Intelligent fault diagnosis and rapid isolation method for power distribution protection cabinet

Abstract

The invention relates to the technical field of power distribution protection cabinet fault diagnosis, and discloses an intelligent fault diagnosis and rapid isolation method for a power distribution protection cabinet, and the method comprises the steps: synchronously collecting electrical parameters, environment states and equipment motion signals in the cabinet as multi-source monitoring data when the power distribution protection cabinet operates; parameter fluctuation, state change and action time sequence features are extracted after preprocessing, cross-data association analysis is carried out through a multi-dimensional association model to generate an association feature set, a fault diagnosis information set is formed based on the identified abnormal types, fault positions and development trends, an isolation strategy is formulated according to severity levels and influence ranges, and a fault diagnosis result is obtained. And generating a rapid isolation instruction and dynamically adjusting along with fault evolution. According to the method, multi-source data fusion analysis is realized, faults can be accurately diagnosed and rapidly isolated, the operation safety and reliability of the power distribution protection cabinet are improved, and the method is suitable for fault processing of the power distribution protection cabinet in a power system.

Description

Technical Field

[0001] This invention relates to the field of fault diagnosis technology for power distribution protection cabinets, specifically to an intelligent fault diagnosis and rapid isolation method for power distribution protection cabinets. Background Technology

[0002] In power systems, distribution protection cabinets are key equipment for power distribution and protection, and their stable operation is crucial to the reliability of the entire power system. With the continuous development of power systems and the improvement of their intelligence level, higher requirements are placed on the fault diagnosis and isolation of distribution protection cabinets.

[0003] Traditional fault diagnosis methods for power distribution protection cabinets have many shortcomings. On the one hand, traditional methods often only analyze a single type of data, such as focusing only on changes in electrical parameters, while ignoring the combined influence of multi-source data such as environmental conditions and equipment action signals. This single-data analysis approach cannot comprehensively and accurately reflect the actual operating status of the protection cabinet, easily leading to omissions or misjudgments in fault diagnosis.

[0004] Traditional fault diagnosis and isolation processes are relatively slow. Once a fault occurs, manual investigation and handling are required, which is not only time-consuming but may also lead to the expansion of the fault's scope, posing a serious threat to the safe operation of the power system. Furthermore, traditional methods lack the ability to effectively predict the development trend of faults, cannot detect potential fault hazards in advance, and are difficult to implement preventative maintenance for protection cabinets.

[0005] With the development of smart grids, distribution protection cabinets face more complex operating environments and higher reliability requirements. Fluctuations in internal electrical parameters, changes in ambient temperature and humidity, and the frequency of equipment operation can all affect the normal operation of the protection cabinet. Therefore, there is an urgent need for a method that can perform real-time, accurate fault diagnosis and rapid isolation of distribution protection cabinets to improve the safety and reliability of power systems.

[0006] While some existing intelligent fault diagnosis methods have improved diagnostic accuracy to some extent, they still have shortcomings in areas such as multi-source data fusion processing, rapid fault location, and dynamic adjustment of isolation strategies. For example, in the process of multi-source data fusion, the lack of effective time alignment mechanisms and correlation analysis models makes it difficult to accurately identify causal relationships between data; in terms of fault location, the accuracy and speed of location need to be improved; and in terms of isolation strategy formulation, the lack of comprehensive consideration of the severity level and scope of impact of faults makes it difficult to achieve rapid and effective isolation. Summary of the Invention

[0007] The purpose of this invention is to provide an intelligent fault diagnosis and rapid isolation method for power distribution protection cabinets, so as to solve the problems mentioned in the background art.

[0008] To achieve the above objectives, the present invention provides an intelligent fault diagnosis and rapid isolation method for power distribution protection cabinets, the method comprising: During the operation of the power distribution protection cabinet, electrical parameters, environmental conditions and equipment action signals inside the cabinet are collected simultaneously as multi-source monitoring data. Real-time preprocessing is performed on the collected multi-source monitoring data, including calibration of electrical parameters to measure deviations, filtering of environmental interference and noise, and marking of time nodes for equipment action signals. Feature extraction is performed on the preprocessed electrical parameters, environmental conditions, and equipment action signals to form parameter fluctuation features, state change features, and action timing features. The parameter fluctuation characteristics, state change characteristics, and action time sequence characteristics are analyzed across data through a multi-dimensional correlation model to generate a set of correlation features containing causal relationships. Based on the set of associated features, the abnormality type, fault location and development trend of the protection cabinet are identified, and a fault diagnosis information set is formed. Isolation strategies are formulated based on the severity level and impact range of the fault diagnosis information set, generating rapid isolation instructions for specific fault points, and dynamically adjusting the content of the instructions as the fault evolves.

[0009] Preferably, the collected multi-source monitoring data is preprocessed in real time. The specific steps are as follows: Electrical parameters are corrected for sensor measurement errors through a dimension conversion module, and the acquisition timestamp of each parameter is recorded synchronously. The environmental status is filtered by a noise filtering algorithm to remove electromagnetic interference signals from the temperature and humidity sensors, while retaining valid status data related to equipment operation. The equipment action signals are filtered to remove falsely triggered instantaneous pulse signals, and the remaining signals are labeled with the start and end times of the corresponding circuit breaker or contactor action.

[0010] Preferably, feature extraction is performed on the preprocessed electrical parameters, environmental conditions, and equipment action signals. The specific steps are as follows: Electrical parameters are calculated using a sliding window to determine the root mean square values ​​and rates of change of voltage and current, thereby extracting the fluctuation characteristics of parameters that reflect operational stability. The environmental status is calculated using a trend analysis algorithm to determine the slope of continuous temperature and humidity changes, and to extract characteristics of abnormal environmental changes. The equipment action signals are statistically analyzed by arranging the action interval duration and the number of consecutive actions in a time sequence to extract the action time sequence features that characterize operational abnormalities.

[0011] Preferably, parameter fluctuation characteristics, state change characteristics, and action time sequence characteristics are analyzed across data using a multi-dimensional correlation model. The specific steps are as follows: Establish a time alignment mechanism to align parameter fluctuation characteristics, state change characteristics, and action timing characteristics to the same time axis according to timestamps; By mining association rules, the correlation strength between different data features is calculated, with a focus on associating electrical parameters with corresponding environmental conditions and equipment actions within the same time period. The associated features are cross-validated and fused to generate an associated feature set containing information on time, parameters, and actions.

[0012] Preferably, the abnormality type, fault location, and development trend of the protection cabinet are identified based on the associated feature set. The specific steps are as follows: By using a pattern matching model to cluster the associated feature set, the abnormality categories of the protection cabinet and the time periods of occurrence of each abnormality can be determined. Within each time period of an anomaly, the electrical node with the most significant parameter fluctuations is identified as the fault location using the location positioning model; By analyzing the rate of change of characteristics using trend prediction models, we can determine the possible time points and severity at which anomalies evolve into faults.

[0013] Preferably, an isolation strategy is formulated based on the severity level and impact range of the fault diagnosis information set. The specific steps are as follows: Faults are classified into minor anomalies, moderate faults, and severe faults according to their severity level, with each level corresponding to a different isolation priority. Under each fault level, the faults are sorted according to their electrical connections, and critical nodes that directly affect the main circuit are isolated first. The rapid isolation instructions are individually labeled with the name of the isolation device, the order of operations, and the completion time limit, forming a clearly hierarchical policy execution list.

[0014] Preferably, a rapid isolation command is generated for a specific fault point and dynamically adjusted. The specific steps are as follows: The policy execution list is converted into control commands that the device can recognize, and the content is organized using the execution logic of "level-location-operation"; When a new trend emerges in the development of a fault, the steps of multi-source data acquisition, preprocessing, feature extraction, correlation analysis, and fault identification are repeated to obtain new diagnostic information. Insert new diagnostic information into the corresponding fault level or location node, adjust the content and sequence of the original isolation instructions, and maintain the synchronization between the instructions and the fault evolution.

[0015] Preferably, when simultaneously acquiring electrical parameters, environmental conditions, and equipment action signals within the cabinet, the specific acquisition steps are as follows: Electrical parameters are acquired by electronic instrument transformers installed on the busbars and branches, and differential algorithms are used to suppress electromagnetic coupling interference. The environmental status is measured at multiple points by temperature and humidity sensor groups distributed on the top and bottom of the cabinet, and environmental data of different areas inside the cabinet are recorded simultaneously. The equipment action signals are acquired in real time through the digital acquisition module connected to the auxiliary contacts of the circuit breaker, including the equipment number, action time and status change information, forming a multi-source monitoring data set with multiple heterogeneous sources.

[0016] Preferably, when calculating the association strength between different data features through association rule mining, the specific calculation steps are as follows: Based on the timestamp, the parameter fluctuation characteristics, state change characteristics, and action sequence characteristics within the same time period are grouped into feature groups; For each set of characteristics, calculate the synchronicity between electrical parameter fluctuations and changes in environmental conditions; The correlation between the timing of the calculation device's actions and the onset of parameter fluctuations; Assign association strength based on synchronicity and correlation results; higher strength indicates a stronger causal relationship between the data.

[0017] Preferably, when the electrical parameters are corrected for sensor measurement errors via a dimension conversion module, the specific correction steps are as follows: The electrical parameters are divided into sampling segments of fixed duration, and the validity of each segment is checked to identify abnormal sampling points that exceed the normal range. The effective sampling segments are calibrated with reference values, and the sensor output signal is converted into candidate values ​​of actual electrical quantities using preset standard dimensions; Historical data was compared with the candidate values ​​of electrical quantities, and the most reasonable value was selected as the final correction result based on the equipment operating status.

[0018] Compared with the prior art, the beneficial effects of the present invention are: In the data acquisition phase, electrical parameters are collected by installing electronic current transformers on the busbars and branches, and differential algorithms are used to suppress electromagnetic coupling interference. At the same time, environmental conditions are measured at multiple points using temperature and humidity sensor groups distributed on the top and bottom of the cabinet, and equipment action signals are acquired in real time through digital acquisition modules connected to the auxiliary contacts of the circuit breaker. This forms a multi-source heterogeneous monitoring data set. This multi-source synchronous data acquisition method ensures the comprehensiveness and accuracy of the information obtained, laying a solid foundation for subsequent fault diagnosis and isolation.

[0019] During the data preprocessing stage, electrical parameters are corrected for sensor measurement errors and timestamps are recorded through a dimension conversion module. Environmental conditions are filtered to remove electromagnetic interference signals using a noise filtering algorithm. Equipment action signals are filtered to remove false trigger pulses and time nodes are marked. These preprocessing measures effectively improve data quality and reduce the impact of interference and errors on subsequent analysis.

[0020] During feature extraction, electrical parameters are extracted by calculating the root mean square values ​​of voltage and current and the rate of change through a sliding window. Environmental conditions are extracted by calculating the slope of continuous temperature and humidity changes through a trend analysis algorithm. Equipment action signals are extracted by statistically arranging action intervals and the number of consecutive actions through time sequence arrangement. By deeply mining data features from multiple dimensions, rich information is provided for fault identification.

[0021] The multi-dimensional association model establishes a time alignment mechanism to align various features according to timestamps, uses association rules to mine and calculate association strength, and then performs cross-validation fusion to generate a set of association features containing multi-dimensional information. This enables cross-data association analysis, accurately identifies causal relationships between data, and improves the accuracy and reliability of fault diagnosis.

[0022] Based on the associated feature set, the abnormality category and time period are determined by the pattern matching model, the fault location is identified by the location model, and the development trend is analyzed by the trend prediction model. This can comprehensively and accurately identify the abnormality type, fault location and development trend of the protection cabinet, providing detailed and accurate information for fault diagnosis.

[0023] In terms of isolation strategy formulation, priority is given according to the severity of the fault, and key nodes are prioritized for isolation based on electrical connection relationships. A strategy execution list containing equipment name, operation sequence and completion time limit is generated to ensure the scientific nature and effectiveness of the isolation strategy. This enables the rapid and accurate isolation of fault points and reduces the impact of faults on the power system.

[0024] When a new trend emerges in the development of a fault, new diagnostic information is obtained by repeatedly executing steps such as data acquisition and preprocessing, and isolation commands are dynamically adjusted to keep the commands synchronized with the evolution of the fault. This improves the adaptability and reliability of the system, enables timely responses to changes in the fault, and further ensures the safe operation of the power system. Attached Figure Description

[0025] Figure 1 This is a schematic diagram illustrating the working principle of the intelligent fault diagnosis and rapid isolation method for the power distribution protection cabinet described in this invention. Figure 2 A flowchart for real-time preprocessing of multi-source monitoring data; Figure 3 A flowchart for extracting features of electrical parameters, environmental conditions, and equipment action signals; Figure 4 A flowchart for cross-data correlation analysis of a multi-dimensional correlation model; Figure 5 A flowchart for calculating association strength in association rule mining. Detailed Implementation

[0026] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0027] Please see Figures 1-5 The present invention provides an intelligent fault diagnosis and rapid isolation method for power distribution protection cabinets, the method comprising:

[0028] During the operation of the power distribution protection cabinet, electrical parameters, environmental conditions and equipment action signals inside the cabinet are collected simultaneously as multi-source monitoring data. Real-time preprocessing is performed on the collected multi-source monitoring data, including calibration of electrical parameters to measure deviations, filtering of environmental interference and noise, and marking of time nodes for equipment action signals. Feature extraction is performed on the preprocessed electrical parameters, environmental conditions, and equipment action signals to form parameter fluctuation features, state change features, and action timing features. The parameter fluctuation characteristics, state change characteristics, and action time sequence characteristics are analyzed across data through a multi-dimensional correlation model to generate a set of correlation features containing causal relationships. Based on the set of associated features, the abnormality type, fault location and development trend of the protection cabinet are identified, and a fault diagnosis information set is formed. Isolation strategies are formulated based on the severity level and impact range of the fault diagnosis information set, generating rapid isolation instructions for specific fault points, and dynamically adjusting the content of the instructions as the fault evolves. Example

[0029] The load data acquisition module divides the power distribution network operating area within a preset range into hierarchical levels and obtains basic monitoring units. The specific implementation method is as follows: In practice, it is necessary to identify the initial coverage area of ​​the power distribution network within a preset range. This process requires comprehensive consideration of multiple factors, such as obtaining the geographical boundaries of the area through a geographic information system, clarifying the route and distribution of lines by combining the topology map of the power grid, and analyzing the load distribution within the area, including the electricity demand and load characteristics of different types of users, so as to accurately define the boundaries of the initial coverage area and provide a reliable basis for subsequent hierarchical division.

[0030] After defining the initial coverage area, a primary segmentation threshold needs to be extracted from its range values. The determination of the primary segmentation threshold is not arbitrary but based on various power grid-related factors. For example, different voltage levels in the power grid have varying requirements for load monitoring accuracy; high-voltage areas may require coarser segmentation, while low-voltage areas may require finer segmentation. Line capacity also affects the segmentation threshold; larger capacity lines may allow for larger segmentation ranges, and vice versa. Load density is a key factor; areas with high load density may require smaller primary segmentation thresholds for more accurate load change monitoring, while areas with low load density can have larger thresholds. Through comprehensive analysis and calculation of these factors, a suitable primary segmentation threshold is determined to perform the first-level segmentation of the power distribution network's operating area.

[0031] The power distribution network operating area is divided into first-level monitoring units based on the primary segmentation threshold. During this division, it is necessary to ensure that each first-level monitoring unit has relative independence in grid operation; that is, the load distribution and line characteristics within each unit should have a certain similarity to facilitate targeted monitoring and analysis later. After the division is completed, each first-level monitoring unit has its specific coverage area and operating characteristics, preparing for further second-level division.

[0032] After completing the first-level segmentation, it is necessary to identify the secondary coverage area of ​​each first-level monitoring unit. This requires a more in-depth analysis of the unit's internal situation, such as examining the specific routing of the wiring within the unit, determining the location and capacity of the distribution transformer, understanding the type of users (e.g., industrial, commercial, residential users) and their electricity consumption patterns, and analyzing load fluctuation characteristics. Through these detailed analyses, the secondary coverage area of ​​each first-level monitoring unit is clarified, providing a basis for extracting secondary segmentation thresholds.

[0033] Secondary segmentation thresholds are extracted from the range values ​​of the secondary coverage area. Setting these secondary segmentation thresholds also requires consideration of various factors. Besides the previously mentioned voltage level, line capacity, and load density, the load fluctuation characteristics within the first-level monitoring unit also need to be considered. For example, areas with frequent load fluctuations require smaller secondary segmentation thresholds to capture load changes more promptly. The importance of the equipment also affects the threshold setting; areas containing important equipment may require more refined divisions to monitor the equipment's operating status more closely. By comprehensively considering these factors, suitable secondary segmentation thresholds are determined for each first-level monitoring unit, which are then used for second-level segmentation.

[0034] The first-level monitoring units are further divided into second-level units based on secondary segmentation thresholds. These second-level units are more granular, enabling more precise detection of localized load changes and anomalies, providing a more detailed foundation for subsequent load data acquisition and analysis. Each second-level unit focuses on a smaller area or specific equipment, making monitoring more accurate and effective.

[0035] After obtaining the operating area of ​​the power distribution network, the first-level monitoring units, and the second-level monitoring units, these are designated as monitoring nodes. The next step is to construct the connection paths between these monitoring nodes. First, the overlap ratio between the power distribution network operating area and the first-level monitoring units is extracted. This requires accurately calculating the overlapping portion in both geographic space and power grid structure. For example, the spatial analysis function of a geographic information system (GIS) can be used to calculate the percentage of overlapping area in the geographic region, while simultaneously analyzing the sharing of power grid lines to determine the overlap ratio. The basic association strength is determined based on the overlap ratio. Generally, a higher overlap ratio indicates a stronger basic association strength, signifying a closer connection between the two.

[0036] Extracting the overlap density of load data between the first-level and second-level monitoring units requires detailed statistical analysis of their load data. This involves collecting load power data, voltage fluctuation data, and current change data for both units over a specific time period. The degree of overlap in load data within the same time frame is calculated; for example, the proportion of data values ​​overlapping within a certain range at the same time point relative to the total number of data points is used to determine the load data overlap density. A higher overlap density indicates a stronger secondary correlation, signifying a higher correlation between the load data of the two monitoring units.

[0037] Path weights are set based on the strength of the primary and secondary correlations, and these weights are used to measure the importance of the correlation between monitoring nodes. The path from the power distribution network operating area to the first-level monitoring unit is designated as the primary correlation path, as this is the main correlation direction from the whole to the part, and its weight is relatively large. The path from the first-level monitoring unit to the second-level monitoring unit is designated as the secondary correlation path, and its weight is relatively small. By appropriately setting the path weights, the primary and secondary correlation paths are constructed, thus completing the construction of the correlation path.

[0038] Basic monitoring units are generated based on monitoring nodes and associated paths. Each basic monitoring unit is a hierarchical and tightly interconnected monitoring network, where each monitoring node has a clear hierarchical relationship and associated path, enabling comprehensive and systematic monitoring of the power distribution network's operating area. This hierarchical division and associated path construction allows for comprehensive control of monitoring work from the whole to the part and then from the part back to the whole, improving monitoring efficiency and accuracy, and providing a solid foundation for subsequent multi-source data integration and processing, and state feature construction. The entire process is interconnected, with each step requiring meticulous operation and accurate calculation to ensure the scientific validity and effectiveness of the basic monitoring units. Example

[0039] The multi-source data integration and processing module collects load data from multiple categories of distribution network loads corresponding to the load data acquisition interval according to a preset sampling period and generates time-series integrated data. The specific implementation method is as follows: In practical applications, the initial collection time and time interval of distribution network load data within the load data acquisition range must first be determined based on the preset sampling period. The sampling period setting needs to comprehensively consider the actual operating conditions of the power grid. For example, in areas with drastic load fluctuations, the sampling period may be set relatively short to capture instantaneous load changes in a timely manner; while in areas with more stable load operation, the sampling period can be appropriately extended. The selection of the initial collection time is also crucial. Typically, it combines historical operating patterns of the power grid load to select a time that represents the typical load state of the area as the starting point. For instance, in areas dominated by industrial loads, 8:00 AM on a weekday, a time when the load begins to steadily rise, is chosen to ensure that the subsequently collected data is more representative. The time interval is precisely calculated based on the sampling period. For example, if the sampling period is 15 minutes, then the time interval is set to 15 minutes to ensure the regularity and periodicity of data collection.

[0040] After determining the initial collection time point and time interval, data acquisition nodes can be identified. These nodes are evenly distributed along the timeline, acting like multiple data acquisition "sentinels" to ensure comprehensive and continuous collection of load data throughout the sampling process. To guarantee the accuracy and reliability of data acquisition, comprehensive calibration and debugging of relevant sensors and monitoring equipment are required before determining the data acquisition nodes. For example, the load power sensor undergoes accuracy calibration to ensure that the deviation between its measured and actual values ​​is within acceptable limits; voltage and current monitoring devices undergo functional testing to ensure stable and accurate data acquisition. Only after the equipment has been debugged and confirmed to be working properly can the data acquisition nodes be determined based on the set initial collection time point and time interval, and data collection can then commence.

[0041] Load power data, voltage fluctuation data, and current change data corresponding to the distribution network within the load data acquisition range are collected through data acquisition nodes. During the collection process, a comprehensive data acquisition procedure and standards need to be established to ensure data integrity and consistency. For load power data collection, the scope of data acquisition needs to be clearly defined, including power data for different types of loads (such as industrial loads, commercial loads, and residential loads). For voltage fluctuation data, information such as real-time voltage values, fluctuation amplitudes, and the time points of fluctuation occurrences should be recorded. Current change data should cover the magnitude, phase, and trend of current changes. Simultaneously, to prevent data loss or omission, a data acquisition redundancy mechanism needs to be set up, such as using dual sensors to simultaneously acquire data, allowing for timely verification and correction when discrepancies arise between the two acquisition results.

[0042] The collected load power data, voltage fluctuation data, and current change data are integrated into multi-category distribution network load data corresponding to the load data acquisition interval. During the integration process, the first step is to standardize the format of different data types. For example, load power data, voltage fluctuation data, and current change data are all converted to the same numerical format and unit to avoid data processing difficulties caused by inconsistent formats. Secondly, the data needs to be timestamped to ensure that each set of data corresponds to the accurate acquisition time. This can be achieved by establishing a time synchronization mechanism to keep the clocks of various sensors and monitoring equipment consistent, thereby ensuring the accuracy of the data timestamps. After completing the format standardization and timestamp alignment, the data is classified and stored according to different categories, forming multi-category distribution network load data for further processing and analysis.

[0043] When generating time-series integrated data based on multi-category distribution network load data, the first step is to extract the timestamp identifiers corresponding to the multi-category distribution network load data. The timestamp identifier is crucial information recording the data acquisition time; it accurately reflects the acquisition time of each set of data and is an important basis for time-series arrangement. To ensure the accuracy of the timestamp identifiers, strict quality control is required during the data acquisition process for the generation and recording of timestamps, such as periodically calibrating the time server to ensure timestamp accuracy.

[0044] The multi-category distribution network load data corresponding to the timestamps are arranged in chronological order from morning to evening. During the arrangement process, strict sorting rules and verification mechanisms are required to ensure the accuracy of the data arrangement. Computer programs can be used to automatically perform the sorting, and the results can be manually verified to avoid sorting errors caused by program mistakes or data anomalies. For example, the data can be initially sorted by year, month, and day, and then precisely sorted by hour, minute, and second within the same day to ensure that each set of data is arranged in chronological order of its collection time.

[0045] The sorted multi-category distribution network load data is used as time-series integrated data. Time-series integrated data clearly reflects the changing trends and patterns of load data over time, such as the peak and trough periods of load power, the frequency and amplitude of voltage fluctuations, and the trend of current changes throughout the day. To facilitate the analysis and application of time-series integrated data, a comprehensive data storage and management system needs to be established to store and back up the data according to certain rules, while providing convenient data query and retrieval functions so that the subsequent state feature construction module can quickly and accurately obtain the required data. The entire multi-source data integration and processing process requires strict quality control at every stage, from data acquisition and integration to time-series sorting, ensuring the accuracy, completeness, and consistency of the data to provide reliable data support for subsequent power distribution network load monitoring. Example

[0046] The state feature construction module abstracts the distribution network load status within the load data acquisition interval based on time-series integrated data to obtain a state abstract model, and extracts multi-dimensional load features using a preset feature extraction framework. The specific implementation method is as follows: In practice, when extracting multi-category distribution network load data from different nodes within the same time period from time-series integrated data, it is necessary to first divide the time-series integrated data into time periods. The division of time periods can be determined according to actual monitoring needs, such as 15 minutes, 30 minutes, or 1 hour per period. After dividing the time periods, multi-category distribution network load data is extracted from different monitoring nodes within each time period. This data includes load power data, voltage fluctuation data, and current change data. Simultaneously, load power data for different nodes is extracted from the multi-category distribution network load data. Load power data is a key indicator reflecting the distribution network load status, and in-depth analysis of it helps to understand load changes.

[0047] Key fluctuation points in load power data at different nodes are extracted one by one. Determining these key fluctuation points requires establishing reasonable judgment criteria. For example, a point can be considered a key fluctuation point when the load power change exceeds 10% of its rated value; or when the load power experiences a continuous rise or fall within a short period (e.g., 5 minutes) and the change exceeds a certain threshold. Identifying the triggering conditions and fluctuation results of key fluctuation points requires analysis in conjunction with the power grid's operational status. Triggering conditions can be varied, such as a large industrial user starting equipment leading to a sudden increase in load, or a power grid line tripping due to a fault causing a load transfer. Fluctuation results refer to the specific changes in load power under the triggering conditions, such as the load power rising to a certain value and remaining stable within a short period, or experiencing continuous fluctuations. By analyzing the triggering conditions and fluctuation results, the causal relationships between the events can be extracted.

[0048] The frequency of key fluctuation points occurring within the same time window is statistically analyzed. The size of the time window can be set according to actual needs, such as 10 minutes or 20 minutes. Within the same time window, the number of times key fluctuation points occur is counted to extract co-occurrence relationships. Causal and co-occurrence relationships are used as the core elements of event association. Directed edges connect key fluctuation points with causal relationships, with the direction of the directed edges pointing from the triggering condition to the fluctuation result, to intuitively represent the direction of causal relationship transmission. Undirected edges connect key fluctuation points with co-occurrence relationships, indicating that these key fluctuation points co-occur within the same time window, thus completing the construction of event association.

[0049] Core fluctuation sequences are identified based on event correlations. These sequences are a series of key fluctuation points that play a dominant role in load state changes within the event correlation framework. By analyzing the connections and order of occurrence among these key fluctuation points, the core fluctuation sequences, which significantly impact load state, are identified. Based on the order of occurrence of these core fluctuation sequences, typical patterns of distribution network loads are summarized. These typical patterns include normal operation patterns, load surge patterns, and load fluctuation patterns. For example, if the core fluctuation sequence exhibits a pattern of stable load power increase over a certain period, reaching a peak and maintaining that peak for a time before gradually decreasing, this can be categorized as a typical daily load change pattern.

[0050] A set of state labels for distribution network loads is generated based on typical patterns. These state labels are used to classify and identify different load states. For example, when the load is in a normal and stable operating state, it is labeled as "normal state"; when the load exceeds the rated value by a certain range, it is labeled as "overload state"; when the load exhibits abnormal fluctuations, it is labeled as "abnormal fluctuation state," and so on. A feature vector space for the distribution network loads is generated based on the state label set. This feature vector space is a multi-dimensional space, with each dimension corresponding to a characteristic of the load. The statistical regularities of the load power data are mapped to the feature vector space to obtain an abstract model of the fluctuations in the distribution network loads. This model mathematically describes the fluctuation characteristics of the load power.

[0051] This paper abstracts voltage fluctuation and current variation data from various types of distribution network load data. For voltage fluctuation data, features such as fluctuation amplitude, fluctuation frequency, and voltage deviation are extracted; for current variation data, features such as variation trend, phase difference, and RMS current value are extracted. By analyzing and processing these features, abstract representations are obtained. Finally, based on the abstracted voltage fluctuation data, abstracted current variation data, and the fluctuation abstract model, a state abstract model is determined. This model comprehensively considers the variation characteristics of load power, voltage, and current, and can comprehensively and accurately describe the state of the distribution network load.

[0052] When extracting multidimensional load features corresponding to the state abstract model using a pre-defined feature extraction framework, the load features of the state abstract model are extracted using the feature extraction framework. These load features include capacity carrying capacity, operational stability, and distribution balance. Capacity carrying capacity reflects the maximum load capacity that the distribution network can carry under different load conditions, and can be determined by analyzing the relationship between load power data and the capacity of distribution network equipment. Operational stability reflects the stability of the distribution network operation, and can be measured by calculating indicators such as the variance of load power, voltage, and current fluctuations. Distribution balance indicates the uniformity of load distribution among different nodes, and can be determined by calculating the difference coefficient of load power at each node.

[0053] The fluctuation deviation characteristics and load distribution characteristics of the state abstract model are extracted. The fluctuation deviation characteristics describe the degree of deviation between load fluctuations and the normal state, and can be obtained by calculating the deviation value between the actual load data and the load data under normal conditions. The load distribution characteristics reflect the spatial distribution of loads in the distribution network, and can be determined by analyzing the distribution pattern of load power at each node. These load characteristics, fluctuation deviation characteristics, and load distribution characteristics are defined as multi-dimensional load characteristics. These characteristics comprehensively depict the state and characteristics of the distribution network load from different dimensions, providing rich evidence for subsequent abnormal condition identification and assessment. The entire process of constructing state characteristics requires meticulous data analysis and processing to ensure that the extracted characteristics accurately reflect the actual state of the distribution network load, providing reliable support for load monitoring and anomaly diagnosis in the power distribution network. Example

[0054] The abnormal operating condition identification and assessment module identifies local abnormal operating conditions of the distribution network within the load data acquisition range based on multi-dimensional load characteristics, and determines the overall abnormal operating conditions of the distribution network accordingly. The specific implementation method is as follows: In practical applications, when dealing with a specific power distribution network area, such as the distribution network of a commercial area in a city, it is first necessary to compare the load characteristics in the multidimensional load characteristics with the pre-set standard load characteristics.

[0055] Taking the capacity carrying capacity characteristic in load characteristics as an example, assuming the standard capacity carrying capacity characteristic of the commercial area's distribution network is derived from historical normal operation data, its rated capacity is 10000kVA, and the capacity utilization rate during normal operation is typically between 60% and 80%. In actual monitoring, the current capacity carrying capacity characteristic of the area is extracted. If it is found that the capacity utilization rate consistently reaches 90% during a certain period, this actual load characteristic needs to be compared with the standard load characteristic. By calculating the difference between the two, for example, the difference between the actual utilization rate and the standard upper limit of 80% is 10%, the first comparison coefficient is obtained. The comparison method here can be to calculate the degree of deviation between the actual value and the standard value, such as the proportion of deviation from the standard upper limit, thus obtaining the first comparison coefficient of 0.125 (i.e., 10% / 80%). Then, the first comparison coefficient is converted into the first outlier. The conversion method can be set such that when the comparison coefficient is 0, the outlier value is 0, and for every 0.1 increase in the comparison coefficient, the outlier value increases by 10, so the first outlier value is 12.5.

[0056] Next, the fluctuation deviation characteristic is processed. Assume the standard deviation characteristic specifies a voltage fluctuation standard deviation of 5%. In the real-time monitoring data of the commercial area's power distribution network, the voltage fluctuation standard deviation reached 8% over a certain period. This fluctuation deviation characteristic is compared with the standard deviation characteristic, and the difference between the actual deviation and the standard deviation is calculated to be 3%, resulting in a second comparison coefficient of 0.6 (i.e., 3% / 5%). According to a preset conversion rule, the second comparison coefficient is converted into a second outlier. For example, if the conversion ratio between the comparison coefficient and the outlier is set to 1:20, then the second outlier is 12.

[0057] Next, let's look at the load distribution characteristics. The standard distribution characteristics stipulate that the difference coefficient of load distribution at each node should be less than 0.3. When analyzing the load distribution at each node of the distribution network in this commercial area, the calculated difference coefficient is 0.45. Comparing this load distribution characteristic with the standard distribution characteristic, the third comparison coefficient is 0.5 (i.e., (0.45-0.3) / 0.3). According to the conversion rules, if each 0.1 comparison coefficient corresponds to an outlier of 5, then the third outlier is 25.

[0058] After obtaining the first outlier of 12.5, the second outlier of 12, and the third outlier of 25, their average outlier was calculated. Assuming the arithmetic mean method is used, the average outlier is (12.5 + 12 + 25) / 3 ≈ 16.5. The pre-set outlier threshold is 10. Since the average outlier of 16.5 exceeds the threshold, it can be determined that there is a localized abnormal condition in the distribution network within this load data collection range. Further analysis revealed that this localized abnormal condition was mainly caused by a large shopping mall operating a large number of air conditioning units during a promotional event, leading to a sudden increase in load, increased voltage fluctuations, and an uneven load distribution across nodes.

[0059] When determining the overall abnormal operating conditions of a distribution network based on local abnormal conditions, it is necessary to comprehensively consider the local abnormal operating conditions of each load data collection interval in the commercial area's distribution network. Assuming the commercial area's distribution network is divided into 5 load data collection intervals, and similar local abnormal operating conditions occur in 3 of these intervals, and these local abnormal conditions are related (e.g., all caused by the start-up of large loads on the same power supply line), then these local abnormal conditions may have a systemic impact on the operation of the entire distribution network. In this case, it is necessary to determine whether these local abnormal conditions have led to an overall voltage drop or frequency fluctuations exceeding the allowable range. If it is found that the voltage of the entire distribution network has generally dropped by 3%, and the frequency fluctuation is within 0.2Hz (still within the allowable range), then the distribution network can be determined to be in a state of mild abnormal operation. Appropriate control measures need to be taken, such as adjusting transformer tap changes or starting backup power supplies, to alleviate the abnormal situation and ensure the safe and stable operation of the distribution network.

[0060] Throughout the entire process of identifying and assessing abnormal operating conditions, each step must be strictly performed according to pre-set standards and methods to ensure data accuracy and the reliability of the judgment. Simultaneously, it is necessary to comprehensively analyze various characteristics in conjunction with the actual operating conditions of the distribution network to avoid misjudging the overall operating condition due to anomalies in a single characteristic. For example, if only the load distribution characteristics are abnormal, while the capacity carrying capacity and fluctuation deviation characteristics are within the normal range, it may only be a load adjustment at a local node and will not have a significant impact on the overall operating condition. Only by comprehensively considering the abnormal situations of multiple characteristics and analyzing their correlations can the local and overall abnormal operating conditions of the distribution network be accurately determined, providing an accurate basis for subsequent monitoring and control strategies. Example

[0061] This embodiment details the specific implementation of the monitoring strategy control module, which dynamically adjusts the power distribution network monitoring parameters based on local and overall abnormal operating conditions, and uses the adjusted parameters to perform real-time dynamic monitoring of the power distribution network operating area to obtain the real-time load status of the power distribution network.

[0062] Taking the power distribution network of an industrial park as an example, when the power distribution network experiences an overall abnormal operating condition, such as during the high-temperature period in summer, the load of the entire industrial park power distribution network generally increases due to the start-up of a large number of air conditioning devices. The current of multiple lines exceeds 80% of the rated value, and the voltage drops significantly. At this time, it is necessary to determine the first monitoring density and monitoring period of the first-level monitoring unit based on the overall abnormal operating condition.

[0063] The first-level monitoring units are typically divided according to the power supply zones of the industrial park, with each zone responsible for supplying power to a specific area. Under abnormal operating conditions, to comprehensively understand the operational status of the entire distribution network, it is necessary to increase the initial monitoring density, increasing the monitoring frequency from once every 30 minutes to once every 10 minutes, in order to obtain load data changes more promptly. Simultaneously, the monitoring period should be extended, from the original monitoring only from 8:00 to 22:00 on weekdays to continuous 24-hour monitoring, ensuring that any abnormal load conditions that may occur at night can also be detected promptly.

[0064] For localized abnormal operating conditions, suppose that in a production workshop within the industrial park, a failure of a large piece of equipment causes a sudden increase in load on the second-level monitoring unit, with the current instantly rising to 120% of its rated value, accompanied by severe voltage fluctuations. In this case, it is necessary to determine the monitoring focus of the second-level monitoring unit through localized abnormal operating conditions, namely the area where the production workshop is located and related power supply lines and equipment, such as the transformers and distribution cabinets supplying power to the workshop.

[0065] The monitoring density and depth are configured to prioritize key monitoring areas. The second monitoring density is increased from once every 15 minutes to once every 5 minutes, while the monitoring depth is also increased. In addition to monitoring conventional parameters such as load power, voltage, and current, monitoring of parameters such as equipment temperature, switch status, and harmonic content is also added. For example, a temperature sensor is used to monitor the transformer oil temperature in real time, and a harmonic analyzer is used to detect harmonic components in the current, in order to gain a more comprehensive understanding of the specific circumstances of local abnormal operating conditions.

[0066] The power distribution network monitoring parameters of the first-level monitoring unit are adjusted according to the first monitoring density and monitoring period. Specifically, this includes adjusting the sampling frequency and data transmission frequency of the monitoring equipment. The sampling frequency of the monitoring equipment in the first-level monitoring unit is increased from 100Hz to 300Hz, and the data transmission frequency is adjusted from once per minute to once every 10 minutes to ensure that more intensive monitoring data can be collected in a timely manner.

[0067] The power distribution network monitoring parameters of the second-level monitoring unit were adjusted according to the second monitoring density and monitoring depth. Temporary monitoring equipment, such as portable harmonic detectors and temperature monitoring instruments, were added to the second-level monitoring unit where the production workshop is located. At the same time, the sampling frequency of the original monitoring equipment was increased to 500Hz to meet the higher monitoring density and depth requirements.

[0068] When using adjusted distribution network monitoring parameters to conduct real-time dynamic monitoring of the power distribution network operating area, the adjusted sampling frequency and monitoring equipment are used to collect real-time monitoring data of the power distribution network operating area. For example, the monitoring equipment of the first-level monitoring unit collects data such as load power, voltage, and current of the entire industrial park every 10 minutes, while the monitoring equipment of the second-level monitoring unit collects detailed data of the production workshop every 5 minutes, including load power, voltage, current, transformer oil temperature, and harmonic content.

[0069] The status quantification indicators corresponding to the real-time monitoring data are analyzed. For the data of the first-level monitoring unit, the load change trend, voltage stability, and current distribution of the entire industrial park are analyzed. For the data of the second-level monitoring unit, the focus is on analyzing the abnormal load of the production workshop, the causes of voltage fluctuations, whether the equipment temperature is normal, and whether harmonics exceed the standard.

[0070] Based on the analyzed quantitative indicators, the real-time load status of the distribution network is determined. Assuming that the analysis reveals that while the load across the entire industrial park is generally high, under adjusted monitoring, the load growth trend has slowed. After voltage regulation measures were implemented, the voltage gradually stabilized at approximately 95% of the rated value, and the current was controlled within 85% of the rated value, indicating that the overall abnormal operating conditions have been alleviated to some extent. Meanwhile, after the equipment fault in the production workshop was resolved, the load gradually returned to normal, with the current stabilizing at 70% of the rated value, the voltage returning to normal, the transformer oil temperature dropping to the normal range, and harmonic content meeting standards, indicating that the local abnormal operating conditions have been eliminated. At this point, the real-time load status of the distribution network can be determined to be generally stable, with localized areas returning to normal. However, a high monitoring density is still needed to prevent the recurrence of abnormal conditions.

[0071] Throughout the implementation process, monitoring parameters need to be flexibly adjusted based on actual abnormal operating conditions to ensure that the monitoring accurately reflects the operating status of the distribution network and provides reliable monitoring assurance for the safe and stable operation of the power system. Simultaneously, attention must be paid to the operating status of the monitoring equipment, with regular maintenance and calibration to guarantee the accuracy and reliability of the monitoring data. By dynamically adjusting the monitoring strategy, precise monitoring of the distribution network load and timely detection of anomalies can be achieved, providing strong support for subsequent fault diagnosis and handling.

[0072] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0073] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A method for intelligent fault diagnosis and rapid isolation of a power distribution protection cabinet, characterized in that, Includes the following steps: During the operation of the power distribution protection cabinet, electrical parameters, environmental conditions and equipment action signals inside the cabinet are collected simultaneously as multi-source monitoring data. Real-time preprocessing is performed on the collected multi-source monitoring data, including calibration of electrical parameters to measure deviations, filtering of environmental interference and noise, and marking of time nodes for equipment action signals. Feature extraction is performed on the preprocessed electrical parameters, environmental conditions, and equipment action signals to form parameter fluctuation features, state change features, and action timing features. The parameter fluctuation characteristics, state change characteristics, and action time sequence characteristics are analyzed across data through a multi-dimensional correlation model to generate a set of correlation features containing causal relationships. Based on the set of associated features, the abnormality type, fault location and development trend of the protection cabinet are identified, and a fault diagnosis information set is formed. Isolation strategies are formulated based on the severity level and impact range of the fault diagnosis information set, generating rapid isolation instructions for specific fault points, and dynamically adjusting the content of the instructions as the fault evolves.

2. The intelligent fault diagnosis and rapid isolation method for power distribution protection cabinets according to claim 1, characterized in that, The collected multi-source monitoring data undergoes real-time preprocessing, and the specific steps are as follows: Electrical parameters are corrected for sensor measurement errors through a dimension conversion module, and the acquisition timestamp of each parameter is recorded synchronously. The environmental status is filtered by a noise filtering algorithm to remove electromagnetic interference signals from the temperature and humidity sensors, while retaining valid status data related to equipment operation. The equipment action signals are filtered to remove falsely triggered instantaneous pulse signals, and the remaining signals are labeled with the start and end times of the corresponding circuit breaker or contactor action.

3. The intelligent fault diagnosis and rapid isolation method for power distribution protection cabinets according to claim 1, characterized in that, Feature extraction is performed on the preprocessed electrical parameters, environmental conditions, and equipment action signals. The specific steps are as follows: Electrical parameters are calculated using a sliding window to determine the root mean square values ​​and rates of change of voltage and current, thereby extracting the fluctuation characteristics of parameters that reflect operational stability. The environmental status is calculated using a trend analysis algorithm to determine the slope of continuous temperature and humidity changes, and to extract characteristics of abnormal environmental changes. The equipment action signals are statistically analyzed by arranging the action interval duration and the number of consecutive actions in a time sequence to extract the action time sequence features that characterize operational abnormalities.

4. The intelligent fault diagnosis and rapid isolation method for power distribution protection cabinet according to claim 3, characterized in that, The parameter fluctuation characteristics, state change characteristics, and action time sequence characteristics are analyzed across data using a multi-dimensional correlation model. The specific steps are as follows: Establish a time alignment mechanism to align parameter fluctuation characteristics, state change characteristics, and action timing characteristics to the same time axis according to timestamps; By mining association rules, the correlation strength between different data features is calculated, with a focus on associating electrical parameters with corresponding environmental conditions and equipment actions within the same time period. The associated features are cross-validated and fused to generate an associated feature set containing information on time, parameters, and actions.

5. The intelligent fault diagnosis and rapid isolation method for power distribution protection cabinet according to claim 4, characterized in that, The specific steps for identifying the anomaly type, fault location, and development trend of the protection cabinet based on the associated feature set are as follows: By using a pattern matching model to cluster the associated feature set, the abnormality categories of the protection cabinet and the time periods of occurrence of each abnormality can be determined. Within each time period of an anomaly, the electrical node with the most significant parameter fluctuations is identified as the fault location using the location positioning model; By analyzing the rate of change of characteristics using trend prediction models, we can determine the possible time points and severity at which anomalies evolve into faults.

6. The intelligent fault diagnosis and rapid isolation method for power distribution protection cabinet according to claim 5, characterized in that, Isolation strategies are formulated based on the severity level and impact range of the fault diagnosis information set. The specific steps are as follows: Faults are classified into minor anomalies, moderate faults, and severe faults according to their severity level, with each level corresponding to a different isolation priority. Under each fault level, the faults are sorted according to their electrical connections, and critical nodes that directly affect the main circuit are isolated first. The rapid isolation instructions are individually labeled with the name of the isolation device, the order of operations, and the completion time limit, forming a clearly hierarchical policy execution list.

7. The intelligent fault diagnosis and rapid isolation method for power distribution protection cabinets according to claim 6, characterized in that, Generate and dynamically adjust rapid isolation commands for specific fault points. The specific steps are as follows: The policy execution list is converted into control commands that the device can recognize, and the content is organized using the execution logic of "level-location-operation"; When a new trend emerges in the development of a fault, the steps of multi-source data acquisition, preprocessing, feature extraction, correlation analysis, and fault identification are repeated to obtain new diagnostic information. Insert new diagnostic information into the corresponding fault level or location node, adjust the content and sequence of the original isolation instructions, and maintain the synchronization between the instructions and the fault evolution.

8. The intelligent fault diagnosis and rapid isolation method for power distribution protection cabinets according to claim 1, characterized in that, When simultaneously acquiring electrical parameters, environmental conditions, and equipment action signals within the cabinet, the specific acquisition steps are as follows: Electrical parameters are acquired by electronic instrument transformers installed on the busbars and branches, and differential algorithms are used to suppress electromagnetic coupling interference. The environmental status is measured at multiple points by temperature and humidity sensor groups distributed on the top and bottom of the cabinet, and environmental data of different areas inside the cabinet are recorded simultaneously. The equipment action signals are acquired in real time through the digital acquisition module connected to the auxiliary contacts of the circuit breaker, including the equipment number, action time and status change information, forming a multi-source monitoring data set with multiple heterogeneous sources.

9. The intelligent fault diagnosis and rapid isolation method for power distribution protection cabinet according to claim 4, characterized in that, When calculating the association strength between different data features using association rule mining, the specific calculation steps are as follows: Based on the timestamp, the parameter fluctuation characteristics, state change characteristics, and action sequence characteristics within the same time period are grouped into feature groups; For each set of characteristics, calculate the synchronicity between electrical parameter fluctuations and changes in environmental conditions; The correlation between the timing of the calculation device's actions and the onset of parameter fluctuations; Assign association strength based on synchronicity and correlation results; higher strength indicates a stronger causal relationship between the data.

10. The intelligent fault diagnosis and rapid isolation method for power distribution protection cabinet according to claim 2, characterized in that, When correcting sensor measurement errors using a dimensionless conversion module, the specific correction steps are as follows: The electrical parameters are divided into sampling segments of fixed duration, and the validity of each segment is checked to identify abnormal sampling points that exceed the normal range. The effective sampling segments are calibrated with reference values, and the sensor output signal is converted into candidate values ​​of actual electrical quantities using preset standard dimensions; Historical data was compared with the candidate values ​​of electrical quantities, and the most reasonable value was selected as the final correction result based on the equipment operating status.

Images