Method and system for data transmission optimization of communication units of large-scale power data

Abstract

The application relates to the technical field of data processing, and discloses a communication unit data transmission optimization method and system for large-scale electric energy data. The method comprises the following steps: extracting electric energy data features through a space-time correlation analysis algorithm, constructing an electric energy data time sequence dependency matrix and a space correlation matrix; constructing an EQ-learning intelligent scheduling model based on the matrix, and generating a Q value decision table that fuses power grid constraints; selecting a multi-communication unit path by applying the decision table, and obtaining an optimal transmission path set; realizing multi-network card concurrent transmission by using a P-DPDK processor; and adjusting dynamic load balancing according to the transmission result, and optimizing electric energy data transmission performance. The application solves the problems of intelligent transmission scheduling and dynamic resource configuration lacking space-time correlation perception of electric energy data in the prior art, and improves the real-time performance, reliability and resource utilization efficiency of large-scale electric energy data transmission.

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a method and system for optimizing communication unit data transmission for large-scale electric energy data. Background Art

[0002] In existing technologies, data transmission in power systems primarily utilizes traditional SCADA systems and TCP / IP-based communications, with data transmitted via a single network card and a simple priority queue mechanism to process different types of power data. Traditional methods typically classify power data by device type or geographic location, employing fixed transmission paths and static bandwidth allocation strategies. Data processing primarily relies on general load balancing algorithms and standard operating system scheduling mechanisms to allocate computing resources. These existing technologies perform adequately when processing small-scale power data, but with the advancement of smart grid construction and the increasing digitization of power equipment, traditional data transmission methods are gradually experiencing performance bottlenecks.

[0003] The main deficiencies of the existing technology are reflected in the low efficiency of data transmission and irrational resource utilization. First, the traditional single-network card transmission method cannot fully utilize the multi-path concurrent transmission capabilities of modern communication equipment, resulting in low network bandwidth utilization. Network congestion and data packet loss are prone to occur during large-scale burst transmission of electric energy data. Secondly, the existing data classification method is too simple and does not take into account the temporal correlation and spatial correlation characteristics of the electric energy data itself, resulting in highly correlated data being transmitted in a dispersed manner, affecting the integrity and consistency of the data. Thirdly, the traditional path selection algorithm mainly calculates the shortest path based on the physical distance of the network topology, ignoring the impact of the power grid operation status and the importance of electric energy data on the transmission path selection. It is unable to adaptively adjust the transmission strategy in the event of a power grid failure or abnormal state. Summary of the Invention

[0004] The present application provides a communication unit data transmission optimization method and system for large-scale electric energy data, which is used to solve the problem of intelligent transmission scheduling and dynamic resource allocation of electric energy data lacking temporal and spatial correlation perception in the existing technology, and improves the real-time, reliability and resource utilization efficiency of large-scale electric energy data transmission.

[0005] In the first aspect, the present application provides a communication unit data transmission optimization method for large-scale electric energy data, and the communication unit data transmission optimization method for large-scale electric energy data includes: performing feature extraction processing on the electric energy data through a spatiotemporal correlation analysis algorithm to obtain a time dependency matrix and a spatial correlation matrix of the electric energy data; constructing an EQ-learning intelligent scheduling model based on the time dependency matrix and the spatial correlation matrix of the electric energy data to obtain a Q-value decision table that integrates grid constraints; applying the Q-value decision table to multi-communication unit path selection processing to obtain an optimal transmission path set based on the grid topology; performing multi-network card concurrent transmission processing on the optimal transmission path set through a P-DPDK processor to obtain a classified transmission result of an electric energy data packet; performing dynamic load balancing adjustment processing based on the classified transmission result of the electric energy data packet to obtain optimized electric energy data transmission performance parameters.

[0006] In a second aspect, the present application provides a communication unit data transmission optimization system for large-scale electric energy data, the communication unit data transmission optimization system for large-scale electric energy data comprising:

[0007] The extraction module is used to perform feature extraction processing on the electric energy data through the spatiotemporal correlation analysis algorithm to obtain the temporal dependency matrix and spatial correlation matrix of the electric energy data;

[0008] A construction module is used to construct an EQ-learning intelligent scheduling model based on the time series dependency matrix and spatial correlation matrix of the electric energy data to obtain a Q-value decision table integrating grid constraints;

[0009] A selection module, configured to apply the Q-value decision table to a multi-communication unit path selection process to obtain an optimal transmission path set based on a power grid topology;

[0010] A transmission module, configured to perform multi-network card concurrent transmission processing on the optimal transmission path set through a P-DPDK processor to obtain a classified transmission result of an electric energy data packet;

[0011] The processing module is used to perform dynamic load balancing adjustment processing according to the classification transmission results of the electric energy data packets to obtain optimized electric energy data transmission performance parameters.

[0012] In a third aspect, a communication unit data transmission optimization device for large-scale electric energy data is provided, comprising: a memory and at least one processor, wherein instructions are stored in the memory; the at least one processor calls the instructions in the memory so that the communication unit data transmission optimization device for large-scale electric energy data executes the above-mentioned communication unit data transmission optimization method for large-scale electric energy data.

[0013] In a fourth aspect, a computer-readable storage medium is provided, wherein instructions are stored in the computer-readable storage medium, which, when executed on a computer, enables the computer to execute the above-mentioned communication unit data transmission optimization method for large-scale electric energy data.

[0014] The technical solution provided in this application uses a spatiotemporal correlation analysis algorithm to extract features from power data, generating a temporal dependency matrix and a spatial correlation matrix. This overcomes the prior art's failure to ignore the inherent correlations of power data, transforming power data transmission from simple point-to-point communication to intelligent transmission based on data features. The EQ-learning intelligent scheduling model incorporates grid constraints, generating a Q-value decision table that optimizes data transmission strategies while ensuring safe and stable grid operation, avoiding the operational risks associated with traditional scheduling algorithms. Applying the Q-value decision table to multi-communication unit path selection yields an optimal transmission path set that not only considers network performance metrics but also leverages the unique characteristics of the grid topology, making transmission path selection more aligned with the actual operational requirements of the power system. The P-DPDK processor's multi-NIC concurrent transmission process overcomes the bandwidth limitations of traditional single-NIC transmission. The classified transmission of power data packets ensures differentiated transmission guarantees for data of different priorities. Dynamic load balancing adaptively adjusts transmission parameters based on the actual grid operating status. The resulting optimized power data transmission performance parameters demonstrate the system's ability to rapidly respond to changes in grid operating conditions, addressing the rigidity of transmission strategies in the face of grid failures or abnormalities in the prior art.

[0015] The spatiotemporal correlation analysis algorithm of this application fully considers the domain characteristics of electric energy data, which has strong temporal and spatial coupling. The autocorrelation function calculation and electrical distance weight analysis in the algorithm design are both specially optimized for the characteristics of power system data. Compared with general reinforcement learning algorithms, the EQ-learning intelligent scheduling model specifically incorporates grid flow constraints and safety and stability requirements. The design of the reward function reflects the power system's strict requirements for data transmission reliability and real-time performance. The P-DPDK processor is deeply optimized for power communication protocols such as IEC 61850. The design of the zero-copy memory management mechanism and the multi-network card load balancing scheduler both take into account the burst and periodic characteristics of electric energy data transmission. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0017] Figure 1 A schematic diagram of an embodiment of a method for optimizing communication unit data transmission of large-scale electric energy data in an embodiment of the present application;

[0018] Figure 2 A schematic diagram of an embodiment of a communication unit data transmission optimization system for large-scale electric energy data in an embodiment of the present application;

[0019] Figure 3 It is a schematic block diagram of the structure of a communication unit data transmission optimization device for large-scale electric energy data in an embodiment of the present invention. DETAILED DESCRIPTION

[0020] The embodiments of the present application provide a method and system for optimizing the transmission of communication unit data of large-scale electric energy data. The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of this application and the above-mentioned drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0021] For ease of understanding, the specific process of the embodiment of the present application is described below. Figure 1 In one embodiment of the present application, an embodiment of a communication unit data transmission optimization method for large-scale electric energy data includes:

[0022] Step S101: Perform feature extraction processing on the electric energy data using a spatiotemporal correlation analysis algorithm to obtain a temporal dependency matrix and a spatial correlation matrix of the electric energy data;

[0023] Step S102: constructing an EQ-learning intelligent scheduling model based on the time series dependency matrix and spatial correlation matrix of the electric energy data to obtain a Q-value decision table integrating grid constraints;

[0024] Step S103: Apply the Q-value decision table to the multi-communication unit path selection process to obtain an optimal transmission path set based on the power grid topology;

[0025] Step S104: Perform multi-network card concurrent transmission processing on the optimal transmission path set through the P-DPDK processor to obtain a classified transmission result of the electric energy data packet;

[0026] Step S105: Perform dynamic load balancing adjustment processing according to the classified transmission results of the electric energy data packets to obtain optimized electric energy data transmission performance parameters.

[0027] It is understandable that the execution subject of this application can be a communication unit data transmission optimization system for large-scale electric energy data, or a terminal or a server, which is not limited here. The embodiment of this application is described by taking the server as the execution subject as an example.

[0028] Specifically, the spatiotemporal correlation analysis algorithm first calculates the autocorrelation function of the active power data, reactive power data, voltage data, and current data collected by communication units such as smart meters and substation monitoring equipment. The autocorrelation function quantifies the temporal dependency by calculating the correlation strength between data at different times. The specific calculation process is to multiply the difference between the power value and the average power value at each moment, sum it, and then divide it by the variance to obtain the temporal correlation strength coefficient. These coefficients constitute the temporal dependency matrix of the electric energy data, and each element in the matrix represents the degree of correlation between the corresponding moments. At the same time, the algorithm calculates the electrical distance of each grid node based on the grid topology. The electrical connection strength is determined by analyzing the resistance and reactance values ​​between nodes. These resistance and reactance values ​​are input into the impedance calculation model for weight calculation to obtain the electrical distance weight coefficient. These weight coefficients constitute the spatial correlation matrix, which reflects the spatial correlation between different grid nodes.

[0029] Based on the aforementioned temporal dependency matrix and spatial correlation matrix, an EQ-learning intelligent scheduling model is constructed. The model first defines a state space set, combining the temporal and spatial characteristics of power data with the grid topology and load level to form state parameters. It then constructs an action space set, including actions for transmission path selection, bandwidth allocation, and priority adjustment. The model evaluates state-action pairs using a specially designed grid-constrained reward function. This reward function consists of four components: a reliability reward calculated based on the data transmission success rate; a power flow constraint reward calculated based on the grid power flow equation to ensure the safety margin of the transmission path; a temporal integrity reward that measures the correctness of the packet arrival sequence; and a spatial correlation reward that evaluates the coordinated transmission of data between related grid nodes. The algorithm then weights these four rewards to produce a comprehensive reward function value, which is then fed into a Q-value update algorithm for iterative training. The Q-value update process adjusts the values ​​in the decision table using a weighted combination of the current reward and the expected future reward. After multiple rounds of iterative training, a Q-value decision table that incorporates grid constraints is formed.

[0030] The trained Q-value decision table is applied to the multi-communication unit path selection process. First, a communication unit network topology graph is constructed based on the Q-value decision table, which includes all communication unit nodes and communication links connecting them. The algorithm performs a comprehensive evaluation of the bandwidth, delay, packet loss rate, and stability of each communication link, and obtains a link comprehensive evaluation index value through weighted calculation. Then, these index values are input into the improved Dijkstra algorithm, which adds power data feature constraints to the traditional shortest path algorithm. The algorithm calculates the candidate transmission paths between communication units by traversing all nodes and edges. The algorithm adjusts the priority weight of the candidate paths according to the power data time sequence dependency matrix, converts the data importance into a correction factor of the path cost value, and finally selects the optimal transmission path set based on the power grid topology through path cost value comparison.

[0031] The P-DPDK processor is used for multi-network card concurrent transmission processing of the optimal transmission path set. The P-DPDK processor is a data plane development kit optimized for power communication protocols, with deep optimization capabilities for IEC 61850 protocols. The processor first configures dedicated receive queues and send queues for each network card, establishes a mapping table between the network card and the transmission path, and then classifies the importance of the power data according to the power data time sequence dependency matrix. The data is divided into three levels: first-level critical data such as protection control signals, second-level important data such as real-time monitoring data, and third-level regular data such as historical statistical data. The data packet classifier performs queue allocation according to the classification identifier and the network card queue mapping table. Critical data is allocated to a dedicated high-performance network card queue, important data is allocated to multiple network card queues using redundant transmission, and regular data is allocated using load balancing. The processor constructs a zero-copy memory management mechanism to avoid data duplication overhead by pre-allocating a memory pool. The multi-network card load balancing scheduler dynamically adjusts the data allocation ratio according to the real-time load of each network card, and finally forms the power data packet classification transmission result.

[0032] Dynamic load balancing is performed based on the classified transmission results of power data packets. The algorithm first constructs a power grid operation status monitoring matrix, which records the operating status of each grid node in real time. This matrix includes four types: normal, lightly loaded, heavily loaded, and faulted. Each state corresponds to a different data transmission priority requirement. The algorithm detects and assesses fault severity based on grid node operating parameters. It calculates fault impact factors and severity levels by analyzing indicators such as voltage deviation, frequency fluctuation, and load mutation. The adaptive priority adjustment algorithm recalculates data transmission priority weights based on the fault impact factors and severity levels. When a grid fault is detected, it automatically increases the transmission priority of power data in the faulted area. Based on the adjusted priority weights, the algorithm dynamically allocates CPU core resources, assigning dedicated CPU cores to first-level critical data to ensure exclusive processing power. It establishes CPU affinity binding for second-level critical data, securing processing cores. And it allocates a shared CPU core pool to third-level routine data to achieve resource reuse. A comprehensive transmission performance evaluation model evaluates the performance of resource allocation solutions, deriving optimized power data transmission performance parameters through weighted calculations of throughput ratio, latency ratio, packet loss rate, integrity ratio, and reliability ratio.

[0033] In a specific embodiment, the process of executing step S101 may specifically include the following steps:

[0034] Perform autocorrelation function calculation on the collected active power data, reactive power data, voltage data and current data to obtain the time series correlation intensity coefficient;

[0035] Based on the time series correlation strength coefficient, the time series dependency matrix of electric energy data is constructed to obtain the quantitative results of the time-to-time correlation;

[0036] Calculate the electrical distance of each grid node according to the grid topology to obtain the resistance and reactance values ​​between nodes;

[0037] The resistance and reactance values ​​between nodes are input into the impedance calculation model for weight calculation to obtain the electrical distance weight coefficient;

[0038] A spatial correlation matrix is ​​constructed based on the electrical distance weight coefficient to obtain quantitative data of spatial correlation between power grid nodes.

[0039] Specifically, the autocorrelation function identifies the periodic and trend characteristics of the data by calculating the correlation of the power data at different time delays. The specific calculation process is to multiply each value in the collected active power data series with its time-delayed value. The average of all the multiplication results is then divided by the variance of the data series to eliminate dimensionality effects, thereby obtaining a standardized time series correlation strength coefficient. The algorithm performs the same autocorrelation function calculation on reactive power data, voltage data, and current data simultaneously. Each data type generates a series of correlation strength coefficients reflecting its time series characteristics. These coefficients range from negative one to positive one. The closer the value is to positive one, the stronger the positive correlation is, the closer the value is to negative one, the stronger the negative correlation is, and the closer the value is to zero, the less correlation is. The power data collected by the communication unit is processed by the autocorrelation function to form a multidimensional set of time series correlation strength coefficients, each coefficient corresponding to a specific time delay and data type combination.

[0040] The temporal dependency matrix construction process arranges the temporal correlation strength coefficients calculated above into a matrix according to chronological order and data type. The rows of the matrix represent different time points, and the columns represent different time delays. Each element in the matrix stores the correlation strength coefficient between the corresponding time point and the delayed time point. The matrix construction algorithm first creates a two-dimensional array structure with a dimension equal to the number of time points multiplied by the maximum delay time. Each temporal correlation strength coefficient is then assigned to the corresponding position in the array according to its corresponding time coordinate. Lacking positions are filled using linear interpolation to ensure matrix integrity. The quantification of inter-moment correlation is reflected in the distribution pattern of the matrix elements. High values ​​near the diagonal indicate strong short-term correlation, while high values ​​away from the diagonal indicate long-term periodic correlation. The algorithm further normalizes the matrix, mapping all element values ​​to the range of zero to one, to facilitate numerical calculations and convergence analysis in the subsequent processing algorithm.

[0041] The electrical distance calculation process quantifies the electrical connectivity between different grid nodes based on the electrical characteristics of the grid topology. This calculation requires detailed topological information, including node connectivity, line parameters, and equipment parameters. The algorithm first extracts the coordinates and connectivity of each grid node from the grid dispatching and management system. It then calculates the resistance and reactance of each transmission line based on the physical characteristics of the line. Resistance is primarily determined by the resistivity of the conductor material, the conductor cross-sectional area, and the line length, while reactance is related to the line's inductance and capacitance. The algorithm employs power flow analysis to determine the electrical distance between nodes by solving the grid's node voltage equations. The calculation treats the grid as a complex impedance network composed of resistance and reactance, with the impedance of each line equal to the complex sum of the resistance and reactance values. The resistance between nodes reflects the loss characteristics of active power transmission, while the reactance reflects the difficulty of reactive power transmission and voltage regulation. Together, these two values ​​determine the strength of the electrical coupling between grid nodes.

[0042] The impedance calculation model's weighted calculation process converts the resistance and reactance values ​​between nodes into weight coefficients for data transmission path selection. This calculation model utilizes the inverse relationship between electrical distance and transmission priority. The model first combines the resistance and reactance values ​​into a complex impedance value. The modulus of the impedance value is then calculated as a measure of electrical distance. A smaller electrical distance indicates a closer electrical connection between two nodes, and the corresponding data transmission priority should be higher. The weight calculation formula uses an inverse transformation to convert electrical distance into a weight coefficient: the weight coefficient equals the inverse of the electrical distance multiplied by a normalization constant. This ensures that pairs of nodes with smaller electrical distances receive larger weight coefficients. The algorithm also incorporates a load level correction factor to adjust weight distribution under different operating conditions. When a node is heavily loaded, its corresponding weight coefficient increases to reflect the urgency of data transmission. The range of the electrical distance weight coefficient is constrained to between zero and one through maximum normalization, facilitating unified processing and comparative analysis with parameters from other optimization algorithms.

[0043] The spatial correlation matrix construction process organizes the calculated electrical distance weight coefficients into a matrix based on the topological relationships of the power grid nodes. The rows and columns of the matrix correspond to different power grid nodes, and the matrix elements store the electrical distance weight coefficients between corresponding node pairs. The matrix construction algorithm uses a sparse matrix storage structure to handle the sparse connectivity characteristics of the power grid topology, storing only non-zero weight coefficients to save storage space and computing resources. The algorithm first determines the locations in the matrix that need to be filled with values ​​based on the power grid topological connectivity. It then fills the corresponding locations with the corresponding electrical distance weight coefficients. Matrix elements corresponding to nodes that are not directly connected are set to zero or the weight coefficients of indirect connections are calculated through multi-hop paths. The quantitative data of the spatial correlation between power grid nodes is reflected through the numerical distribution of the matrix. Elements with larger values ​​in the matrix correspond to pairs of nodes with close electrical connections. These node pairs should adopt a coordinated transmission strategy during data transmission to maintain spatial correlation.

[0044] In a specific embodiment, the process of executing step S102 may specifically include the following steps:

[0045] Based on the time series dependency matrix and spatial correlation matrix of electric energy data, a state space set is defined to obtain a state parameter combination including the grid topology state and load level level;

[0046] Construct an action space set based on the combination of state parameters to obtain transmission path selection action, bandwidth allocation action and priority adjustment action;

[0047] The state parameter combination and action space set are input into the power grid constraint reward function for evaluation processing to obtain the reliability reward value, power flow constraint reward value, time sequence integrity reward value and spatial correlation reward value;

[0048] Perform weighted summation based on each reward value to obtain the value of the comprehensive reward function;

[0049] The comprehensive reward function value is input into the Q-value update algorithm for iterative training to obtain the Q-value decision table integrating the power grid constraints.

[0050] Specifically, the state space set definition process uses the time-series dependency matrix and spatial correlation matrix of power data as input data and extracts key characteristic parameters of power grid operation through a data fusion algorithm. The algorithm first extracts the periodic characteristics of the power data from the time-series dependency matrix. By analyzing the distribution pattern of the values ​​in the matrix, it identifies temporal patterns such as daily and weekly load cycles and quantifies these periodic characteristics into time-series state parameters. Simultaneously, the algorithm extracts the connectivity characteristics of the power grid topology from the spatial correlation matrix. By analyzing the distribution of weight coefficients in the matrix, it determines the closeness between power grid nodes and defines groups of closely connected nodes as topological state units. Power grid topological states are defined by enumerating the main operating modes of the power grid, including four basic types: normal operation, maintenance, fault, and load transfer. Each state corresponds to different data transmission requirements and constraints. Load levels are categorized into four levels: light load, medium load, heavy load, and overload, based on the ratio of the real-time load to the rated load. Each level corresponds to different data transmission priorities and bandwidth allocation strategies. The state parameter combination forms a multidimensional state vector by performing Cartesian product operation on the timing state parameters, topological state type and load level grade. Each state vector uniquely describes the operation status and data transmission environment of the power grid at a specific moment.

[0051] The action space set construction process designs corresponding control action types based on the characteristics of state parameter combinations. Each action type corresponds to a specific data transmission optimization strategy. The transmission path selection action defines a set of alternative paths based on the connection topology between communication units. The algorithm traverses all nodes and edges in the power grid communication network to enumerate all possible data transmission paths and classifies and labels the paths based on path length, bandwidth capacity, and reliability indicators. The bandwidth allocation action allocates transmission resources to different data types based on the total bandwidth capacity of network links. The algorithm segments the total bandwidth according to the importance of power data: high-priority bandwidth segments are allocated to critical protection data, medium-priority bandwidth segments are allocated to general monitoring data, and low-priority bandwidth segments are allocated to historical statistical data. The priority adjustment action dynamically adjusts the data transmission priority weights based on changes in the power grid operating state. When the power grid is in a fault state, the algorithm automatically adjusts the data priority of the fault-related area to the highest level. When the power grid is normal, the algorithm executes according to the preset priority allocation scheme. The action space set combines the three basic action types to form composite action vectors. Each composite action vector contains the path selection decision, bandwidth allocation ratio, and priority adjustment coefficient, forming a complete data transmission control strategy.

[0052] The grid constraint reward function evaluation process uses state parameter combinations and action space sets as input parameters and quantifies the performance of different state-action pairs using a multi-objective evaluation system. The reliability reward calculation process measures the success rate of data transmission. The algorithm records the ratio of the number of successfully transmitted packets to the total number of transmitted packets after each state-action pair is executed and uses this ratio as a quantitative measure of reliability. A higher ratio indicates better transmission reliability and a correspondingly larger reward. The power flow constraint reward evaluation process uses the power flow equation to determine the electrical safety of the selected data transmission path. The algorithm solves the power flow equation, including the data transmission load, to calculate the voltage amplitude and phase angle at each node. A positive reward is awarded when all node voltages are within the permitted range and the line power flow does not exceed the thermal limit; otherwise, a negative reward or zero reward is awarded. The timing integrity reward measures the correctness of the time sequence of data packets arriving at the receiving end. The algorithm compares the send and receive timestamps of the packets to determine whether there is out-of-order transmission. A high reward is awarded when the packets arrive in the order they were sent. A penalty reward is imposed based on the degree of out-of-order transmission. The spatial correlation reward value evaluates the collaborative transmission effect of data from related power grid nodes. The algorithm determines whether highly correlated data adopts a collaborative transmission strategy based on the weight coefficient in the spatial correlation matrix. A positive reward value is given when highly correlated data adopts the same or adjacent transmission path, and a negative reward value is given when highly correlated data is transmitted in a dispersed manner.

[0053] The weighted summation process linearly combines four different types of reward values ​​according to preset weight coefficients. The weight coefficients are set based on the actual needs and importance of power grid data transmission. The algorithm first normalizes the four reward values, mapping reward values ​​of different dimensions and numerical ranges to a standard interval between zero and one, eliminating the impact of dimensional differences on the weighted calculation. The algorithm then performs a weighted summation operation in the order of reliability weight, power flow constraint weight, temporal integrity weight, and spatial correlation weight. The reliability weight is set to the highest value to ensure the basic success rate of data transmission, the power flow constraint weight is set to the second highest value to ensure the safety of power grid operation, and the temporal integrity weight and spatial correlation weight are adjusted according to the specific application scenario. The value of the comprehensive reward function is calculated by summing the products of the four component reward values ​​and the corresponding weight coefficients. This value comprehensively reflects the overall performance of a specific state-action pair across multiple evaluation dimensions.

[0054] The Q-value update algorithm uses the temporal difference method (TD) in reinforcement learning to gradually optimize the decision-making strategy. The algorithm uses trial and error and accumulated experience to find the optimal state-action mapping. The algorithm first initializes the Q-value decision table, setting the Q-values ​​of all state-action pairs to zero or a random decimal value, and then begins the iterative training process. In each iteration, the algorithm selects an action based on the current Q-value table. After obtaining the value of the comprehensive reward function, the TD error (TD error) is calculated. This error is equal to the difference between the current Q-value and the maximum future TD value after adding the immediate reward and discount. The algorithm uses a learning rate parameter to control the Q-value update amplitude. The learning rate is set to a small positive number to ensure training stability and convergence. The Q-value update formula adds the current Q-value to the product of the learning rate and the TD error to obtain a new Q-value. This process continues until the change in the Q-value table falls below a preset convergence threshold. After sufficient training, the Q-value decision table, which incorporates grid constraints, can provide the optimal or near-optimal action for each possible state. The magnitude of the TD error in the table directly reflects the expected cumulative reward for the corresponding state-action pair.

[0055] In a specific embodiment, the process of executing step S103 may specifically include the following steps:

[0056] Based on the Q-value decision table, a communication unit network topology graph is constructed to obtain a network structure including a communication unit node set and a communication link set;

[0057] Perform bandwidth, delay, packet loss rate, and stability evaluation on each link in the communication link set to obtain a comprehensive link evaluation index value.

[0058] The link comprehensive evaluation index value is input into the improved Dijkstra algorithm to perform path calculation processing to obtain the candidate transmission path between each communication unit;

[0059] According to the time series dependency matrix of electric energy data, the priority weight of candidate transmission paths is adjusted to obtain the path cost value considering the importance of data;

[0060] The optimal path screening process is performed based on the path cost value to obtain the optimal transmission path set based on the power grid topology.

[0061] Specifically, the communication unit network topology graph construction process identifies the effective connection relationship in the network based on the state-action mapping relationship in the Q-value decision table. The process first extracts all state-action pairs with positive Q-values from the Q-value decision table, which indicates that the corresponding transmission path performs well in historical training and is worth preserving. The algorithm determines the connection relationship between communication units by analyzing the path selection information in each state-action pair. All communication devices involved in data transmission, such as smart meters, substation monitoring devices, and distribution automation terminals, are abstracted as network nodes to form a communication unit node set. Each node in the node set is marked with its device type, geographic location, processing capacity, and access bandwidth, etc. These attribute information comes from the real-time query results of power grid infrastructure database and device management system. The construction process of communication link set determines the available data transmission channels by analyzing the physical connection mode and logical connection relationship between nodes, including optical fiber direct link, Ethernet switch link, wireless communication link, and power carrier communication link, etc. Each communication link records its starting node, terminating node, link type, physical medium, transmission protocol, and capacity limit, etc. to form a complete network structure description.

[0062] The communication link evaluation process performs multi-dimensional performance evaluation on each link in the network topology graph. The evaluation process uses a combination of active measurement and passive monitoring to obtain real-time performance parameters of the link. Bandwidth evaluation determines the actual available bandwidth of the link by sending test data packets and measuring the transmission rate. The algorithm uses the method of gradually increasing the size of the data packet to detect the bandwidth bottleneck point of the link. When the data packet transmission rate no longer increases linearly with the size of the data packet, the bandwidth upper limit is reached. The time delay evaluation uses the round-trip time measurement method. The algorithm deploys time delay measurement modules at both ends of the link. By sending probe packets with timestamps and calculating half of the round-trip time, the one-way transmission time delay is determined. The average value of multiple measurement results is used as the stable estimated value of the link time delay. The packet loss rate evaluation calculates the packet loss rate by continuously sending a certain number of test data packets and counting the number of successfully received packets at the receiving end. The algorithm uses the sequence number mechanism to identify lost packets and distinguish between packet loss and out-of-order phenomena. The stability evaluation quantifies the stability of the link by long-term monitoring of the degree of change in link performance parameters. The algorithm calculates the variance or standard deviation of bandwidth, time delay, and packet loss rate within a certain time window. The smaller the variance, the more stable the link performance. The link comprehensive evaluation index value is calculated by weighted sum of the four component indicators according to the preset weights, where the bandwidth weight is set the highest to ensure the data transmission capacity requirement, the time delay weight is second to ensure the real-time requirement, and the packet loss rate and stability weights are adjusted according to the fault tolerance capability of the application scenario.

[0063] The improved Dijkstra algorithm path calculation process adds special constraints for power data transmission to the traditional shortest path algorithm. The core improvement lies in the redefinition of the path cost function and the addition of constraints. The algorithm first uses the inverse of the link's comprehensive evaluation index as the edge weight. This is because the Dijkstra algorithm seeks the minimum-cost path, and larger evaluation indexes indicate better link performance, corresponding to smaller path costs. During the initialization phase, the algorithm creates a distance array to record the shortest distances from the source node to each node, a predecessor node array to record the predecessor nodes of each node on the shortest path, and a visit flag array to record whether a node has been processed. During the iterative calculation process, the algorithm selects the node with the smallest distance from an unvisited node as the current node. The algorithm then checks whether the path from the current node to its adjacent nodes is better than the known shortest path. If so, the distance array and predecessor node array are updated. The algorithm also improves the path update process by checking grid operational constraints, including load capacity constraints on nodes along the path, transmission capacity constraints on links, and grid security and stability constraints. Only paths that meet all constraints are considered valid candidate paths. The candidate transmission path generation process reconstructs the complete path from the source node to the destination node by backtracking the predecessor node array. The algorithm calculates several candidate paths for each pair of communication units to provide flexibility in path selection.

[0064] The priority weight adjustment process modifies the importance ranking of candidate transmission paths based on the correlation information in the power data timing dependency matrix. This process reflects the impact of timing correlation on path selection in power data transmission. The algorithm first extracts the timing correlation strength coefficient associated with the currently transmitted data from the timing dependency matrix. These coefficients reflect the degree of temporal correlation between the current data and historical data. A high timing correlation strength coefficient indicates that the current data has a strong timing dependency with previously transmitted data and requires the same or similar transmission path to maintain temporal consistency. Based on this, the algorithm assigns a higher priority to candidate paths that share the same path. A low timing correlation strength coefficient indicates that the current data is relatively independent. The algorithm primarily determines the priority weight based on the basic performance indicators of the path. The priority weight is calculated by multiplying the timing correlation coefficient by the path's base cost. A smaller product indicates a lower overall cost for the path after accounting for timing dependencies. The data importance is quantified based on the criticality of the power data in grid operation. Protection and control data is given the highest importance, real-time monitoring data is given medium importance, and historical statistical data is given lower importance. The final calculation result of the path cost value comprehensively considers multiple factors such as the basic transmission performance of the path, timing dependency, and data importance, providing a quantitative basis for path screening.

[0065] The optimal path selection process determines the optimal transmission path between each pair of communication units based on the relationship between path costs. This selection process utilizes a multi-objective optimization approach to balance transmission performance and resource utilization efficiency. The algorithm first sorts all candidate transmission paths in ascending order of path cost. The path with the lowest cost is considered the optimal choice. However, the algorithm also considers resource conflicts and load balancing between paths. When multiple paths share the same link resources, the algorithm uses a load-sharing strategy to distribute data of different priorities to different paths for transmission, avoiding resource waste caused by overloading a single path while other paths remain idle. The path selection process also considers the redundancy requirements of the power grid topology. The algorithm selects two sets of primary and backup paths for critical power data transmission tasks. The backup path can immediately take over data transmission tasks in the event of a failure of the primary path. The optimal transmission path set based on the power grid topology contains the best path selection solutions between all communication unit pairs. Each path solution contains detailed information about all nodes, links, transmission parameters, and backup options along the path.

[0066] In a specific embodiment, the process of executing step S104 may specifically include the following steps:

[0067] Based on the optimal transmission path set, a dedicated receiving queue and sending queue are configured for each network card, and a network card queue mapping relationship table is obtained;

[0068] The importance level of the electric energy data is divided according to the electric energy data time series dependency matrix to obtain the classification identification of the first-level key data, the second-level important data and the third-level regular data;

[0069] Input the classification identifier and the network card queue mapping relationship table into the data packet classifier for queue allocation processing, and obtain the network card queue allocation plan corresponding to each priority data;

[0070] Based on the network card queue allocation scheme, a zero-copy memory management mechanism is built to obtain the pre-allocated memory pool and buffer management parameters;

[0071] The pre-allocated memory pool and buffer management parameters are input into the multi-network card load balancing scheduler for concurrent transmission processing to obtain the classified transmission results of the power data packets.

[0072] Specifically, the network card queue configuration process allocates dedicated data processing queues for each network interface card based on path information in the optimal transmission path set. The configuration process first analyzes the starting node, target node, and intermediate jump node information of each path in the path set, and then determines which paths require data transmission through a specific network card based on the physical connection relationship of the network card. The dedicated receive queue is a buffer area in the network card hardware for temporarily storing incoming data packets. Each receive queue has an independent memory address space and a descriptor ring structure, and the queue depth is usually set to 1,024 descriptors to balance memory occupancy and buffering capacity. The send queue is a buffer area in the network card hardware for temporarily storing outgoing data packets. Its structure is similar to the receive queue but the data flow direction is opposite. The algorithm allocates different numbers of send queues for each network card based on the bandwidth requirements and priority requirements of the transmission path. The network card queue mapping table records the correspondence between the transmission path and the network card queue. Each entry in the table contains fields such as path identifier, network card identifier, queue number, and queue type. This mapping table serves as an important reference for subsequent data packet classification and scheduling. The queue configuration process also involves setting queue priorities. High-priority queues get more processing time slices and bandwidth resources, while low-priority queues yield to high-priority data flows when the network is congested.

[0073] The electric energy data importance level division process determines the transmission priority of different types of electric energy data based on the correlation strength information in the time-dependent matrix. The division process first extracts the correlation coefficient between each type of electric energy data and the system key operating parameters from the time-dependent matrix. First-level critical data includes protection trip signals, emergency control instructions, and fault recording data, which directly affect the safe and stable operation of the power grid. These data have a very strong correlation with key parameters such as power grid frequency and voltage amplitude in the time-dependent matrix, with a correlation coefficient usually exceeding 0.9. Second-level important data includes real-time load monitoring data, power quality data, and device status information, which affect the efficiency of power grid operation and the health of devices. The correlation coefficient of these data with power grid operating parameters is between 0.5 and 0.9, requiring timely transmission but allowing some time delay. Third-level regular data includes historical load statistics, electricity metering data, and maintenance records, which have little impact on real-time operation. The correlation coefficient of these data is usually less than 0.5, with relatively loose transmission time delay requirements. The classification identifier generation process automatically identifies data types by analyzing the message header information and data content characteristics of electric energy data. The algorithm uses a combination of pattern matching and feature extraction methods to determine the importance level of data based on the source address, target address, protocol type, and load content of the data packet. The identifier uses a three-bit binary encoding scheme, where 100 represents first-level critical data, 010 represents second-level important data, and 001 represents third-level regular data. This encoding scheme facilitates fast identification and processing at the hardware level.

[0074] The packet classifier queue allocation process matches the energy data's classification identifier with the network interface card (NIC) queue mapping table to determine which NIC and queue each packet should be processed. The classifier uses a hash table structure to store classification rules, using the packet's classification identifier and transmission path information as the key and the corresponding NIC queue information as the lookup result. This structure supports fast, linear-time lookups. The queue allocation algorithm first determines the packet's priority level based on its classification identifier, then determines the transmission path based on the packet's destination address. It then searches the NIC and queue information corresponding to this path from the NIC queue mapping table. When multiple packets compete for the same queue resources, the algorithm employs a priority scheduling strategy, prioritizing high-priority packets and processing lower-priority packets only when queues are free. The NIC queue allocation scheme details the processing queue arrangements for each type of energy data, including dedicated queues for high-performance NICs for critical data at the first level, multiple queues for redundant transmission for important data at the second level, and shared queues for load-balanced distribution of regular data at the third level. The allocation scheme also takes into account the load balancing problem of queues. The algorithm dynamically monitors the utilization rate of each queue and automatically redirects part of the data flow to a backup queue with a lower load when the load of a queue is too high.

[0075] The zero-copy memory management mechanism is constructed based on the memory requirements of the network card queue allocation scheme to optimize memory operation efficiency during data transmission. The core concept of this mechanism is to avoid repeated data copying between user space and kernel space. A preallocated memory pool is a large, contiguous memory area pre-allocated at program startup. This memory pool is divided into several fixed-size memory blocks, each corresponding to the maximum size of a network packet. Memory pool management utilizes a ring buffer structure, using head and tail pointers to track the location of available memory blocks. When a packet needs to be sent, a memory block is retrieved from the head pointer location; after the packet is sent, the memory block is returned to the tail pointer location. Buffer management parameters include key configuration information such as memory block size, total memory pool capacity, allocation strategy, and reclaim strategy. These parameters are optimized based on the transmission characteristics of power data and the network card hardware capabilities. The algorithm allocates memory pools of different sizes to different data types based on the transmission volume requirements of different priority levels. Larger memory pools are allocated to first-level critical data to ensure transmission continuity, while smaller memory pools are allocated to third-level regular data to conserve memory resources. The zero-copy mechanism uses direct memory access technology to allow the network card hardware to directly read and write pre-allocated memory areas, avoiding the overhead of copying data between different memory areas and significantly reducing CPU usage and transmission latency.

[0076] The concurrent transmission processing of the multi-card load balancing scheduler takes the pre-allocated memory pool and buffer management parameters as the scheduling basis to realize the coordinated work and load sharing among multiple network cards. The scheduler uses a round-robin scheduling algorithm to periodically check the status of each network card queue, including queue length, processing rate, error statistics, and resource occupation, and other key indicators. The scheduling period is usually set to one hundred milliseconds to balance the response speed and system overhead. The load balancing algorithm dynamically adjusts the data distribution ratio according to the real-time load of each network card. When the load of a certain network card is too high, the algorithm automatically reduces the amount of data allocated to that network card, while increasing the amount of data allocated to the network card with lower load. Concurrent transmission control is achieved through multi-threading technology, with each network card corresponding to an independent processing thread. Threads communicate and coordinate through shared memory and semaphore mechanisms. The scheduler also implements fault detection and automatic recovery functions. When a hardware fault or performance degradation is detected in a network card, the algorithm automatically migrates the data stream on that network card to other normally working network cards for continuous transmission. The power data packet classification transmission result records detailed information such as the transmission path, the network card used, the queue number, the transmission time, and the transmission status of each data packet.

[0077] In a specific embodiment, the process of performing step S105 can specifically include the following steps:

[0078] Based on the power data packet classification transmission result, an electric grid operation state monitoring matrix is constructed to obtain electric grid node operation parameters including normal state, light load state, heavy load state, and fault state;

[0079] According to the electric grid node operation parameters, fault detection and severity evaluation processing are performed to obtain a fault influence factor and a severity level value;

[0080] The fault influence factor and the severity level value are input into an adaptive priority adjustment algorithm for weight calculation processing to obtain a dynamically adjusted data transmission priority weight;

[0081] Based on the data transmission priority weight, dynamic allocation processing of CPU core resources is performed to obtain a resource configuration scheme of a key data dedicated core, an important data binding core, and a regular data shared core;

[0082] The resource configuration scheme is input into a transmission performance comprehensive evaluation model for performance calculation processing to obtain an optimized power data transmission performance parameter.

[0083] Specifically, the power grid operation state monitoring matrix construction process analyzes the real-time operation status of each node in the power grid based on the transmission statistics information in the classified transmission results of the power data packets. The matrix infers the load level and operation status of the power grid nodes by analyzing the transmission frequency, data content changes, and transmission path distribution of the data packets. The algorithm first extracts the data packet quantity statistics of each power grid node from the classified transmission results, including the total number of sent data packets, the total number of received data packets, the data packet size distribution, and the transmission time interval, etc. These statistical information directly reflects the data activity level and load change trend of the node. The normal state recognition process judges the periodicity and stability of the node data packet transmission. When the data packet transmission frequency of the node remains within the preset range and the transmission time interval fluctuation is small, the algorithm marks the node as normal operation state. The light load state is identified by detecting that the data packet transmission frequency is lower than the normal threshold but still maintains stable transmission, indicating that the electrical load level of the node is low but the equipment is running normally. The heavy load state is identified by detecting that the data packet transmission frequency is significantly higher than the normal threshold and the data content contains a large amount of load data, indicating that the node bears heavy electrical load and needs close monitoring. The fault state is identified by detecting that the data packet transmission is interrupted, the transmission error rate increases sharply, or the data content contains fault alarm information, indicating that the node has equipment failure or abnormal operation. The power grid node operation parameters form a multi-dimensional parameter vector by combining the four state information with the electrical parameters, geographical location, and device type, etc. attribute information of the node. Each vector comprehensively describes the current operation status of the corresponding node.

[0084] Fault detection and severity assessment processes conduct in-depth analysis of abnormalities in grid node operating parameters. This process uses a combination of rule-based expert knowledge and statistical anomaly detection to identify and quantify fault severity. The fault detection algorithm first establishes a baseline model of normal grid operation. This model uses historical operating data to determine the normal parameter ranges for each node under different time periods and load conditions. Fault detection is triggered when real-time node parameters exceed the baseline range. The algorithm employs a multi-level detection mechanism. It first detects single-node anomalies, comparing the deviation of current node parameters with historical statistical values ​​to determine whether a local anomaly exists. It then performs regional anomaly detection, analyzing the correlation of parameters between adjacent nodes to determine whether a regional fault exists. The severity assessment process quantifies fault severity based on the scope of the fault, its duration, and the threat to grid stability. The algorithm employs a hierarchical assessment strategy, categorizing faults into four levels: device-level faults, line-level faults, substation-level faults, and regional-level faults. The fault impact factor is determined by calculating the electrical correlation between the faulty node and other healthy nodes. A higher correlation indicates a greater fault scope and a higher impact factor. The severity level is rated on a 10-point scale, with levels one to three representing minor faults, four to six representing general faults, seven to nine representing serious faults, and ten representing an extremely serious fault. The rating is based on comprehensive factors including the load capacity affected by the fault, the number of users involved, and the difficulty of repair.

[0085] The adaptive priority adjustment algorithm uses the fault impact factor and severity level as input parameters for weight calculation. It uses a dynamic weighting mechanism to re-determine the transmission priority of different types of power data. The algorithm first establishes a mapping between fault severity and priority adjustment magnitude. This mapping is based on practical experience with power grid dispatch operations and safety and stability requirements. Lower fault severity levels result in smaller priority adjustments, while higher fault severity levels result in significantly larger adjustments. The weight calculation uses an exponential weighting approach, using the fault impact factor as the base and the severity level as the exponent to calculate a priority adjustment coefficient. This coefficient is then multiplied by the original priority weight to produce the adjusted new weight. The algorithm also considers the varying importance of different types of power data to fault response. Protection and control data maintains the highest priority under any fault conditions. The priority of monitoring data is dynamically adjusted based on its relevance to the fault area. Historical data is appropriately lowered during faults to free up transmission resources. The dynamically adjusted data transmission priority weights form a new weighting scheme that automatically adjusts data transmission strategies based on the actual power grid operation, optimizing overall transmission efficiency while ensuring the transmission of critical data. The weight adjustment process also includes a time decay mechanism. As fault handling progresses and the grid status recovers, the adjusted weights gradually return to normal levels, avoiding long-term abnormal weight distribution affecting daily operations.

[0086] Dynamic CPU core resource allocation reconfigures computing resources based on adjusted data transfer priority weights. This allocation process controls CPU core usage for data processing tasks of different priorities through operating system-level resource management interfaces. The dedicated core allocation process for critical data binds first-level critical data processing tasks to specific CPU cores. These cores are isolated from other tasks, ensuring real-time and stable critical data processing. The algorithm uses CPU affinity to restrict critical data processing processes to designated CPU cores, while also implementing a highest-priority scheduling policy to ensure these processes preempt the execution of lower-priority tasks. The bound core allocation process for critical data binds second-level critical data processing tasks to a separate set of CPU cores. These cores primarily process critical data but relinquish some resources when demand for critical data processing surges. This binding mechanism employs a soft binding strategy, allowing for limited task migration and load redistribution when system load is uneven. The shared core allocation process for regular data assigns third-level regular data processing tasks to the remaining CPU core pool. These cores use a time-sliced ​​round-robin scheduling approach to handle multiple regular data tasks. This allows for faster processing when the core pool load is low, while a task queuing mechanism ensures sequential processing when the load is high. The resource allocation plan records in detail the resource usage parameters corresponding to each type of data, such as the CPU core number, scheduling priority, memory allocation quota, and I / O bandwidth limit. The plan can be dynamically adjusted according to the power grid operation status and fault conditions.

[0087] The comprehensive transmission performance evaluation model uses the resource allocation plan as input and quantifies the overall effectiveness of data transmission optimization through the calculation of multi-dimensional performance indicators. The evaluation model encompasses five key dimensions: throughput, latency, reliability, resource utilization, and energy consumption. Each dimension has its own calculation method and evaluation criteria. Throughput is calculated by counting the number of successfully transmitted power packets and the total number of data bytes per unit time. The algorithm calculates the throughput of key data, important data, and regular data separately, then performs a weighted average based on the weights to produce a comprehensive throughput metric. Latency is calculated by measuring the end-to-end transmission time of data packets from sender to receiver. This includes components such as network transmission delay, queue waiting delay, and processing delay. The algorithm uses statistical methods to calculate statistical characteristics such as average delay, maximum delay, and delay jitter. Reliability measures transmission quality by calculating parameters such as data transmission success rate, error rate, and retransmission rate. The algorithm analyzes anomalies such as packet loss, damage, and out-of-order transmission, calculating the frequency and impact of each anomaly. Resource utilization metrics assess the rationality of resource allocation by monitoring system resource consumption, including CPU core utilization, memory usage, and network bandwidth utilization. The optimized power data transmission performance parameter is calculated by comprehensively calculating the indicator values ​​of five dimensions according to preset weights. This parameter comprehensively reflects the overall transmission performance level after dynamic load balancing adjustments.

[0088] In a specific embodiment, the process of dynamically allocating CPU core resources based on data transmission priority weights may specifically include the following steps:

[0089] Perform dedicated core allocation processing on the first-level critical data according to the data transmission priority weight, and obtain the exclusive CPU core identifier corresponding to the protection control data;

[0090] Perform core binding processing on the secondary important data and CPU affinity binding algorithm to obtain the binding relationship table between real-time monitoring data and specific CPU cores;

[0091] Perform shared core allocation based on the load characteristics of the three-level regular data to obtain a shared CPU core pool corresponding to historical statistical data;

[0092] Build a core scheduling priority queue based on the exclusive CPU core identifier, binding relationship table, and shared CPU core pool to obtain the core access timing arrangement for data with different priorities;

[0093] The core access timing arrangement is input into the CPU resource manager for resource configuration processing, and a resource configuration scheme for a key data dedicated core, an important data binding core and a conventional data sharing core is obtained.

[0094] Specifically, the first-level critical data dedicated core allocation process identifies protection and control data requiring exclusive processing resources based on the highest-weighted identifier in the data transmission priority weights. This process first selects data types with a weight of 1.0 from the priority weight table. These data types include critical information directly related to grid safety and stability, such as power grid protection trip signals, emergency load shedding instructions, and fault isolation control commands. The algorithm queries the list of currently available CPU cores through the operating system's processor affinity interface and then selects the best-performing core as the dedicated core for critical data processing. Selection criteria include core frequency, cache size, and instruction execution efficiency. The dedicated core allocation process uses a hard binding method, forcing the critical data processing process to execute on a designated core by setting a CPU mask while simultaneously prohibiting other processes from accessing that core to ensure exclusive resource use. The exclusive CPU core identifier corresponding to the protection and control data is recorded as a combination of core number and binding status. Each identifier contains detailed information such as the core's physical number, logical number, binding process identifier, and resource reservation status. The algorithm also establishes a core health monitoring mechanism to regularly check the operating status of the dedicated core, including parameters such as temperature, load rate, and error count. When a core abnormality is detected, it automatically switches to the backup core to continue processing critical data.

[0095] The secondary critical data core binding process uses a CPU affinity binding algorithm to establish soft bindings between real-time monitoring data and specific CPU cores. Unlike hard binding for dedicated cores, this algorithm allows for limited adjustments and migrations when system load is uneven. The CPU affinity binding algorithm first analyzes the processing characteristics of real-time monitoring data, including key attributes such as data arrival frequency, processing complexity, and memory access patterns. Based on these characteristics, it then selects the most appropriate CPU core for each type of monitoring data. The algorithm employs a load-aware binding strategy, prioritizing computationally intensive monitoring data to high-performance cores and allocating I / O-intensive monitoring data to cores with larger caches. The binding table is constructed using a hash table structure to store the mapping between data types and core numbers. Each entry in the table records parameters such as the data type identifier, bound core number, binding strength, and adjustment permissions. Binding strength reflects the tightness of the binding relationship; higher strength values ​​indicate more stable bindings, making them less susceptible to being broken by the system scheduler. The binding relationship table between real-time monitoring data and specific CPU cores also includes a dynamic adjustment mechanism. When it is detected that the load on a core is too high or the performance has degraded, the algorithm automatically migrates part of the monitoring data bound to the core to continue processing on the core with lower load, maintaining the continuity of data processing and the consistency of the state during the migration process.

[0096] The three-level shared core allocation process for regular data builds a multi-core shared processing resource pool based on the load characteristics of regular data. This process fully considers the processing characteristics of regular data, such as historical statistical data, electricity metering data, and equipment maintenance records. The load characteristic analysis process determines resource requirements by analyzing parameters such as processing time distribution, memory usage patterns, and computational complexity. The algorithm finds that regular data, characterized by short processing times and small memory usage but large data volumes, is well-suited for batch and parallel processing. The shared core allocation algorithm constructs a core pool from the system's remaining CPU cores. The cores in the pool use a combination of time-slicing and priority scheduling to process multiple regular data tasks. The shared CPU core pool corresponding to the historical statistical data includes configuration parameters such as the core pool number, number of cores, scheduling policy, and load balancing algorithm. The core pool utilizes a dynamic scaling mechanism. When regular data processing demand increases, new cores are automatically added from idle cores in the system. When processing demand decreases, some cores are released for use by other tasks. The algorithm also establishes a load balancing mechanism within the core pool. By monitoring the task queue length and processing rate of each core, it dynamically adjusts task allocation to ensure a balanced load across the cores within the core pool.

[0097] The core scheduling priority queue construction process integrates information from dedicated core identifiers, binding tables, and the shared core pool to form a unified CPU core access scheduling framework. The construction algorithm first determines the highest-priority access rights for critical data based on the exclusive CPU core identifiers. These cores prioritize critical data tasks under all circumstances and do not accept scheduling requests from other types of tasks. The algorithm then uses the binding table to determine medium-priority access rights for important data. These data receive priority processing on a specific core but must share core resources with a small number of other tasks. Finally, the algorithm determines low-priority access rights for regular data based on the shared core pool. These data are processed within the core pool on a first-come, first-served basis and with fair scheduling. Core access scheduling for data of different priorities utilizes a multi-level queue scheduling algorithm, which maintains a separate task queue for each priority level. Tasks in high-priority queues always take precedence over those in lower-priority queues. The scheduling also considers the real-time requirements of tasks. For critical data tasks with strict latency requirements, the algorithm adopts an earliest-deadline-first scheduling strategy, while for regular data tasks with more relaxed latency requirements, a fair round-robin scheduling strategy is used. The algorithm establishes a preemption mechanism. When a high-priority task arrives, it can preempt the executing low-priority task. The preempted task suspends execution and saves the state, waiting to resume execution at the next scheduling opportunity.

[0098] The CPU resource manager's resource configuration process uses core access scheduling as configuration instructions and implements CPU core allocation and scheduling policies through the operating system's underlying interfaces. The resource manager first parses the configuration parameters in the scheduling, including information such as the core allocation plan, scheduling policy, and priority settings. It then calls the operating system's processor management interface to perform specific configuration operations. The critical data dedicated core configuration process ensures that a designated core processes only critical data tasks by setting CPU affinity masks and scheduling policies. This configuration involves disabling general scheduling for the core, setting real-time scheduling priorities, and configuring interrupt handling policies. The critical data bound core configuration process establishes a prioritized binding relationship between data and cores through soft affinity settings, while maintaining a certain degree of scheduling flexibility to accommodate system load fluctuations. The general data shared core configuration process establishes a core pool management structure to centrally schedule and manage multiple core resources. This includes operations such as initializing the core pool, configuring the load balancer, and setting the task scheduler. The final output of the resource configuration plan contains complete information about each CPU core's specific purpose, the type of data it processes, scheduling priorities, and performance monitoring parameters. This plan can guide the execution of actual power data processing tasks and optimize resource usage.

[0099] The above describes the communication unit data transmission optimization method for large-scale electric energy data in the embodiment of the present application. The following describes the communication unit data transmission optimization system for large-scale electric energy data in the embodiment of the present application. Figure 2 In one embodiment of the present application, a communication unit data transmission optimization system for large-scale electric energy data includes:

[0100] The extraction module is used to perform feature extraction processing on the electric energy data through the spatiotemporal correlation analysis algorithm to obtain the temporal dependency matrix and spatial correlation matrix of the electric energy data;

[0101] A construction module is used to build an EQ-learning intelligent scheduling model based on the time series dependency matrix and spatial correlation matrix of power data, and obtain a Q-value decision table that integrates grid constraints;

[0102] A selection module, configured to apply the Q-value decision table to a multi-communication unit path selection process to obtain an optimal transmission path set based on a power grid topology;

[0103] A transmission module, configured to perform multi-network card concurrent transmission processing on the optimal transmission path set through a P-DPDK processor to obtain a classified transmission result of an electric energy data packet;

[0104] The processing module is used to perform dynamic load balancing adjustment processing according to the classification transmission results of the electric energy data packets to obtain optimized electric energy data transmission performance parameters.

[0105] above Figure 2 From the perspective of modular functional entities, the communication unit data transmission optimization system for medium and large-scale electric energy data in the embodiment of the present invention is described in detail. The following describes in detail the communication unit data transmission optimization device for large-scale electric energy data in the embodiment of the present invention from the perspective of hardware processing.

[0106] Reference Figure 3 In an embodiment of the present invention, a communication unit data transmission optimization device for large-scale electric energy data is also provided. The communication unit data transmission optimization device for large-scale electric energy data can be a server, and its internal structure can be as follows: Figure 3 As shown. The communication unit data transmission optimization device for large-scale electric energy data includes a processor, a memory, a display screen, an input device, a network interface and a database connected through a system bus. Among them, the computer-designed processor is used to provide computing and control capabilities. The memory of the communication unit data transmission optimization device for large-scale electric energy data includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the communication unit data transmission optimization device for large-scale electric energy data is used to store the corresponding data in this embodiment. The network interface of the communication unit data transmission optimization device for large-scale electric energy data is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, the above method is implemented.

[0107] Those skilled in the art will understand that Figure 3 The structure shown in is only a block diagram of a part of the structure related to the solution of the present invention, and does not constitute a limitation on the communication unit data transmission optimization device for large-scale electric energy data to which the solution of the present invention is applied.

[0108] The present invention also provides a computer-readable storage medium, which may be a non-volatile computer-readable storage medium or a volatile computer-readable storage medium. The computer-readable storage medium stores instructions, which, when executed on a computer, enable the computer to execute the steps of the communication unit data transmission optimization method for large-scale electric energy data.

[0109] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0110] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the existing technology, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for enabling a large-scale electric energy data communication unit data transmission optimization device (which can be a personal computer, server, or network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk, etc., various media that can store program code.

[0111] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A communication unit data transmission optimization method for large-scale electric energy data, characterized in that: The method comprises: The electric energy data is subjected to feature extraction processing by a spatiotemporal correlation analysis algorithm to obtain a time series dependency matrix and a spatial correlation matrix of the electric energy data, including: performing autocorrelation function calculation processing on the collected active power data, reactive power data, voltage data, and current data to obtain a time series correlation strength coefficient; constructing a time series dependency matrix of the electric energy data based on the time series correlation strength coefficient to obtain a time-to-time correlation quantification result; performing electrical distance calculation processing on each grid node according to the grid topology to obtain a resistance value and a reactance value between nodes; inputting the resistance value and reactance value between nodes into an impedance calculation model for weight calculation processing to obtain an electrical distance weight coefficient; constructing a spatial correlation matrix based on the electrical distance weight coefficient to obtain quantified data of spatial correlation between grid nodes; An EQ-learning intelligent scheduling model is constructed based on the time series dependency matrix and spatial correlation matrix of the electric energy data to obtain a Q-value decision table integrating grid constraints; Applying the Q-value decision table to multi-communication unit path selection processing to obtain an optimal transmission path set based on the power grid topology; Performing multi-network card concurrent transmission processing on the optimal transmission path set through the P-DPDK processor to obtain a classified transmission result of the electric energy data packet; Dynamic load balancing adjustment processing is performed according to the classified transmission results of the electric energy data packets to obtain optimized electric energy data transmission performance parameters.

2. The communication unit data transmission optimization method for large-scale electric energy data according to claim 1, characterized in that: The EQ-learning intelligent scheduling model is constructed according to the electric energy data time series dependency matrix and the spatial correlation matrix to obtain a Q-value decision table integrating grid constraints, including: Defining a state space set based on the electric energy data time series dependency matrix and the spatial correlation matrix to obtain a state parameter combination including a grid topology state and a load level grade; Constructing an action space set based on the state parameter combination to obtain a transmission path selection action, a bandwidth allocation action, and a priority adjustment action; Inputting the state parameter combination and the action space set into the power grid constraint reward function for evaluation processing to obtain a reliability reward value, a power flow constraint reward value, a time sequence integrity reward value, and a spatial correlation reward value; Perform weighted summation based on each reward value to obtain the value of the comprehensive reward function; The comprehensive reward function value is input into the Q-value update algorithm for iterative training processing to obtain a Q-value decision table integrating power grid constraints.

3. The communication unit data transmission optimization method for large-scale electric energy data according to claim 1, characterized in that: The applying the Q value decision table to the multi-communication unit path selection process to obtain an optimal transmission path set based on the power grid topology includes: Constructing a communication unit network topology diagram based on the Q-value decision table to obtain a network structure including a communication unit node set and a communication link set; Performing bandwidth, delay, packet loss rate, and stability evaluation on each link in the communication link set to obtain a comprehensive link evaluation index value; Inputting the link comprehensive evaluation index value into the improved Dijkstra algorithm to perform path calculation processing to obtain candidate transmission paths between each communication unit; performing priority weight adjustment processing on the candidate transmission paths according to the electric energy data time sequence dependency matrix to obtain a path cost value considering the importance of the data; An optimal path screening process is performed based on the path cost value to obtain an optimal transmission path set based on the power grid topology.

4. The communication unit data transmission optimization method for large-scale electric energy data according to claim 1, characterized in that: The performing multi-network card concurrent transmission processing on the optimal transmission path set by the P-DPDK processor to obtain a classified transmission result of the electric energy data packet includes: Based on the optimal transmission path set, a dedicated receiving queue and a sending queue are configured for each network card, and a network card queue mapping relationship table is obtained; The importance level of the electric energy data is divided according to the electric energy data time series dependency matrix to obtain classification identifications of first-level key data, second-level important data and third-level regular data; Inputting the classification identifier and the network card queue mapping relationship table into the data packet classifier for queue allocation processing to obtain a network card queue allocation plan corresponding to each priority data; Building a zero-copy memory management mechanism based on the network card queue allocation scheme to obtain pre-allocated memory pool and buffer management parameters; The pre-allocated memory pool and buffer management parameters are input into a multi-network card load balancing scheduler for concurrent transmission processing to obtain a classified transmission result of the electric energy data packet.

5. The communication unit data transmission optimization method for large-scale electric energy data according to claim 4, characterized in that: The dynamic load balancing adjustment process is performed according to the classification transmission result of the electric energy data packet to obtain the optimized electric energy data transmission performance parameters, including: Based on the classification and transmission results of the electric energy data packets, a grid operation status monitoring matrix is ​​constructed to obtain grid node operation parameters including normal state, light load state, heavy load state and fault state; Perform fault detection and severity assessment based on the grid node operating parameters to obtain fault impact factors and severity level values; Inputting the fault impact factor and severity level value into the adaptive priority adjustment algorithm for weight calculation processing to obtain a dynamically adjusted data transmission priority weight; Dynamically allocating CPU core resources based on the data transmission priority weights to obtain a resource configuration scheme for key data dedicated cores, important data bound cores, and conventional data shared cores; The resource configuration scheme is input into the transmission performance comprehensive evaluation model for performance calculation processing to obtain optimized power data transmission performance parameters.

6. The communication unit data transmission optimization method for large-scale electric energy data according to claim 5, characterized in that: The CPU core resources are dynamically allocated based on the data transmission priority weights to obtain a resource configuration scheme for a key data dedicated core, an important data bound core, and a conventional data shared core, including: Perform dedicated core allocation processing on the first-level critical data according to the data transmission priority weight, and obtain an exclusive CPU core identifier corresponding to the protection control data; Perform core binding processing on the secondary important data and the CPU affinity binding algorithm to obtain a binding relationship table between the real-time monitoring data and the specific CPU core; Performing shared core allocation processing based on the load characteristics of the three-level conventional data to obtain a shared CPU core pool corresponding to historical statistical data; Building a core scheduling priority queue based on the exclusive CPU core identifier, the binding relationship table, and the shared CPU core pool to obtain a core access timing arrangement for data of different priorities; The core access timing arrangement is input into a CPU resource manager for resource configuration processing, and a resource configuration scheme for a key data dedicated core, an important data binding core, and a conventional data sharing core is obtained.

7. A communication unit data transmission optimization system for large-scale electric energy data, characterized in that: A method for optimizing communication unit data transmission for large-scale electric energy data according to any one of claims 1 to 6, wherein the communication unit data transmission optimization system for large-scale electric energy data comprises: An extraction module is used to perform feature extraction processing on electric energy data through a spatiotemporal correlation analysis algorithm to obtain a time series dependency matrix and a spatial correlation matrix of the electric energy data, including: performing autocorrelation function calculation processing on the collected active power data, reactive power data, voltage data, and current data to obtain a time series correlation strength coefficient; constructing a time series dependency matrix of the electric energy data based on the time series correlation strength coefficient to obtain a time-to-time correlation quantification result; performing electrical distance calculation processing on each grid node according to the grid topology to obtain a resistance value and reactance value between nodes; inputting the resistance value and reactance value between nodes into an impedance calculation model for weight calculation processing to obtain an electrical distance weight coefficient; and constructing a spatial correlation matrix based on the electrical distance weight coefficient to obtain quantified data of spatial correlation between grid nodes; A construction module is used to construct an EQ-learning intelligent scheduling model based on the time series dependency matrix and spatial correlation matrix of the electric energy data to obtain a Q-value decision table integrating grid constraints; A selection module, configured to apply the Q-value decision table to a multi-communication unit path selection process to obtain an optimal transmission path set based on a power grid topology; A transmission module, configured to perform multi-network card concurrent transmission processing on the optimal transmission path set through a P-DPDK processor to obtain a classified transmission result of an electric energy data packet; The processing module is used to perform dynamic load balancing adjustment processing according to the classification transmission results of the electric energy data packets to obtain optimized electric energy data transmission performance parameters.

8. A communication unit data transmission optimization device for large-scale electric energy data, characterized in that: The method comprises a memory and a processor, wherein the memory stores a computer program that can be run on the processor, and when the processor executes the computer program, the communication unit data transmission optimization method for large-scale electric energy data according to any one of claims 1 to 6 is implemented.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the processor executes the communication unit data transmission optimization method for large-scale electric energy data according to any one of claims 1 to 6.

Images