Power quality data correlation analysis method and system based on intelligent fusion terminal

Abstract

The application is suitable for the technical field of electric power data analysis, and particularly relates to a power quality data correlation analysis method and system based on intelligent fusion terminals. The method comprises receiving power quality time series data uploaded by each intelligent fusion terminal in a transformer area, and constructing a power quality data set based on a preset sequence; for any type of disturbance event, performing feature extraction on the power quality data set, extracting collection error features, and constructing a terminal error feature set; performing data regularization on the to-be-processed data based on the terminal error feature set; performing correlation analysis on the regularized to-be-processed data, and outputting a transformer area state; the application statistically analyzes conventional errors generated in the collection process of the intelligent fusion terminal by constructing a terminal error feature set, corrects the correlation between disturbance events according to the terminal error feature set, and further optimizes the data quality of the collected data, so as to obtain more accurate correlation analysis results.

Description

Technical Field

[0001] This invention relates to the field of power data analysis technology, and in particular to a method and system for power quality data correlation analysis based on intelligent fusion terminals. Background Technology

[0002] Intelligent fusion terminals are crucial edge devices for online power quality sensing in low-voltage distribution areas. They continuously collect operational data such as voltage, current, harmonics, and power factor, providing data support for power quality analysis, anomaly early warning, and disturbance source tracing. However, due to their long-term deployment in complex field environments, intelligent fusion terminals are susceptible to factors such as temperature and humidity changes, electromagnetic interference, device aging, sampling accuracy deviations, and terminal clock asynchrony. This results in errors such as data drift, time offset, and instantaneous noise in the collected data. These errors can be mixed with real power grid operational data, affecting the accuracy of subsequent power quality analysis results. Therefore, the technical problem this invention aims to solve is how to accurately distinguish between data correlations caused by terminal acquisition errors and data correlations caused by real power grid disturbances, thus preventing false correlations from participating in power quality analysis and disturbance source tracing. Summary of the Invention

[0003] The purpose of this invention is to provide a method and system for power quality data correlation analysis based on intelligent fusion terminals, so as to solve the problem of "how to provide continuous navigation based on the user's movement status" mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides the following technical solution:

[0005] A method for power quality data correlation analysis based on a smart fusion terminal, the method comprising:

[0006] Receive power quality time-series data uploaded by various smart converged terminals in the distribution area, and construct a power quality dataset based on a preset order;

[0007] For any type of disturbance event, feature extraction is performed on the power quality dataset to extract acquisition error features and construct a terminal error feature set; the terminal error feature set includes event items and feature items.

[0008] The system generates correlation analysis instructions periodically, extracts data to be processed from the power quality dataset, queries the disturbance events between the current instruction and the previous instruction, and performs data normalization on the data to be processed based on the terminal error feature set.

[0009] The normalized data to be processed is subjected to correlation analysis, and the status of the transformer area is output. The correlation analysis process includes theoretical correlation process and actual correlation process.

[0010] As a further aspect of the present invention, the step of receiving power quality time-series data uploaded by various intelligent converged terminals in the receiving area and constructing a power quality dataset based on a preset order includes:

[0011] Establish connection channels with each intelligent converged terminal in the transformer area, and extract the sampling timestamp, terminal number, phase identifier and environmental parameters corresponding to each connection channel as terminal attribute data;

[0012] The voltage, current, harmonic, voltage fluctuation, and power factor data uploaded by various intelligent converged terminals in the receiving area are used as the collected data.

[0013] By statistically analyzing different types of collected data within the same time series, a data sequence containing type labels is obtained;

[0014] The amplitude of data fluctuation, frequency of data loss, and frequency of data mutation of each data sequence of each intelligent fusion terminal within a continuous time window are obtained as data quality parameters.

[0015] Based on the preset type label sequence, the terminal attribute data, the data sequence containing type labels and its data quality parameters are statistically analyzed to obtain the power quality dataset.

[0016] As a further aspect of the present invention, the step of extracting features from the power quality dataset for any type of disturbance event, extracting acquisition error features, and constructing a terminal error feature set includes:

[0017] Establish a connection channel with the historical database, and extract the data sequence from the power quality dataset corresponding to each disturbance event based on a preset time span;

[0018] For each disturbance event's data sequence, calculate the data change rate; the data change rate includes at least the voltage change rate, current change rate, and harmonic change rate.

[0019] The phase difference is calculated by determining the time point at which the rate of change of the location data reaches a preset rate of change threshold, combined with the time of the disturbance.

[0020] Under this disturbance event, the phase difference, the rate of change including type label, and the original data of all data sequences of all terminals are statistically analyzed and used as the error feature group of the disturbance event; the data in the error feature group includes terminal label, type label, phase difference, rate of change, and original data;

[0021] The disturbance events are classified, and for each type of disturbance event, the error feature groups of each disturbance event are compared to fit a comprehensive feature group for each type of disturbance event. The data in the comprehensive feature group includes type label, phase difference, rate of change range and original data range.

[0022] As a further aspect of the present invention, the steps of generating correlation analysis instructions at regular intervals, extracting data to be processed from the power quality dataset, querying disturbance events between the current instruction and the previous instruction, and performing data normalization on the data to be processed based on the terminal error feature set include:

[0023] Generate correlation analysis commands periodically, query the generation time of the current command and the generation time of the previous command, and determine the time interval;

[0024] A subset of data within a time interval is extracted from the power quality dataset and used as the data to be processed. The extraction process includes preserving terminal attribute data, extracting the data sequence, and extracting the data quality parameters.

[0025] Query disturbance events within a time interval and match them with the corresponding comprehensive feature groups;

[0026] Based on the comprehensive feature set, outlier data is located in the data to be processed. Based on the non-outlier data, a data function is fitted to fit the location of the outlier data, and the data to be processed is obtained after data normalization.

[0027] As a further aspect of the present invention, the step of performing correlation analysis on the normalized data to be processed and outputting the status of the transformer area includes:

[0028] Pair each intelligent converged terminal with another;

[0029] For two paired terminals, obtain their theoretical physical relationship and determine the theoretical correlation between the terminals;

[0030] For two paired terminals, the correlation of each data type is calculated based on the power quality dataset comparison, and then the mean correlation is calculated as the actual correlation.

[0031] Based on theoretical and practical correlation, a correlation analysis process is performed to output the status of the transformer area.

[0032] As a further aspect of the present invention, the step of performing correlation analysis based on theoretical and actual correlation and outputting the status of the transformer area includes:

[0033] The theoretical relevance is compared with the theoretical relevance threshold, and the actual relevance is compared with the actual relevance threshold.

[0034] When the theoretical correlation reaches the preset theoretical correlation threshold and the actual correlation is less than the preset actual correlation threshold, the power quality dataset of the corresponding terminal is regarded as an invalid dataset, a prompt message is generated, and feedback is sent to the management terminal.

[0035] When the theoretical correlation reaches the preset theoretical correlation threshold and the actual correlation reaches the preset actual correlation threshold, the power quality dataset of the corresponding terminal is taken as the valid dataset, and the status of the transformer area is determined and output based on the valid dataset.

[0036] When the theoretical correlation is less than the preset theoretical correlation threshold and the actual correlation is less than the preset actual correlation threshold, the power quality dataset of the corresponding terminal will be used as the relevant dataset, a prompt message will be generated, and the message will be fed back to the management terminal.

[0037] When the theoretical relevance is less than the preset theoretical relevance threshold and the actual relevance is less than the preset actual relevance threshold, the corresponding terminal is marked as a mutually exclusive terminal, and the station status determination process is recursively updated according to the mutually exclusive terminal.

[0038] The process for determining the status of the transformer area is as follows:

[0039] Randomly select valid datasets and repeat the process until the number of selected datasets reaches a preset total threshold, which is considered as one selection scheme.

[0040] For any selected scheme, the selected valid dataset is input into the trained transformer area status recognition model, and the transformer area status is output; the final transformer area status is determined by combining all the output transformer area statuses.

[0041] During the cyclic execution of each selection scheme, a random selection process is first performed in the mutual exclusion terminal until the number of valid datasets selected in the mutual exclusion terminal reaches the preset mutual exclusion number threshold.

[0042] The present invention also provides a power quality data correlation analysis system based on an intelligent fusion terminal, the system comprising:

[0043] The dataset construction module is used to receive power quality time-series data uploaded by various smart converged terminals in the distribution area and construct a power quality dataset based on a preset order.

[0044] The error extraction module is used to extract features from the power quality dataset for any type of disturbance event, extract the acquisition error features, and construct a terminal error feature set; the terminal error feature set includes event items and feature items.

[0045] The data shaping module is used to generate correlation analysis instructions on a timed basis, extract data to be processed from the power quality dataset, query the disturbance events between the current instruction and the previous instruction, and shape the data to be processed based on the terminal error feature set.

[0046] The correlation analysis module is used to perform correlation analysis on the normalized data to be processed and output the status of the transformer area; the correlation analysis process includes theoretical correlation process and actual correlation process.

[0047] As a further aspect of the present invention, the dataset construction module includes:

[0048] The attribute data acquisition unit is used to establish connection channels with each smart converged terminal in the transformer area, and extract the sampling timestamp, terminal number, phase identifier and environmental parameters corresponding to each connection channel as terminal attribute data.

[0049] The power data acquisition unit is used to receive voltage data, current data, harmonic data, voltage fluctuation data, and power factor data uploaded by various smart converged terminals in the distribution area based on the connection channel, as the collected data;

[0050] The sequence generation unit is used to statistically analyze different types of collected data based on the same time series to obtain a data sequence containing type labels.

[0051] The data quality analysis unit is used to obtain the data fluctuation amplitude, data missing frequency and data mutation frequency of each data sequence of each intelligent fusion terminal within a continuous time window, as data quality parameters.

[0052] The data statistics unit is used to statistically analyze terminal attribute data, data sequences containing type labels, and their data quality parameters based on a preset type label order, in order to obtain a power quality dataset.

[0053] As a further aspect of the present invention, the error extraction module includes:

[0054] The data extraction unit is used to establish a connection channel with the historical database and extract the data sequence in the power quality dataset corresponding to each disturbance event based on a preset time span.

[0055] The rate of change calculation unit is used to calculate the rate of change of data for each disturbance event data sequence; the rate of change of data includes at least the rate of change of voltage, the rate of change of current, and the rate of change of harmonics.

[0056] Anomaly localization unit is used to locate the time point when the data change rate reaches a preset change rate threshold, and calculate the phase difference in combination with the disturbance time.

[0057] The feature group generation unit is used to statistically analyze the phase difference, the rate of change containing type labels, and the original data of all data sequences of all terminals under the disturbance event, as the error feature group of the disturbance event; the data in the error feature group includes terminal label, type label, phase difference, rate of change, and original data;

[0058] The feature group statistical unit is used to classify disturbance events. For each type of disturbance event, the error feature group of each disturbance event is compared and fitted to obtain a comprehensive feature group for each type of disturbance event. The data in the comprehensive feature group includes type label, phase difference, rate of change range and original data range.

[0059] As a further embodiment of the present invention, the data normalization module includes:

[0060] The time interval determination unit is used to generate correlation analysis instructions at regular intervals, query the generation time of the current instruction and the generation time of the previous instruction, and determine the time interval.

[0061] The subset acquisition unit is used to extract a subset of data within a time interval from the power quality dataset as data to be processed; the extraction process includes retaining terminal attribute data, extracting data sequences, and extracting data quality parameters.

[0062] The feature matching unit is used to query disturbance events within a time interval and match the corresponding comprehensive feature groups.

[0063] The data fitting unit is used to locate outlier data in the data to be processed based on a comprehensive feature set, and to fit a data function based on non-outlier data to fit the location of outlier data, thereby obtaining the data to be processed after data normalization.

[0064] Compared with the prior art, the beneficial effects of the present invention are:

[0065] This invention constructs a terminal error feature set to characterize the data drift, clock offset, and noise features generated during the data acquisition process of intelligent fusion terminals. It then uses the terminal error feature set to correct the reliability of the correlation between disturbance events, thereby optimizing the data quality of the acquired data and obtaining more accurate correlation analysis results. Attached Figure Description

[0066] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention.

[0067] Figure 1 This is a flowchart illustrating the power quality data correlation analysis method based on an intelligent fusion terminal provided in an embodiment of the present invention.

[0068] Figure 2 This is a block diagram illustrating the composition of a power quality data correlation analysis system based on an intelligent fusion terminal, as provided in an embodiment of the present invention. Detailed Implementation

[0069] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0070] Figure 1 The flowchart illustrates a power quality data correlation analysis method based on an intelligent fusion terminal provided by an embodiment of the present invention. The technical solution of the present invention provides a power quality data correlation analysis method based on an intelligent fusion terminal, the method comprising:

[0071] Step S100: Receive power quality time-series data uploaded by each smart converged terminal in the distribution area, and construct a power quality dataset based on a preset order;

[0072] The term "transformer area" refers to a power supply zone. In this invention, it specifically refers to the power supply zone requiring data analysis. The transformer area is equipped with various intelligent detection devices, referred to as intelligent fusion terminals. Each intelligent fusion terminal collects power quality parameters such as voltage, current, harmonic content, power factor, and three-phase imbalance according to a fixed sampling period (frequency). Since the upload time, data format, and sampling start time may differ between terminals, the received data is first uniformly organized according to a preset order. This preset order can be an arrangement of "terminal number—sampling time—data type." For example, if terminal A uploads 220V voltage data at 10:00:01 and terminal B uploads 221V voltage data at 10:00:01, the data is sorted and stored according to a unified timeline. For missing data, it can be handled by filling in the missing data from the previous moment or marking the missing status. After unified sorting, format conversion, and time normalization, a power quality dataset containing fields such as terminal identifier, sampling time, voltage, current, and harmonics is constructed. The data acquisition process is a routine data interaction process and will not be described in detail here.

[0073] Step S200: For any type of disturbance event, extract features from the power quality dataset, extract acquisition error features, and construct a terminal error feature set; the terminal error feature set includes event items and feature items;

[0074] Disturbance events refer to events occurring in a distribution area that can affect the entire power grid, such as environmental changes, demand changes, or equipment start-up and shutdown. The total number of event types is predetermined and finite. After obtaining the power quality dataset, the data change patterns are analyzed to identify acquisition errors that may be generated by the intelligent fusion terminal itself. For example, if the voltage data of a certain terminal is stable at around 220V for a long time, but suddenly changes to 265V at a certain moment and returns to normal at the next sampling moment, then the data has typical instantaneous jump characteristics, which is a type of error. During the identification process, it is easily identified as target data, but it is a regular error caused by the equipment itself, and for the analysis process, it is actually interference. Therefore, the technical solution of this invention needs to eliminate these interferences.

[0075] Specifically, for each type of disturbance event, the power quality dataset corresponding to each event in that type of event is queried, and feature extraction is performed on the power quality dataset. Some data that appear abnormal under normal conditions (situations that would occur under normal disturbance conditions) are extracted. These data are actually errors in the subsequent data analysis process because the data itself is very abrupt. Most existing data analysis processes, especially anomaly identification processes, can certainly identify these data, but these seemingly abnormal data are actually normal. After obtaining this data in step S200, the subsequent data analysis process can be corrected, thereby optimizing the data analysis results. The extracted terminal error feature set includes event items and feature items. Event items are data items composed of disturbance events, and feature items are the collected features (error features) corresponding to the disturbance events.

[0076] Step S300: Generate correlation analysis instructions periodically, extract the data to be processed from the power quality dataset, query the disturbance events between the current instruction and the previous instruction, and perform data normalization on the data to be processed based on the terminal error feature set;

[0077] An analysis command is generated at preset time intervals. Since the analysis command needs to perform comprehensive analysis on multiple smart fusion terminals, and the power quality dataset of each smart fusion terminal needs to be correlated, it is called a correlation analysis command. The correlation analysis command is generated periodically, and the data to be processed within the time span is extracted from the power quality dataset. The disturbance events between the current command and the previous command are queried, that is, the disturbance events within the time span. Based on the already determined terminal error feature set, the data characteristics (error features) brought by these disturbance events are determined, and then the influence of these data features is reduced in the extracted data to be processed, thereby achieving data regularization.

[0078] Step S400: Perform correlation analysis on the normalized data to be processed and output the status of the transformer area; wherein, the correlation analysis process includes theoretical correlation process and actual correlation process.

[0079] The normalized data to be processed has eliminated error features. The status of the substation is determined based on the normalized data to be processed. Specifically, the correlation analysis process is adopted. The correlation analysis process is to associate the data (normalized data to be processed) of different smart converged terminals together and output a comprehensive result. The correlation analysis process includes a theoretical correlation process and a practical correlation process. The theoretical correlation process refers to the data relationship between two smart converged terminals in theory, while the practical correlation process refers to the data relationship between two smart converged terminals in practice.

[0080] Regarding step S100, the step of receiving power quality time-series data uploaded by each smart converged terminal in the receiving area and constructing a power quality dataset based on a preset order includes:

[0081] Establish connection channels with each intelligent converged terminal in the transformer area, and extract the sampling timestamp, terminal number, phase identifier and environmental parameters corresponding to each connection channel as terminal attribute data;

[0082] The voltage, current, harmonic, voltage fluctuation, and power factor data uploaded by various intelligent converged terminals in the receiving area are used as the collected data.

[0083] By statistically analyzing different types of collected data within the same time series, a data sequence containing type labels is obtained;

[0084] The amplitude of data fluctuation, frequency of data loss, and frequency of data mutation of each data sequence of each intelligent fusion terminal within a continuous time window are obtained as data quality parameters.

[0085] Based on the preset type label sequence, the terminal attribute data, the data sequence containing type labels and its data quality parameters are statistically analyzed to obtain the power quality dataset.

[0086] In one example of the technical solution of this invention, the construction process of the power quality dataset is described. A connection channel is established with each smart converged terminal in the distribution area. The sampling timestamp, terminal number, phase identifier, and environmental parameters corresponding to each connection channel are extracted as terminal attribute data, which is used to determine the basic information of the data. Voltage data, current data, harmonic data, voltage fluctuation data, and power factor data uploaded by each smart converged terminal in the distribution area are received through the connection channel as collected data. Collected data is generally power data obtained by the smart converged terminal. Different types of collected data are statistically analyzed based on the same time series to obtain a data sequence containing type labels. This process refers to the fact that the collection frequencies of different types of collected data are actually different, and their data volumes vary greatly, making it difficult to process them together. The solution is to first create a time series, such as a time series with a five-second time interval. For any type of collected data, the most recent data is read every five seconds based on the time series to obtain a data sequence (with possible repetitions), thus ensuring that each type of collected data is in the same dimension as the time series. This process is essentially a time calibration of the collected data.

[0087] Then, the data fluctuation amplitude, data missing frequency, and data mutation frequency of each data sequence of each intelligent fusion terminal within a continuous time window are obtained as data quality parameters. The time window is a smaller time interval, equivalent to a small time window. The data fluctuation amplitude, data missing frequency, and data mutation frequency are obtained by sliding the time window on the data sequence. The data is still statistically analyzed according to type and time order, and the obtained data is used as data quality parameters.

[0088] Finally, based on the preset type label order, the terminal attribute data, the data sequence containing type labels, and their data quality parameters are statistically analyzed to obtain the power quality dataset. The type label order is preset to limit the format of the power quality dataset, so that the power quality datasets of different smart converged terminals are in the same format.

[0089] Regarding step S200, the step of extracting features from the power quality dataset for any type of disturbance event, extracting acquisition error features, and constructing a terminal error feature set includes:

[0090] Establish a connection channel with the historical database, and extract the data sequence from the power quality dataset corresponding to each disturbance event based on a preset time span;

[0091] For each disturbance event's data sequence, calculate the data change rate; the data change rate includes at least the voltage change rate, current change rate, and harmonic change rate.

[0092] The phase difference is calculated by determining the time point at which the rate of change of the location data reaches a preset rate of change threshold, combined with the time of the disturbance.

[0093] Under this disturbance event, the phase difference, the rate of change including type label, and the original data of all data sequences of all terminals are statistically analyzed and used as the error feature group of the disturbance event; the data in the error feature group includes terminal label, type label, phase difference, rate of change, and original data;

[0094] The disturbance events are classified, and for each type of disturbance event, the error feature groups of each disturbance event are compared to fit a comprehensive feature group for each type of disturbance event. The data in the comprehensive feature group includes type label, phase difference, rate of change range and original data range.

[0095] In one example of the technical solution of this invention, the construction process of the terminal error feature set is described. The terminal error feature set consists of parameters for each type of disturbance event. This requires first determining the parameters of each disturbance event, and then classifying and combining them to obtain the parameters of a type of disturbance event. Specifically, a connection channel with the historical database is established, and data sequences from the power quality dataset corresponding to each disturbance event are extracted based on a preset time span. This process generally takes the occurrence time of the disturbance event as the left endpoint and combines it with the preset time span to determine a time range, thereby obtaining a subset of the power quality dataset within the time range. In this case, a subset of each power quality dataset is extracted for each disturbance event. Therefore, each disturbance event actually corresponds to multiple subsets, and each subset contains multiple data sequences. Thus, the data sequences corresponding to each disturbance event are actually very numerous, and the amount of data is sufficient.

[0096] Then, for each data sequence of each disturbance event, the data change rate is calculated; the data change rate includes at least the voltage change rate, current change rate, and harmonic change rate; the time point when the data change rate reaches the preset change rate threshold is located, and the phase difference is calculated in combination with the disturbance time. The phase difference is the time difference, which indicates how long it takes for data to appear on the data sequence after the disturbance event occurs; after processing each data sequence, the phase difference, the change rate containing type labels, and the original data of all data sequences of all terminals under the disturbance event are statistically analyzed as the error feature group of the disturbance event; the data in the error feature group includes terminal labels, type labels, phase differences, change rates, and original data.

[0097] Finally, the disturbance events are classified. For each category of disturbance events, the error feature groups of each disturbance event are compared, and a comprehensive feature group for each category of disturbance events is fitted. The data in the comprehensive feature group includes type label, phase difference, rate of change range, and original data range. This process is actually a merging process. Each category of disturbance events includes multiple disturbance events, and each disturbance event has an error feature group. These are merged to obtain the comprehensive feature group. The comprehensive feature group and the error feature group are actually in the same format, except that the element values ​​are converted into numerical ranges. The process of converting to obtain numerical ranges is not complicated. The simplest way is to select the maximum and minimum values ​​to determine a range. Alternatively, one maximum and minimum value can be removed, and then the range can be determined based on the remaining data.

[0098] Regarding step S300, the steps of periodically generating correlation analysis instructions, extracting data to be processed from the power quality dataset, querying disturbance events between the current instruction and the previous instruction, and performing data normalization on the data to be processed based on the terminal error feature set include:

[0099] Generate correlation analysis commands periodically, query the generation time of the current command and the generation time of the previous command, and determine the time interval;

[0100] A subset of data within a time interval is extracted from the power quality dataset and used as the data to be processed. The extraction process includes preserving terminal attribute data, extracting the data sequence, and extracting the data quality parameters.

[0101] Query disturbance events within a time interval and match them with the corresponding comprehensive feature groups;

[0102] Based on the comprehensive feature set, outlier data is located in the data to be processed. Based on the non-outlier data, a data function is fitted to fit the location of the outlier data, and the data to be processed is obtained after data normalization.

[0103] The above describes the data normalization process. It involves periodically generating correlation analysis commands, querying the generation time of the current command and the previous command, and determining the time interval (the command generation cycle). A subset of the data within this time interval is extracted from the power quality dataset as the data to be processed. This extraction process includes retaining terminal attribute data, extracting the data sequence, and extracting data quality parameters. Disturbance events within the time interval are queried, and their corresponding comprehensive feature groups are matched. This process determines the type of disturbance event and reads the comprehensive feature group corresponding to that type. Based on the comprehensive feature group, outlier data is located in the data to be processed. Data matching elements in the comprehensive feature group is considered outlier and removed. A function, called the data function, is fitted to the remaining non-outlier data. Normalization is achieved by fitting the data function to the locations of outlier data.

[0104] It should be noted that the elements in the comprehensive feature group represent the range (phase difference, rate of change, and specific data) of a certain type of data after a certain period of time when the disturbance event begins. The abnormal data localization process based on the comprehensive feature group is a simple data reading and comparison process.

[0105] Regarding step S400, the step of performing correlation analysis on the normalized data to be processed and outputting the station status includes:

[0106] Pair each intelligent converged terminal with another;

[0107] For two paired terminals, obtain their theoretical physical relationship and determine the theoretical correlation between the terminals;

[0108] For two paired terminals, the correlation of each data type is calculated based on the power quality dataset comparison, and then the mean correlation is calculated as the actual correlation.

[0109] Based on theoretical and practical correlation, a correlation analysis process is performed to output the status of the transformer area.

[0110] The above content provides a specific correlation analysis process, which involves pairing each intelligent fusion terminal in pairs; for each pair of terminals, obtaining their theoretical physical relationship and determining the theoretical correlation between the terminals; for each pair of terminals, comparing the power quality dataset based on data type, calculating the correlation of each data type, and then calculating the average correlation as the actual correlation; executing the correlation analysis process based on the theoretical and actual correlations, and outputting the transformer area status; where, regarding the theoretical correlation, a simplified determination process is to set the theoretical correlation to one if the two terminals have a series-parallel relationship, otherwise to zero; a more complex process could be to determine the theoretical relationship between each type of data, whether they can influence each other, and if they can theoretically influence each other, setting the correlation between the corresponding data to one; finally, calculating the ratio of the number of data with a value of one to the total number of all data, as the theoretical correlation.

[0111] For actual correlation, for two paired terminals, the correlation of each data type is calculated based on the power quality dataset comparison. Each data type corresponds to a data sequence or a single value. For data sequences, the DTW distance is calculated, and for single values, the difference is calculated to obtain the correlation of the data types. Finally, the mean correlation of all data types is calculated as the actual correlation. After both theoretical and actual correlations are calculated, the correlation analysis process is executed based on the theoretical and actual correlations to output the status of the transformer area.

[0112] As a preferred embodiment of the technical solution of the present invention, the step of performing correlation analysis based on theoretical correlation and actual correlation and outputting the status of the transformer area includes:

[0113] The theoretical relevance is compared with the theoretical relevance threshold, and the actual relevance is compared with the actual relevance threshold.

[0114] When the theoretical correlation reaches the preset theoretical correlation threshold and the actual correlation is less than the preset actual correlation threshold, the power quality dataset of the corresponding terminal is regarded as an invalid dataset, a prompt message is generated, and feedback is sent to the management terminal.

[0115] When the theoretical correlation reaches the preset theoretical correlation threshold and the actual correlation reaches the preset actual correlation threshold, the power quality dataset of the corresponding terminal is taken as the valid dataset, and the status of the transformer area is determined and output based on the valid dataset.

[0116] When the theoretical correlation is less than the preset theoretical correlation threshold and the actual correlation is less than the preset actual correlation threshold, the power quality dataset of the corresponding terminal will be used as the relevant dataset, a prompt message will be generated, and the message will be fed back to the management terminal.

[0117] When the theoretical relevance is less than the preset theoretical relevance threshold and the actual relevance is less than the preset actual relevance threshold, the corresponding terminal is marked as a mutually exclusive terminal, and the station status determination process is recursively updated according to the mutually exclusive terminal.

[0118] In one embodiment of the technical solution of this invention, a specific determination process is provided. This process applies theoretical and practical correlation, comparing these two correlations with corresponding thresholds to determine their magnitude, thereby executing different processes accordingly. Specifically, as follows:

[0119] When the theoretical correlation is high but the actual correlation is low, the power quality dataset of the corresponding terminal is treated as an invalid dataset, a prompt message is generated, and feedback is sent to the management end; if the theory is relevant but the actual data is irrelevant, it indicates that the actual data is likely disordered, and an abnormal state is handled. In this case, the corresponding power quality dataset is unusable.

[0120] When the theoretical correlation is high and the actual correlation is also high, the power quality dataset of the corresponding terminal is used as the effective dataset. The status of the transformer area is determined and output based on the effective dataset. If the data is theoretically relevant and also practically relevant, it is standard data and needs to be extracted to determine the status of the transformer area.

[0121] When the theoretical correlation is small and the actual correlation is small, the power quality dataset of the corresponding terminal is used as the relevant dataset to generate a prompt message and feed it back to the management end. If they are theoretically unrelated but actually connected, it means that there may be a relationship between them that cannot be described by conventional theoretical relationships. This synergy usually has some additional causes, such as being mistakenly connected to the same line due to a line fault. This needs to be analyzed in detail. Therefore, a prompt message is generated pointing to the two smart converged terminals.

[0122] When the theoretical correlation is small and the actual correlation is small, the corresponding terminal is marked as a mutually exclusive terminal, and the station status determination process is recursively updated based on the mutually exclusive terminals. When the theoretical correlation is small and the actual correlation is small, it means that the two intelligent fusion terminals are independent of each other. When making a status determination, it is best to select mutually exclusive terminals. This allows for a wider range of determination results when analyzing the same number of terminals. The reason is that it is best to use as few terminals as possible that have a correlation, because when there is a correlation, the data of the other terminal can be obtained with higher accuracy. Therefore, when the theoretical correlation is small and the actual correlation is small, this situation is used to recursively update the determination process.

[0123] Furthermore, the process for determining the status of the transformer area is as follows:

[0124] Randomly select valid datasets and repeat the process until the number of selected datasets reaches a preset total threshold, which is considered as one selection scheme.

[0125] For any selected scheme, the selected valid dataset is input into the trained transformer area status recognition model, and the transformer area status is output; the final transformer area status is determined by combining all the output transformer area statuses.

[0126] During the cyclic execution of each selection scheme, a random selection process is first performed in the mutual exclusion terminal until the number of valid datasets selected in the mutual exclusion terminal reaches the preset mutual exclusion number threshold.

[0127] The above content actually includes the judgment process and the judgment adjustment process. Valid datasets are randomly selected and executed cyclically until the number of selected datasets reaches a preset total threshold. As a selection scheme, since there are many intelligent fusion terminals, selecting a portion of the terminals is sufficient to determine the status of the transformer area. There are many selection schemes. Finally, the transformer area statuses from multiple selection schemes are combined to obtain the final transformer area status. This process essentially applies the idea of ​​random forests, randomly performing status judgments and outputs under low resource conditions, and then combining multiple status judgment results to obtain the final transformer area status.

[0128] Furthermore, during the cyclic execution of each selection scheme, a random selection process is first performed on the mutually exclusive terminals until the number of valid datasets selected from the mutually exclusive terminals reaches the preset mutual exclusion number threshold. This ensures that the number of mutually exclusive terminals selected is as large as possible, but not too large, because selecting all mutually exclusive terminals will also result in some data loss. The recursive update process is actually updating the total number of mutually exclusive terminals, and the update can be done in a selectable manner.

[0129] Figure 2 The diagram illustrates the structural composition of a power quality data correlation analysis system based on an intelligent fusion terminal, as provided in an embodiment of the present invention. In one example of the technical solution of the present invention, a power quality data correlation analysis system 10 based on an intelligent fusion terminal is also provided. The power quality data correlation analysis system 10 based on the intelligent fusion terminal includes:

[0130] The dataset construction module 11 is used to receive power quality time-series data uploaded by various smart converged terminals in the distribution area and construct a power quality dataset based on a preset order.

[0131] Error extraction module 12 is used to extract features from the power quality dataset for any type of disturbance event, extract acquisition error features, and construct a terminal error feature set; the terminal error feature set includes event items and feature items;

[0132] Data shaping module 13 is used to generate correlation analysis instructions at regular intervals, extract data to be processed from the power quality dataset, query the disturbance events between the current instruction and the previous instruction, and shape the data to be processed based on the terminal error feature set.

[0133] The correlation analysis module 14 is used to perform correlation analysis on the normalized data to be processed and output the status of the transformer area; the correlation analysis process includes theoretical correlation process and actual correlation process.

[0134] Furthermore, the dataset construction module 11 includes:

[0135] The attribute data acquisition unit is used to establish connection channels with each smart converged terminal in the transformer area, and extract the sampling timestamp, terminal number, phase identifier and environmental parameters corresponding to each connection channel as terminal attribute data.

[0136] The power data acquisition unit is used to receive voltage data, current data, harmonic data, voltage fluctuation data, and power factor data uploaded by various smart converged terminals in the distribution area based on the connection channel, as the collected data;

[0137] The sequence generation unit is used to statistically analyze different types of collected data based on the same time series to obtain a data sequence containing type labels.

[0138] The data quality analysis unit is used to obtain the data fluctuation amplitude, data missing frequency and data mutation frequency of each data sequence of each intelligent fusion terminal within a continuous time window, as data quality parameters.

[0139] The data statistics unit is used to statistically analyze terminal attribute data, data sequences containing type labels, and their data quality parameters based on a preset type label order, in order to obtain a power quality dataset.

[0140] Specifically, the error extraction module 12 includes:

[0141] The data extraction unit is used to establish a connection channel with the historical database and extract the data sequence in the power quality dataset corresponding to each disturbance event based on a preset time span.

[0142] The rate of change calculation unit is used to calculate the rate of change of data for each disturbance event data sequence; the rate of change of data includes at least the rate of change of voltage, the rate of change of current, and the rate of change of harmonics.

[0143] Anomaly localization unit is used to locate the time point when the data change rate reaches a preset change rate threshold, and calculate the phase difference in combination with the disturbance time.

[0144] The feature group generation unit is used to statistically analyze the phase difference, the rate of change containing type labels, and the original data of all data sequences of all terminals under the disturbance event, as the error feature group of the disturbance event; the data in the error feature group includes terminal label, type label, phase difference, rate of change, and original data;

[0145] The feature group statistical unit is used to classify disturbance events. For each type of disturbance event, the error feature group of each disturbance event is compared and fitted to obtain a comprehensive feature group for each type of disturbance event. The data in the comprehensive feature group includes type label, phase difference, rate of change range and original data range.

[0146] Furthermore, the data straightening module 13 includes:

[0147] The time interval determination unit is used to generate correlation analysis instructions at regular intervals, query the generation time of the current instruction and the generation time of the previous instruction, and determine the time interval.

[0148] The subset acquisition unit is used to extract a subset of data within a time interval from the power quality dataset as data to be processed; the extraction process includes retaining terminal attribute data, extracting data sequences, and extracting data quality parameters.

[0149] The feature matching unit is used to query disturbance events within a time interval and match the corresponding comprehensive feature groups.

[0150] The data fitting unit is used to locate outlier data in the data to be processed based on a comprehensive feature set, and to fit a data function based on non-outlier data to fit the location of outlier data, thereby obtaining the data to be processed after data normalization.

[0151] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. 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 power quality data correlation analysis based on an intelligent fusion terminal, characterized in that, The method includes: Receive power quality time-series data uploaded by various smart converged terminals in the distribution area, and construct a power quality dataset based on a preset order; For any type of disturbance event, feature extraction is performed on the power quality dataset to extract acquisition error features and construct a terminal error feature set; the terminal error feature set includes event items and feature items. The system generates correlation analysis instructions periodically, extracts data to be processed from the power quality dataset, queries the disturbance events between the current instruction and the previous instruction, and performs data normalization on the data to be processed based on the terminal error feature set. The normalized data to be processed is subjected to correlation analysis, and the status of the transformer area is output. The correlation analysis process includes theoretical correlation process and actual correlation process.

2. The power quality data correlation analysis method based on an intelligent fusion terminal according to claim 1, characterized in that, The steps for constructing a power quality dataset based on a preset order from the power quality time-series data uploaded by each smart converged terminal in the receiving area include: Establish connection channels with each intelligent converged terminal in the transformer area, and extract the sampling timestamp, terminal number, phase identifier and environmental parameters corresponding to each connection channel as terminal attribute data; The voltage, current, harmonic, voltage fluctuation, and power factor data uploaded by various intelligent converged terminals in the receiving area are used as the collected data. By statistically analyzing different types of collected data within the same time series, a data sequence containing type labels is obtained; The amplitude of data fluctuation, frequency of data loss, and frequency of data mutation of each data sequence of each intelligent fusion terminal within a continuous time window are obtained as data quality parameters. Based on the preset type label sequence, the terminal attribute data, the data sequence containing type labels and its data quality parameters are statistically analyzed to obtain the power quality dataset.

3. The power quality data correlation analysis method based on an intelligent fusion terminal according to claim 1, characterized in that, The steps of extracting features from the power quality dataset for any type of disturbance event, extracting acquisition error features, and constructing a terminal error feature set include: Establish a connection channel with the historical database, and extract the data sequence from the power quality dataset corresponding to each disturbance event based on a preset time span; For each disturbance event's data sequence, calculate the data change rate; the data change rate includes at least the voltage change rate, current change rate, and harmonic change rate. The phase difference is calculated by determining the time point at which the rate of change of the location data reaches a preset rate of change threshold, combined with the time of the disturbance. Under this disturbance event, the phase difference, the rate of change including type label, and the original data of all data sequences of all terminals are statistically analyzed and used as the error feature group of the disturbance event; the data in the error feature group includes terminal label, type label, phase difference, rate of change, and original data; The disturbance events are classified, and for each type of disturbance event, the error feature groups of each disturbance event are compared to fit a comprehensive feature group for each type of disturbance event. The data in the comprehensive feature group includes type label, phase difference, rate of change range and original data range.

4. The power quality data correlation analysis method based on intelligent fusion terminal according to claim 1, characterized in that, The steps of generating correlation analysis instructions at regular intervals, extracting data to be processed from the power quality dataset, querying disturbance events between the current instruction and the previous instruction, and performing data normalization on the data to be processed based on the terminal error feature set include: Generate correlation analysis commands periodically, query the generation time of the current command and the generation time of the previous command, and determine the time interval; A subset of data within a time interval is extracted from the power quality dataset and used as the data to be processed. The extraction process includes preserving terminal attribute data, extracting the data sequence, and extracting the data quality parameters. Query disturbance events within a time interval and match them with the corresponding comprehensive feature groups; Based on the comprehensive feature set, outlier data is located in the data to be processed. Based on the non-outlier data, a data function is fitted to fit the location of the outlier data, and the data to be processed is obtained after data normalization.

5. The power quality data correlation analysis method based on an intelligent fusion terminal according to claim 1, characterized in that, The step of performing correlation analysis on the normalized data to be processed and outputting the status of the transformer area includes: Pair each intelligent converged terminal with another; For two paired terminals, obtain their theoretical physical relationship and determine the theoretical correlation between the terminals; For two paired terminals, the correlation of each data type is calculated based on the power quality dataset comparison, and then the mean correlation is calculated as the actual correlation. Based on theoretical and practical correlation, a correlation analysis process is performed to output the status of the transformer area.

6. The power quality data correlation analysis method based on an intelligent fusion terminal according to claim 5, characterized in that, The steps for performing correlation analysis based on theoretical and actual correlation, and outputting the status of the transformer area, include: The theoretical relevance is compared with the theoretical relevance threshold, and the actual relevance is compared with the actual relevance threshold. When the theoretical correlation reaches the preset theoretical correlation threshold and the actual correlation is less than the preset actual correlation threshold, the power quality dataset of the corresponding terminal is regarded as an invalid dataset, a prompt message is generated, and feedback is sent to the management terminal. When the theoretical correlation reaches the preset theoretical correlation threshold and the actual correlation reaches the preset actual correlation threshold, the power quality dataset of the corresponding terminal is taken as the valid dataset, and the status of the transformer area is determined and output based on the valid dataset. When the theoretical correlation is less than the preset theoretical correlation threshold and the actual correlation is less than the preset actual correlation threshold, the power quality dataset of the corresponding terminal will be used as the relevant dataset, a prompt message will be generated, and the message will be fed back to the management terminal. When the theoretical relevance is less than the preset theoretical relevance threshold and the actual relevance is less than the preset actual relevance threshold, the corresponding terminal is marked as a mutually exclusive terminal, and the station status determination process is recursively updated according to the mutually exclusive terminal. The process for determining the status of the transformer area is as follows: Randomly select valid datasets and repeat the process until the number of selected datasets reaches a preset total threshold, which is considered as one selection scheme. For any selected scheme, the selected valid dataset is input into the trained transformer area status recognition model, and the transformer area status is output; the final transformer area status is determined by combining all the output transformer area statuses. During the cyclic execution of each selection scheme, a random selection process is first performed in the mutual exclusion terminal until the number of valid datasets selected in the mutual exclusion terminal reaches the preset mutual exclusion number threshold.

7. A power quality data correlation analysis system based on an intelligent fusion terminal, characterized in that, The system includes: The dataset construction module is used to receive power quality time-series data uploaded by various smart converged terminals in the distribution area and construct a power quality dataset based on a preset order. The error extraction module is used to extract features from the power quality dataset for any type of disturbance event, extract the acquisition error features, and construct a terminal error feature set; the terminal error feature set includes event items and feature items. The data shaping module is used to generate correlation analysis instructions on a timed basis, extract data to be processed from the power quality dataset, query the disturbance events between the current instruction and the previous instruction, and shape the data to be processed based on the terminal error feature set. The correlation analysis module is used to perform correlation analysis on the normalized data to be processed and output the status of the transformer area; the correlation analysis process includes theoretical correlation process and actual correlation process.

8. The power quality data correlation analysis system based on an intelligent fusion terminal according to claim 7, characterized in that, The dataset construction module includes: The attribute data acquisition unit is used to establish connection channels with each smart converged terminal in the transformer area, and extract the sampling timestamp, terminal number, phase identifier and environmental parameters corresponding to each connection channel as terminal attribute data. The power data acquisition unit is used to receive voltage data, current data, harmonic data, voltage fluctuation data, and power factor data uploaded by various smart converged terminals in the distribution area based on the connection channel, as the collected data; The sequence generation unit is used to statistically analyze different types of collected data based on the same time series to obtain a data sequence containing type labels. The data quality analysis unit is used to obtain the data fluctuation amplitude, data missing frequency and data mutation frequency of each data sequence of each intelligent fusion terminal within a continuous time window, as data quality parameters. The data statistics unit is used to statistically analyze terminal attribute data, data sequences containing type labels, and their data quality parameters based on a preset type label order, in order to obtain a power quality dataset.

9. The power quality data correlation analysis system based on an intelligent fusion terminal according to claim 7, characterized in that, The error extraction module includes: The data extraction unit is used to establish a connection channel with the historical database and extract the data sequence in the power quality dataset corresponding to each disturbance event based on a preset time span. The rate of change calculation unit is used to calculate the rate of change of data for each disturbance event data sequence; the rate of change of data includes at least the rate of change of voltage, the rate of change of current, and the rate of change of harmonics. Anomaly localization unit is used to locate the time point when the data change rate reaches a preset change rate threshold, and calculate the phase difference in combination with the disturbance time. The feature group generation unit is used to statistically analyze the phase difference, the rate of change containing type labels, and the original data of all data sequences of all terminals under the disturbance event, as the error feature group of the disturbance event; the data in the error feature group includes terminal label, type label, phase difference, rate of change, and original data; The feature group statistical unit is used to classify disturbance events. For each type of disturbance event, the error feature group of each disturbance event is compared and fitted to obtain a comprehensive feature group for each type of disturbance event. The data in the comprehensive feature group includes type label, phase difference, rate of change range and original data range.

10. The power quality data correlation analysis system based on an intelligent fusion terminal according to claim 7, characterized in that, The data straightening module includes: The time interval determination unit is used to generate correlation analysis instructions at regular intervals, query the generation time of the current instruction and the generation time of the previous instruction, and determine the time interval. The subset acquisition unit is used to extract a subset of data within a time interval from the power quality dataset as data to be processed; the extraction process includes retaining terminal attribute data, extracting data sequences, and extracting data quality parameters. The feature matching unit is used to query disturbance events within a time interval and match the corresponding comprehensive feature groups. The data fitting unit is used to locate outlier data in the data to be processed based on a comprehensive feature set, and to fit a data function based on non-outlier data to fit the location of outlier data, thereby obtaining the data to be processed after data normalization.

Images