CN-121711369-B - Multi-dimensional collaborative scheduling method based on power communication network

CN121711369BCN 121711369 BCN121711369 BCN 121711369BCN-121711369-B

Abstract

The invention discloses a multidimensional collaborative scheduling method based on an electric power communication network, which relates to the technical field of electric power communication network scheduling and comprises the steps of establishing a space-time evolution sequence by sensing a service flow multidimensional index, and analyzing sequence data to form association characteristics; the method comprises the steps of constructing a multi-level topological structure with different grid resolutions according to correlation characteristics and regional monitoring density differences, analyzing a service flow time mode based on the structure, deducing spectrum information of potential network demands, carrying out virtual network deduction by utilizing the spectrum information, solving core node flow distribution to generate a global load state portrait, and evaluating available processing allowance by combining node overload frequency and a preset equipment performance degradation track to form a global cooperative scheduling scheme. The method realizes the dynamic matching of the topology characterization precision and the monitoring capability, brings the time-varying attenuation of the equipment performance into capacity evaluation, and improves the accuracy and the foresight of scheduling decisions.

Inventors

YIN NA
LIU TINGTING
CHU JINGJING

Assignees

山东宏业发展集团有限公司

Dates

Publication Date: 20260505
Application Date: 20260212

Claims (8)

1. The multidimensional collaborative scheduling method based on the power communication network is characterized by comprising the following steps of: Continuously sensing multidimensional operation indexes of the whole network service flow, and establishing sequence data reflecting the space-time evolution of the service quality according to the sensing data; analyzing the sequence data to capture the internal association between the service load and the network operation situation, and forming association characteristics; Constructing a multi-level topological structure with different grid resolutions according to the correlation characteristics and referring to the monitoring densities of different areas of the network; Analyzing the mode change of the service flow in the time dimension based on the multi-level topological structure, and deducing spectrum information representing the potential requirement of the network; Carrying out virtual network deduction by using the frequency spectrum information, solving the flow distribution condition of the network core hub node, and further generating a network global load state portrait; Counting the frequency of each core hub node exceeding a load threshold according to the network global load state portrait, evaluating the available processing allowance of each node by combining a preset equipment performance degradation track, and synthesizing the available processing allowance of each node to form a global cooperative scheduling scheme; the method for constructing the multi-level topological structure with the differential grid resolution according to the correlation characteristics and by referring to the monitoring density of different areas of the network specifically comprises the following steps: dividing the power communication network domain into a plurality of analysis blocks which are logically independent and have different monitoring densities; each analysis block is allocated with an independent calculation thread, and the calculation thread is used for processing the structure construction task in the corresponding analysis block; Setting grid construction parameters for each analysis block according to the monitoring density corresponding to the analysis block, wherein the grid construction parameters comprise the physical space span of grid cells, the total number of the grid cells and the space arrangement density of the grid cells; Driving a calculation thread corresponding to each analysis block, and respectively constructing an initial grid framework of each analysis block by using grid construction parameters set for the analysis block; Integrating the initial grid frames and the associated features of all analysis blocks to generate the multi-level topological structure; The driving the calculation thread corresponding to each analysis block, using the grid construction parameters set for the analysis blocks, respectively constructing an initial grid frame of each analysis block, including: driving each calculation thread to generate two-dimensional grid coordinate information representing the planar layout according to grid construction parameters of the corresponding analysis block; based on the two-dimensional grid coordinate information, generating a three-dimensional grid of each analysis block through spatial interpolation and elevation mapping, and extracting three-dimensional structure information of the three-dimensional grid; carrying out systematic identification on nodes and grid elements of each three-dimensional grid, endowing each node with node identity codes, and endowing each grid element with grid element identity codes; Starting a coordination communication process among all computing threads, so that the computing threads corresponding to geographically adjacent analysis blocks exchange boundary grid elements and grid element identity codes thereof to achieve grid boundary alignment consensus; and carrying out boundary fusion and morphological normalization processing on the three-dimensional grids of each analysis block according to grid boundary alignment consensus, and finally obtaining an initial grid frame of each analysis block.
2. The method for multidimensional collaborative scheduling based on the power communication network according to claim 1, wherein the continuous perception of multidimensional operation index of the whole network service flow establishes sequence data reflecting the space-time evolution of the service quality according to the perception data, and comprises the following specific steps: the method comprises the steps of collecting original monitoring values of operation indexes in a service transmission process, wherein the original monitoring values comprise end-to-end time delay change, data packet arrival interval fluctuation and data packet loss; carrying out dimension unification and numerical range standardization treatment on the original monitoring value to obtain standardized index data; and excavating feature vectors reflecting time delay trend, jitter law and packet loss mode from standardized index data by using a depth feature extraction network which is trained in advance, and arranging the feature vectors in time sequence to form the sequence data.
3. The multidimensional collaborative scheduling method based on the power communication network according to claim 2, wherein the dimension unification and the numerical range standardization processing are performed on the original monitoring value to obtain the standardized index data, and the specific steps are as follows: for the collected original monitoring values of different operation indexes, converting the units of the original monitoring values into reference measurement units according to physical meanings, wherein the end-to-end time delay is uniform in milliseconds, the arrival interval of data packets is uniform in milliseconds, and the loss condition of the data packets is uniform in percentage; for each index data after the dimension unification is completed, a min-max normalization algorithm is adopted to linearly map all monitoring values of each index into a standard value interval of [0,1 ]; Checking the mapped data sequence, identifying and removing abnormal values beyond the reasonable range of normal distribution, and filling the missing values by adopting linear interpolation of the data at the front and rear moments; And integrating and aligning all index data subjected to dimension unification, numerical range standardization and data cleaning again according to the acquisition time stamp to form the standardized index data.
4. The method of multi-dimensional collaborative scheduling based on a power communication network according to claim 1, wherein the parsing the sequence data to capture an intrinsic association between traffic load and network operational situation to form association features includes: identifying and combining an initial feature group related to the traffic intensity, the network link congestion degree and the network equipment resource occupancy rate from the sequence data; When the dimension of the initial feature set exceeds a preset complexity limit, performing feature screening and compression operation on the initial feature set to obtain a simplified feature set; And carrying out mode mining on time sequence information carried by the simplified feature set, and carrying out key enhancement extraction on dynamic interaction features which can reveal the mutual restriction relation between the service flow peak value and the network path availability, wherein the dynamic interaction features are the association features.
5. The method for multidimensional collaborative scheduling based on a power communication network according to claim 1, wherein the driving each calculation thread first generates two-dimensional grid coordinate information representing a planar layout according to grid construction parameters of a corresponding analysis block, and the specific steps are as follows: Each calculation thread reads grid construction parameters of the corresponding analysis block, and calculates the number of coordinate sampling points to be set on an X axis and a Y axis of the two-dimensional projection of the analysis block according to the physical space span and the space arrangement density degree of grid units in the parameters; based on the calculated number of coordinate sampling points, equidistant sampling is carried out on the boundary line of the two-dimensional projection of the analysis block, and an initial point set formed by the boundary points is generated; In a two-dimensional area surrounded by the initial point set, adopting a regularized bilinear interpolation algorithm according to the total number of grid cells, filling the inside of the area to generate internal grid points, and combining all boundary points with the internal grid points to form a complete two-dimensional grid point set; And sequencing and indexing each point in the two-dimensional grid point set according to the position relation of the two-dimensional grid point set in a two-dimensional plane, and finally outputting the two-dimensional grid point set as structured two-dimensional grid coordinate information.
6. The method for multidimensional collaborative scheduling based on a power communication network according to claim 1, wherein the method for enabling coordinated communication among all computing threads to enable computing threads corresponding to geographically adjacent analysis blocks to exchange boundary grid elements and grid element identity codes thereof, thereby achieving grid boundary alignment consensus comprises the following specific steps: Each calculation thread identifies all grid elements positioned on the geographic boundary of the analysis block according to the constructed initial grid frame of the analysis block, extracts the space position coordinates of the boundary grid elements and the grid element identity codes thereof, and encapsulates the space position coordinates and the grid element identity codes into a boundary data packet; the system coordinator initiates a boundary alignment coordination instruction to the calculation thread groups corresponding to all analysis blocks with geographic adjacent relations according to a preset network area adjacent relation table; Synchronous exchange of respective boundary data packets is carried out between each pair of adjacent computing threads receiving the coordination instruction through a message passing interface; Each calculation thread compares and matches the received boundary grid metadata of the adjacent analysis blocks with the boundary grid metadata of the calculation thread in space position and geometric shape; If the comparison finds that the boundaries are not matched, node coordinates or grid division densities of grid elements at the respective boundaries are adjusted through negotiation, and multiple rounds of iterative communication and adjustment are performed until the grids of the boundaries of the two sides are completely connected in space; When all the adjacent computing thread groups report that the boundary matching is successful, the system coordinator confirms and broadcasts that the grid boundary alignment consensus is achieved, and each computing thread updates and locks the grid boundary structure of the analysis block according to the consensus.
7. The method for multidimensional collaborative scheduling based on the power communication network according to claim 1, wherein the method for analyzing the mode change of the service flow in the time dimension based on the multilevel topology structure derives the spectrum information representing the potential requirement of the network, and comprises the following specific steps: Extracting a multidimensional operation index time sequence of the service flow in a preset history period from each grid unit of a multilevel topological structure to serve as an original time sequence sample set; Respectively carrying out trending and seasonal preprocessing on each time sequence in the original time sequence sample set to obtain stabilized time sequence data; inputting the stabilized time series data into a multi-resolution time-frequency analysis module, and decomposing the time-frequency energy distribution of the service flow, the time delay and the packet loss index on a plurality of time scales through continuous wavelet transformation; from the time-frequency energy distribution of each index, frequency components with the energy significantly higher than background noise and periodic or quasi-periodic characteristics are identified and used as candidate characteristic frequencies; fusing and clustering candidate characteristic frequencies corresponding to different grid units and different indexes, filtering out isolated and random fluctuation frequencies, and screening out a core frequency set with continuity in spatial distribution and stability in time evolution; And constructing a multidimensional vector based on the time-frequency energy intensity and the space-time distribution range thereof corresponding to each frequency component in the core frequency set, wherein the multidimensional vector is the spectrum information for representing the potential demand mode of the network service.
8. The method of claim 1, wherein the performing virtual network deduction by using spectrum information, solving a traffic distribution condition of a core hub node of a network, and further generating a network global load state representation comprises: The power communication network comprises core data exchange equipment; establishing a data packet forwarding logic model of each core data exchange device, embedding all the data packet forwarding logic models into the network skeleton model to form a simulation reference model; Loading the dynamic service flow into a simulation reference model, performing iterative simulation operation under the preset network resource limitation condition, and calculating the flow load distribution of each network core hub node in a simulation period; analyzing the flow load distribution of each network core hub node, and summarizing the evolution curve of the load pushing along with time and the distribution gradient of the load in the network space; And (3) integrating the load evolution curves and the distribution gradients of all the network core hub nodes, and drawing a dynamic map reflecting the space-time variation of the overall load of the network, wherein the dynamic map is the overall load state representation of the network.

Description

Multi-dimensional collaborative scheduling method based on power communication network Technical Field The invention belongs to the technical field of power communication network scheduling, and particularly relates to a multidimensional collaborative scheduling method based on a power communication network. Background The service load of the current power communication network is more and more complex, and the service flows of video monitoring, data acquisition, protection control and the like show dynamic heterogeneous characteristics on space-time distribution. The traditional scheduling method generally performs load analysis and resource allocation according to static partition of network topology and globally uniform monitoring strategy. These methods rely on fixed network views and transient performance index monitoring, and it is difficult to characterize the deep correlation of the continuous evolution law of the quality of service in the space-time dimension with the internal state of the network. The existing scheme generally regards network topology as a homogeneous or static layered structure, and cannot be characterized according to the situation that actual monitoring capability and business importance of different areas are differentiated, so that resource visualization and management and control granularity are rough. Meanwhile, the evaluation of the processing capacity of the core node is mostly based on the simple comparison of the current load and the fixed threshold value, and the objective rule of natural decay of the performance of the communication equipment along with the running time is ignored, so that the deviation exists between capacity judgment and long-term scheduling decision. A method capable of matching network topology characterization accuracy with regional monitoring capability is needed to improve accuracy of whole network state sensing under limited monitoring resources. In addition, there is also a need for a mechanism that incorporates device performance variability into node capacity assessment so that scheduling decisions are not only based on current load, but rather predictive of node sustainable service capabilities. These are key issues that current power communication networks need to address to achieve refined, predictive scheduling. Disclosure of Invention The present invention aims to solve at least one of the technical problems existing in the prior art; Therefore, the invention provides a multidimensional collaborative scheduling method based on an electric power communication network, which comprises the following steps: Continuously sensing multidimensional operation indexes of the whole network service flow, and establishing sequence data reflecting the space-time evolution of the service quality according to the sensing data; analyzing the sequence data to capture the internal association between the service load and the network operation situation, and forming association characteristics; Constructing a multi-level topological structure with different grid resolutions according to the correlation characteristics and referring to the monitoring densities of different areas of the network; Analyzing the mode change of the service flow in the time dimension based on the multi-level topological structure, and deducing spectrum information representing the potential requirement of the network; Carrying out virtual network deduction by using the frequency spectrum information, solving the flow distribution condition of the network core hub node, and further generating a network global load state portrait; And counting the frequency of each core hub node exceeding a load threshold according to the network global load state portrait, evaluating the available processing allowance of each node by combining a preset equipment performance degradation track, and integrating the available processing allowance of each node to form a global cooperative scheduling scheme. Preferably, the multi-dimensional operation index of the continuous sensing whole network service flow establishes sequence data reflecting the space-time evolution of the service quality according to the sensing data, and the specific steps are as follows: Acquiring original monitoring values of operation indexes such as end-to-end time delay change, data packet arrival interval fluctuation and data packet loss condition in the service transmission process; carrying out dimension unification and numerical range standardization treatment on the original monitoring value to obtain standardized index data; and excavating feature vectors reflecting time delay trend, jitter law and packet loss mode from standardized index data by using a depth feature extraction network which is trained in advance, and arranging the feature vectors in time sequence to form the sequence data. Preferably, the dimension unification and the numerical range standardization processing are performed on the original monitoring