Power Grid Data Collaborative Optimization Method Based on Big Data Analysis

By constructing a time-series correlation graph of power grid data, identifying core data nodes and extracting collaborative influence subgraphs from them, and using graph structure entropy to quantify the complexity of subgraphs, the problems of resource waste and poor data anomaly correction in existing technologies are solved, and efficient optimization and consistency assurance of power grid data are achieved.

CN123069306APending Publication Date: 2026-11-10GUANGZHOU SMART TECHNOLOGY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611270710.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-08-20
Publication Date
2026-11-10

AI Technical Summary

Technical Problem

Existing technologies, when processing power grid data collaborative optimization, ignore the temporal correlation and linkage changes between data items, resulting in resource waste and poor data anomaly correction effects. They lack effective means to locate core driving nodes from the temporal correlation network of power grid data itself and quantify the complexity of the influence subgraph structure, making it difficult to meet the data timeliness requirements of real-time scheduling.

Method used

By constructing a time-series correlation graph of power grid data, identifying core data nodes and extracting collaborative influence subgraphs from them, the complexity of the subgraphs is quantified using graph structure entropy, and the optimization order is determined according to the complexity level, thus enabling precise data collaborative optimization processing.

Benefits of technology

It has enabled improvements in the quality and consistency of power grid data within limited computing resources, ensuring high reliability and accuracy of the data foundation for power dispatching decisions, reducing processing overhead, and improving data timeliness.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123069306A_ABST
    Figure CN123069306A_ABST
Patent Text Reader

Abstract

This invention discloses a power grid data collaborative optimization method based on big data analysis, belonging to the field of power grid data processing technology. The method includes: acquiring diverse and heterogeneous power grid data within a target power grid region and unifying the format; constructing a time-series correlation graph of the power grid data; extracting a set of core data nodes based on the data fluctuation activity and data update frequency of nodes; traversing along directed edges starting from each core data node to obtain a collaborative influence subgraph; calculating the graph structure entropy of each collaborative influence subgraph; and performing data collaborative optimization processing sequentially according to the magnitude of the graph structure entropy. This method can determine the key objects and execution order of optimization processing based on the dynamic characteristics and topological influence range of the power grid data itself, improving data quality while reducing processing overhead.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of power grid data processing technology, specifically a collaborative optimization method for power grid data based on big data analysis. Background Technology

[0002] The power grid dispatch center server needs to aggregate and process massive amounts of power grid data uploaded from different regions and equipment terminals. This data differs in communication protocols, encapsulation formats, time scales, and data types, constituting a diverse and heterogeneous data source. For collaborative optimization of such data, existing methods often employ unified cleaning rules to perform a full scan of all data, improving data availability through redundancy removal, missing data filling, and consistency checks. However, in big data scenarios, full-data collaborative optimization faces challenges such as high processing latency and high computational resource consumption, making it difficult to meet the real-time dispatch requirements for data timeliness.

[0003] Existing technologies, when handling collaborative optimization of power grid data, typically ignore the structured dependencies formed between data items due to temporal correlations and interconnected changes, treating all data items as independent optimization objects. This indiscriminate optimization approach fails to identify key data sets that have a dominant impact on the regional power grid's operational status, and also fails to perceive the extent of data anomalies caused by the propagation of influence between data items. This results in optimization operations being evenly distributed across a large amount of low-value data, leading to resource waste, and its effectiveness in correcting deeply coupled data anomalies is poor.

[0004] Multivariate power grid data exhibits varying fluctuation characteristics and update frequencies within continuous scheduling cycles. While some data items are sampled frequently, their values ​​remain stable, while others, even with occasional fluctuations, can have a cascading impact on downstream related data along dependency links. Identifying data nodes with both high-frequency changes and strong propagation capabilities, and characterizing the synergistic influence range originating from these nodes, is a prerequisite for targeted optimization. Furthermore, the complexity of data organization varies within different influence ranges. Prioritizing optimization processes according to the complexity of the influence range allows for the resolution of the most structurally chaotic and conflict-ridden subsets of data within a limited computational budget. However, existing methods lack effective means to locate core driving nodes within the temporal correlation network of power grid data and measure the complexity of their influence subgraph structures, resulting in a lack of clear objectives and a reasonable execution order for collaborative optimization. Summary of the Invention

[0005] This invention provides a power grid data collaborative optimization method based on big data analysis. The purpose is to locate core data nodes from a time-series correlation network composed of massive power grid data, based on the data fluctuation activity and update frequency, and extract collaborative influence subgraphs along the time-series dependency direction starting from the core data nodes. The structural complexity of each collaborative influence subgraph is quantitatively evaluated using graph structure entropy, and the processing order of collaborative optimization is determined according to the complexity level. This achieves accurate optimization of large-scale, strongly coupled power grid data, reduces processing overhead, and improves data consistency and timeliness.

[0006] To achieve the above objectives, the present invention provides the following technical solution: The present invention provides a power grid data collaborative optimization method based on big data analysis, applied to a power dispatch center server, the method comprising:

[0007] The process involves acquiring multi-dimensional heterogeneous power grid data uploaded by data acquisition terminals within a target power grid area during a continuous scheduling cycle, and then standardizing the data format of this data to obtain a standard power grid data set. Preferably, the process involves identifying the terminal type identifier and communication protocol version number carried by each data acquisition terminal's uploaded multi-dimensional heterogeneous power grid data, thereby determining the original data encapsulation format of each data set. A standard encapsulation format corresponding to each original data encapsulation format is then queried from a preset format mapping rule base. Data fields are rearranged and data types are converted for each multi-dimensional heterogeneous power grid data set, transforming it into standard power grid data entries with a unified field order and unified data type definition. All the converted standard power grid data entries are then arranged according to the chronological order of their respective acquisition timestamps to form a standard power grid data set. Through this process, multi-dimensional heterogeneous data generated by different terminals and different protocols can be integrated into a time-series data set with a unified structure, providing a consistent data foundation for subsequent time-series correlation analysis.

[0008] Based on the acquisition timestamps and data item type identifiers of each data item in the standard power grid data set, a time-series correlation graph of the power grid data for the target power grid region is constructed. Nodes in the time-series correlation graph correspond to different data item types, and directed edges between nodes correspond to the temporal dependencies between different data item types. As a technical solution of this invention, data items in the standard power grid data set that are at the same acquisition timestamp and belong to different data item type identifiers are identified as data item groups in the same time slice. Synchronous co-occurrence relationships between data item type identifiers are extracted from each of these time-slice data item groups. Each data item type identifier is used as a node in the time-series correlation graph of the power grid data. The direction of the directed edge between two nodes is determined based on the number of consecutive occurrences of the synchronous co-occurrence relationship. If the number of synchronous co-occurrences of the first data item type identifier and the second data item type identifier in two consecutive time slices is greater than a preset threshold, a directed edge is constructed from the node corresponding to the first data item type identifier to the node corresponding to the second data item type identifier. Preferably, the number of consecutive occurrences of the synchronous co-occurrence relationship is obtained by statistically analyzing the co-occurrence frequency of different data item type identifiers in the same time slice data item group through a sliding time window, and the preset threshold number is dynamically adjusted according to the total number of data acquisition frames within the continuous scheduling cycle. The time-series correlation graph constructed in this way accurately depicts the dynamic correlation and transmission relationships between power grid data items in the time dimension.

[0009] The data fluctuation activity and data update frequency of each node in the power grid data time-series correlation graph are determined, and a core data node set is extracted from the power grid data time-series correlation graph based on the data fluctuation activity and the data update frequency. As a preferred embodiment, for each node, the sequence of all data values ​​for the corresponding data item type within a continuous scheduling period is obtained; the absolute value of the difference between any two adjacent data values ​​is calculated, and the average of all absolute differences is taken as the data fluctuation activity of that node; the total number of data updates for the corresponding data item type within the continuous scheduling period is counted, and the ratio between the total number of updates and the total duration of the continuous scheduling period is taken as the data update frequency of that node; the product of the data fluctuation activity and the data update frequency of that node is calculated, and all nodes are sorted in descending order of the product value. The top N nodes are extracted as the core data node set, where N is a positive integer. Preferably, the value of N is determined by multiplying the total number of nodes in the time-series correlation graph of the power grid data by a preset core node proportion coefficient, and the preset core node proportion coefficient is related to the load level of the target power grid area. The core data nodes extracted in this way can reflect the key data items that have the greatest impact on the power grid's operating status and are the most actively changing, providing a clear starting point for subsequent collaborative optimization.

[0010] Starting from each core data node in the core data node set, the power grid data time-series correlation graph is traversed according to the direction of the directed edges to obtain the collaborative influence subgraph corresponding to each core data node. Specifically, a core data node is selected from the core data node set as the current starting node, and the node list and edge list of the current collaborative influence subgraph are initialized. Using the current starting node as the traversal starting point, a depth-first traversal is performed along the direction of the directed edges. During the traversal, each visited node is added to the node list, and each directed edge traversed is added to the edge list. When the depth-first traversal starting from the current starting node cannot reach a new node, the subgraph formed by the current node list and edge list is taken as a collaborative influence subgraph corresponding to a core data node, and then the next unprocessed core data node is selected to repeat the above steps. As a further improvement, when a previously visited node is encountered during depth-first traversal, it is marked as a repeatedly visited node. The existing access paths for this repeatedly visited node in the node list are obtained, and the current traversal path is merged with the existing access paths to form a closed-loop path containing the repeatedly visited node. All nodes in the closed-loop path are marked as strongly related nodes, and the depth-first traversal continues from the next adjacent node of the repeatedly visited node. The resulting collaborative influence subgraph can fully cover the influence range and feedback loops of each core data node on the temporal dependency chain.

[0011] Calculate the graph structure entropy of each collaborative influence subgraph, and sort all collaborative influence subgraphs according to the magnitude of the graph structure entropy. Perform data collaborative optimization processing on each collaborative influence subgraph sequentially according to the sorting results. Preferably, for each collaborative influence subgraph, obtain the out-degree values ​​of all nodes in the subgraph, and determine the out-degree distribution probability sequence of the subgraph based on the out-degree values ​​of all nodes. Accumulate the product of each out-degree probability distribution probability with its logarithmic form, and take the negative value of the accumulation result to obtain the graph structure entropy of the collaborative influence subgraph. Arrange all collaborative influence subgraphs in descending order of graph structure entropy, and perform data redundancy cleaning and data consistency correction on each data item contained in each collaborative influence subgraph sequentially according to the arrangement order. A larger graph structure entropy indicates a more complex internal correlation structure within the subgraph and a more urgent need for collaborative optimization; prioritizing these subgraphs can quickly eliminate inconsistencies and redundancy issues in the power grid data. During data collaborative optimization, the data item type identifiers corresponding to each node are extracted from the node list of the current collaborative impact subgraph to be processed, and the corresponding data values ​​are read from the standard power grid data set according to the data item type identifiers. When it is determined that there are multiple data values ​​with the same timestamp under the same data item type identifier, only the data value with the earliest timestamp is retained, and the remaining data values ​​with the same timestamp are marked as redundant data items and deleted. It is determined whether the data values ​​of the two ends of the directed edge of the current collaborative impact subgraph meet the preset numerical range constraint relationship. If not, the data values ​​of the two ends of the node pointing to the endpoint are corrected according to the preset numerical range constraint relationship. Through orderly optimization oriented towards the collaborative impact subgraph, the quality and consistency of power grid data can be systematically improved, ensuring that the data foundation on which power dispatch decisions depend has high reliability and high accuracy.

[0012] The present invention has the following effects: Based on the acquisition of standard power grid datasets and the construction of a time-series correlation graph of power grid data, for each node in the graph, the average absolute value of the adjacent differences in its data value sequence within a continuous scheduling cycle is calculated as the data fluctuation activity level. The ratio of the total number of data updates for that node to the total duration of the continuous scheduling cycle is calculated as the data update frequency. The product of the data fluctuation activity level and the data update frequency is used as the comprehensive influence index of the node, and the top N nodes with the largest product values ​​are selected to form the core data node set. The data fluctuation activity level reflects the oscillation amplitude of the data item's own value, while the data update frequency reflects the density of its changes. Multiplying the two can identify key data nodes that not only have drastic value changes but also frequently change their states. These nodes often correspond to the collection indicators of core state parameters such as load and voltage in power grid operation, and their fluctuation and refresh characteristics will spread to related data items through time-series dependencies. Using this product value to select core data nodes avoids misjudgments caused by using only fluctuation amplitude or update frequency, making the extracted node set more closely match the data sources with strong driving effects in actual power grid operation, providing an accurate starting point for subsequent analysis of the scope of collaborative influence.

[0013] After obtaining the set of core data nodes, a depth-first traversal is performed along the directed edges of the power grid data time-series correlation graph, starting from each core data node, until no new nodes can be accessed. All nodes and directed edges traversed constitute the collaborative influence subgraph corresponding to that core data node. This subgraph depicts the complete range of influence that the fluctuations and changes of the core data node can propagate along the time-series dependency path. For each collaborative influence subgraph, the out-degree values ​​of all nodes are obtained, and an out-degree distribution probability sequence is constructed. The graph structure entropy is obtained by multiplying each out-degree distribution probability by its logarithm, summing the results, and taking the negative value. Graph structure entropy quantifies the degree of unevenness and structural complexity of the node connection methods within the subgraph: the more uneven the out-degree distribution and the more chaotic the information organization of the collaborative influence subgraph, the greater its graph structure entropy. All collaborative influence subgraphs are arranged in descending order of graph structure entropy, and the subgraph with the largest graph structure entropy is processed first. This means that limited computing resources are first invested in the data item set corresponding to the subgraph with the most complex internal dependency structure and the most serious data collaboration conflicts. In collaborative optimization operations, for data within a subgraph, redundancy cleaning and consistency correction based on directed edge numerical range constraints are performed. This effectively eliminates cascading data anomalies caused by core data nodes and restores the logical consistency within the dataset. This processing strategy, driven by graph structure entropy sorting, changes the traditional indiscriminate full-scale optimization approach. By allocating optimization computing power based on the complexity of the subgraph, it can achieve more significant data quality improvements within the same processing time.

[0014] The above and other objects, advantages and features of the present invention will become more apparent to those skilled in the art from the following detailed description of specific embodiments of the invention in conjunction with the accompanying drawings. Attached Figure Description

[0015] The following sections will describe some specific embodiments of the invention in a detailed manner by way of example and not limitation, with reference to the accompanying drawings. The same reference numerals in the drawings denote the same or similar parts or portions. Those skilled in the art should understand that these drawings are not necessarily drawn to scale. In the drawings: Figure 1 This is a flowchart of a power grid data collaborative optimization method based on big data analysis; Figure 2 This is a flowchart of the unified processing of data formats from diverse and heterogeneous power grids; Figure 3 This is a flowchart of the collaborative influence subgraph data optimization processing based on graph structure entropy. Detailed Implementation

[0016] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0017] See Figure 1 This invention provides a power grid data collaborative optimization method based on big data analysis, which is applied to the power dispatch center server. This method acquires multi-variable heterogeneous power grid data uploaded by data acquisition terminals within a target power grid area during a continuous scheduling cycle, and performs data format unification processing on the multi-variable heterogeneous power grid data to obtain a standard power grid data set. Based on the acquisition timestamp and data item type identifier of each data item in the standard power grid data set, a time-series correlation graph of power grid data for the target power grid area is constructed. Nodes in the time-series correlation graph correspond to different data item types, and directed edges between nodes correspond to the time-series dependencies between different data item types. The data fluctuation activity and data update frequency of each node in the time-series correlation graph are determined, and a core data node set is extracted from the time-series correlation graph based on the data fluctuation activity and data update frequency. Starting from each core data node in the core data node set, the time-series correlation graph of power grid data is traversed according to the direction of the directed edges to obtain the collaborative influence subgraphs corresponding to each core data node. The graph structure entropy of each collaborative influence subgraph is calculated, and all collaborative influence subgraphs are sorted according to the size of the graph structure entropy. Data collaborative optimization processing is then performed on each collaborative influence subgraph according to the sorting results.

[0018] In specific implementation, please refer to Figure 2After acquiring the multi-dimensional heterogeneous power grid data uploaded by each data acquisition terminal within the target power grid area during a continuous scheduling cycle, the multi-dimensional heterogeneous power grid data is processed to unify the data format to obtain a standard power grid data set.

[0019] The system identifies the terminal type identifier and communication protocol version number carried in the multi-source heterogeneous power grid data uploaded by each data acquisition terminal. The terminal type identifier distinguishes different models or categories of data acquisition terminals, and the communication protocol version number identifies the specific version of the communication protocol used by the data acquisition terminal. The power dispatch center server pre-stores a mapping table between terminal type identifiers, communication protocol version numbers, and raw data encapsulation formats. Different combinations of terminal type identifiers and different communication protocol version numbers correspond to a specific raw data encapsulation format. The power dispatch center server queries the mapping table based on the terminal type identifier and communication protocol version number parsed from the multi-source heterogeneous power grid data to determine the raw data encapsulation format of the multi-source heterogeneous power grid data.

[0020] The power dispatch center server has a pre-set format mapping rule library. This library stores one-to-one mapping relationships between various raw data encapsulation formats and standard encapsulation formats. Each mapping relationship defines the order of data fields in the raw data encapsulation format within the standard encapsulation format, as well as data type conversion rules. After determining the raw data encapsulation format of the diverse and heterogeneous power grid data, the power dispatch center server queries the format mapping rule library for the corresponding standard encapsulation format.

[0021] Optionally, the construction process of the format mapping rule base is as follows: Collect the types and communication protocol version numbers of all data acquisition terminals managed by the power dispatch center server to determine all possible raw data encapsulation formats; for each raw data encapsulation format, define a unified standard encapsulation format. The standard encapsulation format has a fixed field order, which is: acquisition timestamp field, data item type identifier field, data value field, terminal type identifier field, and communication protocol version number field; define the mapping relationship between each field in the raw data encapsulation format and its corresponding field in the standard encapsulation format, including field name mapping, field position mapping, and data type mapping. Data type mapping is used to convert the data type of the fields in the raw data encapsulation format to the unified data type definition specified in the standard encapsulation format. For example, converting floating-point voltage values ​​to double-precision floating-point values ​​and integer frequency values ​​to 32-bit signed integer values.

[0022] Based on the standard encapsulation format obtained from the query, the data fields of each multi-element heterogeneous power grid data are rearranged and their data types are converted. Data field rearrangement refers to extracting the corresponding data fields from the original data encapsulation format and rearranging them according to the field order defined in the standard encapsulation format. Data type conversion refers to converting the original data type of the extracted data fields to the unified data type specified by the standard encapsulation format according to the data type mapping relationship defined in the format mapping rule base. For example, if the voltage value field in the original data encapsulation format is a single-precision floating-point number, while the corresponding field in the standard encapsulation format is specified as a double-precision floating-point number, then the single-precision floating-point number is converted to a double-precision floating-point number; if the timestamp field in the original data encapsulation format is a string type, while the standard encapsulation format specifies it as a 64-bit integer type of UNIX timestamp, then the string is parsed according to the preset time format and converted to a 64-bit integer. Through the above processing, each multi-element heterogeneous power grid data is converted into a standard power grid data entry with a unified field order and a unified data type definition. Each standard power grid data entry includes a collection timestamp, data item type identifier, data value, terminal type identifier, and communication protocol version number.

[0023] All the converted standard power grid data entries are arranged according to the chronological order of their respective acquisition timestamps. During sorting, the timestamp values ​​are arranged in ascending order. If multiple standard power grid data entries have the exact same timestamp, their relative order is maintained, or a secondary sort is performed according to the lexicographical order of the data item type identifier. After sorting, a standard power grid data set is formed, which is an ordered sequence of data entries whose arrangement reflects the chronological order of the power grid data generation.

[0024] In practice, a time-series correlation map of power grid data for the target power grid region is constructed based on a standard power grid dataset. The standard power grid dataset contains multiple standard power grid data entries, each carrying a collection timestamp and a data item type identifier.

[0025] Data items belonging to different data item type identifiers at the same acquisition time stamp in the standard power grid data set are identified as time-slice data item groups. For each acquisition time stamp, all data item type identifiers appearing at that time stamp are collected, and the set of these data item type identifiers constitutes a time-slice data item group. The co-occurrence relationship between the data item type identifiers in each time-slice data item group is extracted. The co-occurrence relationship refers to any two different data item type identifiers appearing together once within the same time-slice data item group.

[0026] Each distinct data item type identifier appearing in the standard power grid dataset is used as a node in the power grid data time-series correlation graph. The direction of the directed edge between two nodes in the power grid data time-series correlation graph is determined based on the number of consecutive occurrences of the synchronous co-occurrence relationship. For any two distinct data item type identifiers, namely the first data item type identifier and the second data item type identifier, the number of synchronous co-occurrences of the first data item type identifier and the second data item type identifier in two consecutive time slices is counted.

[0027] In practice, the number of consecutive occurrences of synchronous co-occurrence relationships is obtained by statistically analyzing the co-occurrence frequency of different data item type identifiers within the same time slice data item group using a sliding time window. The length of the sliding time window is set to... A time slice, The value is half the total number of timestamps collected within a continuous scheduling period. The sliding time window slides along the time axis with a step size of one time slice. Within each sliding time window, for the first data item type identifier and the second data item type identifier, the number of events satisfying the following condition is counted: there exists a... The first collection timestamp and the first All collected timestamps are located within the sliding time window, and the first data item type identifier and the second data item type identifier appear simultaneously in the first time window. In the same time slice data item group corresponding to each collection timestamp, and the first data item type identifier and the second data item type identifier appear simultaneously in the first time slice data item group. Within each time-slice data item group corresponding to a collection timestamp, each time the above event occurs, the co-occurrence count of the first data item type identifier and the second data item type identifier in two consecutive time slices increases by one. By traversing all sliding time windows, the total co-occurrence count of the first data item type identifier and the second data item type identifier is obtained.

[0028] Optionally, the preset threshold number of times can be dynamically adjusted based on the total number of data acquisition frames within a continuous scheduling cycle. The total number of data acquisition frames is denoted as... , This represents the total number of data item groups in the same time slice within a consecutive scheduling period. The formula for calculating the preset number threshold is: ; in, Indicates rounding up. For the preset number of times threshold, This is the proportionality coefficient. Total number of data acquisition frames. Scale factor. The value of this value is related to the load level of the target power grid area, which is divided into Level 1 load, Level 2 load, and Level 3 load. When the load level of the target power grid area is Level 1 load... The value is 0.1; when the load level of the target power grid area is level two load, The value is 0.15; when the load level of the target power grid area is level three load, The value is set to 0.2. The corresponding value is determined based on the load level of the target power grid area. After obtaining the value, combine it with the total number of data acquisition frames. Calculate the preset number of times threshold .

[0029] When the total number of times the first data item type identifier and the second data item type identifier co-occurs exceeds a preset threshold. At that time, a directed edge is constructed from the node corresponding to the first data item type identifier to the node corresponding to the second data item type identifier. This process is repeated for all different data item type identifier pairs to complete the establishment of directed edges in the power grid data time-series correlation graph.

[0030] In practice, after constructing the time-series correlation graph of power grid data, the data fluctuation activity and data update frequency of each node in the time-series correlation graph are determined, and the core data node set is extracted from the time-series correlation graph of power grid data based on the data fluctuation activity and data update frequency.

[0031] For each node in the power grid data time-series correlation graph, the sequence of all data values ​​for the corresponding data item type within a continuous scheduling period is obtained. A data item type corresponds to a node in the power grid data time-series correlation graph, and all data values ​​for that data item type within a continuous scheduling period originate from a standard power grid data set. For each standard power grid data entry in the standard power grid data set, if its data item type identifier matches the data item type corresponding to the node, then the data value in that standard power grid data entry belongs to a data value of that data item type. All data values ​​belonging to the same data item type are arranged sequentially according to the chronological order of their acquisition timestamps to form the data value sequence for that data item type within a continuous scheduling period.

[0032] For the data value sequence, calculate the absolute value of the difference between every two adjacent data values. Sum the absolute values ​​of all adjacent data values ​​in the data value sequence, then divide by the total number of absolute differences to obtain the average of all absolute differences. This average is used as the data volatility activity of the node. Data volatility activity is used... The total number of data values ​​in the data value sequence is denoted as . The data value sequence of the first The data values ​​are denoted as , No. The data values ​​are denoted as The calculation process for data fluctuation activity is expressed as follows: ; in, The data fluctuation activity of the node. This refers to the index of the data value in the data value sequence. The range of values ​​is arrive Integers between [a certain number] Indicates the first value in the data value sequence The specific numerical value of each data point. Indicates the first value in the data value sequence The specific numerical value of each data point. This represents the total number of data values ​​in the data value sequence. Indicates the first The value of the data and the first The absolute value of the difference between the data values. This indicates that all terms from p=1 to p=M-1 are summed.

[0033] The total number of data updates for the data item type corresponding to the node within a continuous scheduling period is counted. A data update refers to the number of times the value of the same data item type changes between two adjacent data collection timestamps. By iterating through each pair of adjacent data values ​​in the data value sequence, if the values ​​of the two adjacent data values ​​are not equal, a data update is determined, and the total number of data updates is incremented by one. The total number of data updates is divided by the total duration of the continuous scheduling period, and the resulting ratio is taken as the data update frequency of the node. The total duration of the continuous scheduling period is the time difference between the last data collection timestamp and the first data collection timestamp within the continuous scheduling period. The data update frequency is... express.

[0034] Calculate the data fluctuation activity of the node. With data update frequency The product value, denoted as Calculate the corresponding product value for each node in the time-series correlation graph of the power grid data. Sort all nodes in descending order of their product values. Then, select the nodes that appear first. The nodes of each bit are extracted into a core data node set, where It is a positive integer.

[0035] In the core data node set The value of is determined by multiplying the total number of nodes in the power grid data time-series correlation graph by the preset core node ratio coefficient. The total number of nodes in the power grid data time-series correlation graph is denoted as . The preset core node ratio coefficient is denoted as , The value of is determined in the following way: and After multiplying, round the product up. If the rounded value is greater than 1, then the product is considered complete. Then Set as Preset core node ratio coefficient It is associated with the load level of the target power grid area. The load level of the target power grid area is divided into Level 1 load, Level 2 load, and Level 3 load. When the load level of the target power grid area is Level 1 load, The value is 0.2; when the load level of the target power grid area is level two load, The value is 0.25; when the load level of the target power grid area is level three, The value is set to 0.3. The basis for this value is that the higher the load level, the more concentrated the coverage of core data nodes is required for power grid operation, and the selection ratio is reduced accordingly to focus on the few most active nodes; the lower the load level, the higher the selection ratio is to expand the coverage of collaborative optimization.

[0036] In the specific implementation, a core data node that has not yet been processed is selected from the core data node set as the current starting node. An empty node list and an empty edge list are initialized. The node list records information about all nodes accessible from the current starting node, and the edge list records information about all directed edges traversed during the traversal. Simultaneously, a local access marker set is established to record the nodes visited during the current collaborative influence subgraph traversal; this local access marker set is initially empty.

[0037] Starting from the current node, perform a depth-first traversal along the directed edges in the power grid data time-series correlation graph. During the depth-first traversal, mark the current node as visited and add it to the node list. Maintain a traversal stack and push the current node onto the stack.

[0038] Pop the top node from the traversal stack and designate it as the current processing node. Obtain all adjacent nodes of the current processing node in the power grid data time-series correlation graph pointed to by directed edges. For each adjacent node, perform the following operations: If the adjacent node is not in the local access marker set, it means that the adjacent node has not been visited yet. Mark the adjacent node as visited, add the adjacent node to the node list, add the directed edge from the current processing node to the adjacent node to the edge list, push the adjacent node onto the traversal stack, and record the parent node of the adjacent node as the current processing node. The parent node information is used for subsequent path backtracking. If the adjacent node is already in the local access marker set, it means that the adjacent node is a repeatedly visited node. Mark the repeatedly visited node as a repeatedly visited node.

[0039] When a repeatedly visited node is encountered, the existing access path corresponding to the repeatedly visited node in the node list is obtained. The existing access path is obtained by backtracking from the repeatedly visited node back to the starting node level by level, based on the parent node information, until the current starting node is reached. All nodes traversed during the backtracking process are arranged in order to form the existing access path. Simultaneously, the current traversal path is obtained. The current traversal path is obtained by backtracking from the current processing node back to the starting node level by level, based on the parent node information, until the current starting node is reached. All nodes traversed during the backtracking process are arranged in order to form the current traversal path. The nodes contained in the current traversal path are merged with the nodes contained in the existing access path. The merged node set constitutes the node set corresponding to the closed-loop path. All nodes in the node set corresponding to the closed-loop path are marked as strongly related nodes. The marking information of strongly related nodes can be stored in the additional attributes of the collaborative influence subgraph.

[0040] After processing the repeatedly visited nodes, skip further depth-first traversal of the adjacent nodes and proceed to process the next adjacent node of the currently processed node, resuming the depth-first traversal process from the next adjacent node. If all adjacent nodes of the currently processed node have been processed, pop a new top node from the traversal stack as the current processed node and repeat the above adjacent node processing process.

[0041] When the traversal stack is empty and all reachable nodes corresponding to the current starting node have been visited, and no new unvisited nodes can be reached, the depth-first traversal process corresponding to the current starting node ends. At this point, the current node list and edge list together form a collaborative influence subgraph, which is then associated and stored with the current starting node.

[0042] Next, select the next unprocessed core data node from the core data node set as the new current starting node, reinitialize an empty node list, an empty edge list, and an empty local access marker set, and generate a new collaborative influence subgraph using the same depth-first traversal process described above. Repeat the above steps until all core data nodes in the core data node set have been processed, resulting in the collaborative influence subgraph corresponding to each core data node.

[0043] In specific implementation, please refer to Figure 3 Calculate the graph structure entropy of each collaborative influence subgraph, sort all collaborative influence subgraphs according to the size of the graph structure entropy, and perform data collaborative optimization processing on each collaborative influence subgraph in turn according to the sorting results.

[0044] For each collaborative influence subgraph, obtain the out-degree values ​​of all nodes in the subgraph. The out-degree value of a node refers to the total number of directed edges originating from that node and pointing to other nodes in the collaborative influence subgraph. Count the out-degree values ​​of each node in the collaborative influence subgraph to form an out-degree value set. Determine the out-degree distribution probability sequence of the collaborative influence subgraph based on the out-degree value set. The out-degree distribution probability sequence is determined by: counting the frequency of each different out-degree value in the out-degree value set, dividing the frequency of each different out-degree value by the total number of nodes in the collaborative influence subgraph to obtain the out-degree distribution probability corresponding to each different out-degree value. Arrange the out-degree distribution probabilities corresponding to all different out-degree values ​​in ascending order of out-degree value to form the out-degree distribution probability sequence.

[0045] The graph structure entropy of the collaborative influence subgraph is obtained by summing the product of each out-degree probability in the out-degree probability sequence and its logarithmic form, and then taking the negative value of the summation result. The calculation process of graph structure entropy is expressed as follows: ; in, The graph structure entropy represents the co-influence subgraph. This indicates the number of different out-degree values ​​in the collaborative influence subgraph. This is the index of the out-degree value type. The range of values ​​is from arrive integers, Represents the probability sequence of the out-degree distribution. The probability distribution of the out-degree corresponding to the out-degree value. Indicates to Take Logarithm to base 0. When The value is When, define The value is 0. This represents the summation of all terms from u=1 to u=R. All cooperative influence subgraphs are arranged in descending order of graph structure entropy. A higher graph structure entropy indicates a more uneven distribution of node out-degrees and higher structural complexity in the cooperative influence subgraph; conversely, a lower graph structure entropy indicates a more uniform distribution of node out-degrees and lower structural complexity. During the sorting process, if two or more cooperative influence subgraphs have the same graph structure entropy value, they are then arranged in descending order of the number of nodes they contain.

[0046] Perform data redundancy cleaning and data consistency correction on each data item in each collaborative effect subgraph in the order they are arranged. For the current collaborative effect subgraph to be processed, perform the following operations.

[0047] Extract the data item type identifiers corresponding to each node from the node list of the currently pending collaborative impact subgraph. Based on each data item type identifier, read all data values ​​corresponding to the data item type identifier and the collection timestamp corresponding to each data value from the standard power grid dataset.

[0048] For all read data values, determine whether multiple data values ​​with the same acquisition timestamp exist under the same data item type identifier. If multiple data values ​​with the same acquisition timestamp exist under the same data item type identifier, determine the data value with the earliest acquisition timestamp based on the order of these data values ​​in the standard power grid data set. Only retain the data value with the earliest acquisition timestamp, and mark the remaining data values ​​with the same acquisition timestamp as redundant data items, then delete the redundant data items from the standard power grid data set.

[0049] After data redundancy cleaning, it is determined whether the data values ​​of the two endpoints of the directed edge in the current collaborative influence subgraph satisfy a preset numerical range constraint. For each directed edge, there is a starting node and an ending node. The starting node corresponds to one data item type identifier, and the ending node corresponds to another data item type identifier. The preset numerical range constraint defines the mathematical relationship that should be satisfied between the data values ​​corresponding to the starting node and the data values ​​corresponding to the ending node. Optionally, the preset numerical range constraint includes: the data value corresponding to the ending node should be within the numerical range obtained by multiplying the data value corresponding to the starting node by a preset proportional coefficient. When the preset numerical range constraint is not satisfied, the data values ​​corresponding to the endpoint nodes are corrected according to the preset numerical range constraint. The correction method is to adjust the data value corresponding to the ending node to the result value obtained by multiplying the data value corresponding to the starting node by the preset proportional coefficient.

[0050] After completing the data redundancy cleaning and data consistency correction for the current collaborative impact subgraph, select the next collaborative impact subgraph in the order of arrangement, and repeat the above data redundancy cleaning and data consistency correction process until all collaborative impact subgraphs have been processed.

[0051] Therefore, those skilled in the art should recognize that although numerous exemplary embodiments of the present invention have been shown and described in detail herein, many other variations or modifications conforming to the principles of the present invention can be directly determined or derived from the disclosure of the present invention without departing from the spirit and scope of the invention. Thus, the scope of the present invention should be understood and construed as covering all such other variations or modifications.

Claims

1. A power grid data collaborative optimization method based on big data analysis, characterized in that, The method is applied to a power dispatch center server, and the method includes: The multi-dimensional heterogeneous power grid data uploaded by each data acquisition terminal in the target power grid area during a continuous scheduling cycle is obtained, and the multi-dimensional heterogeneous power grid data is processed to unify the data format to obtain a standard power grid data set. Based on the collection timestamp and data item type identifier of each data item in the standard power grid data set, a power grid data time-series correlation graph of the target power grid area is constructed. The nodes in the power grid data time-series correlation graph correspond to different data item types, and the directed edges between nodes correspond to the time-series dependencies between different data item types. The data fluctuation activity and data update frequency of each node in the power grid data time series correlation graph are determined, and the core data node set is extracted from the power grid data time series correlation graph based on the data fluctuation activity and the data update frequency. Starting from each core data node in the core data node set, the power grid data time-series correlation graph is traversed according to the direction of the directed edges to obtain the collaborative influence subgraph corresponding to each core data node. Calculate the graph structure entropy of each collaborative influence subgraph, sort all collaborative influence subgraphs according to the magnitude of the graph structure entropy, and perform data collaborative optimization processing on each collaborative influence subgraph in turn according to the sorting results.

2. The power grid data collaborative optimization method based on big data analysis according to claim 1, characterized in that, The obtained standard power grid data set includes: Identify the terminal type identifier and communication protocol version number carried in the multi-dimensional heterogeneous power grid data uploaded by each data acquisition terminal, and determine the original data encapsulation format of each multi-dimensional heterogeneous power grid data according to the terminal type identifier and the communication protocol version number; The standard encapsulation format corresponding to each of the original data encapsulation formats is queried from the preset format mapping rule library, and the data fields of each of the heterogeneous power grid data are rearranged and the data type is converted according to the standard encapsulation format obtained from the query, so that each heterogeneous power grid data is converted into a standard power grid data entry with a unified field order and a unified data type definition. All the converted standard power grid data entries are arranged in chronological order according to their respective acquisition timestamps to form a standard power grid data set.

3. The power grid data collaborative optimization method based on big data analysis according to claim 1, characterized in that, The step of constructing a time-series correlation map of power grid data for the target power grid region based on the collection timestamp and data item type identifier of each data item in the standard power grid data set includes: Each data item in the standard power grid data set that is under the same acquisition timestamp and belongs to different data item type identifiers is identified as a time-slice data item group, and the synchronous co-occurrence relationship between each data item type identifier is extracted from each time-slice data item group. Each data item type identifier is used as a node in the time-series correlation graph of the power grid data. The direction of the directed edge between two nodes in the time-series correlation graph is determined according to the number of consecutive occurrences of the synchronous co-occurrence relationship. If the number of synchronous co-occurrences of the first data item type identifier and the second data item type identifier in two consecutive time slices is greater than a preset threshold, then a directed edge is constructed from the node corresponding to the first data item type identifier to the node corresponding to the second data item type identifier.

4. The power grid data collaborative optimization method based on big data analysis according to claim 1, characterized in that, The process involves determining the data fluctuation activity and data update frequency of each node in the power grid data time-series correlation graph, and extracting a core data node set from the power grid data time-series correlation graph based on the data fluctuation activity and the data update frequency, including: For each node in the power grid data time-series correlation graph, obtain all data value sequences of the data item type corresponding to that node within a continuous scheduling period, calculate the absolute value of the difference between every two adjacent data values ​​in the data value sequence, and take the average of all absolute values ​​of the difference as the data fluctuation activity of that node; The total number of times the data item type corresponding to the node is updated within a continuous scheduling period is counted, and the ratio between the total number of times is increased and the total duration of the continuous scheduling period is taken as the data update frequency of the node. Calculate the product of the data fluctuation activity of the node and the data update frequency. Sort all nodes in descending order of the product value and extract the top N nodes as the core data node set, where N is a positive integer.

5. The power grid data collaborative optimization method based on big data analysis according to claim 1, characterized in that, The process of obtaining the collaborative impact subgraphs corresponding to each core data node includes: Select a core data node from the core data node set as the current starting node, and initialize the node list and edge list of the current collaborative influence subgraph; Starting from the current starting node, a depth-first traversal is performed along the direction of the directed edges in the time-series correlation graph of the power grid data. During the traversal, each visited node is added to the node list, and each directed edge traversed is added to the edge list. When a depth-first traversal starting from the current starting node fails to reach a new node, the subgraph formed by the current node list and edge list is treated as a collaborative influence subgraph corresponding to a core data node. Then, the next unprocessed core data node is selected from the set of core data nodes, and the above steps are repeated.

6. The power grid data collaborative optimization method based on big data analysis according to claim 1, characterized in that, The process involves calculating the graph structure entropy of each collaboratively affected subgraph, sorting all collaboratively affected subgraphs according to their entropy values, and then performing data collaborative optimization processing on each subgraph in sequence according to the sorting results. This includes: For each collaborative influence subgraph, obtain the out-degree value of all nodes in the collaborative influence subgraph, and determine the out-degree distribution probability sequence of the collaborative influence subgraph based on the out-degree values ​​of all nodes; The graph structure entropy of the cooperative influence subgraph is obtained by summing the product of each out-degree probability in the out-degree probability sequence with its logarithmic form and taking the negative value of the summation result. All collaborative influence subgraphs are arranged in descending order of graph structure entropy, and data redundancy cleaning and data consistency correction are performed on each data item contained in each collaborative influence subgraph in the order of arrangement.

7. The power grid data collaborative optimization method based on big data analysis according to claim 3, characterized in that, The number of consecutive occurrences of the synchronous co-occurrence relationship is obtained by statistically analyzing the co-occurrence frequency of different data item type identifiers in the same time slice data item group through a sliding time window, and the preset number threshold is dynamically adjusted according to the total number of data acquisition frames within the continuous scheduling period.

8. The power grid data collaborative optimization method based on big data analysis according to claim 4, characterized in that, The value of N in the core data node set is determined by multiplying the total number of nodes in the power grid data time-series correlation graph by a preset core node ratio coefficient, and the preset core node ratio coefficient is related to the load level of the target power grid area.

9. The power grid data collaborative optimization method based on big data analysis according to claim 5, characterized in that, The process of performing a depth-first traversal along the direction of the directed edges in the time-series correlation graph of the power grid data includes: When a node that has already been visited is encountered during the depth-first traversal, the node is marked as a repeated visit node, and the existing visit path of the repeated visit node in the node list is obtained. The current traversal path is merged with the existing access path to form a closed-loop path that includes the repeatedly accessed node. All nodes in the closed-loop path are marked as strongly related nodes. The depth-first traversal is then resumed from the next adjacent node of the repeatedly accessed node.

10. The power grid data collaborative optimization method based on big data analysis according to claim 6, characterized in that, The step of performing data redundancy cleaning and data consistency correction on each data item contained in each collaborative influence subgraph in the order of arrangement includes: Extract the data item type identifiers corresponding to each node from the node list of the currently pending collaborative influence subgraph, and read the corresponding data values ​​from the standard power grid data set according to the data item type identifiers; When determining that there are multiple data values ​​with the same timestamp under the same data item type identifier, only the data value with the earliest timestamp is retained, and the remaining data values ​​with the same timestamp are marked as redundant data items and deleted. Determine whether the data values ​​of the two ends of the directed edge of the current collaborative influence subgraph satisfy a preset numerical range constraint relationship. If not, correct the data values ​​of the two ends pointing to the endpoint according to the preset numerical range constraint relationship.