segment addresses, socket port tables, and concurrent access quotas; the resource status feature matrix includes congestion risk coefficients, connectivity probability values, and device forwarding loads; the bandwidth resource allocation strategy includes peak guaranteed rate, queuing priority weight, and drop / interception threshold; and the resource scheduling matching result includes packet encapsulation labels, frame sequence levels, and bridge isolation identifiers.

[0013] The present invention is improved in that the steps for obtaining the business data are specifically as follows:

[0014] S111: Obtain feedback messages from pre-set nodes of network probes, perform protocol layer segment extraction and parsing on the feedback messages to read the topology of the power dispatch data network, the total bandwidth of physical links, the remaining available bandwidth of the network, and the node memory occupancy rate, and call the basic computing component to perform same-level subtraction logic calculation on the total bandwidth of physical links based on the remaining available bandwidth of the network to generate the bandwidth consumed by the links.

[0015] S112: Monitor the service request data frames reported by the underlying network interface, extract service voltage level parameters, terminal device type parameters, service application scenario parameters, service guarantee bandwidth threshold and maximum allowable communication latency for the request data frames, and perform feature vector association and splicing calculation based on the basic attribute dimensions for service voltage level parameters, terminal device type parameters, service application scenario parameters, service guarantee bandwidth threshold and maximum allowable communication latency to establish a service requirement feature set;

[0016] S113: Based on the bandwidth consumption of the link, perform mapping architecture topology construction for the service requirement feature set, and perform internal data dimension alignment and reorganization according to the network communication node identification level and the distribution format of the underlying mapping nodes to obtain the service data.

[0017] The present invention is improved in that the step of obtaining the matching resource pool is specifically as follows:

[0018] S211: Call the business data, perform field extraction, read the business voltage level parameters, terminal device type parameters and business application scenario parameters, perform matrix splicing transformation on the business voltage level parameters, terminal device type parameters and business application scenario parameters, and generate attribute feature vectors;

[0019] S212: Obtain the node resource attribute identifier, link available bandwidth, node response latency, and network protocol overhead associated with the fourth-level Internet Protocol resource pool. Combine the attribute feature vector and evaluate the obtained attribute distance value by considering the difference between the attribute feature vector and the node resource attribute identifier, and the influence of link available bandwidth and node response latency.

[0020] S213: Perform sequence comparison for the Level 4 Internet Protocol resource pool, sort in ascending order according to the attribute distance values, generate a node sorting sequence, and perform truncation and filtering on the node sorting sequence to obtain the matching resource pool.

[0021] The present invention is improved in that the steps for obtaining the resource status feature matrix are as follows:

[0022] S311: Based on the aforementioned business data, perform communication node identifier extraction and connection segment routing analysis on the power dispatch data network topology, arrange the nodes according to the connection form of the underlying equipment, and establish a node connection matrix;

[0023] S312: Obtain the total bandwidth of the communication link, the remaining available bandwidth of the network, the bandwidth occupied by the link, and the memory usage rate of the node. Based on the communication node identifier, perform attribute value splicing mapping on the total bandwidth of the communication link, the remaining available bandwidth of the network, the bandwidth occupied by the link, and the memory usage rate of the node to establish a network topology node matrix.

[0024] S313: Call the node connection matrix and network topology node matrix, perform feature dimension multiplication and addition transformation operation on the network topology node matrix based on the node connection matrix, call the weight parameter set to perform nonlinear transformation on the transformation output value, aggregate neighbor distribution attributes, and generate resource status feature matrix.

[0025] The present invention is improved in that the specific steps for obtaining the bandwidth resource allocation strategy are as follows:

[0026] S411: Call the matching resource pool and resource status feature matrix, extract the node access address and destination network segment identifier for the matching resource pool, extract the link capacity vector according to the resource status feature matrix, perform addressing logic association for the node access address, destination network segment identifier and link capacity vector, write association information to the protocol action instruction field, and establish a network forwarding flow table.

[0027] S412: Obtain the service guarantee bandwidth threshold and the maximum allowable communication delay, read the connectivity topology branch for the network forwarding flow table and define it as an environment observation sequence, define the node switching action as a system action set, construct the state reward evaluation function based on the service guarantee bandwidth threshold and the maximum allowable communication delay, integrate the environment observation sequence, the system action set and the state reward evaluation function, and establish a Markov inference architecture.

[0028] S413: Invoke the Markov inference architecture, perform multi-round simulation inference calculations, calculate the cumulative reward value for the system action set based on the state reward evaluation function, perform gradient update of network weight parameters based on the cumulative reward value, and perform inverse mapping decoding on the converged state network policy output matrix to obtain the data transmission routing path and bandwidth resource allocation strategy.

[0029] The present invention is improved in that the step of obtaining the resource scheduling matching result is specifically as follows:

[0030] S511: Perform network layer traversal scan to read node attributes for the data transmission routing path, call preset protocol features to perform feature matching and comparison operations on node attributes, eliminate access terminals and extract relay device addresses, perform hash encoding and compression operations on relay device addresses, and generate transmission node device identifiers.

[0031] S512: Based on the transmission node device identifier, retrieve the underlying asset database, execute a pull command to read the interface model, network throughput, queue capacity, and mapping rules for the underlying asset database, perform parameter matrix reorganization for the interface model, network throughput, queue capacity, and mapping rules, and establish a network forwarding device configuration list;

[0032] S513: Invoke the bandwidth resource allocation strategy and network forwarding device configuration list, extract the rate threshold and service queue quota recorded in the bandwidth resource allocation strategy, perform resource capacity constraint judgment on the network forwarding device configuration list based on the rate threshold and service queue quota, allocate service forwarding labels, isolation identifiers and transmission queuing sequences to the network interface according to the judgment matrix, aggregate the service forwarding labels, isolation identifiers and transmission queuing sequences, and obtain the resource scheduling matching result.

[0033] A power dispatch data network resource intelligent matching system is provided, the power dispatch data network resource intelligent matching system being used to implement the above-mentioned power dispatch data network resource intelligent matching method, the system comprising:

[0034] The data extraction and processing module obtains the topology of the power dispatch data network, the total bandwidth of physical links, the remaining available bandwidth of the network, and the node memory occupancy rate through network probes. It calculates the bandwidth consumed by the links, collects service voltage level parameters, terminal equipment type parameters, service application scenario parameters, service guarantee bandwidth thresholds, and maximum allowable communication latency, and obtains service processing data.

[0035] The resource pool allocation module, based on the business data, splices together the business voltage level parameters, terminal device type parameters and business application scenario parameters to construct a feature vector, calculates the distance value between the feature vector and the node attribute, and filters the matching resource pools in ascending order according to the attribute distance value.

[0036] The resource status analysis module reorganizes the data based on the business data, constructs a network topology node matrix, and inputs the network topology node matrix into a graph convolutional neural network model to aggregate node feature information to obtain a resource status feature matrix.

[0037] The path bandwidth adjustment module performs flow table rule mapping on the matching resource pool and resource status feature matrix, constructs a network forwarding flow table, and inputs the network forwarding flow table, service guarantee bandwidth threshold and maximum allowable communication latency into the deep reinforced network model to perform Markov decision process deduction to obtain data transmission routing path and bandwidth resource allocation strategy.

[0038] The resource matching and analysis module uses the data transmission routing path to filter nodes and obtain the transmission node device identifier. Based on the bandwidth resource allocation strategy, it matches the device identifier with the configuration list and generates a resource scheduling matching result.

[0039] A computer device includes a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the intelligent matching system for power dispatch data network resources as described above.

[0040] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the intelligent matching method for power dispatch data network resources as described above.

[0041] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0042] In this invention, a deep fusion analysis is performed on multi-source service requests and network topology features. A candidate resource space is constructed based on the geometric distance evaluation mechanism between feature vectors to eliminate redundant interference terms. A graph convolutional architecture is used to perform neighborhood aggregation on node feature information to extract global network state constraints. Multi-dimensional environmental observation sequences are injected into an enhanced inference framework to explore optimal solutions. Resource allocation strategies and routing directions are dynamically adjusted based on Markov evolution processes. The passive fixed threshold judgment framework and reliance on manual inspection are completely abandoned. The optimal mapping relationship between service communication boundaries and node forwarding configurations is directly calculated, reducing the risk of congestion and disconnection caused by delays in manual intervention and limited local visibility, and ensuring that the transmission link is in a robust connectivity state. Attached Figure Description

[0043] Figure 1 This is a flowchart of the method of the present invention;

[0044] Figure 2 This is a flowchart illustrating the process of acquiring business data in this invention;

[0045] Figure 3 This is a flowchart illustrating how the present invention obtains a matching resource pool;

[0046] Figure 4 This is a flowchart illustrating how the present invention obtains the resource status feature matrix;

[0047] Figure 5 This is a flowchart illustrating the bandwidth resource allocation strategy of the present invention;

[0048] Figure 6 This is a flowchart illustrating how the resource scheduling matching results are obtained in this invention. Detailed Implementation

[0049] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0050] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0051] Please see Figure 1 This invention provides a technical solution, a method for intelligent matching of power dispatch data network resources, comprising the following steps:

[0052] S1: Obtain the topology of the power dispatch data network, total physical link bandwidth, remaining available network bandwidth, and node memory usage rate through network probes. Calculate the bandwidth consumed by the link based on the difference between the total physical link bandwidth and the remaining available network bandwidth. Call the system management interface to collect service voltage level parameters, terminal device type parameters, service application scenario parameters, service guarantee bandwidth threshold, and maximum allowable communication latency to obtain service processing data.

[0053] S2: Based on the business data, perform data splicing operations on business voltage level parameters, terminal device type parameters, and business application scenario parameters to construct attribute feature vectors. Use the Euclidean distance algorithm to evaluate the distance of node resource attribute identifiers within the Level 4 Internet Protocol resource pool based on the attribute feature vectors to obtain attribute distance values. Sort and filter in ascending order based on attribute distance values ​​to obtain matching resource pools.

[0054] S3: Based on the power dispatch data network topology, total physical link bandwidth, remaining available network bandwidth, link bandwidth consumption, and node memory utilization, a network topology node matrix is ​​constructed by data reorganization. The network topology node matrix is ​​then input into a graph convolutional neural network model to aggregate node feature information and obtain a resource status feature matrix.

[0055] S4: Perform flow table rule mapping operations on the matching resource pool and resource status feature matrix to construct a network forwarding flow table. Input the network forwarding flow table, service guarantee bandwidth threshold and maximum allowable communication latency into the deep enhanced network model to perform Markov decision process inference and calculation to obtain the data transmission routing path and bandwidth resource allocation strategy.

[0056] S5: Filter network link nodes for the data transmission routing path to obtain the transmission node device identifier. Based on the bandwidth resource allocation policy, perform matching logic on the network forwarding device configuration list pointed to by the transmission node device identifier to generate resource scheduling matching results.

[0057] The business data includes business load throughput, protocol handshake overhead, and reserved margin; the matching resource pool includes host network segment addresses, socket port tables, and concurrent access quotas; the resource status feature matrix includes congestion risk coefficients, connectivity probability values, and device forwarding load; the bandwidth resource allocation strategy includes peak guaranteed rate, queuing priority weight, and drop / interception threshold; and the resource scheduling matching results include packet encapsulation labels, frame sequence levels, and bridge isolation identifiers.

[0058] Please see Figure 2 The specific steps for obtaining business data are as follows:

[0059] S111: Obtain feedback messages from pre-set nodes of network probes, perform protocol layer segment extraction and parsing on the feedback messages to read the topology of the power dispatch data network, the total bandwidth of physical links, the remaining available bandwidth of the network, and the node memory occupancy rate, and call the basic computing component to perform same-level subtraction logic calculation on the total bandwidth of physical links based on the remaining available bandwidth of the network to generate the bandwidth consumed by the links.

[0060] Data retrieval commands based on Simple Network Management Protocol version 3 (SMMP3) are issued to 50 network probe devices deployed in the aggregation layer of the power dispatch data network. The probe devices send User Datagram Protocol (UDP) data packets containing status feedback messages from underlying physical nodes to the network management control plane at fixed intervals of 500 milliseconds. For the received feedback messages, header stripping is performed to extract Extensible Markup Language (XML) format data from the application layer payload. Tag tree parsing is then performed on this XML data to read the tags. <topology>Using the link interconnection identifier array, construct a directed graph data structure for the network topology, with node media access control addresses as vertices and link physical connections as edges. (Location) <totalbandwidth>Tags and <availablebandwidth>The tags are used to extract the total physical link bandwidth and remaining available network bandwidth for each interface. This is achieved through parsing... <memusage>The tags retrieve the memory usage percentage of each network node. The extracted data needs to be cleaned to remove null and outlier values, and the bandwidth data is converted to the standard unit of megabits per second (Mbps). The pre-set subtraction logic register in the basic arithmetic component is called, inputting the total physical link bandwidth of 10,000 Mbps and the remaining network bandwidth of 7,500 Mbps into the arithmetic logic unit to perform a subtraction operation of the same unit. The calculation is 10,000 - 7500 = 2500, yielding the current link's bandwidth consumption of 2,500 Mbps. The acquired network node characteristic data is written to a specific state database, with specific parameters shown in Table 1.

[0061] Table 1. Initial characteristic parameters of network nodes:

[0062]

[0063] As shown in Table 1, by extracting and quantifying the feedback data from each probe, a deterministic environmental baseline state reference can be provided for the subsequent allocation of network resources.

[0064] S112: Monitor the service request data frames reported by the underlying network interface, extract service voltage level parameters, terminal device type parameters, service application scenario parameters, service guarantee bandwidth threshold and maximum allowable communication latency for the request data frames, and perform feature vector association and splicing calculation based on the basic attribute dimensions for service voltage level parameters, terminal device type parameters, service application scenario parameters, service guarantee bandwidth threshold and maximum allowable communication latency to establish a service requirement feature set;

[0065] The packet capture interface of the data plane development kit is invoked to continuously intercept uplink service request Ethernet data frames in promiscuous mode on the physical port of the access switch in the power dispatching network. Deep packet inspection technology is used to locate the offset of the transmission control protocol payload of the service request data frame, and the first 128 bytes of the payload are extracted as the service identification field. From the byte range of offset 16 to 19, a 16-bit integer value representing the service voltage level is read; for example, a value of 220 represents a 220 kV voltage level. From the byte range of offset 20 to 23, the terminal device type parameter is parsed, quantizing and mapping the measurement and control device, protection relay, and environmental monitoring terminal to integer scalars 1, 2, and 3, respectively. From the byte range of offset 24 to 27, the service application scenario parameter is read, quantizing differential protection, automated acquisition, and video inspection to discrete values ​​of 10, 20, and 30, respectively. From the bytes of offset 28 to 35, the service guarantee bandwidth threshold is extracted, for example, a value of 50, in megabits per second. The maximum allowable communication latency, e.g., 20, is extracted from bytes with offsets from 36 to 43, in milliseconds. The extracted scalar data is then pushed sequentially into a one-dimensional array memory space of a preset length of 5, based on the fundamental attribute dimensions of voltage level, equipment type, application scenario, bandwidth threshold, and latency constraint. Taking a 220 kV differential protection service as an example, the specific elements pushed into the array are 220, 2, 10, 50, 20, generating a feature vector of the form [220, 2, 10, 50, 20]. The feature vectors corresponding to 1000 consecutive service requests are written into a two-dimensional matrix cache in chronological order according to their timestamps, forming a 1000×5 dimension service requirement feature set, providing a structured demand-side input source for subsequent resource scheduling and attribute matching.

[0066] S113: Based on the bandwidth consumption of the link, the mapping architecture topology is constructed according to the feature set of business requirements. According to the network communication node identification level, the internal data dimension alignment and reorganization are performed on the distribution format of the underlying mapping nodes to obtain the business data.

[0067] A 1000×5 service requirement feature set is used as the source node requirement payload, and the 2500 Mbps link bandwidth consumed by the S111 algorithm is used as the network edge weight constraint. A variant of the shortest path first algorithm is used to traverse the directed graph data structure of the network topology. During the traversal, a condition is determined for each candidate link: if the available capacity of the candidate link, after subtracting the 2500 Mbps bandwidth consumed by the total bandwidth of 10000 Mbps, is less than the 50 Mbps service guarantee bandwidth threshold in the feature vector, then the connectivity of the link is set to blocked. For the remaining physical topology paths after filtering by available capacity, based on the hierarchical relationship from the network layer to the data link layer in the Open Systems Interconnection model, the 32-bit Internet Protocol addresses of all communication nodes along the path are extracted as node identifiers. According to the distribution format from the source node, relay node to the destination node, the attribute fields in the service requirement feature set and the bandwidth status fields of the passing nodes are aligned to the internal dimension byte. The original discrete business requirement feature vectors are horizontally concatenated with the corresponding mapping node bandwidth arrays using a matrix, eliminating redundant header check bits and padding fields. The merged data blocks are serialized into a general object representation format text, and finally solidified in memory as a business organized data object containing business requirements and topology mapping relationships.

[0068] Please see Figure 3 The specific steps for obtaining the matching resource pool are as follows:

[0069] S211: Call the business data to perform field extraction, read the business voltage level parameters, terminal device type parameters and business application scenario parameters, perform matrix splicing transformation on the business voltage level parameters, terminal device type parameters and business application scenario parameters, and generate attribute feature vectors;

[0070] The generated generic object representation format business data object is retrieved from the memory address pool, and a key-value pair parsing function is used to perform high-frequency scanning of the object's internal fields. The business voltage level parameter (key: VoltageLevel), the terminal device type parameter (key: DeviceType), and the business application scenario parameter (key: SceneClass) are accurately extracted using key-name matching. For the extracted set of values, such as voltage level parameter 110, device type parameter 1, and application scenario parameter 20, these three scalar values ​​are placed in a temporary register and injected into a 3x3 column vector data structure according to fixed dimension coordinates (x, y, z). The matrix transpose function is used to convert this column vector into a 1×3 row vector format, completing the matrix concatenation transformation. Finally, an attribute feature vector with constant dimensions and containing specific element values ​​[110, 1, 20] is generated in the cache area, serving as the core retrieval parameter for subsequent resource pool matching.

[0071] S212: Obtain the node resource attribute identifier, link available bandwidth, node response latency, and network protocol overhead rate associated with the Level 4 Internet Protocol resource pool, and combine them with the attribute feature vector using the following formula:

[0072] ;

[0073] Calculate and obtain the attribute distance value;

[0074] in, Represents the distance value of the attribute. The normalized scalar value representing the attribute feature vector is obtained by performing matrix concatenation transformation on the business voltage level parameter, terminal device type parameter, and business application scenario parameter, and then using the min-max normalization method to convert and read the matrix feature values. The normalized value of the feature mapping representing the node resource attribute identifier is obtained by extracting the node resource attribute identifier from the underlying node database associated with the Level 4 Internet Protocol resource pool, and then performing numerical feature mapping and range standardization processing. The normalized value representing the available bandwidth of the link is obtained by monitoring the underlying physical link's bandwidth margin data through network probes, and then converting it using a normalization function to eliminate the influence of broadband unit dimensions. The normalized value representing the node response latency is obtained by calling the system management interface to monitor the historical round-trip time statistics of network nodes, and then performing a linear scaling transformation to truncate the data and eliminate the time unit dimension. Represents the network protocol overhead rate, which is a dimensionless percentage value. It is obtained by parsing the data frame packets fed back by the network transmission node devices and extracting the ratio of the packet header byte length to the total data packet length.

[0075] The system connects to the relational database corresponding to the Level 4 Internet Protocol (IP) resource pool using Structured Query Language (SCL) commands, and performs a full table scan on the active node registry within the database. A unique 64-bit hash string is extracted from each underlying node as its resource attribute identifier. Simultaneously, the system calls the traffic monitoring interface to obtain the measured available bandwidth of the physical ports corresponding to each node, and uses the network control protocol (NCP) response probe to obtain the round-trip time of data packets as the node response latency. Furthermore, by analyzing captured NCP packets, the ratio of the 20-byte header length to the 1500-byte total data packet length (20 / 1500 = 0.0133) is calculated to obtain the network protocol overhead rate. The specific parameter mapping results obtained from the Level 4 IP resource pool are shown in Table 2.

[0076] Table 2. Physical performance parameters of resource pool nodes:

[0077]

[0078] For the generated attribute feature vector [110,1,20], the maximum-min normalization formula (x-min) / (max-min) is used for conversion. The maximum and minimum voltage levels for the entire network are set to 1000 and 10 respectively. The normalized result is recorded as 0.82, which represents... After extracting the aforementioned hash identifiers from the underlying node database, they are transformed into numerical features using a specific hash value modulo rule, and then normalized values ​​are obtained using the same range standardization process. For example, the calculated value is 0.65. The available link bandwidth in Table 2 is 8500 megabits per second. Using a fixed division function with a maximum bandwidth of 10000 to eliminate the megabits per second unit, the calculated value is... =0.85. The system management interface is called to obtain the node response latency of 12 milliseconds. This latency is then divided by the maximum tolerable latency of 50 milliseconds to perform a linear scaling truncation, resulting in the calculated value. =0.24. The obtained network protocol overhead rate is 0.0133, which is directly recorded as... .

[0079] Import the above specific parameter values ​​into the formula for calculation. The calculation process is as follows:

[0080] ;

[0081] The numerical result of 0.7892 represents the multidimensional spatial adaptation distance between the current service attribute requirements and the resource status of a specific network node. The smaller this distance value, the higher the degree of fit between the node and the current service's transmission requirements. This distance value will directly serve as the sole quantitative criterion for the node ranking and selection process described below. The advantage of the formula lies in incorporating network protocol overhead. The joint nonlinear operation with the latency-to-bandwidth ratio significantly amplifies the penalty weight of high-latency, low-bandwidth nodes in distance calculation.

[0082] S213: Perform sequence comparison for the Level 4 Internet Protocol resource pool, sort in ascending order based on attribute distance values, generate a node sorting sequence, and perform truncation and filtering on the node sorting sequence to obtain the matching resource pool.

[0083] A quicksort algorithm is applied to the calculated distance values ​​(D) of 1000 nodes within the Level 4 Internet Protocol (IP) resource pool. These 1000 nodes are sorted in ascending order based on their D values, generating a node sorting sequence containing the hash identifiers of 1000 nodes and their corresponding D values. The sequence truncation hyperparameter is set to 30% of the first digits. This 30% ratio was determined through Monte Carlo simulations of 100,000 historical network congestion scenarios, achieving an optimal balance between computational overhead and matching success rate. This 30% ratio is applied to the node sorting sequence, removing the last 700 nodes with large distance values ​​and retaining the first 300 network communication nodes with the smallest D values ​​and highest fit. The information of these 300 selected nodes is packaged and written into a dedicated high-priority isolated area of ​​random access memory (RAM), forming a matching resource pool to provide high-density resource entities for the subsequent construction of the topology state matrix.

[0084] Please see Figure 4 The specific steps for obtaining the resource status feature matrix are as follows:

[0085] S311: Based on business data processing, perform communication node identifier extraction and connection segment routing analysis for the power dispatch data network topology, arrange the nodes according to the connection form of the underlying equipment, and establish a node connection matrix;

[0086] The system parses the business data objects in memory and extracts the internally encapsulated global route tracing record fields. Regular expression matching is performed on the record fields to accurately extract all 32-bit Internet Protocol addresses as communication node identifiers. For the extracted set of node identifiers, the system sequentially reads the next-hop forwarding table entries between each node to parse the physical fiber optic or twisted-pair cable connection routes. Based on the actual physical connection forms of the underlying devices in the network access layer, aggregation layer, and core layer, a 300×300 zero matrix is ​​established in the system backend. When a direct connection is found between node A (identified as 192.168.10.1) and node B (identified as 192.168.10.2), the values ​​of the elements in row A, column B and row B, column A of the matrix are changed from 0 to 1. This process is repeated for all connections, thus constructing a complete undirected node connection matrix structure that maps the underlying physical connectivity.

[0087] S312: Obtain the total bandwidth of the communication link, the remaining available bandwidth of the network, the bandwidth occupied by the link, and the memory usage rate of the node. Based on the communication node identifier, perform attribute value splicing mapping on the total bandwidth of the communication link, the remaining available bandwidth of the network, the bandwidth occupied by the link, and the memory usage rate of the node to establish a network topology node matrix.

[0088] A simple network management protocol is used to initiate status polling requests to 300 nodes within the matching resource pool. Accurate data is collected on each device's total communication link bandwidth (e.g., 10,000 Mbps), remaining available network bandwidth (e.g., 6,500 Mbps), bandwidth consumed by the link (e.g., 3,500 Mbps), and node memory usage (e.g., 42%) returned through the operating system performance monitoring interface. These data, with their different dimensions, are standardized using Z-scores. For each communication node identifier, the corresponding four standardized values ​​are sequentially placed into a one-dimensional tensor of length 4, and attribute value concatenation mapping is performed. This concatenation operation is repeated for all 300 nodes, stacking them in memory to generate a 300×4 two-dimensional floating-point array. This array rigorously records the real-time physical resource capacity and load distribution of key nodes across the entire network, thereby establishing the network topology node matrix.

[0089] S313: Call the node connection matrix and network topology node matrix, perform feature dimension multiplication and addition transformation operation on the network topology node matrix based on the node connection matrix, call the weight parameter set to perform nonlinear transformation on the transformation output value, aggregate neighbor distribution attributes, and generate resource status feature matrix;

[0090] A 300×300 node connection matrix is ​​loaded into the high-speed shared memory of the graphics processor, and a 300×4 network topology node matrix is ​​simultaneously loaded into the associated memory segment. Utilizing the aggregation operation mechanism of graph convolutional networks, matrix multiplication is performed between the node connection matrix and the network topology node matrix. Specifically, each row of the adjacency matrix is ​​multiplied by the corresponding column element of the topology node matrix, and the results are accumulated. This dimension-wise multiplication-addition transformation ensures that the current node absorbs the resource load data of all its first-degree physically directly connected neighbors. A pre-trained and converged set of weight parameters (a 4×8 floating-point matrix where the initial weight values ​​follow a Gaussian distribution with a mean of 0 and a variance of 0.01) is called, and multiplied by the 300×4 intermediate result matrix output from the above multiplication-addition transformation to obtain a 300×8 transformation matrix. A modified linear unit activation function is applied to each element of this transformation matrix; that is, when the matrix element is less than 0, it is set to 0, and when it is greater than 0, the original value is retained, performing a nonlinear transformation. This process effectively filters redundant negative noise information propagated from neighboring nodes, deeply aggregates the bottleneck distribution attributes of the surrounding network, and finally generates a resource status feature matrix with a stable dimension of 300×8 at the output layer. This matrix not only includes the node's own attributes, but also integrates the comprehensive load status of the local topology environment.

[0091] Please see Figure 5 The specific steps for obtaining the bandwidth resource allocation strategy are as follows:

[0092] S411: Call the matching resource pool and resource status feature matrix, extract the node access address and destination network segment identifier for the matching resource pool, extract the link capacity vector based on the resource status feature matrix, perform addressing logic association for the node access address, destination network segment identifier, and link capacity vector, write association information to the protocol action instruction field, and establish a network forwarding flow table.

[0093] The system requests read access to memory and loads a matching resource pool containing 300 candidate nodes and a pre-generated 300×8 dimension resource status feature matrix. It iteratively extracts the node access address (e.g., 10.0.1.1) and corresponding destination network segment identifier (e.g., 10.0.2.0 / 24) for each entry from the matching resource pool. Based on the row index number corresponding to the extracted access address, it precisely slices the resource status feature matrix to extract the corresponding 1×8 feature array, directly defining it as the link capacity vector of that node. Using the node access address and destination network segment identifier as keys and the link capacity vector as values, it performs an addressing logic association operation in a hash table. Subsequently, it calls the northbound application programming interface of the software-defined network controller, using an extensible markup language script to batch write the above key-value pair association information into the protocol action instruction field.<action_list> Within the tag, an update command is ultimately issued to all 300 open flow table switches across the network, establishing specific network forwarding flow table entries in the high-speed tri-state content-addressable memory of the underlying hardware.

[0094] S412: Obtain the service guarantee bandwidth threshold and the maximum allowable communication latency, read the connectivity topology branch for the network forwarding flow table and define it as an environmental observation sequence, define the node switching action as a set of system actions, construct the state reward evaluation function based on the service guarantee bandwidth threshold and the maximum allowable communication latency, integrate the environmental observation sequence, the set of system actions and the state reward evaluation function, and establish a Markov inference architecture.

[0095] Extract the 50 Mbps guaranteed bandwidth threshold and the 20 ms maximum allowable communication latency contained in the service request data frame. Scan the network forwarding flow table already distributed to the hardware and extract all available connectivity topology branch path combinations from the start point to the end point using a depth-first search algorithm. Define the topology branch combination data space containing the current network load state variables of each node (such as queue length and port flow rate) as the environment observation sequence S for reinforcement learning. Simultaneously, define the system action set A as an enumeration of all possible route switching actions for data packets selecting adjacent outgoing ports at the current node. Construct a state reward evaluation function R, whose calculation logic is as follows: when the action is executed, if the remaining available bandwidth of the selected path combination is greater than 50 Mbps and the cumulative transmission latency is less than 20 ms, a reward value of +10 is assigned; if either condition is not met, a penalty value of -10 is assigned; if packet loss occurs, an extreme penalty value of -100 is assigned. By combining the environmental observation sequence S defined by the aforementioned dimensions, the system action set A containing switching actions, and the state reward evaluation function R with a set of definite numerical trigger thresholds, a discrete-time Markov inference architecture model that fully defines state transition probabilities and immediate rewards is constructed.

[0096] S413: Call the Markov inference architecture, perform multi-round simulation inference calculations, calculate the cumulative reward value for the system action set based on the state reward evaluation function, perform gradient update of network weight parameters based on the cumulative reward value, and perform inverse mapping decoding on the converged state network policy output matrix to obtain the data transmission routing path and bandwidth resource allocation strategy.

[0097] The established Markov inference architecture was invoked using the Deep Q-Network algorithm framework. The simulation was performed on a computing cluster with 50,000 rounds. The exploration rate parameter epsilon was initially set to 1.0 and gradually decreased to 0.05 with each round at a decay rate of 0.995. In each round, the agent outputs an exploration action based on the current environmental observation sequence, and the environment provides immediate rewards according to the +10, -10, or -100 rules of the state reward evaluation function. The Bellman equation was used to discount the future rewards of the action sequence (with a fixed discount factor gamma of 0.95), calculating the cumulative reward value corresponding to the system's action set in that round. The mean squared error loss function was used to calculate the deviation between the predicted Q-value and the target Q-value. An adaptive moment estimation optimizer was used, with a learning rate hyperparameter set to 0.001. Backward gradient updates were performed on the weight parameter matrices of each hidden layer and fully connected layer of the Deep Q-Network based on the calculated cumulative reward deviation. After 50,000 iterations, the network loss value tended to be below 0.01, indicating convergence. The system takes the latest environmental state as input from the converged deep Q-network and obtains the optimal action Q-value sequence matrix. It then performs maximum index inverse mapping decoding on this output matrix, translating the action sequence with the highest Q-value into a connected path containing the hop count of a specific Internet Protocol address. This yields the final data transmission routing path and a bandwidth resource allocation strategy accurate to a specific megabit value.

[0098] Please see Figure 6 The specific steps for obtaining resource scheduling matching results are as follows:

[0099] S511: Perform network layer traversal scan to read node attributes for the data transmission routing path, call the preset protocol features to perform feature matching and comparison operations on the node attributes, eliminate access terminals and extract relay device addresses, perform hash encoding and compression operations on the relay device addresses, and generate transmission node device identifiers.

[0100] The route tracing tool is invoked to perform a hop-by-hop IP layer network hierarchy traversal scan of the data transmission route path generated by S413. Using the Internet Control Message Protocol echo request, a response is triggered at each hop, and the Time-to-Live (TTL) field and device identification number attribute are read from the returned message. A binary bitwise AND feature matching operation is performed between the read device attributes and the system's pre-configured terminal media access control address prefix feature library. If the matching result indicates that the current address is an end-user access terminal, it is discarded from the address sequence; if the matching fails, it is identified as a relay device address in the core network's internal data forwarding link. The extracted 32-bit relay device address (e.g., 192.168.100.5) is fed into the SHA-256 secure hash algorithm operation module to generate a 64-character hexadecimal hash code. This hash code is XORed and compressed in 4-byte blocks to finally generate a 16-bit unique transmission node device identification string, significantly reducing the bit comparison overhead in subsequent hardware addressing processes.

[0101] S512: Retrieve the underlying asset database based on the transmission node device identifier, execute pull commands to read the interface model, network throughput, queue capacity, and mapping rules for the underlying asset database, and perform parameter matrix reorganization for the interface model, network throughput, queue capacity, and mapping rules to establish a network forwarding device configuration list;

[0102] A persistent connection using the Transmission Control Protocol (TCP) points the session pointer to the underlying relational asset management database of the data center. Using the 16-bit transmission node device identifier string obtained from the S511 as the primary key, a precise data query with an inner join retrieval command is executed within the database. The device interface model (e.g., Gigabit Ethernet optical port), hardware rated network throughput (e.g., 10,000 megabits per second), hardware buffer queue capacity (e.g., 2048 packet depth), and the device's virtual route forwarding mapping rule code are directly read from the matching rows. For these extracted heterogeneous parameters, a 1×4 data structure block is partitioned in memory. The string-type interface model is converted into a numeric code, and then combined with the throughput value, queue depth value, and rule code to perform parameter matrix reassembly. This operation is repeated for all relay devices, and each reassembled parameter matrix is ​​pushed vertically onto a stack to establish a network forwarding device configuration list containing the configuration baseline of all physical relay nodes globally. The relevant list parameter composition is shown in Table 3.

[0103] Table 3. Low-level configuration parameters of forwarding devices:

[0104]

[0105] S513: Call the bandwidth resource allocation policy and network forwarding device configuration list, extract the rate threshold and service queue quota recorded in the bandwidth resource allocation policy, perform resource capacity constraint judgment on the network forwarding device configuration list based on the rate threshold and service queue quota, allocate service forwarding labels, isolation identifiers and transmission queuing sequences to the network interface based on the judgment matrix, aggregate the service forwarding labels, isolation identifiers and transmission queuing sequences, and obtain the resource scheduling matching result;

[0106] Within the central processing unit of the control plane, the bandwidth resource allocation policy file derived from S413 and the network forwarding device configuration list containing the parameters in Table 3 are synchronously cross-called. A regular expression matching tool is used to extract the 50 Mbps service rate threshold and the specified strict priority service queue quota parameters recorded in the bandwidth resource allocation policy file. A device in the configuration list is selected with a hardware network throughput of 10,000 Mbps and a buffer queue capacity of 2048. The extracted 50 Mbps rate threshold is compared with the device's current available throughput remaining quota. Simultaneously, the number of 100 packets required for the service queue quota depth is compared with the remaining buffer capacity of 2048, performing a strict inequality resource capacity constraint judgment. If both constraints are satisfied (i.e., required is less than remaining), a state truth value of 1 is written to the device coordinate position corresponding to the system judgment matrix. Based on the truth value 1, the network protocol stack configuration interface is invoked to automatically assign a VLAN service forwarding label with a value of 1024 to the specified physical network interface of the device. The 0x8100 isolation identifier code for slice identification is written, and the service flow is mapped to the first high-priority transmission queuing sequence in the weighted fair queue. Using an extensible markup language text editor, the 1024 forwarding label, 0x8100 isolation identifier, and the first queuing sequence parameter are aggregated within the same text block to generate a message file containing specific hardware instructions. This yields the final resource scheduling matching result that can be directly sent to the switch data plane. This execution process ensures that service demands are completely confined within the safe range of the physical device's actual carrying capacity, effectively reducing port micro-burst congestion and packet loss faults caused by over-issuing instructions.

[0107] A power dispatch data network resource intelligent matching system is provided to implement the aforementioned power dispatch data network resource intelligent matching method. The system includes:

[0108] The data extraction and processing module obtains the topology of the power dispatch data network, the total bandwidth of physical links, the remaining available bandwidth of the network, and the node memory occupancy rate through network probes. It calculates the bandwidth consumed by the links, collects service voltage level parameters, terminal equipment type parameters, service application scenario parameters, service guarantee bandwidth thresholds, and maximum allowable communication latency, and obtains service processing data.

[0109] The resource pool allocation module, based on business data, splices together business voltage level parameters, terminal device type parameters, and business application scenario parameters to construct a feature vector, calculates the distance between the feature vector and node attributes, and filters the matching resource pools in ascending order according to the attribute distance values.

[0110] The resource status analysis module reorganizes business data, constructs a network topology node matrix, and inputs the network topology node matrix into a graph convolutional neural network model to aggregate node feature information and obtain a resource status feature matrix.

[0111] The path bandwidth adjustment module performs flow table rule mapping based on the matching resource pool and resource status feature matrix, constructs a network forwarding flow table, and inputs the network forwarding flow table, service guarantee bandwidth threshold and maximum allowable communication latency into the deep enhanced network model to perform Markov decision process deduction to obtain data transmission routing path and bandwidth resource allocation strategy.

[0112] The resource matching and analysis module uses data transmission routing paths to filter nodes and obtain the device identifiers of transmission nodes. Based on the bandwidth resource allocation strategy, it matches the device identifiers with the configuration list and generates resource scheduling matching results.

[0113] A computer device includes a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement the intelligent matching system for power dispatch data network resources as described above.

[0114] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the intelligent matching method for power dispatch data network resources as described above.

[0115] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.< / memusage> < / availablebandwidth> < / totalbandwidth> < / topology>

Claims

1. A method for intelligent matching of resources in a power dispatch data network, characterized in that, Includes the following steps: S1: Obtain the topology of the power dispatch data network, total bandwidth of physical links, remaining available bandwidth of the network, and node memory occupancy rate through network probes; calculate the bandwidth consumed by the links; collect service voltage level parameters, terminal equipment type parameters, service application scenario parameters, service guarantee bandwidth threshold, and maximum allowable communication latency to obtain service processing data. S2: Based on the business data, splice together the business voltage level parameters, terminal device type parameters and business application scenario parameters to construct a feature vector, calculate the distance between the feature vector and the node attribute values, and filter the matching resource pool in ascending order according to the attribute distance values. S3: Based on the business data, perform data reorganization, construct a network topology node matrix, input the network topology node matrix into a graph convolutional neural network model to aggregate node feature information, and obtain a resource status feature matrix; S4: Perform flow table rule mapping on the matching resource pool and resource status feature matrix to construct a network forwarding flow table. Input the network forwarding flow table, service guarantee bandwidth threshold and maximum allowable communication latency into the deep enhanced network model to perform Markov decision process deduction to obtain data transmission routing path and bandwidth resource allocation strategy. S5: Use the data transmission routing path to filter nodes and obtain the transmission node device identifier. Match the device identifier to the configuration list according to the bandwidth resource allocation strategy and generate a resource scheduling matching result.

2. The intelligent matching method for power dispatch data network resources according to claim 1, characterized in that, The service processing data includes service load throughput, protocol handshake overhead, and reserved margin; the matching resource pool includes host network segment addresses, socket port tables, and concurrent access quotas; the resource status feature matrix includes congestion risk coefficient, connectivity probability value, and device forwarding load; the bandwidth resource allocation strategy includes peak guaranteed rate, queuing priority weight, and drop / interception threshold; and the resource scheduling matching result includes packet encapsulation label, frame sequence level, and bridge isolation identifier.

3. The intelligent matching method for power dispatch data network resources according to claim 1, characterized in that, The specific steps for obtaining the business data are as follows: S111: Obtain feedback messages from pre-set nodes of network probes, perform protocol layer segment extraction and parsing on the feedback messages to read the topology of the power dispatch data network, the total bandwidth of physical links, the remaining available bandwidth of the network, and the node memory occupancy rate, and call the basic computing component to perform same-level subtraction logic calculation on the total bandwidth of physical links based on the remaining available bandwidth of the network to generate the bandwidth consumed by the links. S112: Monitor the service request data frames reported by the underlying network interface, extract service voltage level parameters, terminal device type parameters, service application scenario parameters, service guarantee bandwidth threshold and maximum allowable communication latency for the request data frames, and perform feature vector association and splicing calculation based on the basic attribute dimensions for service voltage level parameters, terminal device type parameters, service application scenario parameters, service guarantee bandwidth threshold and maximum allowable communication latency to establish a service requirement feature set; S113: Based on the bandwidth consumption of the link, perform mapping architecture topology construction for the service requirement feature set, and perform internal data dimension alignment and reorganization according to the network communication node identification level and the distribution format of the underlying mapping nodes to obtain the service data.

4. The intelligent matching method for power dispatch data network resources according to claim 3, characterized in that, The specific steps for obtaining the matching resource pool are as follows: S211: Call the business data, perform field extraction, read the business voltage level parameters, terminal device type parameters and business application scenario parameters, perform matrix splicing transformation on the business voltage level parameters, terminal device type parameters and business application scenario parameters, and generate attribute feature vectors; S212: Obtain the node resource attribute identifier, link available bandwidth, node response latency, and network protocol overhead associated with the fourth-level Internet Protocol resource pool. Combine the attribute feature vector and evaluate the obtained attribute distance value by considering the difference between the attribute feature vector and the node resource attribute identifier, and the influence of link available bandwidth and node response latency. S213: Perform sequence comparison for the Level 4 Internet Protocol resource pool, sort in ascending order according to the attribute distance values, generate a node sorting sequence, and perform truncation and filtering on the node sorting sequence to obtain the matching resource pool.

5. The intelligent matching method for power dispatch data network resources according to claim 4, characterized in that, The specific steps for obtaining the resource status feature matrix are as follows: S311: Based on the aforementioned business data, perform communication node identifier extraction and connection segment routing analysis on the power dispatch data network topology, arrange the nodes according to the connection form of the underlying equipment, and establish a node connection matrix; S312: Obtain the total bandwidth of the communication link, the remaining available bandwidth of the network, the bandwidth occupied by the link, and the memory usage rate of the node. Based on the communication node identifier, perform attribute value splicing mapping on the total bandwidth of the communication link, the remaining available bandwidth of the network, the bandwidth occupied by the link, and the memory usage rate of the node to establish a network topology node matrix. S313: Call the node connection matrix and network topology node matrix, perform feature dimension multiplication and addition transformation operation on the network topology node matrix based on the node connection matrix, call the weight parameter set to perform nonlinear transformation on the transformation output value, aggregate neighbor distribution attributes, and generate resource status feature matrix.

6. The intelligent matching method for power dispatch data network resources according to claim 5, characterized in that, The specific steps for obtaining the bandwidth resource allocation strategy are as follows: S411: Call the matching resource pool and resource status feature matrix, extract the node access address and destination network segment identifier for the matching resource pool, extract the link capacity vector according to the resource status feature matrix, perform addressing logic association for the node access address, destination network segment identifier and link capacity vector, write association information to the protocol action instruction field, and establish a network forwarding flow table. S412: Obtain the service guarantee bandwidth threshold and the maximum allowable communication delay, read the connectivity topology branch for the network forwarding flow table and define it as an environment observation sequence, define the node switching action as a system action set, construct the state reward evaluation function based on the service guarantee bandwidth threshold and the maximum allowable communication delay, integrate the environment observation sequence, the system action set and the state reward evaluation function, and establish a Markov inference architecture. S413: Invoke the Markov inference architecture, perform multi-round simulation inference calculations, calculate the cumulative reward value for the system action set based on the state reward evaluation function, perform gradient update of network weight parameters based on the cumulative reward value, and perform inverse mapping decoding on the converged state network policy output matrix to obtain the data transmission routing path and bandwidth resource allocation strategy.

7. The intelligent matching method for power dispatch data network resources according to claim 6, characterized in that, The specific steps for obtaining the resource scheduling matching result are as follows: S511: Perform network layer traversal scan to read node attributes for the data transmission routing path, call preset protocol features to perform feature matching and comparison operations on node attributes, eliminate access terminals and extract relay device addresses, perform hash encoding and compression operations on relay device addresses, and generate transmission node device identifiers. S512: Based on the transmission node device identifier, retrieve the underlying asset database, execute a pull command to read the interface model, network throughput, queue capacity, and mapping rules for the underlying asset database, perform parameter matrix reorganization for the interface model, network throughput, queue capacity, and mapping rules, and establish a network forwarding device configuration list; S513: Invoke the bandwidth resource allocation strategy and network forwarding device configuration list, extract the rate threshold and service queue quota recorded in the bandwidth resource allocation strategy, perform resource capacity constraint judgment on the network forwarding device configuration list based on the rate threshold and service queue quota, allocate service forwarding labels, isolation identifiers and transmission queuing sequences to the network interface according to the judgment matrix, aggregate the service forwarding labels, isolation identifiers and transmission queuing sequences, and obtain the resource scheduling matching result.

8. A power dispatch data network resource intelligent matching system, characterized in that, The system is used to implement the intelligent matching method for power dispatch data network resources as described in any one of claims 1-7, and the system includes: The data extraction and processing module obtains the topology of the power dispatch data network, the total bandwidth of physical links, the remaining available bandwidth of the network, and the node memory occupancy rate through network probes. It calculates the bandwidth consumed by the links, collects service voltage level parameters, terminal equipment type parameters, service application scenario parameters, service guarantee bandwidth thresholds, and maximum allowable communication latency, and obtains service processing data. The resource pool allocation module, based on the business data, splices together the business voltage level parameters, terminal device type parameters and business application scenario parameters to construct a feature vector, calculates the distance value between the feature vector and the node attribute, and filters the matching resource pools in ascending order according to the attribute distance value. The resource status analysis module reorganizes the data based on the business data, constructs a network topology node matrix, and inputs the network topology node matrix into a graph convolutional neural network model to aggregate node feature information to obtain a resource status feature matrix. The path bandwidth adjustment module performs flow table rule mapping on the matching resource pool and resource status feature matrix, constructs a network forwarding flow table, and inputs the network forwarding flow table, service guarantee bandwidth threshold and maximum allowable communication latency into the deep reinforced network model to perform Markov decision process deduction to obtain data transmission routing path and bandwidth resource allocation strategy. The resource matching and analysis module uses the data transmission routing path to filter nodes and obtain the transmission node device identifier. Based on the bandwidth resource allocation strategy, it matches the device identifier with the configuration list and generates a resource scheduling matching result.

9. A computer device, comprising a memory and a processor, characterized in that, The memory stores a computer program, and when the processor executes the computer program, it implements the intelligent matching system for power dispatch data network resources as described in claim 8.

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 intelligent matching method for power dispatch data network resources as described in any one of claims 1 to 7.

Images