Method and system for real-time analysis and fault early warning based on digital electric port box variable data

Abstract

This invention discloses a method and system for real-time data analysis and fault early warning of digital power substations, belonging to the technical field of power equipment monitoring and control systems. The method includes: real-time acquisition of environmental parameters of various components of the substation through a monitoring network; obtaining parameter sequences after data smoothing and noise reduction; constructing a dynamic storage framework based on the sequences and spatiotemporal attributes; extracting historical records of target components for multi-dimensional sorting and rate of change calculation to identify potential anomalies; generating an alarm sequence and determining preliminary anomaly markers when the rate of change exceeds limits and the distribution boundary is breached; verifying accuracy by comparing the markers with historical records and integrating influence weight analysis to obtain the final early warning classification result; and pushing parameter summary information to the control system based on the classification result and obtaining response feedback to form a closed-loop control of the operating status. This invention solves the control lag problem caused by data noise and delay, improving the timeliness, reliability, and automation level of substation control.

Description

Technical Field

[0001] This invention relates to the field of power equipment monitoring and control system technology, specifically to a method and system for real-time data analysis and fault early warning of digital power port transformer substations. Background Technology

[0002] In the field of power system operation and maintenance, prefabricated transformers (referred to as "prefabricated transformers") are key nodes in power distribution, and their safe and stable operation is of paramount importance to ensuring the reliability of regional power supply. With the popularization of the Internet of Things and digital technologies, real-time status monitoring based on sensor data has become a mainstream means to improve the efficiency of power equipment management.

[0003] Currently, most existing substation monitoring solutions achieve data acquisition and basic display of environmental parameters (such as temperature and humidity). However, from the perspective of control systems, these existing technologies have significant shortcomings in achieving proactive fault warning and forming an automated control closed loop. The core problems are mainly reflected in the following aspects:

[0004] The existing solutions suffer from weak data processing capabilities, resulting in poor input signal quality for control decisions. Environmental parameter data within the transformer substation is susceptible to various electromagnetic noises and field interferences during real-time acquisition. Current technologies lack effective dynamic filtering and denoising mechanisms, leading to a large amount of invalid information in the data transmitted to the control system. When the control system makes judgments based on this impure data, it is highly prone to false alarms or missed alarms, severely impacting the accuracy of control execution. The rigid data storage and analysis models of existing technologies result in significant control response delays. Current solutions often employ static threshold alarm mechanisms, whose data storage structures fail to fully consider the dynamic correlation of parameters in time and space. When equipment exhibits initial abnormalities (such as a slow temperature rise in a component), the system cannot identify minute deviations through comparison with historical trends and the status of adjacent components, preventing early warnings. Alarms are only triggered when parameter values ​​exceed fixed safety thresholds, by which time the equipment may have progressed from a sub-optimal state to an irreversible failure, causing the control system to miss the optimal early intervention opportunity.

[0005] The existing technological architecture fundamentally separates "monitoring" from "control," failing to form an intelligent closed loop of perception-analysis-decision-execution. Most existing monitoring systems merely act as data dashboards, and their alarm outputs often lack risk classification and specific trend guidance, failing to provide sufficiently accurate and forward-looking decision-making basis for higher-level control systems (such as load dispatching systems and cooling control systems). This results in control behavior remaining in a passive "remedial" mode rather than a proactive "prevention" mode, severely restricting the automation and intelligence level of power system operation and maintenance.

[0006] In summary, existing transformer substation monitoring solutions suffer from poor data quality and delayed analysis, resulting in slow control response and an inability to proactively intervene in faults. Summary of the Invention

[0007] This invention discloses a method and system for real-time analysis and fault early warning of digital power port transformer substation data. Specifically, it is a method and system that collects and processes environmental parameters of multiple components of the transformer substation in real time, and automatically generates graded early warning signals based on dynamic anomaly identification results to drive the control system to respond. It can effectively solve the technical problem that existing transformer substation monitoring schemes are slow to control response due to poor data quality and analysis lag, and cannot achieve proactive intervention in faults.

[0008] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is as follows:

[0009] The method for real-time data analysis and fault early warning of digital power port transformer substations includes the following steps:

[0010] Step 1: Collect environmental parameter data from various components of the power equipment in real time through the monitoring device network, perform preliminary screening based on parameter deviation indicators and distribution balance, and use data smoothing methods to denoise the collected data to obtain a smoothed parameter sequence;

[0011] Step 2: Based on the smoothed parameter sequence and combined with spatiotemporal attribute information, analyze the characteristics of deviation persistence and fluctuation dynamic indicators, construct a data organization structure, group the environmental parameter data according to frequency distribution and anomaly ratio indicators, and determine the initial storage container;

[0012] Step 3: Obtain parameter data within the initial storage container, calculate spatial correlation indicators and consistency indicators. If the trend changes exceed preset conditions or the time correlation is lower than the standard, trigger dynamic adjustment of the initial storage container to obtain the adjusted storage framework.

[0013] Step 4: Extract historical parameter records of the target component from the adjusted storage framework, sort them in multiple dimensions based on peak comparison and benchmark comparison, and determine potential anomaly locations and time information;

[0014] Step 5: For potential anomaly locations, calculate the rate of change index and evaluate it in conjunction with triggering conditions and risk assessment factors. If the rate of change exceeds the condition and the distribution boundary is breached, generate an alarm output sequence to determine the preliminary anomaly marker.

[0015] Step 6: By comparing the initial anomaly markers with historical parameter records, and integrating the influence weight analysis and anomaly ratio assessment, the accuracy of the alarm output is verified, and the final early warning classification result is obtained.

[0016] Step 7: Based on the final early warning classification result, push parameter summary information including trend changes and spatial consistency to the control system and obtain response feedback records.

[0017] Step 1 is as follows:

[0018] Real-time environmental parameter data is acquired from various components of the power equipment through a monitoring device network and continuously recorded at a preset acquisition frequency to obtain an initial environmental parameter dataset. For this initial dataset, a data smoothing method is used to denoise the data, eliminating noise interference and obtaining a smoothed parameter sequence. Based on the smoothed parameter sequence, the specific deviations of the parameters are analyzed. If a parameter value exceeds a preset threshold range, it is marked as an abnormal parameter point, resulting in a marked abnormal dataset. From the marked abnormal dataset, feature information of the balanced distribution state is extracted to determine whether there is an uneven distribution, obtaining an evaluation result of the distribution state. Based on the evaluation result, if an uneven distribution is found, the weights of the abnormal parameter points are adjusted to determine the adjusted parameter distribution set. The adjusted parameter distribution set is obtained, and combined with the update frequency of real-time data, the changing trend of the parameter sequence is continuously monitored to determine whether there are potential abnormal fluctuations, obtaining the final monitoring and analysis result. Based on the final monitoring and analysis result, a corresponding parameter adjustment strategy is generated to dynamically optimize the operating status of the power equipment, resulting in an optimized operating parameter configuration.

[0019] Step 2 is as follows: Using the smoothed parameter sequence and spatiotemporal attribute information, analyze the correlation pattern between the persistence of deviations and the dynamics of fluctuations. Preliminary grouping of the data is performed using preset classification rules to obtain a classified dataset. Based on the classified dataset, a logical framework for data organization is constructed based on the characteristics of frequency distribution and anomaly ratio. A preset threshold is used to filter the anomaly ratio, determining an anomalous data subset. The anomalous data subset is then analyzed, and the fluctuation pattern of the parameter sequence over different time periods is analyzed, using statistical tools to quantify the fluctuation dynamics, resulting in a fluctuation feature set. Based on the fluctuation feature set, a corresponding storage unit allocation scheme is constructed for the frequency distribution pattern. If the anomaly ratio of a certain distribution pattern exceeds a preset threshold, the data under that pattern is prioritized, determining the marked data units. Based on the marked data units, and combining attribute information and the mapping logic of spatiotemporal attributes, the rationality of the data classification is analyzed. The data organization framework is then used to optimize and adjust the classification results, resulting in an optimized classification structure. The optimized classification structure is obtained, and the distribution balance of parameter sequences under different classifications is analyzed based on the allocation of storage units. If the distribution balance under a certain classification is lower than the preset standard, the data under that classification is regrouped to determine the final storage allocation scheme. Based on the final storage allocation scheme, combined with the fluctuation feature set and the dynamic trend, the update status of the parameter sequences is continuously tracked, and automated tools are used to dynamically adjust the data classification and storage units to obtain the adjusted data management framework.

[0020] Step 3 is as follows:

[0021] Using the parameter data in the initial storage container, the fluctuation of trend changes is analyzed based on the calculation results of spatial correlation indicators and consistency indicators. If the trend changes exceed the preset threshold, the data in the storage container is prioritized and sorted to obtain a sorted data set.

[0022] Based on the sorted dataset, obtain the time-related evaluation results. If the time correlation is lower than the predetermined standard, the dataset is stratified to determine the data grouping after stratification.

[0023] After obtaining the hierarchical data groups, the dynamically adjusted allocation logic is analyzed based on the capacity limit of the storage container. If the data volume of a certain group exceeds the container limit, the group is split to obtain the split data units.

[0024] By splitting the data units and combining them with the spatially related distribution pattern, the update rules of the storage framework are constructed. The data units are repositioned using a preset allocation strategy to determine the updated storage layout.

[0025] The updated storage layout is obtained. Based on the stability requirements of consistency, the balance of data units in the storage framework is analyzed. If the balance does not meet the predetermined standard, a secondary adjustment mechanism is triggered to obtain the optimized storage structure.

[0026] Based on the optimized storage structure and dynamically adjusted execution records, the subsequent impact of trend changes is continuously tracked, and automated tools are used to monitor the storage framework in real time to determine the final storage configuration.

[0027] Based on the final storage configuration, a long-term data storage management solution is built to ensure the persistence of the adjustment results. Log recording tools are used to track changes to the storage framework and obtain a complete change profile.

[0028] Step 4 is as follows:

[0029] By adjusting the storage framework, relevant historical parameter data of the target component is obtained, and the data is initially classified using preset filtering rules to obtain a set of classified parameters.

[0030] Based on the classified parameter set, the peak difference and the baseline value are compared. If the peak difference exceeds the preset threshold, the data with peak differences exceeding the threshold are marked to determine the marked abnormal dataset.

[0031] Starting from the labeled abnormal dataset, and combining the logic of multidimensional sorting, the data is prioritized and sorted to obtain the sorted data sequence;

[0032] For the sorted data sequence, analyze the specific distribution of abnormal locations, use a preset positioning tool to accurately identify the abnormal locations, and determine the specific abnormal coordinate points;

[0033] By combining abnormal coordinate points with the recorded data of time nodes, the time correlation information in historical parameters can be traced to obtain the corresponding time period distribution.

[0034] Based on the time period distribution and the results of the difference analysis, a correspondence between abnormal locations and time nodes is constructed to determine the final abnormal distribution pattern.

[0035] Obtain the final anomaly distribution pattern, combine it with the adjustment structure of the storage framework, generate parameter optimization strategies for the target components, and determine the direction of subsequent processing.

[0036] Step 5 is as follows:

[0037] For monitoring the rate of change, the rate change data of the target object is obtained from historical data records and compared with a preset threshold.

[0038] If the rate change exceeds the threshold range, an initial abnormal signal is generated to identify the initial point of interest.

[0039] Based on the initial anomaly signal, data is extracted from the anomaly location, and the location information is refined using a preset positioning tool to obtain the specific anomaly coordinate distribution.

[0040] By analyzing the abnormal coordinate distribution and combining it with the boundary data of the distribution range, we can determine whether there is a significant offset.

[0041] If an offset phenomenon is detected, a corresponding offset identifier is generated to determine the abnormal distribution characteristics;

[0042] For the offset identifier, relevant trigger condition data is obtained, and combined with the preset rules of risk assessment for comprehensive processing to obtain the risk level classification result;

[0043] Based on the risk level classification results, a corresponding alarm sequence is generated, and the alarms are sorted according to a preset priority rule to determine the output order of the alarms;

[0044] By combining the output sequence of the alarms with the recorded data of the initial markings, the abnormal locations are confirmed a second time to obtain the final set of abnormal markings;

[0045] For the final set of anomaly labels, a logistic regression model is used to analyze the correlation between the anomaly labels, identify the potential relationships between them, and determine the key directions for subsequent processing.

[0046] Step 6 is as follows:

[0047] By comparing the initial labels with historical parameters, the difference information between the labeled data and the parameter records is obtained. A preset threshold range is used for preliminary screening to obtain a set of labels with significant differences.

[0048] Based on the significantly different set of labels, data is extracted for the anomaly ratio, and weighted calculation is performed in combination with the preset rules that affect the weight to determine the weight value of the anomaly ratio.

[0049] By using the weighted value of the abnormal ratio, data records related to alarm output are obtained. If the weighted value exceeds the preset threshold range, a corresponding alarm signal is generated, and the priority of the alarm signal is determined.

[0050] Based on the priority of alarm signals and combined with the logical rules of proportional evaluation, the signal data is processed in layers to obtain the layered signal classification.

[0051] By classifying the signals after stratification, and based on the data requirements for early warning classification, a logistic regression model is used to map the classified data to determine the preliminary results of early warning classification.

[0052] Based on the preliminary results of the early warning classification and combined with the processing rules for accurate verification, the classification data is compared a second time to obtain the final output results.

[0053] Based on the final output, the business logic for grading is used to match the result data with the preset grading standards to determine the final warning grading status.

[0054] Step 7 is as follows:

[0055] Based on the final early warning classification results, a summary of parameter data containing trend changes and spatial consistency is compiled and sent to the control system to obtain the interaction log after transmission.

[0056] Based on the interaction logs after sending, the response feedback data returned by the control system is extracted, and the data is parsed using preset field rules to determine the parsed feedback content;

[0057] Based on the parsed feedback content, the key fields in the feedback record are classified and processed to obtain the classified data groups and determine the integrity of the grouped data.

[0058] If the integrity of the grouped data meets the preset threshold range, the classified data groups are prioritized and sorted to obtain the sorted feedback sequence.

[0059] Based on the sorted feedback sequence, data is filtered according to the needs of result analysis. Logistic regression model is used to predict the trend of the filtered data and determine the predicted trend.

[0060] By comparing the predicted trend with the spatially consistent parameter summary data, the matching results are obtained, and the degree of correlation of the matching results is judged.

[0061] If the correlation of the matching results reaches the preset standard, the matching results and response feedback data will be integrated and stored to obtain the final business data archive.

[0062] In addition, this invention also discloses a real-time data analysis and fault early warning system based on digital power port transformer substations, including:

[0063] The data acquisition module is used to collect environmental parameter data from various components of power equipment in real time through a monitoring device network;

[0064] The data processing module is used to smooth and denoise the collected data to obtain a smoothed parameter sequence;

[0065] The storage management module is used to construct a dynamic storage framework based on the smoothed parameter sequence and spatiotemporal attribute information.

[0066] The anomaly detection module is used to extract historical parameter records of target components from the adjusted storage framework, perform multi-dimensional sorting and change rate calculation, and identify potential anomalies.

[0067] The early warning generation module is used to generate alarm sequences based on the rate of change index and the distribution boundary breakthrough, and to verify the accuracy by comparing with historical parameter records to obtain the early warning classification results;

[0068] The control interface module is used to push parameter summary information to the control system and obtain response feedback based on the early warning classification results.

[0069] Furthermore, the data acquisition module includes multiple sensor nodes arranged in various components of the transformer substation to collect environmental parameters such as temperature and humidity. The sampling frequency is once per minute, and the data is transmitted via a ZigBee wireless network.

[0070] Compared with the prior art, the present invention has the following beneficial effects:

[0071] This invention decomposes the real-time analysis and fault early warning process of transformer substation data into basic data processing units, solving the technical problems of data noise interference, low storage efficiency and anomaly detection delay in the prior art.

[0072] This invention employs a data smoothing method to denoise real-time collected environmental parameter data, obtaining a smoothed parameter sequence. This effectively eliminates random noise and interference signals, improving data quality and reliability, and providing an accurate foundation for subsequent analysis. Secondly, it constructs a dynamically adjustable storage framework by combining spatiotemporal attribute information. By grouping and prioritizing data through frequency distribution and anomaly ratio indicators, it achieves efficient classification and storage of environmental parameter data, solving the problem of low retrieval efficiency caused by chaotic data organization in existing technologies. This enables rapid location of historical parameter records of target components, supporting continuous trend tracking and anomaly tracing.

[0073] This invention accurately identifies potential anomaly locations and time information through multi-dimensional sorting and change rate index evaluation, and generates alarm output sequences by combining triggering conditions and risk assessment factors, thus ensuring the timeliness and accuracy of anomaly detection.

[0074] This invention verifies the accuracy of alarm output by comparing preliminary anomaly markers with historical parameter records, integrating influence weight analysis and anomaly ratio assessment, obtaining the final early warning classification result, and pushing parameter summary information including trend changes and spatial consistency to the control system to obtain response feedback records. This enables real-time monitoring and dynamic optimization of the operating status of power equipment, significantly improving the response speed, accuracy, and operation and maintenance efficiency of fault early warning, and ensuring the safe and stable operation of the power system. Attached Figure Description

[0075] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained from these drawings without creative effort.

[0076] Figure 1 This is the overall flowchart of the present invention.

[0077] Figure 2 This is a detailed flowchart of step 3 of the present invention. Detailed Implementation

[0078] In the following description, only certain exemplary embodiments are briefly described. As those skilled in the art will recognize, the described embodiments can be modified in various ways without departing from the spirit or scope of the embodiments of the invention. Therefore, the drawings and description are considered to be exemplary in nature and not restrictive.

[0079] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings.

[0080] Example 1: See Figure 1 and Figure 2 This embodiment discloses a method for real-time analysis and fault early warning based on digital power port transformer substation data, which specifically includes the following steps:

[0081] Step 1: Collect environmental parameter data from various components of the power equipment in real time through the monitoring device network, perform preliminary screening based on parameter deviation indicators and distribution balance, and use data smoothing methods to denoise the collected data to obtain a smoothed parameter sequence;

[0082] Step 2: Based on the smoothed parameter sequence and combined with spatiotemporal attribute information, analyze the characteristics of deviation persistence and fluctuation dynamic indicators, construct a data organization structure, group the environmental parameter data according to frequency distribution and anomaly ratio indicators, and determine the initial storage container;

[0083] Step 3: Obtain parameter data within the initial storage container, calculate spatial correlation indicators and consistency indicators. If the trend changes exceed preset conditions or the time correlation is lower than the standard, trigger dynamic adjustment of the initial storage container to obtain the adjusted storage framework.

[0084] Step 4: Extract historical parameter records of the target component from the adjusted storage framework, sort them in multiple dimensions based on peak comparison and benchmark comparison, and determine potential anomaly locations and time information;

[0085] Step 5: For potential anomaly locations, calculate the rate of change index and evaluate it in conjunction with triggering conditions and risk assessment factors. If the rate of change exceeds the condition and the distribution boundary is breached, generate an alarm output sequence to determine the preliminary anomaly marker.

[0086] Step 6: By comparing the initial anomaly markers with historical parameter records, and integrating the influence weight analysis and anomaly ratio assessment, the accuracy of the alarm output is verified, and the final early warning classification result is obtained.

[0087] Step 7: Based on the final early warning classification result, push parameter summary information including trend changes and spatial consistency to the control system and obtain response feedback records.

[0088] Step 1 is as follows:

[0089] Real-time environmental parameter data is acquired from various components of the power equipment through a monitoring device network and continuously recorded at a preset acquisition frequency to obtain an initial environmental parameter dataset. For this initial dataset, a data smoothing method is used to denoise the data, eliminating noise interference and obtaining a smoothed parameter sequence. Based on the smoothed parameter sequence, the specific deviations of the parameters are analyzed. If a parameter value exceeds a preset threshold range, it is marked as an abnormal parameter point, resulting in a marked abnormal dataset. From the marked abnormal dataset, feature information of the balanced distribution state is extracted to determine whether there is an uneven distribution, obtaining an evaluation result of the distribution state. Based on the evaluation result, if an uneven distribution is found, the weights of the abnormal parameter points are adjusted to determine the adjusted parameter distribution set. The adjusted parameter distribution set is obtained, and combined with the update frequency of real-time data, the changing trend of the parameter sequence is continuously monitored to determine whether there are potential abnormal fluctuations, obtaining the final monitoring and analysis result. Based on the final monitoring and analysis result, a corresponding parameter adjustment strategy is generated to dynamically optimize the operating status of the power equipment, resulting in an optimized operating parameter configuration.

[0090] In practice:

[0091] For example, in the scenario of monitoring environmental parameters of power equipment, suppose we are collecting and analyzing real-time data on the operating status of a transformer, focusing on key parameters such as temperature and humidity.

[0092] For real-time acquisition of environmental parameters, temperature and humidity data of various components of the transformer are continuously collected at a frequency of once per minute through a sensor network to form an initial dataset.

[0093] Suppose that the temperature data sequence collected over a certain period of time is 38.2, 39.1, 45.3, 38.9, and 40.2 degrees Celsius, in which there are obvious abnormal fluctuations.

[0094] In one possible implementation, a moving average method is used to smooth the initial dataset for data denoising.

[0095] Taking temperature data as an example, the average of three data points was used as the smoothing result, resulting in a smoothed sequence of 38.8, 41.1, and 41.5 degrees Celsius. This significantly reduced abrupt changes caused by noise interference and improved data stability. The advantage of this approach is that subsequent analysis can more accurately reflect the true trend and avoid misjudgments.

[0096] For example, in the parameter deviation analysis stage, assuming the preset threshold range for temperature is 35.0 to 42.0 degrees Celsius, a smoothed value of 41.5 exceeds the threshold and is therefore marked as an abnormal parameter point, forming an abnormal dataset. This step helps to quickly locate potential risk points and provides a basis for subsequent processing.

[0097] In one possible implementation, the distribution balance of the abnormal dataset is analyzed to extract the distribution characteristics of outliers. This reveals that temperature anomalies are concentrated within a specific time period, exhibiting an uneven distribution. Statistical analysis shows that these anomalies frequently occur during periods of high equipment load, reflecting potential heat dissipation problems. Such assessment results help to reveal deeper issues in equipment operation.

[0098] For example, in the weighting adjustment stage, for anomalies with uneven distribution, the weight of anomalies during high-load periods is increased, for example, by adjusting the weight from 1.0 to 1.5, to highlight their impact on the overall operating status and form an adjusted parameter distribution set. This can more accurately reflect equipment risks and improve the targeting of monitoring.

[0099] In one possible implementation, the temperature sequence is continuously monitored by combining the adjusted parameter distribution set and the real-time data update frequency. Assuming the temperature continues to fluctuate between 41.0 and 42.5 degrees Celsius, indicating a potential risk of abnormal fluctuations, the final monitoring and analysis results show that the performance of the heat dissipation system needs attention. This provides data support for subsequent optimization.

[0100] For example, when generating parameter adjustment strategies, based on monitoring results, it is recommended to increase the operating frequency of the transformer cooling fan from 30 minutes to 45 minutes per hour, while simultaneously reducing the load limit and dynamically optimizing the operating parameter configuration. This not only effectively reduces the risk of abnormal temperatures but also extends the service life of equipment and ensures the stable operation of the power system.

[0101] Further, step 2 is as follows: Using the smoothed parameter sequence and spatiotemporal attribute information, analyze the correlation pattern between the persistence of deviations and the dynamics of fluctuations. Preliminary grouping of the data is performed using preset classification rules to obtain a classified dataset. Based on the classified dataset, a logical framework for data organization is constructed based on the characteristics of frequency distribution and anomaly ratio. A preset threshold is used to filter the anomaly ratio, determining an anomalous data subset. The anomalous data subset is obtained, and combined with the dynamic trend, the fluctuation pattern of the parameter sequence in different time periods is analyzed. Statistical tools are used to quantify the fluctuation dynamics, obtaining a set of fluctuation features. Based on the fluctuation feature set, a corresponding storage unit allocation scheme is constructed for the frequency distribution pattern. If the anomaly ratio of a certain distribution pattern exceeds a preset threshold, the data under that pattern is prioritized, determining the marked data units. Based on the marked data units, combined with the mapping logic of attribute information and spatiotemporal attributes, the rationality of data classification is analyzed. The classification results are optimized and adjusted using the data organization framework to obtain an optimized classification structure. The optimized classification structure is obtained, and the distribution balance of parameter sequences under different classifications is analyzed based on the allocation of storage units. If the distribution balance under a certain classification is lower than the preset standard, the data under that classification is regrouped to determine the final storage allocation scheme. Based on the final storage allocation scheme, combined with the fluctuation feature set and the dynamic trend, the update status of the parameter sequences is continuously tracked, and automated tools are used to dynamically adjust the data classification and storage units to obtain the adjusted data management framework.

[0102] In practical implementation, for example, in the scenario of monitoring environmental parameters of power equipment, in-depth analysis is conducted on the operating data of transformers, and the data collection and processing background in historical information is combined with a focus on the smoothed parameter sequence and its subsequent analysis process.

[0103] By combining the smoothed parameter sequence with spatiotemporal attribute information to analyze the correlation patterns between the persistence and fluctuation dynamics of the deviation, the variation patterns of temperature data can be identified through the dual dimensions of time period and spatial location. Assuming that the temperature data of a transformer exhibits periodic fluctuations at different times of the day, and considering the differences in heat dissipation conditions at different geographical locations, the persistence characteristics of temperature deviation at specific times and component locations can be analyzed.

[0104] For example, when using preset classification rules to initially group data, the data can be divided into three categories based on temperature range: normal, slightly high, and abnormal. Assuming the normal range is 35.0 to 40.0 degrees Celsius, slightly high is 40.0 to 42.0 degrees Celsius, and abnormal is above 42.0 degrees Celsius, this rule can be used to classify the data within a day, resulting in a categorized dataset.

[0105] A data organization logic framework is constructed based on the frequency distribution and anomaly ratio characteristics. It is assumed that periods with an anomaly ratio exceeding 10% are marked as high-risk periods. A subset of abnormal data is filtered out using a preset threshold. For example, if the anomaly ratio reaches 15% in a certain period, it is classified as a key focus object.

[0106] For example, when analyzing the fluctuation patterns of outlier subsets of data, statistical tools can be used to quantify the fluctuation dynamics. Suppose that temperature data fluctuates frequently between 41.0 and 43.0 degrees Celsius over a certain period. Quantification reveals a fluctuation frequency of 5 times per hour, exceeding the normal range by 2 times, thus obtaining a set of fluctuation characteristics. A storage unit allocation scheme is then constructed based on the frequency distribution pattern. If the outlier proportion of a certain pattern exceeds a threshold of 15%, data in that pattern is prioritized. For example, data units from high-load periods are marked as priority 1 to ensure focused attention during subsequent analysis.

[0107] For example, by combining labeled data units and spatiotemporal attribute mapping logic to optimize the classification structure, the classification weights can be adjusted according to the location of equipment components. For instance, if the data of components near areas with weak heat dissipation are given higher attention, the optimized classification structure will be more in line with the actual operating environment.

[0108] The distribution balance of storage unit allocation is analyzed. If the data distribution is uneven within a certain category, for example, if the data volume of a certain category accounts for only 5%, which is lower than the preset standard of 10%, then the data is regrouped to ensure that the storage scheme is reasonable. For example, by continuously tracking the update status of parameter sequences through the final storage allocation scheme, assuming that temperature data remains consistently high during high-load periods, automated tools are used to dynamically adjust the categories and storage units, such as reassigning the data during that period to high-risk categories, forming an adjusted data management framework. This process ensures the real-time and targeted nature of data analysis, providing a reliable basis for subsequent equipment maintenance.

[0109] Furthermore, step 3 is detailed below:

[0110] Using the parameter data in the initial storage container, the fluctuation of trend changes is analyzed based on the calculation results of spatial correlation indicators and consistency indicators. If the trend changes exceed the preset threshold, the data in the storage container is prioritized and sorted to obtain a sorted data set.

[0111] Based on the sorted dataset, obtain the time-related evaluation results. If the time correlation is lower than the predetermined standard, the dataset is stratified to determine the data grouping after stratification.

[0112] After obtaining the hierarchical data groups, the dynamically adjusted allocation logic is analyzed based on the capacity limit of the storage container. If the data volume of a certain group exceeds the container limit, the group is split to obtain the split data units.

[0113] By splitting the data units and combining them with the spatially related distribution pattern, the update rules of the storage framework are constructed. The data units are repositioned using a preset allocation strategy to determine the updated storage layout.

[0114] The updated storage layout is obtained. Based on the stability requirements of consistency, the balance of data units in the storage framework is analyzed. If the balance does not meet the predetermined standard, a secondary adjustment mechanism is triggered to obtain the optimized storage structure.

[0115] Based on the optimized storage structure and dynamically adjusted execution records, the subsequent impact of trend changes is continuously tracked, and automated tools are used to monitor the storage framework in real time to determine the final storage configuration.

[0116] Based on the final storage configuration, a long-term data storage management solution is built to ensure the persistence of the adjustment results. Log recording tools are used to track changes to the storage framework and obtain a complete change profile.

[0117] In the scenario of monitoring environmental parameters of transformer substations, for the processing and analysis of transformer temperature data, the system first extracts parameter data from the initial storage container, such as obtaining the temperature data of a transformer substation in the past 24 hours. Data points are collected once per minute, totaling 1440 data points, with the temperature range between 40.0 and 85.0 degrees Celsius.

[0118] Next, spatial correlation indicators are calculated using an algorithm. The Pearson correlation coefficient method is employed to analyze the correlation between temperature data at different component locations. Assuming the calculation results show a correlation coefficient of 0.85 between component A and component B, higher than the preset threshold of 0.7, this indicates a strong spatial correlation between the two. Simultaneously, a consistency index is calculated, using standard deviation to analyze the stability of the temperature data. If the standard deviation is 2.5°C, lower than the preset standard of 3.0°C, the data consistency is considered high. Subsequently, the system detects trend changes, using a linear regression algorithm to predict the rate of temperature change. If the prediction shows an increase of 6.0°C per hour, exceeding the preset condition of 4.0°C per hour, the trend is deemed abnormal. Furthermore, temporal correlation is calculated, analyzing the temporal dependence between data points based on the autocorrelation function. If the correlation coefficient is 0.4, lower than the standard value of 0.6, it indicates insufficient temporal correlation.

[0119] Under the aforementioned conditions, the system automatically activates a dynamic adjustment mechanism for storage containers. Using preset rules, it reallocates abnormal trend data to high-priority storage units. For example, periods where 20% of data points exceed 75.0°C are marked as abnormal periods, stored separately, and the storage framework is adjusted. This period's data is then associated with and archived with similar historical abnormal data, forming a new storage structure. To ensure logical integrity, transformer load rate data is also used as an auxiliary business correlation. If the load rate exceeds 80% of the rated value, the storage priority of the data for that period is further increased to ensure the comprehensiveness of subsequent analysis.

[0120] Furthermore, step 4 is detailed below:

[0121] By adjusting the storage framework, relevant historical parameter data of the target component is obtained, and the data is initially classified using preset filtering rules to obtain a set of classified parameters.

[0122] Based on the classified parameter set, the peak difference and the baseline value are compared. If the peak difference exceeds the preset threshold, the data with peak differences exceeding the threshold are marked to determine the marked abnormal dataset.

[0123] Starting from the labeled abnormal dataset, and combining the logic of multidimensional sorting, the data is prioritized and sorted to obtain the sorted data sequence;

[0124] For the sorted data sequence, analyze the specific distribution of abnormal locations, use a preset positioning tool to accurately identify the abnormal locations, and determine the specific abnormal coordinate points;

[0125] By combining abnormal coordinate points with the recorded data of time nodes, the time correlation information in historical parameters can be traced to obtain the corresponding time period distribution.

[0126] Based on the time period distribution and the results of the difference analysis, a correspondence between abnormal locations and time nodes is constructed to determine the final abnormal distribution pattern.

[0127] Obtain the final anomaly distribution pattern, combine it with the adjustment structure of the storage framework, generate parameter optimization strategies for the target components, and determine the direction of subsequent processing.

[0128] In the scenario of monitoring environmental parameters of power equipment, for the analysis of the operating status of key components of generators, the historical parameter records of the target components are first extracted from the adjusted storage framework. For example, the temperature data of a generator spindle over the past 7 days is obtained, collected once per hour, for a total of 168 data points, with the temperature range between 40.5 and 75.8 degrees Celsius.

[0129] Peak values ​​were compared using these data, and a maximum value detection algorithm was used to filter out temperature peak data. Assuming that the highest temperature was found to be 75.8 degrees Celsius, which occurred in the 12th hour of the 5th day, it was compared with historical peak records. If the historical average peak value was 70.0 degrees Celsius, then the current peak value exceeded the normal range by 5.8 degrees Celsius and was marked as a potential anomaly.

[0130] A benchmark comparison analysis was conducted, using the mean plus standard deviation method to set the baseline. The calculation result was a baseline temperature of 55.0 degrees Celsius plus a standard deviation of 8.0 degrees Celsius, resulting in an upper limit of 63.0 degrees Celsius. However, 30 data points in the current data exceeded this upper limit, accounting for approximately 17.9%, indicating abnormal fluctuations.

[0131] Using a multi-dimensional sorting algorithm, temperature data is comprehensively sorted according to three dimensions: time, numerical value, and deviation from the baseline. A weighted scoring method is used, with a weight of 0.3 for the time dimension, 0.5 for the numerical value dimension, and 0.2 for the deviation dimension. The calculation shows that the period of concentrated abnormality is from day 4 to day 5. Among them, the comprehensive score of the 12th hour of day 5 is 0.92, which is much higher than the threshold of 0.6, and is identified as a key abnormal time point.

[0132] By combining generator speed data with business correlation, if the speed exceeds 90% of the rated value during a certain period, for example, reaching 3200 revolutions per minute while the rated value is 3000 revolutions per minute, then the abnormal importance of that time point is further confirmed.

[0133] Based on spatial distribution analysis, potential anomaly locations were identified. By comparing historical data, it was found that the proportion of high data from the temperature sensor at the front end of the spindle was 25%, higher than the 10% in other locations. It was inferred that the anomaly might be concentrated at the front end of the spindle, forming a complete analysis chain.

[0134] Furthermore, step 5 is detailed below:

[0135] For monitoring the rate of change, the rate change data of the target object is obtained from historical data records and compared with a preset threshold.

[0136] If the rate change exceeds the threshold range, an initial abnormal signal is generated to identify the initial point of interest.

[0137] Based on the initial anomaly signal, data is extracted from the anomaly location, and the location information is refined using a preset positioning tool to obtain the specific anomaly coordinate distribution.

[0138] By analyzing the abnormal coordinate distribution and combining it with the boundary data of the distribution range, we can determine whether there is a significant offset.

[0139] If an offset phenomenon is detected, a corresponding offset identifier is generated to determine the abnormal distribution characteristics;

[0140] For the offset identifier, relevant trigger condition data is obtained, and combined with the preset rules of risk assessment for comprehensive processing to obtain the risk level classification result;

[0141] Based on the risk level classification results, a corresponding alarm sequence is generated, and the alarms are sorted according to a preset priority rule to determine the output order of the alarms;

[0142] By combining the output sequence of the alarms with the recorded data of the initial markings, the abnormal locations are confirmed a second time to obtain the final set of abnormal markings;

[0143] For the final set of anomaly labels, a logistic regression model is used to analyze the correlation between the anomaly labels, identify the potential relationships between them, and determine the key directions for subsequent processing.

[0144] In the scenario of monitoring environmental parameters of power equipment, an in-depth analysis was conducted on the potential abnormal locations of key components of the transformer substation. First, the focus was on the abnormal areas identified in the transformer windings, and the temperature change rate index was calculated. Specifically, temperature data was collected once per minute over the past 5 days, totaling 7200 data points, with the temperature range between 40.0 and 85.0 degrees Celsius. The average change rate per hour was calculated using a sliding window algorithm with a window size of 60 minutes. It was found that the change rate at the 15th hour of the 3rd day was 0.35 degrees Celsius per hour, which was significantly higher than the historical average change rate threshold of 0.15 degrees Celsius per hour.

[0145] The evaluation was conducted in conjunction with the triggering conditions, and the rate of change threshold was set at 0.2 degrees Celsius per hour. At the same time, it was analyzed whether the distribution boundary was breached. The 95th quantile of temperature was calculated using the statistical quantile method, which is 75.0 degrees Celsius. However, 12% of the data points in the current data exceeded this boundary, indicating an abnormal distribution.

[0146] Further risk assessment factors were introduced, historical fault databases were accessed, and a correlation model between temperature and equipment aging was used to calculate the risk index corresponding to the current rate of temperature change as 0.78, which is higher than the safety threshold of 0.5. At the same time, related business data such as cooling fan speed were used to find that the speed value dropped to 1200 rpm during this period, which is lower than the normal range of 1500 rpm, thus enhancing the reliability of the risk assessment.

[0147] If the rate of change exceeds the condition and the distribution boundary is breached, an alarm output sequence is automatically generated, including information such as timestamp, 15th hour of day 3, peak temperature of 85.0 degrees Celsius, and risk index of 0.78. The alarm is then marked as high priority using a priority sorting algorithm to determine the preliminary anomaly marker and push it to the subsequent analysis module.

[0148] Furthermore, step 6 is detailed below:

[0149] By comparing the initial labels with historical parameters, the difference information between the labeled data and the parameter records is obtained. A preset threshold range is used for preliminary screening to obtain a set of labels with significant differences.

[0150] Based on the significantly different set of labels, data is extracted for the anomaly ratio, and weighted calculation is performed in combination with the preset rules that affect the weight to determine the weight value of the anomaly ratio.

[0151] By using the weighted value of the abnormal ratio, data records related to alarm output are obtained. If the weighted value exceeds the preset threshold range, a corresponding alarm signal is generated, and the priority of the alarm signal is determined.

[0152] Based on the priority of alarm signals and combined with the logical rules of proportional evaluation, the signal data is processed in layers to obtain the layered signal classification.

[0153] By classifying the signals after stratification, and based on the data requirements for early warning classification, a logistic regression model is used to map the classified data to determine the preliminary results of early warning classification.

[0154] Based on the preliminary results of the early warning classification and combined with the processing rules for accurate verification, the classification data is compared a second time to obtain the final output results.

[0155] Based on the final output, the business logic for grading is used to match the result data with the preset grading standards to determine the final warning grading status.

[0156] In the scenario of monitoring environmental parameters of power equipment, for the detection of anomalies in key parts of generators, the preliminary anomaly markers are first compared with historical parameter records. Specifically, temperature data records of similar equipment over the past 30 days are extracted, totaling approximately 43,200 data points, with temperatures ranging from 45.2 to 78.9 degrees Celsius. Using a time series comparison algorithm, the deviation between the temperature fluctuation value corresponding to the current anomaly marker and the historical average is calculated. It is found that the peak temperature during the current marked period is 85.0 degrees Celsius, exceeding the historical average of 70.0 degrees Celsius by 15.0 units, with a deviation rate of 21.4%, which is higher than the set deviation threshold of 15% and 8%.

[0157] The combined impact weight analysis adopted a multi-factor weighted model, taking temperature fluctuation, runtime and load rate as the main influencing factors, and assigning weights of 0.5, 0.3 and 0.2 respectively. The weighted calculation yielded a comprehensive impact index of 0.82, which is higher than the safety threshold of 0.6, indicating that the current anomaly has a high degree of impact.

[0158] An anomaly rate assessment was conducted, and the frequency of similar temperature anomalies over the past 7 days was statistically analyzed. The anomaly rate was found to be 9.5%, which is higher than the normal range of 5%. Based on the correlation model between the anomaly rate and failure rate in historical data, the probability of the current anomaly developing into a failure was predicted to be 0.67, which exceeds the warning value of 0.4.

[0159] To further verify the accuracy of the alarm output, the coolant flow record in the related business data was checked, and it was found that the current flow rate was 3.1 liters per minute, which was lower than the standard value of 4.0 liters per minute, confirming the potential risk of abnormal temperature.

[0160] Based on the above analysis results, the system uses a tiered algorithm to classify the early warning result as a Level 2 early warning, generating an early warning report containing a timestamp, a peak temperature of 76.3 degrees Celsius, a comprehensive impact index of 0.82, and a failure probability of 0.67. This report is automatically pushed to the monitoring platform, forming a complete logical chain from comparison and analysis to tiering. The logistic regression model is trained using historical operation and maintenance data. Feature engineering includes: temperature deviation (difference between current value and historical mean), rate of change (calculated using a sliding window), load rate (ratio of current load to rated load), and environmental temperature compensation. Labels represent the actual failure status (0-normal, 1-abnormal, 2-failure). The training samples are based on 2000 historical operation records, covering normal, abnormal, and failure states. Five-fold cross-validation ensures that the model accuracy is not less than 80%, and the model is retrained quarterly using the latest data to maintain its effectiveness.

[0161] Furthermore, step 7 is detailed below:

[0162] Based on the final early warning classification results, a summary of parameter data containing trend changes and spatial consistency is compiled and sent to the control system to obtain the interaction log after transmission.

[0163] Based on the interaction logs after sending, the response feedback data returned by the control system is extracted, and the data is parsed using preset field rules to determine the parsed feedback content;

[0164] Based on the parsed feedback content, the key fields in the feedback record are classified and processed to obtain the classified data groups and determine the integrity of the grouped data.

[0165] If the integrity of the grouped data meets the preset threshold range, the classified data groups are prioritized and sorted to obtain the sorted feedback sequence.

[0166] Based on the sorted feedback sequence, data is filtered according to the needs of result analysis. Logistic regression model is used to predict the trend of the filtered data and determine the predicted trend.

[0167] By comparing the predicted trend with the spatially consistent parameter summary data, the matching results are obtained, and the degree of correlation of the matching results is judged.

[0168] If the correlation of the matching results reaches the preset standard, the matching results and response feedback data will be integrated and stored to obtain the final business data archive.

[0169] The early warning classification thresholds are determined based on the safety standards for transformer substation operation and historical fault statistical analysis.

[0170] Level 1 Warning (Low Risk): Trend change rate 20%-35%, spatial consistency 0.6-0.8;

[0171] Level 2 Warning (Medium Risk): Trend change rate 35%-50%, spatial consistency 0.4-0.6;

[0172] Level 3 warning (high risk): Trend change rate > 50%, spatial consistency < 0.4;

[0173] The above thresholds were determined through optimization using 500 historical fault backtracking tests, with the false alarm rate controlled within 5%.

[0174] In the process of pushing and recording the results of early warning classification, the data analysis system extracts parameters from the early warning classification results. Assuming that the current early warning classification is level three, the trend change parameter shows that the risk index has increased from 2.5 to 3.8 in the past 24 hours, an increase of 52%, and the calculation formula is (3.8-2.5) / 2.5*100%. At the same time, the spatial consistency parameter shows that the coverage of the affected area has expanded from 30% to 45%. The consistency index calculated by the spatial analysis algorithm is 0.75, indicating that the regional risk distribution is relatively concentrated.

[0175] The above parameters are summarized into structured information, including a trend change rate of 52% and a spatial consistency index of 0.75. Push messages are automatically generated and sent to the control system in JSON format via the API interface. An example message is {"level":3,"trend":52, "spatial":0.75}. The push process records timestamps and sending status to ensure traceability.

[0176] After receiving the message, the parameters are automatically parsed based on preset rules. If the trend change rate is greater than 50% and the consistency index is greater than 0.7, the secondary response mechanism is triggered to generate a response instruction, such as adjusting the monitoring frequency to once per hour. The response record is stored in the database. The record content includes the response level, trigger time and execution status. Assume that the record ID is 20231001001 and the time is 2023-10-01 14:30:00.

[0177] The system queries feedback records every 30 minutes via a scheduled task to analyze the response execution rate. Assuming the current execution rate is 95%, if it falls below 98%, it automatically generates optimization suggestions, such as increasing the resource allocation ratio to 20%. The analysis results are then pushed to the management platform, forming a closed-loop management logic to ensure efficient connection between early warning and response.

[0178] Example 2: This example discloses a real-time data analysis and fault early warning system based on digital power port transformer substations, including a data acquisition module, a data processing module, a storage management module, an anomaly detection module, an early warning generation module, and a control interface module.

[0179] In practice:

[0180] The data acquisition module consists of temperature and humidity sensors deployed on various components of the transformer substation. It collects environmental parameter data in real time via a ZigBee wireless network, sampling once per minute. For example, inside the transformer enclosure, the sensor nodes collect temperature data ranging from -10℃ to 100℃ with an accuracy of ±0.5℃, and humidity data ranging from 0% to 100% with an accuracy of ±2%. The sensor network covers key components of the transformer substation, such as transformer windings, switchgear, and cable joints, ensuring data comprehensiveness.

[0181] The data processing module is embedded in the edge computing gateway and uses a moving average algorithm to smooth the collected raw data. The moving average window size is set to 5 data points (corresponding to 5 minutes) based on the box-type thermal constant, ensuring that random noise (such as electromagnetic interference) is effectively filtered out while retaining the true temperature trend changes. For example, for the temperature data sequence [38.2, 39.1, 45.3, 38.9, 40.2]°C, a 5-point moving average is taken to obtain the smoothed sequence [38.8, 41.1, 41.5]°C, effectively eliminating abrupt changes. The smoothed parameter sequence is uploaded to the cloud platform through the gateway, reducing network transmission load.

[0182] The storage management module is built on a cloud platform and uses the time-series database InfluxDB to store smoothed parameter sequences. It automatically partitions the data based on spatiotemporal attributes (such as device location coordinates and data timestamps) and sets an anomaly percentage threshold (e.g., 10%). When the anomaly percentage exceeds the threshold, the storage partitioning strategy is automatically adjusted, storing the abnormal data in high-priority partitions to improve retrieval efficiency.

[0183] The anomaly detection module runs on the server side, extracting historical parameter records of the target component from the storage management module and using multi-dimensional sorting algorithms (such as weighted sorting based on peak value, baseline deviation, and time) to identify potential anomaly locations. For example, for temperature data, the rate of change is calculated; if the rate of change exceeds 0.2℃ / minute and the distribution boundary (such as the historical 95th percentile) is breached, it is marked as a preliminary anomaly.

[0184] The early warning generation module receives the output from the anomaly detection module and uses a pre-trained logistic regression model to validate and classify anomaly markers. Model input features include: temperature deviation value, rate of change, load rate, and spatial consistency index; the output is a three-level early warning probability. Decision rules: a level three early warning is triggered when the high-risk probability > 0.7, a level two early warning when the medium-risk probability > 0.6, and a level one early warning when the low-risk probability > 0.5. The model is automatically updated quarterly, retrained using the most recent three months of running data. For example, based on influence weights (e.g., temperature deviation weight 0.6, humidity deviation weight 0.4), a comprehensive risk index is calculated, classifying early warnings into three levels: Level One (low risk), Level Two (medium risk), and Level Three (high risk). Alarm sequences are prioritized and then pushed to the control interface module.

[0185] The control interface module integrates with the upper-level control system (such as a power distribution automation system) via a RESTful API. The pushed parameter summary information includes trend change rate, spatial consistency index, and warning level, in JSON format. For example, a push message might be: {"warning_level": 2, "trend_change": 15%, "spatial_consistency": 0.8}. The control system returns response feedback, such as adjusting cooling system operating parameters, and records feedback logs for closed-loop optimization.

[0186] This embodiment, through modular design, achieves real-time data analysis, fault early warning, and automatic control response for transformer substations. The data flow between system modules is as follows: the data acquisition module sends raw environmental parameters to the edge computing gateway (data processing module) via a ZigBee network; the data processing module performs smoothing and noise reduction, then uploads the parameter sequence to the cloud platform storage management module; the storage management module dynamically partitions and stores data, and provides historical records to the anomaly detection module; the anomaly detection module outputs preliminary anomaly markers to the early warning generation module; after verification and classification, the early warning generation module pushes the early warning results to the control system through the control interface module. Feedback from the control system (such as adjusting cooling system parameters) is parsed by the control interface module and stored in the business database to optimize subsequent early warning strategies. This solves the control lag problem caused by data noise and latency, improving the timeliness, reliability, and automation level of transformer substation control.

[0187] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0188] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention. It should be noted that any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A method for real-time data analysis and fault early warning of digital power port transformer substations, characterized in that: Specifically, the steps include the following: Step 1: Collect environmental parameter data from various components of the power equipment in real time through the monitoring device network, and use a data smoothing method to denoise the collected data to obtain a smoothed parameter sequence; perform preliminary screening on the smoothed parameter sequence based on parameter deviation index and distribution equilibrium state to obtain a screened parameter sequence. Step 2: Based on the smoothed parameter sequence and combined with spatiotemporal attribute information, analyze the characteristics of deviation persistence and fluctuation dynamic indicators, construct a data organization structure, group the environmental parameter data according to frequency distribution and anomaly ratio indicators, and determine the initial storage container; Step 3: Obtain parameter data within the initial storage container, calculate spatial correlation indicators and consistency indicators. If the trend changes exceed preset conditions and the time correlation is lower than the standard, trigger dynamic adjustment of the initial storage container to obtain the adjusted storage framework. Step 4: Extract historical parameter records of the target component from the adjusted storage framework, sort them in multiple dimensions based on peak comparison and benchmark comparison, and determine the location and time information of potential anomalies; Step 5: For potential anomaly locations, calculate the rate of change index and evaluate it in conjunction with triggering conditions and risk assessment factors. If the rate of change exceeds the condition and the distribution boundary is breached, generate an alarm output sequence to determine the preliminary anomaly marker. Step 6: By comparing the initial anomaly markers with historical parameter records, and integrating the influence weight analysis and anomaly ratio assessment, the accuracy of the alarm output is verified, and the final early warning classification result is obtained. Step 7: Based on the final early warning classification result, push parameter summary information including trend changes and spatial consistency to the control system and obtain response feedback records.

2. The method for real-time analysis and fault early warning of digital power port transformer substation data according to claim 1, characterized in that, Step 1 is as follows: Real-time environmental parameter data is acquired from various components of power equipment through a monitoring device network and continuously recorded at a preset acquisition frequency to obtain an initial environmental parameter dataset. For this initial dataset, a data smoothing method is used to denoise the data, eliminating noise interference and obtaining a smoothed parameter sequence. Based on the smoothed parameter sequence, the specific deviations of the parameters are analyzed. If a parameter value exceeds a preset threshold range, it is marked as an abnormal parameter point, resulting in a marked abnormal dataset. From the marked abnormal dataset, feature information of the balanced distribution state is extracted to determine whether there is an uneven distribution phenomenon, obtaining an evaluation result of the distribution state. Based on the evaluation result of the distribution state, if an uneven distribution phenomenon is found, the weights of the abnormal parameter points are adjusted to determine the adjusted parameter distribution set. The adjusted parameter distribution set is obtained, and combined with the update frequency of real-time data, the changing trend of the parameter sequence is continuously monitored to determine whether there are potential abnormal fluctuations, obtaining the final monitoring and analysis result. Based on the final monitoring and analysis results, corresponding parameter adjustment strategies are generated to dynamically optimize the operating status of power equipment and obtain optimized operating parameter configurations.

3. The method for real-time analysis and fault early warning of digital power port transformer substation data according to claim 1, characterized in that, Step 2 is as follows: Using the smoothed parameter sequence and spatiotemporal attribute information, analyze the correlation pattern between the persistence of deviations and the dynamics of fluctuations. Preliminary grouping of the data is performed using preset classification rules to obtain a classified dataset. Based on the classified dataset, a logical framework for data organization is constructed based on the characteristics of frequency distribution and anomaly ratio. A preset threshold is used to filter the anomaly ratio, determining an anomalous data subset. The anomalous data subset is then analyzed, and the fluctuation pattern of the parameter sequence over different time periods is analyzed, using statistical tools to quantify the fluctuation dynamics, resulting in a fluctuation feature set. Based on the fluctuation feature set, a corresponding storage unit allocation scheme is constructed for the frequency distribution pattern. If the anomaly ratio of a certain distribution pattern exceeds a preset threshold, the data under that pattern is prioritized, determining the marked data units. Based on the marked data units, and combining attribute information and the mapping logic of spatiotemporal attributes, the rationality of the data classification is analyzed. The data organization framework is then used to optimize and adjust the classification results, resulting in an optimized classification structure. The optimized classification structure is obtained. Based on the allocation of storage units, the distribution balance of parameter sequences under different classifications is analyzed. If the distribution balance under a certain classification is lower than the preset standard, the data under that classification is regrouped to determine the final storage allocation scheme. By combining the final storage allocation scheme with the set of fluctuation characteristics and the dynamic trend, continuously tracking the update status of the parameter sequence, and using automated tools to dynamically adjust the data classification and storage units, the adjusted data management framework is obtained.

4. The method for real-time analysis and fault early warning of digital power port transformer substation data according to claim 1, characterized in that, Step 3 is as follows: Using the parameter data in the initial storage container, the fluctuation of trend changes is analyzed based on the calculation results of spatial correlation indicators and consistency indicators. If the trend changes exceed the preset threshold, the data in the storage container is prioritized and sorted to obtain a sorted data set. Based on the sorted dataset, obtain the time-related evaluation results. If the time correlation is lower than the predetermined standard, the dataset is stratified to determine the data grouping after stratification. After obtaining the hierarchical data groups, the dynamically adjusted allocation logic is analyzed based on the capacity limit of the storage container. If the data volume of a certain group exceeds the container limit, the group is split to obtain the split data units. By splitting the data units and combining them with the spatially related distribution pattern, the update rules of the storage framework are constructed. The data units are repositioned using a preset allocation strategy to determine the updated storage layout. The updated storage layout is obtained. Based on the stability requirements of consistency, the balance of data units in the storage framework is analyzed. If the balance does not meet the predetermined standard, a secondary adjustment mechanism is triggered to obtain the optimized storage structure. Based on the optimized storage structure and dynamically adjusted execution records, the subsequent impact of trend changes is continuously tracked, and automated tools are used to monitor the storage framework in real time to determine the final storage configuration. Based on the final storage configuration, a long-term data storage management solution is built to ensure the persistence of the adjustment results. Log recording tools are used to track changes to the storage framework and obtain a complete change profile.

5. The method for real-time analysis and fault early warning of digital power port transformer substation data according to claim 1, characterized in that, Step 4 is as follows: By adjusting the storage framework, relevant historical parameter data of the target component is obtained, and the data is initially classified using preset filtering rules to obtain a set of classified parameters. Based on the classified parameter set, the peak difference and the baseline value are compared. If the peak difference exceeds the preset threshold, the data with peak differences exceeding the threshold are marked to determine the marked abnormal dataset. Starting from the labeled abnormal dataset, and combining the logic of multidimensional sorting, the data is prioritized and sorted to obtain the sorted data sequence; For the sorted data sequence, analyze the specific distribution of abnormal locations, use a preset positioning tool to accurately identify the abnormal locations, and determine the specific abnormal coordinate points; By combining abnormal coordinate points with the recorded data of time nodes, the time correlation information in historical parameters can be traced to obtain the corresponding time period distribution. Based on the time period distribution and the results of the difference analysis, a correspondence between abnormal locations and time nodes is constructed to determine the final abnormal distribution pattern. Obtain the final anomaly distribution pattern, combine it with the adjustment structure of the storage framework, generate parameter optimization strategies for the target components, and determine the direction of subsequent processing.

6. The method for real-time analysis and fault early warning of digital power port transformer substation data according to claim 1, characterized in that, Step 5 is as follows: For monitoring the rate of change, the rate change data of the target object is obtained from historical data records and compared with a preset threshold. If the rate change exceeds the threshold range, an initial abnormal signal is generated to identify the initial point of interest. Based on the initial anomaly signal, data is extracted from the anomaly location, and the location information is refined using a preset positioning tool to obtain the specific anomaly coordinate distribution. By analyzing the abnormal coordinate distribution and combining it with the boundary data of the distribution range, we can determine whether there is a significant offset. If an offset phenomenon is detected, a corresponding offset identifier is generated to determine the abnormal distribution characteristics; For the offset identifier, relevant trigger condition data is obtained, and combined with the preset rules of risk assessment for comprehensive processing to obtain the risk level classification result; Based on the risk level classification results, a corresponding alarm sequence is generated, and the alarms are sorted according to a preset priority rule to determine the output order of the alarms; By combining the output sequence of the alarms with the recorded data of the initial markings, the abnormal locations are confirmed a second time to obtain the final set of abnormal markings; For the final set of anomaly labels, a logistic regression model is used to analyze the correlation between the anomaly labels, identify the potential relationships between them, and determine the key directions for subsequent processing.

7. The method for real-time analysis and fault early warning of digital power port transformer substation data according to claim 1, characterized in that, Step 6 is as follows: By comparing the initial labels with historical parameters, the difference information between the labeled data and the parameter records is obtained. A preset threshold range is used for preliminary screening to obtain a set of labels with significant differences. Based on the significantly different set of labels, data is extracted for the anomaly ratio, and weighted calculation is performed in combination with the preset rules that affect the weight to determine the weight value of the anomaly ratio. By using the weighted value of the abnormal ratio, data records related to alarm output are obtained. If the weighted value exceeds the preset threshold range, a corresponding alarm signal is generated, and the priority of the alarm signal is determined. Based on the priority of alarm signals and combined with the logical rules of proportional evaluation, the signal data is processed in layers to obtain the layered signal classification. By classifying the signals after stratification, and based on the data requirements for early warning classification, a logistic regression model is used to map the classified data to determine the preliminary results of early warning classification. Based on the preliminary results of the early warning classification and combined with the processing rules for accurate verification, the classification data is compared a second time to obtain the final output results. Based on the final output, the business logic for grading is used to match the result data with the preset grading standards to determine the final warning grading status.

8. The method for real-time analysis and fault early warning of digital power port transformer substation data according to claim 1, characterized in that, Step 7 is as follows: Based on the final early warning classification results, a summary of parameter data containing trend changes and spatial consistency is compiled and sent to the control system to obtain the interaction log after transmission. Based on the interaction logs after sending, the response feedback data returned by the control system is extracted, and the data is parsed using preset field rules to determine the parsed feedback content; Based on the parsed feedback content, the key fields in the feedback record are classified and processed to obtain the classified data groups and determine the integrity of the grouped data. If the integrity of the grouped data meets the preset threshold range, the classified data groups are prioritized and sorted to obtain the sorted feedback sequence. Based on the sorted feedback sequence, data is filtered according to the needs of result analysis. Logistic regression model is used to predict the trend of the filtered data and determine the predicted trend. By comparing the predicted trend with the spatially consistent parameter summary data, the matching results are obtained, and the degree of correlation of the matching results is judged. If the correlation of the matching results reaches the preset standard, the matching results and response feedback data will be integrated and stored to obtain the final business data archive.

9. A real-time data analysis and fault early warning system based on digital power port transformer substation data, used to execute the real-time data analysis and fault early warning method based on digital power port transformer substation data as described in any one of claims 1-8; characterized in that, include: The data acquisition module is used to collect environmental parameter data from various components of power equipment in real time through a monitoring device network; The data processing module is used to smooth and denoise the collected data to obtain a smoothed parameter sequence; The storage management module is used to construct a dynamic storage framework based on the smoothed parameter sequence and spatiotemporal attribute information. The anomaly detection module is used to extract historical parameter records of target components from the adjusted storage framework, perform multi-dimensional sorting and change rate calculation, and identify potential anomalies. The early warning generation module is used to generate alarm sequences based on the rate of change index and the distribution boundary breakthrough, and to verify the accuracy by comparing with historical parameter records to obtain the early warning classification results; The control interface module is used to push parameter summary information to the control system and obtain response feedback based on the early warning classification results.

10. The real-time data analysis and fault early warning system based on digital power port transformer substations according to claim 9, characterized in that: The data acquisition module includes multiple sensor nodes, which are arranged in various components of the transformer substation to collect environmental parameters such as temperature and humidity. The sampling frequency is once per minute, and the data is transmitted through a ZigBee wireless network.

Images