Substation inspection double-layer inspection data model based on multi-source data fusion and application thereof
Through the correlation algorithm of multi-source data fusion and power event-oriented, the traditional substation inspection system has solved the shortcomings in dynamic adaptability and fault diagnosis accuracy, achieved higher system stability and security, and reduced false alarms and The missed rate improves the adaptability and response speed of the substation intelligent patrol system.
Patent Information
- Application Number
- CN202410959043.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-17
- Publication Date
- 2024-11-12
AI Technical Summary
Traditional substation intelligent patrol systems are difficult to dynamically adapt to environmental changes and complex power events, resulting in false alarms or missed reports, and a single data source analysis cannot accurately diagnose complex faults. Traditional prediction maintenance methods may lead to premature or too late maintenance, increasing costs and risk.
Using a correlation algorithm based on multi-source data fusion and power event-oriented approach, through dynamic statistical analysis and multi-source data fusion technology, the monitoring threshold is adjusted in real time, equipment or system failures are identified, false alarms and missed reports are reduced, and system stability and safety are improved. sex.
It improves the dynamic adaptability and fault diagnosis accuracy of intelligent inspection data processing of substations, reduces false alarms and missed reports, reduces unnecessary maintenance costs and operational interference, quickly identify potential faults, and improves system stability and reliability.
Abstract
Description
Technical Field
[0001] The present invention relates to the field of intelligent inspection of substations, and in particular to a substation inspection data model based on multi-source data fusion and a power event-oriented association algorithm and its application. Background Art
[0002] In modern power systems, substations are key nodes for power transmission and distribution, and their stability and efficiency are crucial to the safe operation of the entire power grid. With the development of technology, a variety of monitoring devices are installed inside substations, such as temperature, voltage, current and vibration sensors, to monitor equipment status and environmental changes in real time. Traditional substation intelligent inspection data processing methods usually rely on setting static thresholds to determine whether the equipment status is normal, but this method cannot effectively adapt to environmental changes and complex power events, and is prone to false alarms or missed alarms.
[0003] In traditional substation inspection systems, a fixed threshold is set for the output of each sensor. For example, if the temperature of the transformer exceeds 100°C, the system will trigger an alarm. This method is simple and intuitive, but it cannot reflect the actual operating status of the equipment under different environmental and load conditions, which may lead to frequent false alarms or missed alarms. Many traditional systems rely on data from a single source for analysis and judgment. For example, they only judge whether the equipment is faulty based on the abnormal current, while ignoring data related to voltage fluctuations or temperature rise. This method ignores the potential correlation between data and may not be able to accurately diagnose complex power events. Some systems use simple trend analysis tools, such as linear regression analysis, to predict the future status of equipment. Although this method can predict some conventional equipment degradation trends, it has limited predictive capabilities for emergencies and nonlinear complex changes. Traditional predictive maintenance methods mainly rely on historical data and empirical formulas to set maintenance cycles, rather than based on the actual operating conditions of the equipment. This method may lead to maintenance too early or too late, increasing maintenance costs or risks.
[0004] Based on the above-mentioned prior art in this technical field, it can be seen that with the increasing complexity and automation of power systems, traditional substation inspection monitoring and maintenance methods have been unable to meet the real-time, accuracy and intelligence requirements of modern power systems. Therefore, it is particularly important to develop and apply a data model that can improve the dynamic adaptability of substation intelligent inspection data processing and the accuracy of fault diagnosis to improve the stability and safety of the system. Summary of the invention
[0005] The technical problem to be solved by the present invention is to provide a substation inspection data model and its application based on multi-source data fusion and power event-oriented correlation algorithm, in order to improve the dynamic adaptability of smart substation inspection data processing and the accuracy of fault diagnosis and apply it, thereby improving the stability and safety of the substation smart inspection system.
[0006] In order to solve the above technical problems, the technical solutions adopted by the present invention are as follows:
[0007] The invention discloses a substation inspection data processing method based on multi-source data fusion and power event-oriented correlation algorithm, which is characterized in that the data processing method collects and analyzes historical monitoring data, and dynamically adjusts the statistical range to adapt to environmental changes, and performs preliminary substation abnormal data screening based on the statistical comparison of current data and historical data indicators; corely, the initial screening abnormal data is further compared with the pre-built multi-source fusion sub-data model to determine whether the initial screening abnormal data is related to the power event; and according to the matching of the initial screening abnormal data with the known events, it is confirmed whether the initial screening abnormal data indicates equipment or system failure.
[0008] Further preferred, the data processing method improves the dynamic adaptability of data processing and the accuracy of fault diagnosis by introducing dynamic statistical analysis models and multi-source data fusion technology, adjusts the monitoring threshold in real time, comprehensively analyzes the relationship between various monitoring data, identifies subtle differences between normal operation and actual faults, reduces false alarms and missed alarms, and improves the stability and security of the system, providing strong technical support for the future development of substation intelligent inspection systems, so that substation intelligent inspection systems can better adapt to the rapidly changing operating environment and the growing automation needs.
[0009] Further preferably, the method comprises the following steps:
[0010] a. Establish dynamic statistical analysis model:
[0011] Data collection: Collect historical data from various monitoring devices in the substation;
[0012] Statistical analysis: Use statistical methods to analyze historical data and establish a dynamic statistical range for each type of monitoring data. The dynamic statistical range will be dynamically adjusted according to changes in historical data to adapt to changes in environmental and operating conditions;
[0013] Abnormal detection: Real-time monitoring data is compared with the dynamic statistical range. If the monitoring data exceeds the dynamic statistical range, it is marked as initial screening abnormal data;
[0014] b. Construction and application of multi-source fusion sub-data model:
[0015] Data annotation and model building: Annotate the monitoring data related to power events and build a multi-source fusion sub-data model that includes the correlation between different monitoring data, especially the data response pattern under specific power events;
[0016] Correlation analysis: comparing the initial screening abnormal data with the multi-source fusion sub-data model; analyzing whether the initial screening abnormal data is consistent with the data pattern associated with the power events marked in the multi-source fusion sub-data model;
[0017] Final anomaly determination: If the abnormal data pattern does not conform to any known data change pattern caused by a power event, it is confirmed that the abnormal data indicates a potential equipment or system failure; if the abnormal data is consistent with the data pattern of a power event, the abnormality is regarded as a normal event response and no fault alarm is issued.
[0018] The substation inspection data model based on multi-source data fusion and power event-oriented correlation algorithm is characterized by including: the data model includes the following data connotations: data collection and preprocessing, statistical analysis and dynamic range determination, anomaly detection, multi-source data fusion and correlation modeling, anomaly verification and final detection.
[0019] Further preferably, the data model specifically includes the following data connotations:
[0020] 1-1. Data collection and preprocessing:
[0021] The initial data set constructs a discrete data set containing multiple monitoring data points (sensors) collected separately. This data set has a two-layer structure. The first layer is the data set collected by the sensor at any substation data monitoring point according to its sampling frequency. This data set is naturally sorted in chronological order; the second layer is the secondary set of data sets collected corresponding to different monitoring points. This second layer of data sets is sorted according to the monitoring point or sensor equipment number or according to the artificial number;
[0022] Data preprocessing: Perform necessary data preprocessing algorithms according to the needs of subsequent data statistical processing; the important point is to set up redundant data space in the data structure to facilitate the expansion of the data model due to the increase of substation monitoring sites or sensor sites, thereby establishing the scalability of the data model;
[0023] 1-2. Statistical analysis:
[0024] For the bottom-level data in the initial data set, that is, the sampling data set sorted by time in the first-level structure under its two-level structure, the following statistics are constructed as the statistical data set: mean μ; standard deviation σ; first quartile Q1: the (n+1) / 4th data point after data sorting; third quartile Q3: the 3rd (n+1) / 4th data point after data sorting; interquartile range IQR: the difference between the third quartile and the first quartile, that is, IQR = Q3-Q1; other reference statistics: expanded and constructed as needed;
[0025] 1-3. Dynamic range determination:
[0026] The Normal Range is obtained based on the statistical processing of historical data. This Normal Range serves as a reference standard for whether the real-time sensor monitoring data of the substation is abnormal.
[0027] Specifically, the data construction process of Normal Range is determined according to the standard deviation method, quartile method or other mathematical methods. When the standard deviation method is used, the Normal Range is obtained. σ , when using the quartile method, we get NormalRange IQR In addition, for special cases, it can also be manually specified according to the sampling frequency of the sensor, the operating cycle of the equipment or other related factors. In this case, the Normal Range is obtained. g ;
[0028] 1-4. Anomaly Detection:
[0029] For the real-time sensor monitoring data of the substation, check whether it falls within the Normal Range obtained in the previous step; the Normal Range is selected from the Normal Range according to the specific situation. σ 、Normal Range IQR 、Normal Range g Or other selected Normal Range types; data falling outside the Normal Range is used as a preliminary screening abnormal data set;
[0030] 2-1. Multi-source data:
[0031] For k types of substation sensor monitoring data, its data representation is {x ij}; the i subscript is sorted in the order of acquisition of any sensor data type in the selected time period, and j represents different types of sensor data; such data subscript setting is consistent with the double-layer data structure of the initial data set constructed in "1-1. Data collection and preprocessing"; at the same time, for any known power event, the data representation is {Ei}; Here i is the number of types of known power events;
[0032] 2-2. Correlation Modeling:
[0033] Associated with any power event E i , will be in this power event E i When the occurrence of different substation sensor monitoring data {x ij} to characterize the change characteristics of data, and the optional data extension of the change characteristics includes but is not limited to: each x ij The range of change of each x ij The percentage change intensity, each x ij The correlation between the changes in ij The correlation between the changes is characterized by a cross-data correlation algorithm;
[0034] 2-3. Abnormal verification:
[0035] For the initial screening abnormal data set obtained in 1-4, import it into the correlation modeling data in 2-2; specifically:
[0036] If the correlation model is based on each x ij The range of change of each x ij The percentage change intensity of these two indicators is the data connotation, and the overall variation pattern of the initial screening abnormal data set is compared with the known power events pre-stored in the correlation modeling data under x ij Whether the variation pattern of the data corresponds to that of the data, thereby identifying whether the initial screening abnormal data comes from a specific power event or a real fault;
[0037] Core, if the correlation model each x ij The correlation between the changes between the two is the data connotation. For each abnormal data point, its correlation with other monitoring data is examined to check whether these abnormal data conform to the data correlation under the known non-fault power events. If so, it is considered that the anomaly is caused by the known power event rather than the fault. In this way, it is determined whether the initial screening abnormal data comes from a specific power event or a real fault.
[0038] The substation inspection data structure based on multi-source data fusion and power event-oriented association algorithm is characterized by:
[0039] a. Data collection and preprocessing;
[0040] b. Statistical analysis and dynamic range determination: construct the calculate_statistical_ranges function to calculate the mean, standard deviation, quartile and interquartile range of each monitoring data; determine the normal range: use the mean ± 3 times the standard deviation and interquartile range;
[0041] c. Anomaly detection: Construct the detect_anomalies function to compare the real-time monitoring data with the normal range and detect abnormal data;
[0042] d. Multi-source data fusion and correlation modeling: Based on the following data content: each x ij The build_association_model function is used to construct a correlation matrix based on known power events. The correlation matrix represents the correlation between different monitoring data under a specific power event type.
[0043] e. Anomaly verification and final detection: Construct the validate_anomalies_with_association_model function to verify whether the detected anomaly data conforms to the association model; construct the is_correlated function to check whether the anomaly data conforms to the association.
[0044] Further preferably, the database structure and operation structure corresponding to the above data structure are: historical monitoring data, real-time monitoring data and power event data are saved as historical_data.csv, current_data.csv and event_data.csv respectively; in operation, the main function main will sequentially execute data collection and preprocessing, statistical analysis and dynamic range determination, anomaly detection, multi-source data fusion and correlation modeling, as well as anomaly verification and final detection, and finally output the abnormal data verified by the double-layer data model.
[0045] The beneficial effects of adopting the above technical solution are:
[0046] The present invention constructs a brand-new two-layer data model, which monitors data through dynamic statistical analysis and automatically adjusts according to changes in environmental and operating conditions, thereby improving the flexibility and accuracy of data processing; in the data processing process of substation intelligent inspection, multi-source data fusion and correlation analysis help distinguish normal power event responses from real equipment or system failures, reduce false alarm rates, reduce invalid data storage and its negative interference effects on data models, and reduce unnecessary maintenance costs and operational interference; compared with the single-variable threshold model in the traditional substation inspection data processing process, the present invention constructs a brand-new two-layer data model that integrates the actual impact of power events on multi-source and multi-dimensional data of substations, which can quickly and accurately identify potential faults and anomalies, speed up response speed, and improve the stability and reliability of substation intelligent inspection systems. DETAILED DESCRIPTION
[0047] The following examples describe the present invention in detail. In the description of the following examples, specific details such as specific system structures, technologies, etc. are proposed for illustration rather than for limitation, so as to thoroughly understand the present application embodiments. However, it should be clear to those skilled in the art that the present application can also be implemented in other embodiments without these specific details. In other cases, the detailed description of well-known systems, devices, circuits and methods is omitted to prevent unnecessary details from hindering the description of the present application. It should be understood that when used in the present application specification and the attached claims, the term "comprising" indicates the presence of the described features, wholes, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or their collections. It should also be understood that the term "and / or" used in the present application specification and the attached claims refers to any combination of one or more of the items listed in association and all possible combinations, and includes these combinations.
[0048] In addition, in the description of the present application specification and the attached claims, the terms "first", "second", "third", etc. are only used to distinguish the description and cannot be understood as indicating or implying relative importance. The reference to "one embodiment" or "some embodiments" described in the present application specification means that one or more embodiments of the present application include specific features, structures or characteristics described in conjunction with the embodiment. Therefore, the statements "in one embodiment", "in some embodiments", "in some other embodiments", "in some other embodiments", etc. that appear in the different parts of this specification do not necessarily refer to the same embodiment, but mean "one or more but not all embodiments", unless otherwise specifically emphasized in other ways. The terms "including", "comprising", "having" and their variations all mean "including but not limited to", unless otherwise specifically emphasized in other ways.
[0049] First, the technical concept of the present invention is summarized as follows.
[0050] Step 1 (this step can be compatible with and linked to the existing substation data monitoring model), for the various existing data sensor monitoring parameters of the smart substation, collect and analyze their historical data, and then build an algorithm. This algorithm evaluates these historical data through statistical analysis and gives a dynamic statistically reasonable value range. Then, when the monitored data exceeds this range, it is considered to be abnormal data, thereby indicating that the substation is in a potential fault state;
[0051] Step 2, here we further improve the technical content of the above step 1. The reason for the improvement is that the technical idea given in step 1 is to compare the current data with the historical data, and consider it abnormal when it exceeds the range threshold; however, this algorithm has inherent defects. The reason is that when performing abnormal inspection on the substation monitoring data, for example, the following situations: season, weather, industrial operation daily cycle, busy farming, etc. will cause unconventional jumps in substation monitoring parameters, such as peak and valley electricity consumption generated over time, and electric vehicles are usually connected to the power grid at 10 o'clock in the evening on a large scale under peak and valley electricity prices, causing a surge in load, etc. In the initial substation inspection data model we constructed (as well as some existing substation monitoring data models), jumps in power monitoring data under these situations will trigger data alarms, but this is actually a false data alarm and should be avoided. The improvement method is: Based on the above improvement reasons, it can be considered probabilistically that it is impossible for two unrelated data to be stored abnormally at the same time, and the sudden increase, decrease or other data mutation caused by normal power events will inevitably cause multiple sets of related data to mutate at the same time. Therefore, the improved data model should integrate the actual impact of various known typical or atypical power events on the multi-source and multi-dimensional data of the substation, and pre-construct the related changes of various data under known typical or atypical power events, so as to form a new two-layer data model, and then compare the abnormal data with the multi-source fusion model. According to the matching of the abnormal data with the known events, it is confirmed whether the abnormal data indicates equipment or system failure, so as to avoid false data alarms such as false alarms and reduce the probability of unwarranted errors.
[0052] Example 1
[0053] The technical framework of the substation inspection data processing method planned by the inventor's technical team is as follows: Based on multi-source data fusion and power event-oriented correlation algorithm, historical monitoring data is collected and analyzed to dynamically adjust the statistical range to adapt to environmental changes; then, the abnormal data is compared with the multi-source fusion model to determine whether the abnormal data is related to the power event; according to the matching of the abnormal data with the known events, it is confirmed whether the abnormal data indicates equipment or system failure. The specific steps are as follows:
[0054] a. Establish dynamic statistical analysis model:
[0055] Data collection: Collect historical data from various monitoring devices in the substation;
[0056] Statistical analysis: Use statistical methods to analyze historical data and establish a dynamic statistical range for each type of monitoring data (as a reference range for substation monitoring data under normal conditions);
[0057] Anomaly detection: Real-time monitoring data is compared with the dynamic statistical range. If the monitoring data exceeds the dynamic statistical range, it is initially marked as abnormal data;
[0058] b. Construction and application of multi-source fusion sub-data model:
[0059] Data annotation and model building: Annotate the monitoring data related to power events and build a multi-source fusion sub-data model that includes the correlation between different monitoring data, especially the data response pattern under specific power events;
[0060] Correlation analysis: comparing the data initially marked as abnormal with the multi-source fusion sub-data model; analyzing whether the abnormal data is consistent with the data pattern associated with the power events marked in the multi-source fusion sub-data model;
[0061] Final anomaly determination: If the abnormal data pattern does not conform to any known data change pattern caused by a power event, it is confirmed that the abnormal data indicates a potential equipment or system failure; if the abnormal data is consistent with the data pattern of a power event, the abnormality is regarded as a normal event response and no fault alarm is issued.
[0062] It can be seen that the innovative technical solution constructed above improves the dynamic adaptability of data processing and the accuracy of fault diagnosis by introducing dynamic statistical analysis models and multi-source data fusion technology. It has the data analysis capability to adjust the monitoring threshold in real time. More importantly, the correlation we introduced in the data model and data process can comprehensively analyze the relationship between various monitoring data, identify the subtle differences between normal operation and actual faults, reduce false alarms and missed alarms, improve the stability and security of the system, and provide strong technical support for the future development of substation intelligent inspection systems, so that substation intelligent inspection systems can better adapt to the rapidly changing operating environment and the growing automation needs.
[0063] Example 2
[0064] Based on the substation inspection data processing method described in Example 1, a corresponding data model can be constructed, namely, a substation inspection data model (based on multi-source data fusion and power event-oriented association algorithm).
[0065] The main structure of the model includes: data collection and preprocessing, statistical analysis and dynamic range determination, anomaly detection, multi-source data fusion and correlation modeling, anomaly verification and final detection.
[0066] The process-oriented data architecture of the model is described below.
[0067] Step 1,
[0068] 1-1. Data collection and preprocessing: The initial data set is constructed as a time series data set containing multiple monitoring data points, where the sampling time points based on the substation monitoring sensor equipment contain the substation intelligent inspection monitoring data; and data preprocessing is performed as needed (common data cleaning, standardization and structuring and other data preprocessing methods are used as needed).
[0069] Specifically, the initial data set constructs a discrete data set containing n monitoring data points (sensors) collected by each of them. This data set has a two-layer structure. The first layer is the data set collected by the sensor of any substation data monitoring point according to its sampling frequency. This data set is naturally sorted in chronological order; the second layer is the secondary set of data sets collected corresponding to different monitoring points. This second layer of data sets is sorted according to the monitoring point or sensor equipment number or according to the artificial number;
[0070] For data preprocessing, in addition to selecting conventional data preprocessing algorithms according to the needs of subsequent data statistical processing, an important point is to set up redundant data space in the data structure to facilitate the subsequent expansion of the data model due to the increase of substation monitoring sites or sensor sites, thereby establishing the scalability of the data model.
[0071] 1-2. Statistical analysis: For each monitoring data x, calculate its statistics: Mean (Mean): μ; Standard Deviation (Standard Deviation): σ; First Quartile (First Quartile, Q1): the (n+1) / 4th data point after data sorting; Third Quartile (Third Quartile, Q3): the 3rd (n+1) / 4th data point after data sorting; Interquartile Range (Interquartile Range, IQR): the difference between the third quartile and the first quartile, that is, IQR = Q3-Q1.
[0072] 1-3. Dynamic Range Determination: Using Standard Deviation Method: Normal Range σ =[μ-3σ, μ+3σ]; using quartile method: Normal Range IQR =[Q1-1.5·IQR, Q3+1.5·IQR]; In addition, for special cases, it can also be manually specified according to the sampling frequency of the sensor, the operating cycle of the equipment or other related factors. In this case, the NormalRange is obtained. g :
[0073] 1-4. Anomaly Detection:
[0074] For the real-time sensor monitoring data of the substation, check whether it falls within the Normal Range obtained in the previous step; the Normal Range is selected from the Normal Range according to the specific situation. σ 、Normal Range IQR 、Normal Range g Or other selected Normal Range types; data falling outside the Normal Range is used as the initial screening abnormal data set {y i};
[0075] Step 2,
[0076] 2-1. Multi-source data:
[0077] For k types of substation sensor monitoring data, its data representation is {x ij}; where the i subscript is sorted in the order of acquisition of any sensor data type in the selected time period, and j represents different types of sensor data, with a maximum value of k; at the same time, for any known power event, the data representation is {E i}; Here i is the number of types of known power events;
[0078] 2-2. Correlation Modeling:
[0079] Associated with any power event E i , will be in this power event E i When the occurrence of different substation sensor monitoring data {x ij} to characterize the change characteristics of data, the optional data edge of the change characteristics includes but is not limited to: each x ij的 Variation range, each x ij The percentage change intensity, each x ij The correlation between the changes in ij The correlation between the changes is characterized by a cross-data correlation algorithm;
[0080] Specifically, for the data content: each x ij The variation range and each x ij The percentage change intensity, generalized data operation is to each power event E i Next x ij The variation range and each x ij The percentage change intensity is pre-stored and constructed (to facilitate the comparison and verification of the initial screening abnormal data during the secondary abnormality test); for the data connotation: each x ij The correlation between the changes in ijThe correlation between the changes is represented by a cross-data correlation algorithm, for example, using a correlation matrix;
[0081] 2-3. Abnormal verification:
[0082] For the initial screening abnormal data set {y i}, import it into the association modeling data in 2-2; specifically:
[0083] If the correlation model is based on each x ij The range of change of each x ij The percentage change intensity of these two indicators is the data connotation, then compare the initial screening abnormal data set {y i} Overall variation pattern and correlation modeling data pre-stored in the known power events x ij Whether the variation pattern of the data corresponds to that of the data, thereby identifying whether the initial screening abnormal data comes from a specific power event or a real fault;
[0084] Core, if the correlation model each x ij The correlation between the changes is the data connotation, then for each abnormal data point y i , examine its correlation with other monitoring data, and check whether these abnormal data conform to the data correlation under known non-fault power events. If so, it is considered that the anomaly is caused by a known power event rather than a fault, thereby identifying whether the initial screening abnormal data comes from a specific power event or a real fault.
[0085] Example 3
[0086] For the software platform implementation of the above algorithms and data models, we further constructed the corresponding data structure in computer language, including the following contents:
[0087] (Here, in order to facilitate the non-computer professionals in the technical team to read, the description of the data calculation structure is combined with the data model and data process mentioned above.)
[0088] 3-1: Dominant Data Structure
[0089] a. Data collection and preprocessing;
[0090] b. Statistical analysis and dynamic range determination: construct the calculate_statistical_ranges function to calculate the mean, standard deviation, quartile and interquartile range of each monitoring data; determine the normal range: use the mean ± 3 times the standard deviation and interquartile range;
[0091] c. Anomaly detection: Construct the detect_anomalies function to compare the real-time monitoring data with the normal range and detect abnormal data;
[0092] d. Multi-source data fusion and association modeling: Construct the build_association_model function to construct an association matrix based on known power events; the association matrix represents the correlation between different monitoring data under a specific power event type;
[0093] e. Anomaly verification and final detection: Construct the validate_anomalies_with_association_model function to verify whether the detected anomaly data conforms to the association model; construct the is_correlated function to check whether the anomaly data conforms to the association;
[0094] f. Database structure and operation instructions corresponding to the above computer process: save historical monitoring data, real-time monitoring data and power event data as historical_data.csv, current_data.csv and event_data.csv respectively; in operation, the main function main will sequentially perform data collection and preprocessing, statistical analysis and dynamic range determination, anomaly detection, multi-source data fusion and correlation modeling, as well as anomaly verification and final detection. Finally, the verified anomaly data is output.
[0095] 3-2. The data structure is expanded and constructed as follows.
[0096] a. Initial data tool import: link and import and use corresponding data tools according to the specific data environment to process and analyze data;
[0097] b. Data collection and preprocessing: The initial data set is constructed as a collection of data collected by multiple sensor monitoring sites of the substation (a two-layer data structure is constructed according to different sensor monitoring sites and different collection time points);
[0098] Data preprocessing: Choose whether to perform data preprocessing according to the actual situation, and select the corresponding data preprocessing tool; set up redundant data space in the data structure to establish the extensibility of the data model, such as setting up redundant data space (redundant data bits, blank data bits, etc.) in the data structure, and then use it to add detection parameters as needed in the future, so as to achieve the expansion of the original data model at the basic parameter object level;
[0099] c. Statistical range analysis:
[0100] Construct the following data structure:
[0101] data.mean(): Calculate the mean of the data;
[0102] data.std(): Calculate the standard deviation of the data;
[0103] data.quantile(0.25): Calculate the first quartile (Q1) of the data;
[0104] data.quantile(0.75): Calculate the third quartile (Q3) of the data;
[0105] iqr = q3 - q1: Calculate the interquartile range (IQR);
[0106] normal_range_std: Calculate the normal range (mean ± 3 standard deviations) using the mean and standard deviation;
[0107] normal_range_iqr: Calculate the normal range (from Q1 - 1.5IQR to Q3 + 1.5IQR) using quartiles;
[0108] d. Detect anomalies:
[0109] Construct the following data structure:
[0110] normal_ranges: The normal ranges of each monitored data;
[0111] data[column] < normal_range_std[0]: Detect whether the data is below the lower limit of the normal range;
[0112] data[column] > normal_range_std[1]: Detect whether the data is above the upper limit of the normal range;
[0113] anomalies[column]: Save the anomaly data for each column;
[0114] e. Construct a correlation model (under the connotation of the correlation data of the changes between each x ij ):
[0115] Construct the following data structure:
[0116] event_data['event_type'].unique(): Obtain the unique power event types;
[0117] event_data[event_data['event_type'] == event]: Filter out the subset of the detected data correlation change data under the power event type;
[0118] event_subset.corr(): Calculate the correlation coefficient matrix of the data subset;
[0119] association_model[event]: saves the correlation coefficient matrix of each event type;
[0120] f. Verification exception:
[0121] Construct the following data structure:
[0122] anomalies.items(): traverse the abnormal data of each column;
[0123] anomaly_data.iterrows(): traverse the abnormal data row by row;
[0124] event_data.loc[index,'event_type']: Get the event type corresponding to the abnormal data;
[0125] association_model[event_type]: Get the power event type correlation coefficient matrix;
[0126] is_correlated(row,correlation_matrix): Checks whether the abnormal data is correlated.
[0127] Example 4
[0128] Common monitoring data in smart substation inspections include but are not limited to (these monitoring data are collected in real time through sensors and monitoring equipment, and analyzed and processed by the monitoring system of the smart substation to ensure the safe, stable and efficient operation of the substation):
[0129] (1) Electrical parameters:
[0130] Voltage
[0131] Current
[0132] Power
[0133] Power Factor
[0134] Frequency
[0135] EnergyConsumption
[0136] Harmonics
[0137] (2) Environmental parameters:
[0138] Temperature
[0139] Humidity
[0140] Atmospheric Pressure
[0141] Wind Speed and Direction
[0142] Rainfall
[0143] (3) Equipment status parameters:
[0144] Transformer Oil Temperature
[0145] Transformer Oil Level
[0146] Circuit Breaker Status
[0147] Isolator Status
[0148] Cable Temperature
[0149] Equipment Vibration
[0150] (4) Safety parameters:
[0151] SF6 Gas Pressure
[0152] SF6 Gas Density
[0153] Fire Alarm
[0154] Smoke Density
[0155] Security Door Status
[0156] (5) Communication and network parameters:
[0157] Communication Link Status
[0158] Data Transmission Rate
[0159] Network Latency
[0160] Packet Loss Rate
[0161] (6) Auxiliary parameters:
[0162] Battery Voltage
[0163] Battery Capacity
[0164] UPS Status
[0165] Equipment Running Time
[0166] (7) Other parameters: New monitoring parameters may be added as needed after certain devices or sensors are added (that is, other monitoring parameters may be added as needed later, for example, some monitoring parameters may be added after certain devices or sensors are added).
[0167] A related point is that, at the data structure level, attention should be paid to setting up redundant data space in the data structure, such as redundant data bits, blank data bits, etc., which can then be used to add new detection parameters as needed, thereby realizing the expansion of the original data model at the basic parameter object level.
[0168] Example 5: Electric vehicle centralized charging event
[0169] Taking the centralized charging event of electric vehicles as an example, a simplified data model and data process instance of intelligent inspection of substations are constructed. This data instance is based on the power event of centralized charging of electric vehicles. Under peak and valley electricity prices, electric vehicles are usually connected to the power grid on a large scale at 10 pm, causing a surge in load. This will cause many power detection parameters to exceed the dynamic range threshold in the "first step, i.e., statistical analysis based on historical data". Then, in the "second step, i.e., improved algorithm based on multi-source data fusion", since the multi-source fusion sub-data model has pre-annotated and constructed the correlation variability of substation operating parameters under the power event of centralized charging of electric vehicles, the abnormal data falsely reported in step one can be eliminated through correlation comparison, thereby improving the accuracy and efficiency of intelligent inspection of substations, and avoiding subsequent data processing and manual verification operations caused by false data reports. Specific data process examples are as follows:
[0170] Assume that the following simplified monitoring data is used: voltage (Voltage); current (Current); power (Power); transformer oil temperature (Transformer Oil Temperature). The corresponding data instance set is constructed as follows:
[0171] Generates and constructs datasets: historical_data, current_data, and event_data.
[0172] **historical_data**: historical data, containing x records, simulating the historical operation of the power system.
[0173] Data structure:
[0174] Voltage: voltage, mean 220, standard deviation 5.
[0175] Current: current, mean 10, standard deviation 2.
[0176] Power: power, mean 2200, standard deviation 100.
[0177] Transformer_Oil_Temperature: Transformer oil temperature, mean 60, standard deviation 5.
[0178] historical_data=pd.DataFrame({'Voltage':np.random.normal(220,5,1000),'Current':np.random.normal(10,2,10 00),'Power':np.random.normal(2200,100,1000),'Transformer_Oil_Temperature':np.random.normal(60,5,1000)}).
[0179] **current_data**: Current data, including y records, simulating the current operation of the power system.
[0180] current_data=pd.DataFrame({'Voltage':np.random.normal(220,5,100),'Current':np.random.normal(10,2,100 ),'Power':np.random.normal(2200,100,100),'Transformer_Oil_Temperature':np.random.normal(60,5,100)}).
[0181] **event_data**: event data, containing z records, simulating the operation of the power system under a specific event (here specified as the centralized charging event of electric vehicles at 10 pm).
[0182] Data structure:
[0183] event_type: event type, all of which are EV_Charging (electric vehicle charging).
[0184] Voltage: voltage, mean 210, standard deviation 5.
[0185] Current: current, mean 15, standard deviation 2.
[0186] Power: power, mean 3150, standard deviation 100.
[0187] Transformer_Oil_Temperature: Transformer oil temperature, mean 65, standard deviation 5.
[0188] event_data=pd.DataFrame({'event_type':['EV_Charging']*100,'Voltage':np.random.normal(210,5,100),'Current':np.random .normal(15,2,100),'Power':np.random.normal(3150,100,100),'Transformer_Oil_Temperature':np.random.normal(65,5,100)}).
[0189] On this basis, the data processing process constructed above is introduced in turn for data processing, including: data preprocessing, historical_data = preprocess_data (historical_data); current_data = preprocess_data (current_data); calculate the range based on the standard deviation and interquartile range (IQR), normal_ranges = {}; for column in historical_data.columns:normal_ranges[column]=calculate_statistical_ranges(historical_data[column]);Detect anomalies, anomalies=detect_anomalies(current_data,normal_ranges);Build an association model based on the 10 pm electric vehicle centralized charging event dataassociation_model=build_association_model(event_data);Use the association model to verify the detected anomalies and check whether these anomalies meet the expectations of the association modelvalidated_anomalies=validate_anomalies_with_association_model(anomalies,association_model,event_data);Until the final substation anomaly data with high authenticity and credibility is output, print("Validated Anomalies:",validated_anomalies).
[0190] The present invention analyzes monitoring data through dynamic statistics and automatically adjusts according to changes in environmental and operating conditions, thereby improving the flexibility and accuracy of data processing; multi-source data fusion and correlation analysis help distinguish normal power event responses from real equipment or system failures, reduce false alarm rates, reduce invalid data storage and its negative interference effects on data models, and reduce unnecessary maintenance costs and operational interference; quickly and accurately identify potential faults and anomalies, speed up response, and improve the stability and reliability of the substation intelligent inspection system.
[0191] Finally, it should be noted that the above is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein with equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the scope of protection of the claims of the present invention.
Claims
1. A substation inspection data processing method based on multi-source data fusion and power event-oriented correlation algorithm is characterized in that: This data processing method collects and analyzes historical monitoring data, and dynamically adjusts the statistical range to adapt to environmental changes. It performs preliminary screening of substation abnormal data based on the statistical comparison of current data and historical data indicators. Core, it further compares the initial screening abnormal data with the pre-built multi-source fusion sub-data model to determine whether the initial screening abnormal data is related to power events; based on the matching of the initial screening abnormal data with known events, it is confirmed whether the initial screening abnormal data indicates equipment or system failure.
2. The substation inspection data processing method based on multi-source data fusion and power event-oriented correlation algorithm according to claim 1 is characterized in that: This data processing method introduces dynamic statistical analysis models and multi-source data fusion technology to improve the dynamic adaptability of data processing and the accuracy of fault diagnosis, adjust the monitoring threshold in real time, comprehensively analyze the relationship between various monitoring data, identify the subtle differences between normal operation and actual faults, reduce false alarms and missed alarms, and improve the stability and security of the system. It provides strong technical support for the future development of substation intelligent inspection systems, enabling substation intelligent inspection systems to better adapt to the rapidly changing operating environment and the growing automation needs.
3. The substation inspection data processing method based on multi-source data fusion and power event-oriented correlation algorithm according to claim 1 is characterized in that: The steps include: a. Establish dynamic statistical analysis model: Data collection: Collect historical data from various monitoring devices in the substation; Statistical analysis: Use statistical methods to analyze historical data and establish a dynamic statistical range for each type of monitoring data. The dynamic statistical range will be dynamically adjusted according to changes in historical data to adapt to changes in environmental and operating conditions; Abnormal detection: Real-time monitoring data is compared with the dynamic statistical range. If the monitoring data exceeds the dynamic statistical range, it is marked as initial screening abnormal data; b. Construction and application of multi-source fusion sub-data model: Data annotation and model building: Annotate the monitoring data related to power events and build a multi-source fusion sub-data model that includes the correlation between different monitoring data, especially the data response pattern under specific power events; Correlation analysis: compare the initial screening abnormal data with the multi-source fusion sub-data model; analyzing whether the primary screening abnormal data is consistent with the data pattern associated with the power events marked in the multi-source fusion sub-data model; Final anomaly determination: If the abnormal data pattern does not conform to any known data change pattern caused by a power event, it is confirmed that the abnormal data indicates a potential equipment or system failure; if the abnormal data is consistent with the data pattern of a power event, the abnormality is regarded as a normal event response and no fault alarm is issued.
4. The substation inspection data model based on multi-source data fusion and power event-oriented association algorithm is characterized by: include: The data model includes the following data contents: data collection and preprocessing, statistical analysis and dynamic range determination, anomaly detection, multi-source data fusion and correlation modeling, anomaly verification and final detection.
5. The substation inspection data model based on multi-source data fusion and power event-oriented association algorithm according to claim 4 is characterized in that: The data model specifically includes the following data content: 1-1. Data collection and preprocessing: The initial data set constructs a discrete data set containing multiple monitoring data points (sensors) collected separately. This data set has a two-layer structure. The first layer is the data set collected by the sensor at any substation data monitoring point according to its sampling frequency. This data set is naturally sorted in chronological order; the second layer is the secondary set of data sets collected corresponding to different monitoring points. This second layer of data sets is sorted according to the monitoring point or sensor equipment number or according to the artificial number; Data preprocessing: Perform necessary data preprocessing algorithms according to the needs of subsequent data statistical processing; the important point is to set up redundant data space in the data structure to facilitate the expansion of the data model due to the increase of substation monitoring sites or sensor sites, thereby establishing the scalability of the data model; 1-2. Statistical analysis: For the bottom-level data in the initial data set, that is, the sampling data set sorted by time in the first-level structure under its two-level structure, the following statistics are constructed as the statistical data set: mean μ; standard deviation σ; first quartile Q1: the (n+1) / 4th data point after data sorting; third quartile Q3: the 3rd (n+1) / 4th data point after data sorting; interquartile range IQR: the difference between the third quartile and the first quartile, that is, IQR = Q3-Q1; other reference statistics: expanded and constructed as needed; 1-3. Dynamic range determination: The NormalRange is obtained based on the statistical processing of historical data, and this NormalRange is used as a reference standard for whether the real-time sensor monitoring data of the substation is abnormal; Specifically, the data construction process of NormalRange is determined according to the standard deviation method, quartile method or other mathematical methods. When the standard deviation method is used, NormalRange is obtained. σ , when using the quartile method, we get the Normal Range IQR In addition, for special cases, it can also be manually specified according to the sampling frequency of the sensor, the operating cycle of the equipment or other related factors. In this case, the NormalRange is obtained. g ; 1-4. Anomaly Detection: For the real-time sensor monitoring data of the substation, check whether it falls within the NormalRange obtained in the previous step; the NormalRange is selected from the NormalRange according to the specific situation. σ 、NormalRange IQR 、NormalRange g Or other selected NormalRange types; data falling outside the NormalRange is used as a preliminary screening abnormal data set; 2-1. Multi-source data: For k types of substation sensor monitoring data, its data representation is {x ij }; the i subscripts are arranged in the order of acquisition of any sensor data type in the selected time period, and j represents different types of sensor data; such data subscript settings are consistent with the double-layer data structure of the initial data set constructed in "1-1. Data collection and preprocessing"; at the same time, for any known power event, the data representation is {E i }; Here i is the number of types of known power events; 2-2. Correlation Modeling: Associated with any power event E i , will be in this power event E i When the occurrence of different substation sensor monitoring data {x ij } to characterize the change characteristics of data, and the optional data extension of the change characteristics includes but is not limited to: each x ij The range of change of each x ij The percentage change intensity, each x ij The correlation between the changes in ij The correlation between the changes is characterized by a cross-data correlation algorithm; 2-3. Abnormal verification: For the initial screening abnormal data set obtained in 1-4, import it into the correlation modeling data in 2-2; specifically: If the correlation model is based on each x ij The range of change of each x ij The percentage change intensity of these two indicators is the data connotation, and the overall variation pattern of the initial screening abnormal data set is compared with the known power events pre-stored in the correlation modeling data under x ij Whether the variation pattern of the data corresponds to that of the data, thereby identifying whether the initial screening abnormal data comes from a specific power event or a real fault; Core, if the correlation modeling each x ij The correlation between the changes between the two is the data connotation. For each abnormal data point, its correlation with other monitoring data is examined to check whether these abnormal data conform to the data correlation under the known non-fault power events. If so, it is considered that the anomaly is caused by the known power event rather than the fault. In this way, it is determined whether the initial screening abnormal data comes from a specific power event or a real fault.
6. The substation inspection data structure based on multi-source data fusion and power event-oriented association algorithm is characterized by: a. Data collection and preprocessing; b. Statistical analysis and dynamic range determination: construct the calculate_statistical_ranges function to calculate the mean, standard deviation, quartile and interquartile range of each monitoring data; determine the normal range: use the mean ± 3 times the standard deviation and interquartile range; c. Anomaly detection: Construct the detect_anomalies function to compare the real-time monitoring data with the normal range and detect abnormal data; d. Multi-source data fusion and correlation modeling: Based on the following data content: each x ij The build_association_model function is used to construct a correlation matrix based on known power events. The correlation matrix represents the correlation between different monitoring data under a specific power event type. e. Anomaly verification and final detection: Construct the validate_anomalies_with_association_model function to verify whether the detected anomaly data conforms to the association model; construct the is_correlated function to check whether the anomaly data conforms to the association.
7. The substation inspection data structure based on multi-source data fusion and power event-oriented association algorithm according to claim 6 is characterized by: The database structure and operation structure corresponding to the above data structure are as follows: historical monitoring data, real-time monitoring data and power event data are saved as historical_data.csv, current_data.csv and event_data.csv respectively; in operation, the main function main will sequentially execute data collection and preprocessing, statistical analysis and dynamic range determination, anomaly detection, multi-source data fusion and correlation modeling, as well as anomaly verification and final detection, and finally output the abnormal data verified by the double-layer data model.
Citation Information
Cited By
Monitoring and early warning system for power network communication equipment
CN119253864A
A monitoring and early warning system for power network communication equipment
CN119253864B
Intelligent monitoring method and system for air environment monitoring station
CN119397458A
Property security inspection method and system and storage medium
CN120071546A
Distribution network load transaction early warning method and device
CN120377269A
