CN-122019980-A - Data analysis method and system for power grid data abnormality

CN122019980ACN 122019980 ACN122019980 ACN 122019980ACN-122019980-A

Abstract

The invention discloses a data analysis method and a system for power grid data abnormality, which relate to the technical field of data analysis and are used for solving the problem that original abnormal characteristics are weakened, monitoring the abnormal level and data retransmission characteristics of power grid data, judging whether a repair analysis mechanism is introduced or not by combining the number of abnormal values and the retransmission behaviors, analyzing time sequence records of repairing completed returned data under the repair analysis mechanism, calculating repair time delay and repair aggregation frequency, describing the concentration degree of the data repair behaviors from a time dimension, performing interval aggregation on repair type information in a characteristic analysis window, extracting repair repeated characteristics, comprehensively evaluating the data repair state, combining repair amplitude data with power grid operation data to perform correlation detection when the repair state is abnormal, identifying the abnormal repair behaviors and outputting alarm results, and providing a new technical means for power grid data reliability evaluation and abnormality tracing under the condition that the original abnormality is smoothed or interpolation covered.

Inventors

ZHANG KUN
YAO XU
YANG YUFEI
CHEN YUE
YE ZIMING
ZHANG ZHIXIN

Assignees

广东电网有限责任公司

Dates

Publication Date: 20260512
Application Date: 20260203

Claims (10)

1. The data analysis method for the power grid data abnormality is characterized by comprising the following steps of: Step S1, monitoring power grid data, counting the number of abnormal values of the power grid data, judging whether to access a data processing platform according to the number of abnormal values, executing repair processing, and acquiring the data retransmission times of the data processing platform before executing the repair processing; Step S2, selecting and triggering a conventional processing mechanism or a repair analysis mechanism based on the number of data retransmission times, acquiring time sequence records of the power grid data after the power grid data is repaired and returned in the repair analysis mechanism, counting the repair time delay of each time sequence record and calculating the repair aggregation frequency; step S3, setting a feature analysis window, accessing the repair type information of each time sequence record in the feature analysis window, performing interval aggregation analysis on the repair type information and generating repair repeated features, and evaluating the data repair state of the time sequence record by combining the repair aggregation frequency; And S4, when the data restoration state is abnormal, restoring amplitude data and power grid operation data are called, detection results are obtained after the restoring amplitude data and the power grid operation data are subjected to association detection, whether abnormal restoration exists or not is judged based on the detection results, and an abnormal alarm signal is output.
2. A data analysis method for grid data anomalies as recited in claim 1, wherein: In step S1, continuously monitoring power grid data, wherein the power grid data is original data formed by real-time acquisition of power grid side sensing equipment in the operation process; the method comprises the steps that a time sequence data sequence arranged in time sequence is formed by power grid data according to a preset sampling period, each data point in the time sequence data sequence is compared one by one based on a preset abnormality judgment condition, and when any data point meets the abnormality judgment condition, the data point is judged to be an abnormal value; the data points meeting the abnormality judgment conditions are accumulated and counted to obtain the abnormal value quantity of the power grid data; When the number of the abnormal values reaches or exceeds a preset abnormal intervention threshold value, judging that the abnormal degree of the power grid data exceeds the local processing capacity, and accessing a data processing platform and executing repair processing; when the number of the abnormal values is smaller than a preset abnormal intervention threshold value, judging that the abnormal degree of the current power grid data is in an acceptable range, and not triggering the repair processing flow of the data processing platform.
3. A data analysis method for grid data anomalies according to claim 2, wherein: In step S1, under the condition that it is determined that the data processing platform needs to be accessed and repair processing is executed, before the grid data is formally submitted to the data processing platform, acquiring the number of data retransmission times in the current data transmission process corresponding to the data processing platform; The data retransmission times are accumulated values of the repeated data transmission times caused by communication abnormality, transmission verification failure or platform side request retransmission in the historical process of transmitting the power grid data to the data processing platform.
4. A data analysis method for grid data anomalies as recited in claim 1, wherein: In step S2, when the number of data retransmissions is less than a preset retransmission decision threshold, triggering a conventional processing mechanism; triggering a repair analysis mechanism when the number of data retransmission times reaches or exceeds a retransmission judgment threshold value; After triggering a repair analysis mechanism, after completing repair processing of the power grid data by a third party data processing platform and returning the repaired power grid data, acquiring a time sequence record corresponding to the power grid data; The time sequence record is a continuous data sequence formed by taking a timestamp as an index, wherein each data point is associated with an original acquisition time and a repair completion return time.
5. A data analysis method for grid data anomalies as recited in claim 4, wherein: in step S2, based on the time sequence record, a repair delay is calculated for each data point, where the repair delay is defined as a time difference between a repair completion return time and a corresponding original acquisition time; Performing time positioning on the repair completion events based on the repair time delay, and performing aggregation statistics on the repair completion events in a preset time statistics window to obtain the repair completion number corresponding to the time statistics window; and taking the number of the repair completion counted in the unit time length as the repair aggregation frequency.
6. A data analysis method for grid data anomalies as recited in claim 1, wherein: In step S3, presetting a characteristic analysis window, and acquiring repair type information corresponding to each time sequence record through a repair log of a third-party data processing platform in the characteristic analysis window; The repair type information refers to a repair type identifier adopted by the third party data processing platform when the third party data processing platform executes repair processing on the power grid data; Performing interval aggregation analysis on the repair type information, dividing adjacent time sequence records with the same repair type identifier in a characteristic analysis window into the same repair type interval, and counting the continuous length and the occurrence frequency of each repair type interval; and for the same repair type interval, taking the ratio result of the occurrence times and the continuous length of the repair type interval as a repair type coverage coefficient, and taking the maximum value of the repair type coverage coefficients as a repair repetition characteristic.
7. A data analysis method for grid data anomalies as recited in claim 1, wherein: In step S3, the repair repetition characteristic and the repair aggregation frequency are respectively standardized to obtain a repair repetition coefficient and a repair aggregation coefficient; taking the product result of the repair repetition coefficient and the repair aggregation coefficient as a comprehensive repair index; If the comprehensive repair index is larger than the preset repair threshold, judging that the data repair state of the time sequence record is an abnormal state; Otherwise, the data restoration state of the time sequence record is judged to be a normal state.
8. A data analysis method for grid data anomalies as recited in claim 1, wherein: In step S4, when the data repair state is abnormal, obtaining repair amplitude data comprising a repair amplitude value through a repair log of the data processing platform; The power grid operation data refer to operation state data which are collected and formed by a power grid side monitoring unit in a characteristic analysis window, and the operation state data comprise voltage variation of the repaired power grid data; The repair amplitude value refers to a numerical value difference between a data value after repair and a data value before repair when the third party data processing platform executes repair processing on the power grid data, and the voltage variation is a numerical value difference between voltage data corresponding to adjacent time sequence records in a characteristic analysis window.
9. A data analysis method for grid data anomalies as recited in claim 8, wherein: in step S4, if the repair amplitude value and the sign direction of the voltage variation at the same time are the same, determining that the repair behavior of the time sequence record is consistent with the change direction of the power grid running state, and recording as consistent repair; otherwise, judging that the repair behavior of the time sequence record is inconsistent with the change direction of the running state of the power grid, and recording as inconsistent repair; Counting the number of the time sequence records of inconsistent repair and the total time sequence record number, and taking the ratio of the number of the time sequence records of inconsistent repair to the total time sequence record number as a correlation detection result; if the correlation detection result is larger than a preset detection result threshold value, judging and outputting an abnormal alarm signal; otherwise, judging that the abnormal alarm signal is not output.
10. A data analysis system for power grid data abnormality, for implementing a data analysis method for power grid data abnormality according to any one of claims 1 to 9, characterized by comprising an abnormality monitoring module, a mechanism judging module, a state evaluating module and an abnormality alarming module, each module having the following functions: The abnormal monitoring module is used for monitoring the power grid data, counting the abnormal value quantity of the power grid data, judging whether to access the data processing platform according to the abnormal value quantity, executing repair processing, acquiring the data retransmission times of the data processing platform before executing the repair processing, and transmitting the data retransmission times to the mechanism judging module; the mechanism judging module selects to trigger a conventional processing mechanism or a repair analysis mechanism based on the data retransmission times, in the repair analysis mechanism, after the repair of the power grid data is completed and returned, the time sequence records of the power grid data are obtained, the repair time delay of each time sequence record is counted, the repair aggregation frequency is calculated, and the repair aggregation frequency is transmitted into the state evaluation module; The state evaluation module sets a feature analysis window, accesses the repair type information of each time sequence record in the feature analysis window, performs interval aggregation analysis on the repair type information and generates repair repeated features, evaluates the data repair state of the time sequence records in combination with the repair aggregation frequency, and transmits the data repair state to the abnormal alarm module; And when the data restoration state is abnormal, the abnormality alarming module invokes restoration amplitude data and power grid operation data, obtains a detection result after carrying out association detection on the restoration amplitude data and the power grid operation data, judges whether abnormality restoration exists or not based on the detection result, and outputs an abnormality alarming signal.

Description

Data analysis method and system for power grid data abnormality Technical Field The invention relates to the technical field of data analysis, in particular to a data analysis method and system for power grid data abnormality. Background With the rapid development of novel power systems and digital power grids, a dispatching automation system, a distribution automation system, a power consumption information acquisition system and various on-line monitoring devices continuously generate massive power grid operation data, and in order to meet application requirements such as situation awareness, operation analysis, state evaluation and auxiliary decision making, a unified data processing platform is generally required to be accessed to execute processing operations such as data cleaning, missing complement and the like so as to improve data integrity and usability. The prior art has the following defects: At present, the prior art focuses on quality evaluation and anomaly determination on original data or repaired data results of a power grid, lacks systematic analysis means for data processing processes such as data retransmission behaviors, repair triggering mechanisms, repair type distribution and timing repair features, and the like, and once the original anomalies of the power grid data are passively corrected in the repair processes such as cleaning, interpolation or smoothing, the original anomalies are weakened or even covered, and the anomaly repair behaviors are difficult to effectively identify and quantitatively evaluate, so that the data analysis method and system for the anomalies of the power grid data are provided. The above information disclosed in the background section is only for enhancement of understanding of the background of the disclosure and therefore it may include information that does not form the prior art that is already known to a person of ordinary skill in the art. Disclosure of Invention In order to overcome the above-mentioned drawbacks of the prior art, embodiments of the present invention provide a data analysis method and system for power grid data anomaly, which solve the problems set forth in the above-mentioned background art by applying a repair mechanism based on the number of outliers and the number of data retransmissions to trigger a decision strategy, and combining a time sequence analysis model of repair delay and repair aggregation frequency, a repair type interval aggregation and repeated feature extraction mechanism, and a correlation detection method of repair amplitude and power grid operation data. In order to achieve the above object, the present invention provides a data analysis method for power grid data anomaly, comprising the following steps: Step S1, monitoring power grid data, counting the number of abnormal values of the power grid data, judging whether to access a data processing platform according to the number of abnormal values, executing repair processing, and acquiring the data retransmission times of the data processing platform before executing the repair processing; Step S2, selecting and triggering a conventional processing mechanism or a repair analysis mechanism based on the number of data retransmission times, acquiring time sequence records of the power grid data after the power grid data is repaired and returned in the repair analysis mechanism, counting the repair time delay of each time sequence record and calculating the repair aggregation frequency; step S3, setting a feature analysis window, accessing the repair type information of each time sequence record in the feature analysis window, performing interval aggregation analysis on the repair type information and generating repair repeated features, and evaluating the data repair state of the time sequence record by combining the repair aggregation frequency; And S4, when the data restoration state is abnormal, restoring amplitude data and power grid operation data are called, detection results are obtained after the restoring amplitude data and the power grid operation data are subjected to association detection, whether abnormal restoration exists or not is judged based on the detection results, and an abnormal alarm signal is output. In a preferred embodiment, in step S1, the power grid data is continuously monitored, where the power grid data is raw data collected in real time by the power grid side sensing device during operation; the method comprises the steps that a time sequence data sequence arranged in time sequence is formed by power grid data according to a preset sampling period, each data point in the time sequence data sequence is compared one by one based on a preset abnormality judgment condition, and when any data point meets the abnormality judgment condition, the data point is judged to be an abnormal value; the data points meeting the abnormality judgment conditions are accumulated and counted to obtain the abnormal value quantity of the power grid data;