CN-121980207-A - Big data-based electricity load analysis method, equipment and medium

CN121980207ACN 121980207 ACN121980207 ACN 121980207ACN-121980207-A

Abstract

The invention discloses a power load analysis method, equipment and medium based on big data, and relates to the technical field of power big data analysis, comprising the steps of collecting a multi-source data set of power grid operation and preprocessing; the method comprises the steps of carrying out node-level time sequence feature analysis on a preprocessed power grid operation multisource data set, extracting demand change features and environment association features, generating node management feature vectors, creating nodes and connecting edges based on an organization structure and operation association relation of the power grid, constructing a dynamic space-time management diagram, carrying out influence propagation path backtracking and association reasoning analysis on an abnormal management event sequence based on the dynamic space-time management diagram, identifying root management events associated with the abnormal management events in time and space, and generating an abnormal operation management report. The invention realizes the evolution from the surface layer abnormality identification to the deep root diagnosis through the abnormality tracing based on the dynamic space-time management diagram, and achieves the optimization guidance of the accurate identification of the abnormality root and the operation and maintenance response.

Inventors

CHEN JIE
YU XINWEI
XU LINGXIANG
Jin Xiaofu
YOU XIAOBO
GONG YU
JIN RUIMIN

Assignees

杭州百富电子技术有限公司

Dates

Publication Date: 20260505
Application Date: 20260409

Claims (10)

1. The power load analysis method based on big data is characterized by comprising the following steps of, Collecting a power grid operation multisource data set and preprocessing; Carrying out node-level time sequence feature analysis on the preprocessed power grid operation multisource data set, extracting demand change features and environment association features, generating node management feature vectors, creating nodes and connecting edges based on the organization structure and operation association relation of the power grid, and constructing a dynamic space-time management diagram; Inputting a dynamic space-time management diagram into a space-time diagram neural network model, and deducing the expected load demand of each node in a future period through node characteristic aggregation and cross-node characteristic propagation calculation to generate a power grid load operation plan; acquiring actual operation data at corresponding time of execution of the power grid load operation plan, comparing and analyzing the actual operation data with the power grid load operation plan, and executing abnormality detection and judgment according to deviation judgment rules and deviation duration time conditions to generate an abnormality management event sequence; Based on the dynamic space-time management diagram, the abnormal management event sequence is subjected to influence propagation path backtracking and association reasoning analysis, the root management event associated with the abnormal management event space-time is identified, and an abnormal operation management report is generated.
2. The electrical load analysis method based on big data of claim 1, wherein the multi-source data set for power grid operation comprises historical and real-time electrical load data, power grid topology data, operational environment meteorological data and planned operational event data; the preprocessing includes data cleaning, spatiotemporal alignment, and feature normalization.
3. The method for analyzing electricity load based on big data according to claim 1, wherein the steps of extracting the demand change feature and the environment correlation feature, generating a node management feature vector are as follows, Taking a power grid node as a unit, and carrying out time sequence recombination on historical and real-time power consumption load data according to a node identifier to form a node load time sequence data set; Based on the node load time sequence data set, extracting the demand change characteristics of each power grid node through a time sequence pattern mining algorithm, and generating a node internal demand characteristic vector; and performing feature intersection and fusion on the node intrinsic demand feature vector, the operation environment meteorological data and the planning operation event data to generate a node management feature vector.
4. The electrical load analysis method based on big data of claim 3, wherein the organization structure based on the power grid and the operation association relation create nodes and connecting edges, and a dynamic space-time management diagram is constructed by the steps of, The node management feature vector is used as node attribute, physical connection relation among grid nodes is mapped by combining the grid topological structure data, and association weight is calculated to generate a static management association map; and in the static management association graph, performing association update on the change of the node management feature vector in a continuous time window to generate a dynamic space-time management graph.
5. The electricity load analysis method based on big data of claim 1, wherein the dynamic space-time management diagram is input into a space-time diagram neural network model, the expected load demand of each node in the future period is deduced through node characteristic aggregation and cross-node characteristic propagation calculation, and a power grid load operation plan is generated by the following steps, Inputting the dynamic space-time management graph into a space-time graph neural network model, carrying out characteristic convolution aggregation and attention weight distribution of multi-order neighborhood nodes on the node management characteristic vector along the connecting edge through a graph convolution layer, and generating a node operation state portrait; Inputting the node operation state portraits to a time sequence network layer of a space-time diagram neural network model, capturing the dynamic evolution rule of the node operation states of each power grid along with time through a gating circulation mechanism, and generating node time sequence operation coding vectors; And carrying out nonlinear regression calculation on the node time sequence operation coding vector through a full-connection layer of the space-time diagram neural network model, mapping out the expected load demand of each power grid node in a future period, and integrating to generate a power grid load operation plan.
6. The electrical load analysis method based on big data of claim 1, wherein the step of collecting actual operation data at the corresponding time of the execution of the power grid load operation plan and comparing the actual operation data with the power grid load operation plan is as follows, Collecting actual electricity load data of each power grid node at each execution time point in a power grid load operation plan, and generating a real-time operation performance data set; And comparing the real-time operation performance data set with the expected load demand in the power grid load operation plan, calculating absolute deviation and relative deviation rate, and generating an operation performance deviation data set.
7. The electrical load analysis method based on big data according to claim 6, wherein the abnormality detection and determination is performed according to a deviation determination rule and a deviation duration condition, an abnormality management event sequence is generated as follows, Scanning the operation performance deviation data set according to the deviation judging rule to identify a primary abnormal signal set; And applying deviation duration time conditions to the primary abnormal signal set to judge the persistence and the stability, confirming the abnormal signals meeting the persistence requirements as operation abnormal events, and integrating and generating an abnormal management event sequence.
8. The method for analyzing electrical loads based on big data according to claim 1, wherein the step of generating the abnormal operation management report is as follows, Extracting power grid nodes and occurrence time corresponding to each abnormal management event in the abnormal management event sequence, and defining associated space-time influence ranges in a dynamic space-time management diagram to form an abnormal influence space-time range set; based on the connection edge relation in the dynamic space-time management diagram, carrying out inverse tracking on the abnormal influence space-time range set, and identifying a potential influence propagation path set; Matching and causal association analysis are carried out on the potential influence propagation path set and the dynamic space-time management graph, the root cause with the highest space-time coupling degree with the abnormal management event is identified, and the root cause is determined as the root management event; and integrating the abnormal management event sequence, the potential influence propagation path set and the root management event to generate an abnormal operation management report.
9. A computer device comprises a memory and a processor, wherein the memory stores a computer program, and the computer device is characterized in that the processor executes the computer program to realize the steps of the big data-based power load analysis method according to any one of claims 1-8.
10. A computer-readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the big data based electricity load analysis method according to any one of claims 1 to 8.

Description

Big data-based electricity load analysis method, equipment and medium Technical Field The invention relates to the technical field of power big data analysis, in particular to a power load analysis method, equipment and medium based on big data. Background Under the background of rapid development of smart power grids and energy Internet, the power load analysis field faces challenges of data scale proliferation and obvious multi-source heterogeneous characteristics, the existing method mainly relies on statistical time sequence analysis (such as seasonal differential autoregressive moving average model) and traditional machine learning algorithms (such as support vector regression, gradient lifting decision tree and the like), a single-point or multi-point load prediction model is built after historical load data are subjected to cleaning, noise reduction and normalization preprocessing, a daily or short-term load prediction curve is provided for a power grid dispatching center, and data support is provided for power grid basic operation by mining the time sequence rule of the load. However, the existing method has room for improvement in the aspects of high-dimensional heterogeneous data fusion and anomaly traceability, integration of multi-source data such as weather, topology and events is stopped at a characteristic splicing level, a deep coupling relation among unified space-time frame mining data cannot be established, the completeness of model input information is affected, the anomaly analysis link lacks dynamic traceability of a fault propagation path, root factors are difficult to position from space-time dimensions, and the accuracy of operation and maintenance decisions is restricted. Disclosure of Invention The present invention has been made in view of the above-described problems occurring in the prior art. Therefore, the invention provides a power load analysis method based on big data, which solves the problems that multi-source data fusion does not go deep and the abnormal traceability is weak. In order to solve the technical problems, the invention provides the following technical scheme: The invention provides an electricity load analysis method based on big data, which comprises the steps of collecting a power grid operation multisource data set and preprocessing the power grid operation multisource data set, carrying out node level time sequence feature analysis on the preprocessed power grid operation multisource data set, extracting demand change features and environment association features, generating node management feature vectors, creating nodes and connection edges based on an organization structure and operation association relation of a power grid, constructing a dynamic space-time management graph, inputting the dynamic space-time management graph into a space-time graph neural network model, deducing the expected load demands of each node in the future period through node feature aggregation and cross-node feature propagation calculation, generating a power grid load operation plan, collecting actual operation data at corresponding time executed by the power grid load operation plan, carrying out anomaly detection and judgment according to deviation judgment rules and deviation duration time conditions, generating an anomaly management event sequence, carrying out influence propagation path backtracking and association reasoning analysis on the anomaly management event sequence based on the dynamic space-time management graph, identifying root management events associated with the anomaly management event time and generating an anomaly operation management report. As a preferable scheme of the big data-based electricity load analysis method, the power grid operation multisource data set comprises historical and real-time electricity load data, power grid topological structure data, operation environment meteorological data and planning operation event data; the preprocessing includes data cleaning, spatiotemporal alignment, and feature normalization. The invention is an optimal scheme of the electricity load analysis method based on big data, wherein the method extracts the demand change characteristic and the environment association characteristic, generates a node management characteristic vector, comprises the following steps, Taking a power grid node as a unit, and carrying out time sequence recombination on historical and real-time power consumption load data according to a node identifier to form a node load time sequence data set; Based on the node load time sequence data set, extracting the demand change characteristics of each power grid node through a time sequence pattern mining algorithm, and generating a node internal demand characteristic vector; and performing feature intersection and fusion on the node intrinsic demand feature vector, the operation environment meteorological data and the planning operation event data to generate a node management feature ve