An intelligent power distribution room sensor multi-dimensional data fusion abnormal event detection method, system and computer readable storage medium

Abstract

This invention discloses a method, system, and computer-readable storage medium for detecting abnormal events using multi-dimensional data fusion from sensors in an intelligent power distribution room. The method includes: S1: acquiring normal multi-dimensional sensor data from the power distribution room and establishing a normal sample database; S2: preprocessing the multi-dimensional data in the normal sample database and establishing a multi-dimensional data correlation feature model; S3: inputting the preprocessed real-time operating data of the power distribution room into the multi-dimensional data correlation feature model, and analyzing the real-time operating data to determine whether the real-time operating data is abnormal. If the data is abnormal, proceed to step S4; if the data is normal, end the current detection process and proceed to the next data detection step; S4: comparing the abnormal data with a preset abnormal event database. If the abnormal data belongs to an abnormal event, output the abnormal event type; otherwise, update the model. This invention fully utilizes the correlation of multi-dimensional data, improving data utilization and the efficiency of abnormal event detection.

Description

Technical Field

[0001] This invention relates to the field of intelligent detection technology for power distribution rooms, and more specifically, to a method, system, and computer-readable storage medium for detecting abnormal events by multi-dimensional data fusion from sensors in intelligent power distribution rooms. Background Technology

[0002] Substations are an indispensable and crucial part of the power grid system. Their wide distribution and large number within the grid increase the difficulty of their management. Manual inspections are time-consuming, inefficient, and labor-intensive. Therefore, intelligent online monitoring systems and inspection robots have been extensively researched and applied in substations. However, the data collected by online monitoring systems is rarely effectively utilized. A large amount of multi-dimensional monitoring data often results in uncorrelated data sets, or simply setting normal operating limits for the measured data. The relevant information between multiple sensors' multi-dimensional data is rarely effectively utilized, making it difficult to detect anomalies in substations in a timely manner.

[0003] In the prior art, Chinese invention patent CN110690763A, published on January 14, 2021, discloses an intelligent monitoring device and method for a power system. This device includes: a signal acquisition front-end, a signal processing system, a central main control system, a data storage unit, a network unit, and a power management unit. The signal acquisition front-end includes: a split-type partial discharge sensor, an ambient humidity sensor, and an ambient temperature sensor. While this solution achieves the acquisition of data from multiple sensors, it fails to correlate or fuse the acquired data, thus failing to effectively utilize multi-dimensional data. Summary of the Invention

[0004] To overcome the shortcomings of existing power distribution room detection methods, such as the lack of effective utilization of multi-dimensional monitoring data, the failure to achieve correlation of multi-dimensional data, low data utilization, and low detection efficiency, this invention provides a method, system, and computer-readable storage medium for detecting abnormal events by multi-dimensional data fusion from intelligent power distribution room sensors.

[0005] The primary objective of this invention is to solve the aforementioned technical problems. The technical solution of this invention is as follows:

[0006] The first aspect of this invention provides a method for detecting abnormal events through multi-dimensional data fusion from sensors in a smart power distribution room, comprising the following steps:

[0007] S1: Acquire normal multidimensional sensor data from the power distribution room and establish a normal sample database;

[0008] S2: Preprocess the multidimensional data in the normal sample database to establish a multidimensional data correlation feature model;

[0009] S3: After preprocessing the real-time operation data of the power distribution room, input it into the multi-dimensional data correlation feature model, and analyze the real-time operation data to determine whether the real-time operation data is abnormal. If the real-time operation data is abnormal, proceed to step S4. If the real-time operation data is normal, end the current detection process and proceed to the detection of the data at the next moment.

[0010] S4: Compare the abnormal data with the preset abnormal event library. If the abnormal data belongs to an abnormal event, output the determined abnormal event type. If the abnormal data does not belong to an abnormal event, update the multidimensional data correlation feature model.

[0011] Furthermore, the construction process of the multidimensional data correlation feature model described in step S2 is as follows:

[0012] Define a fixed time window length L for the multidimensional data, and calculate the correlation coefficient between the multidimensional variables within each time window length L.

[0013] A correlation coefficient matrix is ​​constructed using the calculated correlation coefficients, and the threshold values ​​for each correlation coefficient are determined for normal data.

[0014] A multidimensional data correlation feature model is established based on the correlation coefficient matrix and the correlation coefficient threshold.

[0015] Furthermore, the correlation coefficient is the Pearson correlation coefficient.

[0016] Furthermore, step S3, which involves analyzing the real-time operational data to determine whether the data is abnormal, specifically includes:

[0017] The correlation coefficients of multidimensional data within the real-time running data are calculated sequentially using the sliding time window method.

[0018] If the correlation coefficient between multidimensional data is greater than the correlation coefficient threshold in the model, then it is an outlier data point;

[0019] If the correlation coefficient between multidimensional data is less than or equal to the correlation coefficient threshold in the model, then a judgment is made for the next time window.

[0020] Furthermore, the preset abnormal event library mentioned in step S4 is established based on expert experience, and the event types included in the abnormal event library are: abnormality of the detected device and abnormality of the detection device.

[0021] Furthermore, step S4 also includes manually determining whether the abnormal data point is an abnormal data point if there is no corresponding abnormal event in the abnormal event database; if it is an abnormal data point, the abnormal data point is defined as a new abnormal event and saved to the abnormal event database to complete the update of the abnormal event database; if it is not an abnormal data point, the multidimensional data correlation feature model is updated.

[0022] Furthermore, if the abnormal data is determined not to be an abnormal data point during manual judgment, the specific process for updating the multidimensional data correlation feature model is as follows:

[0023] Add time series windows that were incorrectly identified as abnormal data to the normal monitoring database and update the normal data sample library;

[0024] Recalculate the correlation coefficients among the multidimensional samples;

[0025] Redetermine the correlation coefficient threshold.

[0026] A second aspect of this invention provides an intelligent power distribution room sensor multi-dimensional data fusion anomaly event detection system, characterized by comprising: a data acquisition module, a data processing module, a modeling module, an online analysis module, an anomaly event database, and an evaluation module. The data acquisition module acquires multi-dimensional sensor data from the power distribution room; the data processing module processes the acquired multi-dimensional sensor data; the modeling module establishes and updates a multi-dimensional data correlation feature model; the online analysis module detects abnormal data points and determines the type of anomaly events online; the anomaly event database includes anomalies added during initial construction and those added during model updates; and the evaluation module evaluates the accuracy of the multi-dimensional data correlation feature model and the accuracy of anomaly event pattern judgment.

[0027] Furthermore, the data processing module performs the following data processing methods: classification, cleaning, transformation, time series alignment, standardization, and integration.

[0028] A third aspect of the present invention provides a computer-readable storage medium including a program for detecting abnormal events by multi-dimensional data fusion of sensors in a smart power distribution room. When the program is executed by a processor, it implements the steps of the method for detecting abnormal events by multi-dimensional data fusion of sensors in a smart power distribution room.

[0029] Compared with the prior art, the beneficial effects of the technical solution of the present invention are:

[0030] This invention acquires multidimensional data from sensors in a power distribution room and uses a constructed multidimensional data correlation feature model to detect abnormal events. This fully utilizes the correlation of multidimensional data, improving data utilization and the efficiency of abnormal event detection. Attached Figure Description

[0031] Figure 1 This is a flowchart of an abnormal event detection method for multi-dimensional data fusion of sensors in an intelligent power distribution room according to the present invention.

[0032] Figure 2 This is a block diagram of an intelligent power distribution room sensor multi-dimensional data fusion abnormal event detection system according to the present invention. Detailed Implementation

[0033] To better understand the above-mentioned objectives, features, and advantages of the present invention, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0034] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and therefore the scope of protection of the invention is not limited to the specific embodiments disclosed below.

[0035] Example 1

[0036] like Figure 1 As shown, the first aspect of the present invention provides a method for detecting abnormal events by multi-dimensional data fusion from sensors in a smart power distribution room, comprising the following steps:

[0037] S1: Acquire normal multidimensional sensor data from the power distribution room and establish a normal sample database;

[0038] It should be noted that the normal multidimensional sensing data of the power distribution room mentioned in this invention refers to a continuous period of normal operating data from sensors in the power distribution room. The sensor types mainly include temperature and humidity sensors, noise sensors, ozone sensors, sulfur tetrafluoride tactile sensors, oxygen sensors, etc. The normal sample database is the collection of normal multidimensional sensing data from the power distribution room.

[0039] S2: Preprocess the multidimensional data in the normal sample database to establish a multidimensional data correlation feature model;

[0040] S3: After preprocessing the real-time operation data of the power distribution room, input it into the multi-dimensional data correlation feature model, and analyze the real-time operation data to determine whether the real-time operation data is abnormal. If the real-time operation data is abnormal, proceed to step S4. If the real-time operation data is normal, end the current detection process and proceed to the detection of the data at the next moment.

[0041] S4: Compare the abnormal data with the preset abnormal event library. If the abnormal data belongs to an abnormal event, output the determined abnormal event type. If the abnormal data does not belong to an abnormal event, update the multidimensional data correlation feature model.

[0042] It should be noted that the preprocessing of multidimensional data in the normal sample database and the preprocessing of real-time operation data of the power distribution room in this invention both include the following processes: classification, cleaning, transformation, time series alignment, and standardization. The specific processing steps are as follows:

[0043] Classification: Data is categorized and numbered according to its origin from different sensor terminals. For example, noise data 1 is A1, noise data 2 is A2, noise data 3 is A3, temperature 1 is A4, temperature 2 is A5, temperature 3 is A6, etc.

[0044] Data cleaning: Missing and duplicate data from the sensor are cleaned. Missing values ​​are considered temporary anomalies; if missing values ​​persist for a certain period, they are considered event anomalies. Duplicate data is deleted.

[0045] Conversion: Convert sensor data to the same format. Convert time stamps to the same format, such as 01 / 01 / 2021 11:59:00.

[0046] Time series alignment: Aligning formatted data according to time labels.

[0047] Standardization: Different sensors have different measurement ranges, so all data are standardized. The standardization method is min-max standardization.

[0048] Furthermore, the construction process of the multidimensional data correlation feature model described in step S2 is as follows:

[0049] Define a fixed time window length L for the multidimensional data, and calculate the correlation coefficient between the multidimensional variables within each time window length L.

[0050] A correlation coefficient matrix is ​​constructed using the calculated correlation coefficients, and the threshold values ​​for each correlation coefficient are determined for normal data.

[0051] A multidimensional data correlation feature model is established based on the correlation coefficient matrix and the correlation coefficient threshold.

[0052] In this invention, the correlation coefficient matrix and the threshold constitute a multidimensional data correlation feature model, and the expression of the correlation coefficient matrix is ​​as follows:

[0053] Where AmAn are the correlation coefficients between the sensor data numbered Am and An;

[0054] The correlation coefficient matrix of all samples in the normal sample database is calculated using the above formula, and the threshold of the correlation coefficient matrix is ​​determined.

[0055] Furthermore, the correlation coefficient is the Pearson correlation coefficient.

[0056] Furthermore, step S3, which involves analyzing the real-time operational data to determine whether the data is abnormal, specifically includes:

[0057] The correlation coefficients of multidimensional data within the real-time running data are determined sequentially using a sliding time window method.

[0058] If the correlation coefficient between multidimensional data is greater than the correlation coefficient threshold in the model, then it is an outlier data point;

[0059] If the correlation coefficient between multidimensional data is less than or equal to the correlation coefficient threshold in the model, then a judgment is made for the next time window.

[0060] Furthermore, the preset abnormal event library mentioned in step S4 is established based on expert experience, and the event types included in the abnormal event library are: abnormality of the detected device and abnormality of the detection device.

[0061] It should be noted that the malfunction of the tested equipment, such as a malfunction of the switchgear in the power distribution room, or the malfunction of the detection equipment, such as a malfunction of the sensor itself, are all possible malfunctions. The sensors can be temperature sensors, noise sensors, ozone sensors, sulfur tetrafluoride sensors, oxygen sensors, etc. These sensors can acquire the concentration of the corresponding gas in the power distribution room or environmental parameters such as temperature and humidity. The multidimensional data correlation feature model in this invention can detect anomalies in the temporal change trend of sensor data, as well as anomalies in the correlation between sensor data.

[0062] Furthermore, step S4 also includes manually determining whether the abnormal data point is an abnormal data point if there is no corresponding abnormal event in the abnormal event database; if it is an abnormal data point, the abnormal data point is defined as a new abnormal event and saved to the abnormal event database to complete the update of the abnormal event database; if it is not an abnormal data point, the multidimensional data correlation feature model is updated.

[0063] It should be noted that, in one specific embodiment, both the malfunction of the device being tested and the malfunction of the testing device will generate abnormal data. Further manual judgment will make the abnormality handling more accurate.

[0064] Furthermore, if the abnormal data is determined not to be an abnormal data point during manual judgment, the specific process for updating the multidimensional data correlation feature model is as follows:

[0065] Add time series windows that were incorrectly identified as abnormal data to the normal monitoring database and update the normal data sample library;

[0066] Recalculate the correlation coefficients among the multidimensional samples;

[0067] Redetermine the correlation coefficient threshold.

[0068] like Figure 2 As shown, the second aspect of the present invention provides an intelligent power distribution room sensor multi-dimensional data fusion anomaly event detection system, characterized in that it includes: a data acquisition module, a data processing module, a modeling module, an online analysis module, an anomaly event database, and an evaluation module. The data acquisition module is used to acquire multi-dimensional sensor data of the power distribution room; the data processing module is used to process the acquired multi-dimensional sensor data of the power distribution room; the modeling module is used to establish and update a multi-dimensional data correlation feature model; the online analysis module is used to detect abnormal data points and determine the type of abnormal events online; the anomaly event database includes anomaly events added during the initial construction and anomaly events added during the model update process; and the evaluation module is used to evaluate the accuracy of the multi-dimensional data correlation feature model and the accuracy of the anomaly event pattern judgment.

[0069] Furthermore, the data processing module performs the following data processing methods: classification, cleaning, transformation, time series alignment, standardization, and integration.

[0070] A third aspect of the present invention provides a computer-readable storage medium including a program for detecting abnormal events by multi-dimensional data fusion of sensors in a smart power distribution room. When the program is executed by a processor, it implements the steps of the method for detecting abnormal events by multi-dimensional data fusion of sensors in a smart power distribution room.

[0071] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the implementation of the present invention. Those skilled in the art can make other variations or modifications based on the above description. It is neither necessary nor possible to exhaustively describe all embodiments here. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the scope of protection of the claims of the present invention.

Claims

1. A method for detecting abnormal events by multi-dimensional data fusion from sensors in an intelligent power distribution room, characterized in that, Includes the following steps: S1: Acquire normal multidimensional sensor data from the power distribution room and establish a normal sample database; S2: Preprocess the multidimensional data in the normal sample database to establish a multidimensional data correlation feature model; S3: After preprocessing the real-time operation data of the power distribution room, input it into the multi-dimensional data correlation feature model, and analyze the real-time operation data to determine whether the real-time operation data is abnormal. If the real-time operation data is abnormal, proceed to step S4. If the real-time operation data is normal, end the current detection process and proceed to the detection of the data at the next moment. S4: Compare the abnormal data with the preset abnormal event database. If the abnormal data belongs to an abnormal event, output the determined abnormal event type. If the abnormal data does not belong to an abnormal event, update the multidimensional data correlation feature model. The construction process of the multidimensional data correlation feature model described in step S2 is as follows: Define a fixed time window length L for the multidimensional data, and calculate the Pearson correlation coefficient between the multidimensional variables within each time window length L. A correlation coefficient matrix is ​​constructed using the calculated correlation coefficients, and the threshold values ​​for each correlation coefficient are determined for normal data. A multidimensional data correlation feature model is established based on the correlation coefficient matrix and the correlation coefficient threshold. Specifically, step S3 involves analyzing the real-time operating data to determine whether the data is abnormal. The correlation coefficients of multidimensional data within the real-time running data are calculated sequentially using the sliding time window method. If the correlation coefficient between multidimensional data is greater than the correlation coefficient threshold in the model, then it is an outlier data point; If the correlation coefficient between multidimensional data is less than or equal to the correlation coefficient threshold in the model, then a judgment is made for the next time window.

2. The method for detecting abnormal events by multi-dimensional data fusion of sensors in an intelligent power distribution room according to claim 1, characterized in that, The preset abnormal event library mentioned in step S4 is established based on expert experience, and the event types included in the abnormal event library are: abnormality of the detected device and abnormality of the detection device.

3. The method for detecting abnormal events by multi-dimensional data fusion of sensors in an intelligent power distribution room according to claim 1, characterized in that, Step S4 also includes manually determining whether an abnormal data point is an abnormal data point if there is no corresponding abnormal event in the abnormal event database. If it is an abnormal data point, the abnormal data point is defined as a new abnormal event and saved to the abnormal event database to complete the update of the abnormal event database. If it is not an abnormal data point, the multidimensional data correlation feature model is updated.

4. The method for detecting abnormal events by multi-dimensional data fusion of sensors in an intelligent power distribution room according to claim 3, characterized in that, If, in manual judgment, it is determined that the abnormal data is not an abnormal data point, the specific process of updating the multidimensional data correlation feature model is as follows: add the time series window of the abnormal data that was incorrectly judged to the monitoring normal database, and update the normal data sample library. Recalculate the correlation coefficients among the multidimensional samples; Redetermine the correlation coefficient threshold.

5. A smart power distribution room sensor multi-dimensional data fusion anomaly event detection system, employing the smart power distribution room sensor multi-dimensional data fusion anomaly event detection method as described in any one of claims 1-4, characterized in that, include: The system comprises a data acquisition module, a data processing module, a modeling module, an online analysis module, an anomaly event database, and an evaluation module. The data acquisition module acquires multi-dimensional sensor data from the power distribution room. The data processing module processes the acquired multi-dimensional sensor data. The modeling module establishes and updates a multi-dimensional data correlation feature model. The online analysis module detects abnormal data points and determines the type of abnormal events online. The anomaly event database includes anomalies added during initial model construction and those added during model updates. The evaluation module assesses the accuracy of the multi-dimensional data correlation feature model and the accuracy of anomaly event pattern judgment.

6. The intelligent power distribution room sensor multi-dimensional data fusion abnormal event detection system according to claim 5, characterized in that, The data processing module performs the following data processing methods: classification, cleaning, transformation, time series alignment, standardization, and integration.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a program for detecting abnormal events by multi-dimensional data fusion of sensors in a smart power distribution room. When the program is executed by a processor, it implements the steps of a method for detecting abnormal events by multi-dimensional data fusion of sensors in a smart power distribution room as described in any one of claims 1 to 4.

Images