Substation inspection management system and inspection data classification method

Abstract

The substation inspection management system and the inspection data classification method comprise the following steps: accessing the inspection management system through a single user login mechanism, scanning a two-dimensional code of substation equipment to obtain equipment identification information, loading historical operation data and classification rules of the equipment from a pseudo database, and generating an inspection task identifier; acquiring a state image, an operation parameter and an environment feature of the equipment according to the task identifier, binding the task identifier and the equipment identifier, and adding a timestamp to generate a multi-modal calibration data set; extracting a time migration feature, a frequency domain anomaly feature and a load change feature based on historical data, and constructing a feature set; classifying the calibration data set according to a classification rule, and optimizing the classification rule to generate a preliminary classification result; carrying out consistency comparison on the preliminary classification result and historical data, and if the result is inconsistent, correcting the calibration data and carrying out re-classification; and finally, a classification result is stored in a pseudo database, a temporary cache is cleaned after a task is completed, abnormal data is reserved, and a classification rule is optimized in combination with user feedback.

Description

Technical Field

[0001] The present invention belongs to the technical field of intelligent inspection of power systems, and more specifically, relates to a substation inspection management system and an inspection data classification method. Background Art

[0002] During the operation and maintenance of substations, equipment inspection is an important part of ensuring the safe and stable operation of power systems. However, traditional inspection methods have significant limitations. At present, most inspection methods are mainly manual operations, and data collection is limited to a single form, such as operating parameters or equipment images. There is a lack of integration of multimodal data (including operating status, environmental characteristics and historical data), which makes it difficult for inspection results to fully reflect the true status of the equipment. In addition, the existing classification rules are usually predefined static logic, which cannot adapt to the dynamic changes of equipment operating status and environmental conditions. Especially when the operating mode deviates, the classification accuracy is difficult to guarantee.

[0003] Traditional inspection methods also have significant deficiencies in abnormal data processing and classification rule optimization. On the one hand, the analysis of abnormal data relies on manual intervention and lacks a systematic processing mechanism and automatic verification means, resulting in low inspection efficiency and data reliability; on the other hand, the classification rules are updated lagging and lack real-time performance, and cannot be dynamically optimized for equipment operating status and characteristic value changes. The limitations of such rules may lead to misjudgments or missed judgments. In addition, the management and data storage of different inspection tasks have not formed a unified association mechanism, and the data traceability capability of the inspection process is weak, further limiting the intelligence level and expansion capabilities of the inspection system. Summary of the invention

[0004] In order to solve the deficiencies in the prior art, the present invention provides a substation inspection management system and an inspection data classification method.

[0005] The present invention adopts the following technical solution.

[0006] A first aspect of the present invention provides a method for classifying substation inspection data, comprising the following steps:

[0007] Step 1: Scan the QR code of the substation equipment to obtain the current equipment identification information; load the historical inspection data and classification rules of the current equipment from the pseudo database to generate the inspection task identification;

[0008] Step 2: Collect the inspection data of the current equipment according to the inspection task identifier, the inspection data includes: equipment status image, operating parameters and environmental parameters; bind the collected inspection data with the inspection task identifier and equipment identification information, and add a timestamp to generate calibration data; construct a feature set based on the time offset feature quantity, frequency domain abnormal feature quantity and load change feature quantity between the calibration data and the historical inspection data;

[0009] Step 3: Combine the time offset feature quantity, frequency domain abnormal feature quantity and load change feature quantity in the feature set, adjust the feature weight and classification threshold range of the classification rule stored in the pseudo database, generate an optimized classification rule, and generate a preliminary classification result based on the optimized classification rule;

[0010] Step 4, compare the preliminary classification results with the classification rules and historical inspection data in the pseudo database for consistency; if the comparison results are inconsistent, correct the calibration data, regenerate the classification results, and store the classification results in the pseudo database.

[0011] Preferably, step 1 comprises:

[0012] Scan the QR code of the substation equipment, parse it to obtain the equipment number, functional area and location identification, and format the parsed equipment information into structured data to generate the equipment identification;

[0013] According to the parsed device number, basic information of the current device is collected in the pseudo database, including the type of operation parameters, sensor configuration and environmental feature requirements of the inspection;

[0014] Generate a unique inspection task ID based on the current timestamp, device number and task sequence number to mark the collection target, data source and processing flow of the current task;

[0015] According to the equipment identification and inspection task identification, the historical inspection data related to the equipment is loaded from the pseudo database, including the calibration data, classification rules and abnormal point marks collected in the last inspection task;

[0016] Combined with the classification rules and anomaly records in the historical inspection data, the inspection data that needs to be collected for the current task is clarified, and the collection requirements of the current inspection task are generated, including inspection data, sampling frequency and sensor configuration, as well as the target anomaly points and feature extraction requirements that need to be analyzed for this task. All inspection task collection requirements are integrated into the inspection task identifier.

[0017] Preferably, step 2 comprises:

[0018] According to the inspection task identification, the collection requirements are analyzed and collected in real time through sensors. The collected equipment status images, operating parameters and environmental parameters are calibrated using the timestamp and equipment identification in the inspection task identification, and the data format is unified and bound to the task identification and equipment identification.

[0019] The calibration data is processed according to the inspection task collection requirements in the inspection task identification, and the time offset feature quantity, frequency domain abnormal feature quantity and load change feature quantity are calculated and integrated to construct a feature set;

[0020] The feature set is structured based on a time series and is stored in layers according to time offset feature quantities, frequency domain anomaly feature quantities, and load change feature quantities.

[0021] Preferably, collecting inspection data includes collecting equipment status images, operating parameters and environmental parameters;

[0022] Collecting equipment status images includes using industrial cameras to collect appearance images of the equipment. The resolution and angle are set by the task collection requirements to detect cracks, rust and stains on the equipment surface.

[0023] Collecting operating parameters includes collecting the operating parameters of the equipment in real time through the equipment sensor interface, including voltage, current, temperature and frequency; the operating parameters are recorded by timestamp to generate time series data;

[0024] Collecting environmental parameters includes collecting temperature, humidity, vibration and air pressure data through on-site environmental sensors.

[0025] Preferably, the time offset feature is calculated as follows:

[0026]

[0027] Where Δt(i) represents the time offset feature at the current time point i; N represents the total number of historical inspection data related to the current device status; x r (o) represents the operating parameter value at the oth time point in the current calibration data; x h,j (i) represents the operating parameter value corresponding to the jth time point in the historical inspection data; w j Represents the time weight, which is dynamically adjusted based on the distance between the current time point and the historical time point; t i and t j Represents the timestamp of the current time point and the historical time point; ∈ represents the smoothing parameter;

[0028] The frequency domain abnormal feature quantity is calculated as follows:

[0029]

[0030] In the formula, S f Indicates the current frequency domain abnormal feature quantity; K indicates the total number of frequency points; A k Indicates the amplitude of frequency point k in the current calibration data; A normal,k Indicates the normal amplitude of frequency point k in the historical inspection data; E k Indicates the energy value of the current frequency point k; E j Indicates the energy value of the current task frequency;

[0031] The load change characteristic quantity is calculated as follows:

[0032]

[0033] Where, T load represents the current load change characteristic; n represents the number of time points used for sampling; t0 and t n Indicates the starting point and end point of the sampling time; R(t) indicates the load value at the current time point; R indicates the historical normal threshold value of load change; Indicates the rate of change of the load value.

[0034] Preferably, step 3 comprises:

[0035] Load the classification rules from the pseudo database, combine the generated feature set as input, and dynamically adjust the classification threshold and classification path according to the time offset feature quantity, frequency domain abnormal feature quantity and load change feature quantity in the feature set;

[0036] Dynamically adjust the comprehensive feature weight according to the abnormality of the feature quantity, and optimize the threshold range of the classification rule through feature distribution statistics;

[0037] Use a multi-layer classification algorithm to match the optimized classification rules layer by layer, generate preliminary classification results, and output classification labels and path records;

[0038] The dynamic time warping algorithm is used to compare the consistency of the classification path with the historical path.

[0039] Preferably, the comprehensive feature weight is calculated as follows:

[0040]

[0041] Where α, β, and γ represent the priority weights of the feature quantities.

[0042] Preferably, the classification threshold range is optimized as follows:

[0043] θ f =θ+σW f

[0044] In the formula, θ f represents the optimized classification threshold; θ represents the basic classification threshold; σ represents the adjustment coefficient, which is used to dynamically control the classification sensitivity.

[0045] Preferably, the data with inconsistent comparison results are retained, combined with the optimized classification rules in the pseudo database, including:

[0046] For the comparison results that are found to be inconsistent with the historical classification rules during the consistency check process, the abnormal data will be marked separately and stored in the abnormal data area of ​​the pseudo database. The storage content includes the reasons for the inconsistency of the classification results, the feature extraction anomalies and the task sequence number information of the classification rules; the reasons for the inconsistency of the classification results include feature value deviation and classification path matching failure.

[0047] A second aspect of the present invention provides a substation inspection management system, comprising: a user login module, a data acquisition module, a classification processing module, and a data storage and optimization module;

[0048] The user login module is used to access the inspection management system through a single-user login mechanism, scan the QR code of the substation equipment to obtain the current equipment identification information, and load the historical operation data and classification rules of the current equipment from the pseudo database temporary cache to generate the inspection task identification;

[0049] The data acquisition module is used to collect the operating data of the current equipment according to the inspection task identification, including the equipment status image, operating parameters and environmental characteristics, bind the collected operating data with the inspection task identification and the equipment identification, and add a timestamp to generate a multimodal calibration data set;

[0050] A classification processing module is used to extract time offset features, frequency domain abnormal features and load change features in the calibration data set based on the historical operation data in the pseudo database to construct a feature set, and classify the calibration data set according to the classification rules in the pseudo database, generate a preliminary classification result by optimizing the classification rules using the extracted feature set, and compare the preliminary classification result with the classification rules in the pseudo database and the historical operation data of the equipment for consistency. If the comparison results are inconsistent, the calibration data is corrected and the classification results are regenerated;

[0051] The data storage and optimization module is used to store the regenerated classification results and task identifiers in the pseudo database, clean up the temporary cache in the pseudo database after completing the inspection task, retain the data with inconsistent comparison results, and dynamically optimize the classification rules in the pseudo database based on user feedback.

[0052] Compared with the prior art, the beneficial effects of the present invention include at least: by constructing a multimodal calibration data set, dynamically optimizing classification rules and introducing a consistency check and feedback mechanism, the present invention not only realizes the comprehensive collection and binding of equipment operation status data, but also improves the accuracy and adaptability of classification through feature extraction and rule optimization, and solves the problems of abnormal data processing lag, untimely rule update and task data tracing difficulties in traditional inspection methods, thereby greatly improving the intelligent level and system reliability of substation inspection management, ensuring that equipment anomalies can be discovered and accurately classified in a timely manner, and providing strong support for the stable operation of the power system. . BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 It is a flow chart of a classification method provided according to an embodiment of the present invention. DETAILED DESCRIPTION

[0054] In order to make the purpose, technical scheme and advantages of the present invention clearer, the technical scheme of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. The embodiments described in this application are only part of the embodiments of the present invention, not all of them. Based on the spirit of the present invention, other embodiments obtained by ordinary technicians in this field without creative work are all within the scope of protection of the present invention.

[0055] Example 1 of the present invention provides a substation inspection data classification method, characterized in that it includes the following steps:

[0056] Access the inspection management system through a single-user login mechanism, scan the QR code of the substation equipment to obtain the current equipment identification information; load the historical operation data and classification rules of the current equipment from the pseudo database temporary cache, and generate the inspection task identification;

[0057] Preferably, identity authentication is completed through a single-user login mechanism of the inspection management system, using a username, password and biometrics to restrict one user to perform inspection tasks; if the login is successful, a unique session token is assigned;

[0058] If the login credentials are incorrect, an error message will be returned and you will be allowed to try again (you can set the maximum number of errors, such as 3).

[0059] If the current user attempts to log in again or has already logged in from another terminal, the system automatically rejects the new login request and prompts the user.

[0060] The user scans the QR code of the substation equipment to obtain the current equipment identification information. The QR code information includes the equipment number, functional area and location identification; if the QR code information is missing or the parsing fails, the user is prompted to rescan and an alternative manual input mode is provided to enter the equipment number;

[0061] Loading historical data and classification rules corresponding to the device identification from the temporary cache of the pseudo database, including historical operation data and classification rules; the historical operation data includes the operation status label of the device in the past and the time series record of the operation parameters;

[0062] A unique inspection task ID is generated by combining user identity information, device ID, timestamp and task sequence number.

[0063] Preferably, the data loading rules are as follows:

[0064] Prioritize loading records directly related to the current device, and organize data by the latest records first;

[0065] Data loading adopts asynchronous loading mode to perform multi-task inspection;

[0066] If there is no historical record of the current device in the pseudo database, the user will be prompted with "No historical data", and a default classification rule will be automatically generated, and the current device will be marked as "First inspection".

[0067] According to the task identifier, the operating data of the current equipment is collected, including the equipment status image, operating parameters and environmental characteristics; the collected operating data is bound to the inspection task identifier and the equipment identifier, and a timestamp is added to generate a multimodal calibration data set; based on the historical operating data in the pseudo database, the time offset features, frequency domain abnormal features and load change features in the calibration data set are extracted to construct a feature set;

[0068] Preferably, collecting inspection data includes collecting equipment status images, operating parameters and environmental parameters;

[0069] Collecting equipment status images includes using industrial cameras to collect appearance images of the equipment. The resolution and angle are set by the task collection requirements to detect cracks, rust and stains on the equipment surface.

[0070] Collecting operating parameters includes collecting the operating parameters of the equipment in real time through the equipment sensor interface, including voltage, current, temperature and frequency; the operating parameters are recorded by timestamp to generate time series data;

[0071] Collecting environmental parameters includes collecting temperature, humidity, vibration and air pressure data through on-site environmental sensors.

[0072] Preferably, the time offset feature is calculated as follows:

[0073]

[0074] Where Δt(i) represents the time offset feature at the current time point i; N represents the total number of historical inspection data related to the current device status; x r (i) represents the operating parameter value at the i-th time point in the current calibration data; x h,j (i) represents the operating parameter value corresponding to the jth time point in the historical inspection data; w j Represents the time weight, which is dynamically adjusted based on the distance between the current time point and the historical time point; t i and t j Represents the timestamp of the current time point and the historical time point; ∈ represents the smoothing parameter;

[0075] The frequency domain abnormal feature quantity is calculated as follows:

[0076]

[0077] In the formula, S f Indicates the current frequency domain abnormal feature quantity; K indicates the total number of frequency points; A k Indicates the amplitude of frequency point k in the current calibration data; A normal,k Indicates the normal amplitude of frequency point k in the historical inspection data; E k Indicates the energy value of the current frequency point k; E j Indicates the energy value of the current task frequency;

[0078] The load change characteristic quantity is calculated as follows:

[0079]

[0080] Where, T load represents the current load change characteristic; n represents the number of time points used for sampling; t0 and t n Indicates the starting point and end point of the sampling time; R(t) indicates the load value at the current time point; R indicates the historical normal threshold value of load change; Indicates the rate of change of the load value.

[0081] Preferably, the comprehensive feature weight is calculated as follows:

[0082]

[0083] Where α, β, and γ represent the priority weights of the feature quantities.

[0084] Preferably, the classification threshold range is optimized as follows:

[0085] θ f =θ+σW f

[0086] In the formula, θ f represents the optimized classification threshold; θ represents the basic classification threshold; σ represents the adjustment coefficient, which is used to dynamically control the classification sensitivity.

[0087] Specifically, data collection and operation parameter acquisition

[0088] After the inspection task is started, the system calls the acquisition module of the equipment according to the task identifier, including the image acquisition device and the operating parameter sensor, and connects to the external environmental feature acquisition device at the same time, and performs parallel acquisition operations for three types of data: equipment status, operating parameters and environmental features:

[0089] The acquisition of equipment status images includes connecting industrial cameras with the inspection management system to obtain equipment appearance status images in real time through wireless communication protocols (such as Wi-Fi or Bluetooth). Image acquisition parameters such as resolution and sampling rate are preset according to task requirements.

[0090] The collected image data is compressed in a standard format (such as JPEG, PNG) and then transmitted to the inspection system;

[0091] The image file contains key status information such as the pressure plate status and switch position of the specific equipment, which is used for subsequent feature extraction.

[0092] Operation parameter collection includes:

[0093] Use the device's built-in sensors to read the device's real-time operating parameters, including voltage, current, power, etc., through digital communication protocols (such as Modbus or OPC-UA);

[0094] The data collection frequency is set to once per second to ensure that the parameters can reflect the current status of the equipment and are associated through the task identifier;

[0095] If a sensor returns invalid data or the connection is lost, the system marks the data as “abnormal” and fills the gap with historical data or preset values.

[0096] Environmental feature collection includes:

[0097] External sensors are installed around the equipment to collect environmental parameters such as temperature, humidity, and vibration intensity;

[0098] The data is transmitted to the system through the edge gateway and synchronously bound with the operating parameters and image data to analyze the impact of the environment on the operating status of the equipment. Preferably, the historical operating data in the pseudo database is used to extract the time offset features, frequency domain abnormal features and load change features in the calibration data set to construct a feature set, including:

[0099] The sliding window method is used to analyze the deviation between the operating data at each time point in the calibration data set and the previous N time points, calculate the offset of the equipment operating parameters, identify the trend of equipment status changes over time, and compare the offset value with the standard of historical operating data to extract the time offset feature;

[0100] Fast Fourier transform is applied to the time domain signals in the calibration data set to transform the time series of the operating parameters into the frequency domain. The frequency exceeding the set threshold during equipment operation is set as the abnormal frequency. By analyzing the energy value of the abnormal frequency and comparing it with the normal frequency characteristics in the historical data, the frequency domain abnormal features are extracted to mark potential abnormal behaviors.

[0101] According to the voltage and current operating parameters in the calibration data set, the load fluctuation in each time period is analyzed through differential calculation, and the current load fluctuation amplitude is compared with the historical load curve of the equipment. The amplitude of the load change is marked as exceeding the set normal working range, and the load change characteristics are extracted.

[0102] Preferably, the calibration data set is classified according to the classification rules in the pseudo database, the classification rules are optimized using the extracted feature set, and a preliminary classification result is generated, including:

[0103] The multimodal feature values ​​in the calibration data set are gradually matched according to the classification rules stored in the pseudo database, and the time offset features, frequency domain anomaly features, and load change features are screened in turn according to the hierarchical structure of the classification rules to see whether they meet the classification conditions of the current node;

[0104] If all conditions are met, the corresponding classification labels are output, including "normal", "slightly abnormal" and "severely abnormal";

[0105] If any condition is not met, the next branch will be matched until the path matching of the classification logic is completed and a preliminary classification result is generated.

[0106] Preferably, the consistency comparison of the preliminary classification result with the classification rules in the pseudo database and the historical operation data of the equipment includes:

[0107] By calculating the matching degree of the rule nodes in the classification path, it is determined whether the classification logic is consistent with the historical rules;

[0108] The deviation ratio of the current feature value is calculated by combining the feature mean and standard deviation in the historical operation data. If the rule node matching degree of the classification path is lower than the set threshold or the deviation ratio exceeds the allowable range, the current classification result is marked as inconsistent and the data correction process is triggered.

[0109] Classify the calibration data set according to the classification rules in the pseudo database, optimize the classification rules using the extracted feature set, and generate a preliminary classification result; compare the preliminary classification result with the classification rules in the pseudo database and the historical operation data of the equipment for consistency; if the comparison results are inconsistent, correct the calibration data and regenerate the classification result;

[0110] The regenerated classification results and task identifiers are stored in the pseudo database. After the inspection task is completed, the temporary cache in the pseudo database is cleaned up, the data with inconsistent comparison results are retained, and the classification rules in the pseudo database are optimized based on user feedback.

[0111] Preferably, storing the regenerated classification results and task identifiers in the pseudo database includes:

[0112] After the regenerated classification results are completed, the classification results are bound to the inspection task identifier and the equipment identifier and stored in the pseudo database, including the classification label, feature set and task-related timestamp information, which are stored in the pseudo database asynchronously to completely save the classification results.

[0113] Preferably, retaining data with inconsistent comparison results and optimizing classification rules in the pseudo database in combination with user feedback includes:

[0114] For the comparison results that are found to be inconsistent with the historical classification rules during the consistency check process, the abnormal data are individually marked and stored in the abnormal data area of ​​the pseudo database, and the storage content includes the inconsistency reasons of the classification results, the feature extraction abnormal points and the task sequence number information of the classification rules; the inconsistency reasons of the classification results include feature value deviation and classification path matching failure;

[0115] Preferably, after completing the inspection task, clearing the temporary cache in the pseudo database includes:

[0116] When cleaning the pseudo database temporary cache, a task isolation strategy is used to ensure that the data cleaning scope of each task does not affect other tasks;

[0117] The isolation strategy limits the scope of operations by binding the task identifier, including cleaning up temporary data and intermediate processing results under the current task identifier, while retaining copies of data shared by multiple tasks, including common classification rules and environmental feature data;

[0118] During the cleanup process, the cleanup progress of the current task is dynamically monitored, and a cleanup task log is generated for tracing and troubleshooting.

[0119] Example 2 of the present invention provides a substation inspection management system, including: a user login module, a data acquisition module, a classification processing module, and a data storage and optimization module;

[0120] The user login module is used to access the inspection management system through a single-user login mechanism, scan the QR code of the substation equipment to obtain the current equipment identification information, and load the historical operation data and classification rules of the current equipment from the pseudo database temporary cache to generate the inspection task identification;

[0121] The data acquisition module is used to collect the operating data of the current equipment according to the inspection task identification, including the equipment status image, operating parameters and environmental characteristics, bind the collected operating data with the inspection task identification and the equipment identification, and add a timestamp to generate a multimodal calibration data set;

[0122] A classification processing module is used to extract time offset features, frequency domain abnormal features and load change features in the calibration data set based on the historical operation data in the pseudo database to construct a feature set, and classify the calibration data set according to the classification rules in the pseudo database, generate a preliminary classification result by optimizing the classification rules using the extracted feature set, and compare the preliminary classification result with the classification rules in the pseudo database and the historical operation data of the equipment for consistency. If the comparison results are inconsistent, the calibration data is corrected and the classification results are regenerated;

[0123] The data storage and optimization module is used to store the regenerated classification results and task identifiers in the pseudo database, clean up the temporary cache in the pseudo database after completing the inspection task, retain the data with inconsistent comparison results, and dynamically optimize the classification rules in the pseudo database based on user feedback.

[0124] Compared with the prior art, the beneficial effects of the present invention include at least: by constructing a multimodal calibration data set, dynamically optimizing classification rules and introducing a consistency check and feedback mechanism, the present invention not only realizes the comprehensive collection and binding of equipment operation status data, but also improves the accuracy and adaptability of classification through feature extraction and rule optimization, and solves the problems of abnormal data processing lag, untimely rule update and task data tracing difficulties in traditional inspection methods, thereby greatly improving the intelligent level and system reliability of substation inspection management, ensuring that equipment anomalies can be discovered and accurately classified in a timely manner, and providing strong support for the stable operation of the power system. .

[0125] The present disclosure may be a system, a method and / or a computer program product. The computer program product may include a computer-readable storage medium carrying computer-readable program instructions for causing a processor to implement various aspects of the present disclosure.

[0126] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents, and any modifications or equivalent replacements that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A substation inspection data classification method, characterized in that: The following steps are involved: Step 1: Scan the QR code of the substation equipment to obtain the current equipment identification information; load the historical inspection data and classification rules of the current equipment from the pseudo database to generate the inspection task identification; Step 2: Collect the inspection data of the current equipment according to the inspection task identifier, the inspection data includes: equipment status image, operating parameters and environmental parameters; bind the collected inspection data with the inspection task identifier and equipment identification information, and add a timestamp to generate calibration data; construct a feature set based on the time offset feature quantity, frequency domain abnormal feature quantity and load change feature quantity between the calibration data and the historical inspection data; Step 3: Combine the time offset feature quantity, frequency domain abnormal feature quantity and load change feature quantity in the feature set, adjust the feature weight and classification threshold range of the classification rule stored in the pseudo database, generate an optimized classification rule, and generate a preliminary classification result based on the optimized classification rule; Step 4, compare the preliminary classification results with the classification rules and historical inspection data in the pseudo database for consistency; if the comparison results are inconsistent, correct the calibration data, regenerate the classification results, and store the classification results in the pseudo database.

2. A substation inspection data classification method according to claim 1, characterized in that: Step 1 includes: Scan the QR code of the substation equipment, parse it to obtain the equipment number, functional area and location identification, and format the parsed equipment information into structured data to generate the equipment identification; According to the parsed device number, basic information of the current device is collected in the pseudo database, including the type of operation parameters, sensor configuration and environmental feature requirements of the inspection; Generate a unique inspection task ID based on the current timestamp, device number and task sequence number to mark the collection target, data source and processing flow of the current task; According to the equipment identification and inspection task identification, the historical inspection data related to the equipment is loaded from the pseudo database, including the calibration data, classification rules and abnormal point marks collected in the last inspection task; Combined with the classification rules and anomaly records in the historical inspection data, the inspection data that needs to be collected for the current task is clarified, and the collection requirements of the current inspection task are generated, including inspection data, sampling frequency and sensor configuration, as well as the target anomaly points and feature extraction requirements that need to be analyzed for this task. All inspection task collection requirements are integrated into the inspection task identifier.

3. A substation inspection data classification method according to claim 1, characterized in that: Step 2 includes: According to the inspection task identification, the collection requirements are analyzed and collected in real time through sensors. The collected equipment status images, operating parameters and environmental parameters are calibrated using the timestamp and equipment identification in the inspection task identification, and the data format is unified and bound to the task identification and equipment identification. The calibration data is processed according to the inspection task collection requirements in the inspection task identification, and the time offset feature quantity, frequency domain abnormal feature quantity and load change feature quantity are calculated and integrated to construct a feature set; The feature set is structured based on a time series and is stored in layers according to time offset feature quantities, frequency domain anomaly feature quantities, and load change feature quantities.

4. A substation inspection data classification method according to claim 3, characterized in that: Collecting inspection data includes collecting equipment status images, operating parameters and environmental parameters; Collecting equipment status images includes using industrial cameras to collect appearance images of the equipment. The resolution and angle are set by the task collection requirements to detect cracks, rust and stains on the equipment surface. Collecting operating parameters includes collecting the operating parameters of the equipment in real time through the equipment sensor interface, including voltage, current, temperature and frequency; the operating parameters are recorded by timestamp to generate time series data; Collecting environmental parameters includes collecting temperature, humidity, vibration and air pressure data through on-site environmental sensors.

5. A substation inspection data classification method according to claim 3, characterized in that: The time offset feature is calculated as follows: Where Δt(i) represents the time offset feature at the current time point i; N represents the total number of historical inspection data related to the current device status; x r (i) represents the operating parameter value at the i-th time point in the current calibration data; x h,j (i) represents the operating parameter value corresponding to the jth time point in the historical inspection data; w j Represents the time weight, which is dynamically adjusted based on the distance between the current time point and the historical time point; t i and t j Represents the timestamp of the current time point and the historical time point; ∈ represents the smoothing parameter; The frequency domain abnormal feature quantity is calculated as follows: In the formula, S f Indicates the current frequency domain abnormal feature quantity; K indicates the total number of frequency points; A k Indicates the amplitude of frequency point k in the current calibration data; A normal,k Indicates the normal amplitude of frequency point k in the historical inspection data; E k Indicates the energy value of the current frequency point k; E j Indicates the energy value of the current task frequency; The load change characteristic quantity is calculated as follows: Where, T load represents the current load change characteristic; n represents the number of time points used for sampling; t0 and t n Indicates the starting and ending points of the sampling time; R(t) indicates the load value at the current time point; R indicates the historical normal threshold value of load change; Indicates the rate of change of the load value.

6. A substation inspection data classification method according to claim 1, characterized in that: Step 3 includes: Load the classification rules from the pseudo database, combine the generated feature set as input, and dynamically adjust the classification threshold and classification path according to the time offset feature quantity, frequency domain abnormal feature quantity and load change feature quantity in the feature set; Dynamically adjust the comprehensive feature weight according to the abnormality of the feature quantity, and optimize the threshold range of the classification rule through feature distribution statistics; Use a multi-layer classification algorithm to match the optimized classification rules layer by layer, generate preliminary classification results, and output classification labels and path records; The dynamic time warping algorithm is used to compare the consistency of the classification path with the historical path.

7. A substation inspection data classification method according to claim 6, characterized in that: The comprehensive feature weight is calculated as follows: Where α, β, and γ represent the priority weights of the feature quantities.

8. A substation inspection data classification method according to claim 6, characterized in that: The classification threshold range is optimized as follows: i f =θ+σW f In the formula, θ f represents the optimized classification threshold; θ represents the basic classification threshold; σ represents the adjustment coefficient, which is used to dynamically control the classification sensitivity.

9. A substation inspection data classification method according to claim 1, characterized in that: The preliminary classification result is compared with the classification rules and historical inspection data in the pseudo database for consistency; if the comparison result is inconsistent, the calibration data is corrected, the classification result is regenerated, and the classification result is stored in the pseudo database, including: For the comparison results that are found to be inconsistent with the historical classification rules during the consistency check process, the abnormal data will be marked separately and stored in the abnormal data area of ​​the pseudo database. The storage content includes the reasons for the inconsistency of the classification results, the feature extraction anomalies and the task sequence number information of the classification rules; the reasons for the inconsistency of the classification results include feature value deviation and classification path matching failure.

10. A substation inspection management system, comprising: User login module, data collection module, classification processing module and data storage and optimization module; A substation inspection management system and data classification method thereof as described in any one of claims 1 to 9 is operated, characterized in that: The user login module is used to access the inspection management system through a single-user login mechanism, scan the QR code of the substation equipment to obtain the current equipment identification information, and load the historical operation data and classification rules of the current equipment from the pseudo database temporary cache to generate the inspection task identification; The data acquisition module is used to collect the operating data of the current equipment according to the inspection task identification, including the equipment status image, operating parameters and environmental characteristics, bind the collected operating data with the inspection task identification and the equipment identification, and add a timestamp to generate a multimodal calibration data set; A classification processing module is used to extract time offset features, frequency domain abnormal features and load change features in the calibration data set based on the historical operation data in the pseudo database to construct a feature set, and classify the calibration data set according to the classification rules in the pseudo database, generate a preliminary classification result by optimizing the classification rules using the extracted feature set, and compare the preliminary classification result with the classification rules in the pseudo database and the historical operation data of the equipment for consistency. If the comparison results are inconsistent, the calibration data is corrected and the classification results are regenerated; The data storage and optimization module is used to store the regenerated classification results and task identifiers in the pseudo database, clean up the temporary cache in the pseudo database after completing the inspection task, retain the data with inconsistent comparison results, and dynamically optimize the classification rules in the pseudo database based on user feedback.

Images