A Fault Prediction Method for Substation Equipment Based on Big Data

Abstract

The present invention relates to the technical field of data processing, and particularly relates to a method for predicting faults of substation equipment based on big data. The method includes: obtaining current data, load data, and dissolved gas concentration of the substation equipment in multiple cycles; determining the degree of abnormality in the current cycle; clustering the load data and the dissolved gas concentration based on the correlation relationship between the load data and the dissolved gas concentration to obtain multiple clustering clusters; correcting the degree of abnormality in each cycle according to the load data and the dissolved gas concentration in each clustering cluster to obtain the final corrected degree of abnormality in each cycle, and selecting the cycles with the final degree of abnormality greater than the first threshold as abnormal fluctuation cycles; determining the possibility of a fault occurring in the abnormal fluctuation cycles; and predicting whether there are potential faults in the substation equipment according to the possibility of a fault occurring in the abnormal fluctuation cycles. In this way, the present invention improves the accuracy and reliability of fault prediction for substation equipment.

Description

Technical Field

[0001] The present invention relates to the technical field of data processing, and particularly to a method for predicting faults of substation equipment based on big data. Background Art

[0002] In the construction process of the power system, transformers in the power grid, as key equipment, play an irreplaceable role. Substation equipment such as large main transformers and medium and low voltage distribution transformers are widely distributed in various power supply and distribution scenarios. Especially in complex distribution grids with high penetration of renewable energy, the operation stability and reliability thereof are directly related to the safety of the entire power system.

[0003] In some scenarios, as the operation time of substation equipment accumulates, the internal insulating layer and insulating oil will inevitably be gradually damaged and decomposed, which is likely to cause partial discharge phenomena inside the transformer. Partial discharge will not only further exacerbate the damage of insulating materials, but may also lead to transformer failures, affecting the continuity and stability of power supply. Currently, it is common to monitor the current of substation equipment to predict potential initial faults of substation equipment. However, there are various abnormalities in the potential initial faults of substation equipment. It is difficult to effectively monitor and accurately predict the initial stage of potential faults occurring in substation equipment only based on the method of monitoring current, resulting in low accuracy and reliability of substation equipment fault prediction. Summary of the Invention

[0004] In order to solve the technical problem of low accuracy and reliability of substation equipment fault prediction, the purpose of the present invention is to provide a method for predicting faults of substation equipment based on big data, and the specific technical solutions adopted are as follows:

[0005] In a first aspect, an embodiment of the present invention provides a method for predicting faults of substation equipment based on big data, including: obtaining current data, load data, and dissolved gas concentration of the substation equipment in multiple cycles; determining the abnormality degree of the current cycle according to the amplitude of the maximum value point of the current data in the current cycle, the first quantity of the maximum value points of the current data in the current cycle, and the abnormality degree of the previous cycle adjacent to the current cycle; clustering the load data and the dissolved gas concentration based on the correlation relationship between the load data and the dissolved gas concentration to obtain multiple clustering clusters; correcting the abnormality degree of each cycle according to the load data and the dissolved gas concentration in each clustering cluster to obtain the final abnormality degree after correction of each cycle, and selecting the cycle with the final abnormality degree greater than the first threshold as the abnormal fluctuation cycle; determining the possibility of a fault occurring in the abnormal fluctuation cycle according to the amplitude of each reference cycle in the abnormal fluctuation cycle and the time interval between adjacent reference cycles; predicting whether there are potential faults in the substation equipment according to the possibility of a fault occurring in the abnormal fluctuation cycle.

[0006] Optionally, determining the abnormality degree of the current period according to the amplitude of the maximum value point of the current data in the current period, the first quantity of the maximum value points of the current data in the current period, and the abnormality degree of the previous period adjacent to the current period includes: calculating a first difference between the first quantity and a predetermined value, and a second difference between the predetermined value and the abnormality degree of the adjacent previous period; calculating a third difference between the amplitude of each maximum value point of the current data in the current period and the average value of the amplitudes of all maximum value points of the current data in the adjacent previous period, and superimposing each third difference to obtain a first superimposed value; calculating a first product between the first difference and the second difference and a second product between the first superimposed value and the abnormality degree of the adjacent previous period; determining a first sum value between the first product and the second product as the abnormality degree of the current period.

[0007] Optionally, correcting the abnormality degree of each period according to the load data and dissolved gas concentration in each clustering cluster to obtain the final corrected abnormality degree of each period includes: sorting the load data and dissolved gas concentration in each clustering cluster in chronological order and segmenting them to obtain a plurality of data segments; determining the continuity of the clustering cluster in time according to the second quantity of the data segments in the clustering cluster, the third quantity of the data points in the clustering cluster, the fourth quantity of the data points in the data segment, and the change amount between every two adjacent data points in the data segment; determining the time distribution degree of the clustering cluster according to the time interval between every two adjacent data segments in the clustering cluster and the second quantity; correcting the abnormality degree of each period according to the continuity of the clustering cluster in time and the time distribution degree to obtain the final corrected abnormality degree of each period.

[0008] Optionally, sorting the load data and dissolved gas concentration in each clustering cluster in chronological order and segmenting them to obtain a plurality of data segments includes: determining the time interval between the load data and dissolved gas concentration at the current time point in the clustering cluster and the load data and dissolved gas concentration at the adjacent previous time point; in the case where the time interval is greater than a second threshold, taking the load data and dissolved gas concentration at the current time point as the starting point of a new data segment, and in the case where the time interval is not greater than the second threshold, classifying the load data and dissolved gas concentration at the current time point into the data segment where the adjacent previous load data is located.

[0009] Optionally, determining the temporal continuity of a clustering cluster based on the second quantity of data segments in the clustering cluster, the third quantity of data points in the clustering cluster, the fourth quantity of data points in a data segment, and the change amount between every two adjacent data points in the data segment includes: calculating a first ratio between the second quantity and the third quantity and a second ratio between the fourth quantity and the third quantity; calculating an average change amount of the change amounts between every two adjacent data points in the data segment and a variance of the change amounts between every two adjacent data points in the data segment; calculating a third product between the second ratio, the reciprocal of the average change amount, and the reciprocal of the variance, and superimposing the third products to obtain a second superimposed value; performing normalization processing on a fourth product between the second superimposed value and the first ratio to obtain the temporal continuity of the clustering cluster.

[0010] Optionally, determining the temporal distribution degree of a clustering cluster based on the time interval between every two adjacent data segments in the clustering cluster and the second quantity includes: superimposing the time intervals between every two adjacent data segments in the clustering cluster to obtain a third superimposed value, and calculating the reciprocal of a third difference between the second quantity and a predetermined value; determining a fifth product between the third superimposed value and the reciprocal of the third difference as the temporal distribution degree of the clustering cluster.

[0011] Optionally, correcting the abnormality degree of each period according to the temporal continuity and the temporal distribution degree of the clustering cluster to obtain the final corrected abnormality degree of each period includes: calculating a third ratio between the temporal distribution degree of the clustering cluster where each period is located and the temporal continuity of the clustering cluster where each period is located, and a sixth product between each third ratio and the abnormality degree of each period; performing normalization processing on the sixth products to obtain the final corrected abnormality degree of each period.

[0012] Optionally, determining the possibility of a fault occurring in an abnormal fluctuation period based on the amplitudes of the reference periods of the abnormal fluctuation period and the time intervals between adjacent reference periods includes: taking the periods in the most recent adjacent time period of the abnormal fluctuation period as the reference periods of the abnormal fluctuation period; calculating a fourth difference between the fifth quantity of the reference periods and a predetermined value, and a fifth difference between the amplitudes of two adjacent reference periods; calculating a fourth ratio between the fifth difference and the amplitude of the previous reference period among two adjacent reference periods, and a second sum value between the fourth ratio and the reciprocal of the time interval between adjacent reference periods; superimposing the second sum values to obtain a fourth superimposed value; determining that the fourth superimposed value and the reciprocal of the fourth difference are the possibility of a fault occurring in the abnormal fluctuation period.

[0013] Optionally, predicting whether there is a potential fault in a substation equipment based on the possibility of a fault occurring in an abnormal fluctuation period includes: determining that there is a potential fault in the substation equipment when the possibility of a fault occurring in the most recent abnormal fluctuation period of the substation equipment is greater than a third threshold.

[0014] Optionally, the third threshold is the third difference between the average value of the failure probabilities of multiple power transformation devices in the most recent abnormal fluctuation period in historical data and a preset value.

[0015] The present invention has the following beneficial effects: First, current data, load data, and dissolved gas concentration of a power transformation device in multiple cycles are obtained; then, according to the amplitude of the maximum value point of the current data in the current cycle, the first quantity of the maximum value points of the current data in the current cycle, and the abnormal degree of the previous cycle adjacent to the current cycle, the abnormal degree of the current cycle is determined; and clustering is performed on the load data and the dissolved gas concentration based on the correlation relationship between the load data and the dissolved gas concentration to obtain multiple clustering clusters; secondly, the abnormal degree of each cycle is corrected according to the load data and the dissolved gas concentration in each clustering cluster to obtain the final abnormal degree of each cycle after correction; and the cycles with the final abnormal degree greater than the first threshold are selected as abnormal fluctuation cycles; finally, according to the amplitude of each reference cycle of the abnormal fluctuation cycle and the time interval between adjacent reference cycles, the probability of failure of the abnormal fluctuation cycle is determined; and whether there is a potential failure in the power transformation device is predicted according to the probability of failure of the abnormal fluctuation cycle.

[0016] In this way, the embodiment of the present invention can monitor in real time the current data, load data, and dissolved gas concentration generated during the operation of the power transformation device, and determine the abnormal fluctuation cycle in combination with the above multi-dimensional data. Specifically, the abnormal degree of each cycle is first analyzed through current data, and then the abnormal degree of each cycle is corrected by using the load data and the dissolved gas concentration to obtain the final abnormal degree. Finally, the probability of failure of the abnormal fluctuation cycle is determined, and whether there is a potential failure in the power transformation device is predicted based on this. In this way, through the analysis of multi-dimensional data, the initial stage of potential failures occurring in the power transformation device can be effectively monitored and accurately predicted, improving the accuracy and reliability of power transformation device failure prediction, avoiding the expansion of equipment damage, and improving the reliability of the power system. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings required for use in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present invention, and those of ordinary skill in the art can also obtain other drawings without creative efforts based on these drawings.

[0018] Figure 1 It is a flowchart of a method for predicting power transformation device failures based on big data provided by an embodiment of the present invention;

[0019] Figure 2 Schematic diagram of the current change curve of an oil-immersed transformer under normal conditions provided by an embodiment of the present invention;

[0020] Figure 3 Schematic diagram of the current change curve of an oil-immersed transformer under abnormal conditions provided by an embodiment of the present invention. Detailed implementation manners

[0021] In order to further elaborate on the technical means and effects adopted by the present invention to achieve the intended invention purpose, the following combines the drawings and preferred embodiments to specifically describe a big data-based substation equipment fault prediction method proposed according to the present invention, including its specific implementation manners, structures, features and effects, as follows. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. In addition, the specific features, structures or characteristics in one or more embodiments can be combined in any suitable form.

[0022] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the technical field to which the present invention belongs.

[0023] The following specifically describes the specific solution of a big data-based substation equipment fault prediction method provided by the present invention with reference to the drawings.

[0024] Embodiment 1:

[0025] Please refer to Figure 1 , which shows the flowchart of a big data-based substation equipment fault prediction method provided by an embodiment of the present invention, including:

[0026] S101, obtaining the current data, load data and dissolved gas concentration of the substation equipment in multiple cycles.

[0027] Specifically, the types of substation equipment include, but are not limited to, transformers, circuit breakers, disconnectors, etc. The types of transformers include, but are not limited to, oil-immersed transformers, dry-type transformers, autotransformers, and isolation transformers, etc. In the embodiments of the present invention, when predicting faults of substation equipment, taking the oil-immersed transformer as an example, its common fault is the damage of internal equipment caused by abnormal discharge. Therefore, when analyzing, it is necessary to analyze and collect the common current data and load data in the transformer, and analyze the concentration of the gas decomposed from the insulating oil in the insulating oil during discharge. Therefore, the embodiments of the present invention need to install current sensors and load sensors at the current input and output ports of each oil-immersed transformer, and install dissolved gas analyzers in the insulating oil inside the oil-immersed transformer to continuously and real-time collect real-time data such as current data, load data, and dissolved gas concentration on the oil-immersed transformer. Among them, the current data can be collected in real time, and the collection interval of the dissolved gas concentration and load data can be 1 s. In addition, the embodiments of the present invention also standardize the collected data to facilitate subsequent calculations.

[0028] S102. Determine the abnormality degree of the current cycle according to the amplitude of the maximum value point of the current data in the current cycle, the first quantity of the maximum value points of the current data in the current cycle, and the abnormality degree of the previous cycle adjacent to the current cycle.

[0029] Specifically, the change of the current in the oil-immersed transformer is a smooth sine wave showing periodic changes under normal circumstances. When there is an abnormal discharge or the load changes in the oil-immersed transformer, the waveform of the current in the oil-immersed transformer will fluctuate. Due to the influence of the high-frequency current generated during abnormal discharge or the load change, in the same current cycle as the normal situation, the amplitude of the peak value in its fluctuation image is larger than that in the normal situation, and the number of peak values is also more, showing a high-frequency pulse state. Therefore, the embodiments of the present invention judge the abnormal moment according to the difference between the amplitude and frequency of the current in a fixed cycle and the normal situation. However, due to the interference of the load, it is impossible to accurately judge whether it is caused by abnormal discharge.

[0030] And because the current generated by abnormal discharge will decompose the insulating layer and insulating oil in the local area of the oil-immersed transformer, increasing the total amount of various gases dissolved in the insulating oil. Therefore, according to whether the solubility at each abnormal moment and the total amount increase, the interference caused by the load in the abnormal moments can be screened out. By analyzing the screened abnormal moments, it is judged whether the occurrence frequency increases, and compared with historical data, the final abnormality degree of the oil-immersed transformer is obtained.

[0031] Furthermore, in an oil-immersed transformer, alternating current at different voltage levels is converted through the principle of electromagnetic induction, and the current changes accordingly. Exemplarily, in the embodiments of the present invention, Figure 2 and Figure 3 are used to illustrate the current change situation of the oil-immersed transformer. Figure 2 is a schematic diagram of the current change curve of an oil-immersed transformer under normal conditions provided by an embodiment of the present invention. Figure 3 is a schematic diagram of the current change curve of an oil-immersed transformer under abnormal conditions provided by an embodiment of the present invention. Under normal conditions, as Figure 2 shown, the amplitude, frequency, and waveform of the current in the oil-immersed transformer are relatively stable and change slowly. However, if partial discharge occurs in the oil-immersed transformer, high-frequency pulses and irregular fluctuations will be introduced. As Figure 3 shown, it will cause an instantaneous drastic change in the current, the waveform becomes chaotic, the frequency may fluctuate abnormally, and in addition to partial discharge, changes in the grid load will also cause fluctuations in the current amplitude, waveform, or frequency. By analyzing the current data in each time period of the oil-immersed transformer, abnormal fluctuation points can be identified.

[0032] Furthermore, since the current in the oil-immersed transformer is the superposition of various current fluctuations, the period of its basic current is fixed. However, when abnormal discharge occurs in the oil-immersed transformer or the load in its power grid suddenly changes, the various high-frequency chaotic currents caused by the abnormal discharge or load change will cause the basic current to change after being superimposed with it. However, due to the very short duration of the abnormal discharge, usually between a few milliseconds and several hundred milliseconds, lasting for multiple cycles, the degree of abnormality of the current cycle can be judged according to the number of maximum value points in each cycle of the current in the oil-immersed transformer, the maximum value of the maximum value points in the cycle, and the number and amplitude of the maximum value points in the previous cycle.

[0033] Furthermore, as an optional embodiment of the present invention, determining the degree of abnormality of the current cycle according to the amplitude of the maximum value points of the current data in the current cycle, the first number of the maximum value points of the current data in the current cycle, and the degree of abnormality of the previous cycle adjacent to the current cycle includes: calculating the first difference between the first number and a predetermined value, and the second difference between the predetermined value and the degree of abnormality of the adjacent previous cycle; calculating the third difference between the amplitude of each maximum value point of the current data in the current cycle and the average value of the amplitudes of all maximum value points of the current data in the adjacent previous cycle, and superimposing the third differences to obtain a first superimposed value; calculating the first product between the first difference and the second difference and the second product between the first superimposed value and the degree of abnormality of the adjacent previous cycle; determining the first sum value between the first product and the second product as the degree of abnormality of the current cycle.

[0034] Specifically, in the embodiments of the present invention, the amplitude of the maximum value point of the current data in each period and the number of maximum value points in the period are respectively obtained. For each period, the differences between the amplitude of the maximum value point and the number of maximum value points of the current period and those of the previous period can be calculated. If the previous period is a normal period, and the amplitude of the maximum value point in the current period suddenly increases compared with it, and the number of maximum value points also increases, the degree of abnormality is greater. If the previous period is an abnormal period, and the number and amplitude of the maximum value points in the current period are not much different from those of the previous period, the current period is also an abnormal period, and the degree of abnormality is also greater. If the current period is quite different from it, then the current period may be a normal period. According to the amplitude of the maximum value point and the number of maximum value points of each period and its previous period, the degree of abnormality of each period can be obtained. Among them, in the embodiments of the present invention, the following formula is specifically used to calculate the degree of abnormality of the current period:

[0035]

[0036] In the above formula, is the degree of abnormality of the th period. is the first number of maximum value points in the th period. The larger the number, the more chaotic the frequency in this period, that is, the greater the degree of abnormality of this period. is the th period, that is, the degree of abnormality of the previous period adjacent to the th period. is the third difference between the amplitude of the jth maximum value point in the th period and the average value of the amplitudes of all maximum value points in the previous period.

[0037] It should be noted that when in the first period, its degree of abnormality has no direct relationship with the degree of abnormality of the previous period, because there is no previous period and the amplitude sizes cannot be compared. At this time, is used as the degree of abnormality of the first period. For subsequent periods, the degree of abnormality in the current period is corrected through and . When the degree of abnormality of the previous period is greater, more consideration is given to the similarity between the maximum value points in the current period and it, because abnormal discharges will all cause a rapid increase in current. When the degree of abnormality is smaller, it is necessary to judge whether there is an abnormality through the number of extreme value points.

[0038] S103, cluster the load data and the dissolved gas concentration based on the correlation relationship between the load data and the dissolved gas concentration to obtain a plurality of clustering clusters.

[0039] Specifically, since the equipment connected to the oil-immersed transformer is relatively complex and its load changes frequently and complexly, when its load changes, the frequency and amplitude of the current in the oil-immersed transformer will also change. Among them, a part is similar to abnormal discharge, which will interfere with the detection during abnormal discharge. The moment caused by abnormal discharge cannot be obtained only through the current fluctuation in the oil-immersed transformer. Therefore, it is necessary to exclude the abnormal changes caused by load changes. Since the inside of the oil-immersed transformer is a closed space, the total amount of gas decomposed from the insulating oil under abnormal discharge inside it is constant during two abnormal discharges. Therefore, during two abnormal discharges, the solubility at the same temperature at different times is similar, and the factors that cause the temperature change in the oil-immersed transformer are only the load and abnormal discharge.

[0040] Furthermore, when the oil-immersed transformer is working normally, due to the change of the load, its equipment will heat up, resulting in an increase in its temperature. Since the area of the heat dissipation region is relatively large, the rise of the temperature of the insulating oil is proportional and similar to the rise of the load, resulting in the change of the solubility in the insulating oil corresponding to the change of the load. When abnormal discharge occurs inside the oil-immersed transformer, it will only cause high heat and discharge in a small local area of the oil-immersed transformer, and its impact on the overall temperature in the oil-immersed transformer can be almost ignored. Therefore, the solubility in its insulating oil should still change with the change of the load. However, due to the abnormal discharge, the insulating oil and insulating layer in the local area will decompose and ionize, resulting in an increase in the total amount of various gases (dissolved and undissolved) inside. Therefore, due to the increase in the pressure of the gas in the closed oil-immersed transformer at the corresponding temperature, the corresponding relationship between solubility and temperature changes compared with before the discharge, that is, the change of the load is proportional to the change of the solubility.

[0041] Therefore, according to the analysis of the above embodiments, it can be seen that there is a correlation between the solubility and the load change between two abnormal discharges, and there are differences in this correlation on both sides of each abnormal discharge. Therefore, in the embodiments of the present invention, the correlation relationship between the load data and the dissolved gas concentration at each moment is used as a parameter to perform clustering through the K-means clustering algorithm, and several clustering clusters representing similar data sets are obtained. Among them, the clustering space of the K-means clustering algorithm is constructed by the load and solubility. When performing clustering, the load data and the dissolved gas concentration that are close to each other are grouped into a normal clustering cluster. For some scattered data with a sudden increase in load but little change in solubility caused by irregular abnormal discharge, since the load data of these data points deviate significantly from other data points, they will be clustered into an independent abnormal clustering cluster.

[0042] S104. Correct the abnormality degree of each cycle according to the load data and dissolved gas concentration in each clustering cluster to obtain the final corrected abnormality degree of each cycle, and select the cycle with the final abnormality degree greater than the first threshold as the abnormal fluctuation cycle.

[0043] Specifically, for each clustering cluster obtained by clustering in the above embodiment, including the normal clustering cluster and the abnormal clustering cluster caused by abnormal discharge. In an oil-immersed transformer, abnormal discharge is a small-probability event. Since each abnormal discharge will cause a change in the total amount of various gases in the oil-immersed transformer, the solubility on the same load in different abnormal discharge time periods (between two abnormal discharges) is different. Therefore, when clustering, all the time points during abnormal discharge will be clustered into one clustering cluster. The data in the same load interval during the same abnormal discharge time period will be clustered into one clustering cluster. And because the solubility of gases under the same load in different abnormal discharge time periods is different, there are several clustering clusters in the same load interval corresponding to different abnormal discharge time periods. For each clustering cluster, since the load in the oil-immersed transformer is constantly changing, the data in the clustering cluster is also data of multiple segments under the same load, and the data segments are not connected.

[0044] Further, as an optional embodiment of the present invention, correcting the abnormality degree of each cycle according to the load data and dissolved gas concentration in each clustering cluster to obtain the final corrected abnormality degree of each cycle includes: sorting the load data and dissolved gas concentration in each clustering cluster in chronological order and then segmenting to obtain multiple data segments; determining the temporal continuity of the clustering cluster according to the second quantity of data segments in the clustering cluster, the third quantity of data points in the clustering cluster, the fourth quantity of data points in the data segment, and the change amount between every two adjacent data points in the data segment; determining the temporal distribution degree of the clustering cluster according to the time interval between every two adjacent data segments in the clustering cluster and the second quantity; correcting the abnormality degree of each cycle according to the temporal continuity and temporal distribution degree of the clustering cluster to obtain the final corrected abnormality degree of each cycle.

[0045] Specifically, in the embodiments of the present invention, for a number of obtained clustering clusters, the elements in each clustering cluster are arranged in chronological order. Since each clustering cluster is not completely composed of a continuous segment of data, but is composed of several continuous segments of data. In the abnormal discharge clustering cluster, it is composed of several discrete abnormal discharge time points. Therefore, the embodiments of the present invention need to first divide the data segments in each clustering cluster. However, since the load in the oil-immersed transformer fluctuates, when the load changes and then recovers again within a short period of time, it will not cause a large change in the temperature of the oil-immersed transformer. Therefore, it is necessary to smooth these fluctuations. As an optional embodiment of the present invention, after sorting the load data and the dissolved gas concentration in each clustering cluster in chronological order and segmenting them, a plurality of data segments are obtained, including: determining the time interval between the load data and the dissolved gas concentration at the current time point in the clustering cluster and the load data and the dissolved gas concentration at the previous adjacent time point; in the case where the time interval is greater than the second threshold, taking the load data and the dissolved gas concentration at the current time point as the starting point of a new segment of data points, and in the case where the time interval is not greater than the second threshold, classifying the load data and the dissolved gas concentration at the current time point into the data segment where the previous adjacent load data is located.

[0046] Specifically, the second threshold can be determined according to the actual situation, and the value in the embodiments of the present invention is . For each data point in the clustering cluster, that is, the load data and the dissolved gas concentration at each time point. Determine whether the time interval between it and the previous data point is greater than . If it is greater, take it as the starting point of a new data segment. If it is less, classify it into the data segment where the previous data point is located, so as to obtain several data segments in each clustering cluster. It should be noted that an isolated data point can also be regarded as a data segment.

[0047] Further, for each cluster, the closer the number of data segments is to the number of data points in the cluster, the lower the continuity of the data in the cluster, that is, the more discrete the data is. In an oil-immersed transformer, it is more likely that there have been multiple abnormal discharges in the middle, and the discontinuity of the data in the cluster is more likely to be caused by abnormal discharges. For each data segment, the smaller the difference in the temporal variation of the data within it, the smoother the change of the data segment and the higher its continuity. Therefore, the temporal continuity of each cluster can be obtained based on the data segments in each cluster. As an alternative embodiment of the present invention, determining the temporal continuity of a cluster based on the second number of data segments in the cluster, the third number of data points in the cluster, the fourth number of data points in the data segment, and the variation between every two adjacent data points in the data segment includes: calculating a first ratio between the second number and the third number and a second ratio between the fourth number and the third number; calculating the average variation of the variations between every two adjacent data points in the data segment and the variance of the variations between every two adjacent data points in the data segment; calculating a third product between the second ratio, the reciprocal of the average variation, and the reciprocal of the variance, and superimposing the third products to obtain a second superimposed value; and normalizing the fourth product between the second superimposed value and the first ratio to obtain the temporal continuity of the cluster.

[0048] Specifically, in the embodiment of the present invention, the following formula is specifically used to calculate the temporal continuity of a cluster:

[0049]

[0050] In the above formula, is the temporal continuity of the th cluster. is the second number of data segments in the th cluster. is the th cluster. is the th cluster. is the fourth number of data points in the th data segment of the th cluster. is the mean value of the variations between every two adjacent data points in the th data segment of the th cluster, that is, the average variation. is the variance of the variations between every two adjacent data points in the The function is used to normalize .

[0051] Among them, Measures the relationship between the number of data segments and the total number of data points. The more data segments there are, the worse the continuity of the data, which reflects that the data in the oil-immersed transformer fluctuates for a longer time during this period, the data is broken, and the more data breaks there are in the clustering cluster, the lower the time continuity of the clustering cluster. Can reflect the stability and scale of each data segment in the clustering cluster, that is, the stability of the load change in the oil-immersed transformer during this period. The stability of the operating state of the oil-immersed transformer during this period is represented by the number of data points within the data segment. Whether there are large abnormal fluctuations. The larger the data segment, the more likely the time series within the data segment is stable and has higher continuity. and Measure the mean and variance of the change amount between adjacent data points within the data segment respectively. They jointly reflect the stationarity of the data segment. The smaller the mean and the smaller the variance, the smoother the change of the data segment and the higher the continuity.

[0052] Furthermore, since the data in the clustering cluster is the clustering between each data segment when the total amount of various gases in the oil-immersed transformer remains unchanged, its data only includes the data between two abnormal discharges. The time interval formed by all data segments in each normal clustering cluster is relatively short, only the length between two abnormal discharges. And in the clustering cluster caused by abnormal discharge, it contains the data when abnormal discharge occurs at each time node of the oil-immersed transformer. Therefore, the time interval in the clustering cluster is longer. Therefore, the time distribution degree of each clustering cluster can be determined according to the time interval between each data segment in each clustering cluster. As an optional embodiment of the present invention, determining the time distribution degree of the clustering cluster according to the time interval between every two adjacent data segments in the clustering cluster and the second quantity includes: superimposing the time intervals between every two adjacent data segments in the clustering cluster to obtain a third superimposed value, and calculating the reciprocal of the third difference between the second quantity and a predetermined value; determining that the fifth product between the third superimposed value and the reciprocal of the third difference is the time distribution degree of the clustering cluster.

[0053] Specifically, the predetermined value in the embodiment of the present invention can take the value of 1. The embodiment of the present invention specifically calculates the time distribution degree of the clustering cluster by the following formula:

[0054]

[0055] In the above formula, is the time distribution degree of the th clustering cluster. is the second quantity of the data segments in the th clustering cluster. is the the time interval between the th data segment in a clustering cluster and the next data segment adjacent to the th data segment. The time distribution degree of the clustering cluster is described by the average value of the time intervals of adjacent data segments in the clustering cluster. The smaller this value is, the shorter the time interval between data segments, the more concentrated the time distribution, and the clustering cluster is normal. On the contrary, if the value is larger, it means the time interval is longer, and there may be abnormal discharges.

[0056] Furthermore, after obtaining the temporal continuity of each clustering cluster and the time distribution degree within the clustering cluster, it can be used as a weight to correct the abnormality degree of each obtained cycle, thereby obtaining the final abnormality degree of each cycle. As an optional embodiment of the present invention, correcting the abnormality degree of each cycle according to the temporal continuity and time distribution degree of the clustering cluster, and obtaining the final corrected abnormality degree of each cycle includes: calculating the third ratio between the time distribution degree of the clustering cluster where each cycle is located and the temporal continuity of the clustering cluster where each cycle is located, and the sixth product between each third ratio and the abnormality degree of each cycle; performing normalization processing on each sixth product to obtain the final corrected abnormality degree of each cycle.

[0057] Specifically, the embodiment of the present invention specifically uses the following formula to calculate the final abnormality degree:

[0058]

[0059] In the above formula, is the final corrected abnormality degree of the th cycle. is the abnormality degree of the th cycle. is the temporal continuity of the clustering cluster where the th cycle is located. is the time distribution degree corresponding to the clustering cluster where the th cycle is located. The function is used to perform normalization processing on . Among them, the abnormality degree of the th cycle is corrected by . If is larger, it means that the time distribution within the clustering cluster is wider, there is a longer time interval, indicating that it contains multiple abnormal discharges in the transformer, then the weight of the abnormality at this moment will be increased, indicating that the area with a longer time distribution is more likely to have abnormalities. If Smaller, indicating poor time continuity within the clustering cluster and large data fluctuations, suggesting that there may be abnormal discharges in the transformer in this clustering cluster, then the abnormal degree of this period will be amplified, indicating that a period with large fluctuations and instability is more likely to be an abnormal period.

[0060] Furthermore, the first threshold can be determined according to the actual situation, and the value in the embodiment of the present invention is 0.8. Cycles with the corrected final abnormal degree greater than 0.8 are recorded as abnormal fluctuation cycles.

[0061] S105, Determine the possibility of a fault occurring in the abnormal fluctuation cycle according to the amplitudes of the reference cycles of the abnormal fluctuation cycle and the time intervals between adjacent reference cycles.

[0062] Specifically, when partial discharges occur in an oil-immersed transformer, due to local high heat and ionization, the insulating materials inside the oil-immersed transformer will be further damaged. And along with the use of the oil-immersed transformer, due to the vibration and temperature changes during the operation of the oil-immersed transformer, some parts of the oil-immersed transformer will also be damaged. Due to these reasons, it is easier for partial discharges to occur in this oil-immersed transformer. Therefore, as the oil-immersed transformer is used, the frequency and amplitude of partial discharges in the oil-immersed transformer will become larger and larger. Therefore, the abnormal fluctuation cycles caused by abnormal discharges in the historical data of this oil-immersed transformer can be analyzed.

[0063] Furthermore, as an optional embodiment of the present invention, determining the possibility of a fault occurring in the abnormal fluctuation cycle according to the amplitudes of the reference cycles of the abnormal fluctuation cycle and the time intervals between adjacent reference cycles includes: using the cycles within the nearest adjacent time period of the abnormal fluctuation cycle as the reference cycles of the abnormal fluctuation cycle; calculating the fourth difference between the fifth quantity of the reference cycles and a predetermined value, and the fifth difference between the amplitudes of two adjacent reference cycles; calculating the fourth ratio between the fifth difference and the amplitude of the previous reference cycle among two adjacent reference cycles, and the second sum value between the fourth ratio and the reciprocal of the time interval between adjacent reference cycles; superimposing each second sum value to obtain a fourth superimposed value; determining that the reciprocal of the fourth superimposed value and the fourth difference is the possibility of a fault occurring in the abnormal fluctuation cycle.

[0064] Specifically, the embodiment of the present invention uses the cycles within the nearest short time of the abnormal fluctuation cycle as the reference cycles of the abnormal fluctuation cycle. According to the time points and amplitudes when the current data of the reference cycles of each abnormal fluctuation cycle appear in time series, the possibility of a fault occurring in each abnormal cycle is obtained. Among them, the predetermined value can be taken as 1, and the embodiment of the present invention specifically uses the following formula to calculate the possibility:

[0065]

[0066] In the above formula, is the probability of a fault occurring in the th abnormal fluctuation period. is the fifth quantity of the reference period of the th abnormal fluctuation period. and are respectively the amplitude of the current data of the th reference period and the amplitude of the current data of the th reference period of the th abnormal fluctuation period. is the time interval between the th reference period and its previous reference period of the th abnormal fluctuation period. Among them, is used to represent the degree of amplitude progression. The larger this value is, the more significant the amplitude progression is, and the greater the possibility that it belongs to an abnormal fault. is used to represent the degree of time progression. The larger this value is, the shorter the time interval is, and the higher the occurrence frequency of abnormal fluctuations is, and the greater the possibility that it belongs to an abnormal fault.

[0067] S106. Predict whether there is a potential fault in the substation equipment according to the probability of a fault occurring in the abnormal fluctuation period.

[0068] Specifically, the embodiment of the present invention can compare the probability of a fault occurring in the abnormal fluctuation period with a third threshold to determine whether there is a potential fault in the substation equipment. As an optional embodiment of the present invention, predicting whether there is a potential fault in the substation equipment according to the probability of a fault occurring in the abnormal fluctuation period includes: when the probability of a fault occurring in the most recent abnormal fluctuation period of the substation equipment is greater than the third threshold, it is determined that there is a potential fault in the substation equipment.

[0069] Specifically, for each oil-immersed transformer, its state is detected in real time. When it is monitored that the probability of a fault occurring in the most recent abnormal fluctuation period is greater than the third threshold when the oil-immersed transformer needs to be overhauled, it indicates that the risk of equipment failure is relatively high, maintenance is required, and relevant personnel are notified for overhaul and maintenance.

[0070] Further, the third threshold can be taken according to the actual situation. When the embodiment of the present invention determines the third threshold, the third threshold is the third difference between the mean value of the probability of a fault occurring in the most recent abnormal fluctuation period in the historical data of multiple substation equipment and a preset value.

[0071] Specifically, the preset value can be set according to the actual situation. In the embodiments of the present invention, the value is set to 0.2. After obtaining the probability of a fault occurring in each abnormal fluctuation period, the average value of each probability is calculated based on the probability of a fault occurring in the most recent abnormal fluctuation period before a fault occurred in the historical data of multiple oil-immersed transformers. Since maintenance cannot be carried out until the equipment reaches its limit, the value obtained by subtracting 0.2 from the average value is used as the third threshold for the oil-immersed transformer to require maintenance.

[0072] The embodiments of the present invention can monitor in real time the current data, load data, and dissolved gas concentration generated during the operation of power transformation equipment, and determine the abnormal fluctuation period in combination with the above multi-dimensional data. Specifically, the abnormal degree of each period is first analyzed through current data, and then the abnormal degree of each period is corrected using the load data and dissolved gas concentration to obtain the final abnormal degree. Finally, the probability of a fault occurring in the abnormal fluctuation period is determined, and based on this, it is predicted whether there is a potential fault in the power transformation equipment. In this way, through the analysis of multi-dimensional data, the initial stage of potential faults occurring in power transformation equipment can be effectively monitored and accurately predicted, improving the accuracy and reliability of power transformation equipment fault prediction, avoiding the expansion of equipment damage, and improving the reliability of the power system.

[0073] It should be noted that the above sequence of the embodiments of the present invention is only for description and does not represent the superiority or inferiority of the embodiments. The processes depicted in the drawings do not necessarily require the specific or continuous order shown to achieve the desired results. In some embodiments, multi-tasking and parallel processing are also possible or may be advantageous.

[0074] Each embodiment in this specification is described in a progressive manner. The same or similar parts between the embodiments can be referred to each other, and the key point of each embodiment is to illustrate the differences from other embodiments.

Claims

1. A method for predicting faults of substation equipment based on big data, characterized in that, The big data-based substation equipment fault prediction method includes: Obtaining current data, load data, and dissolved gas concentration of the substation equipment in multiple cycles; Determining the abnormality degree of the current cycle according to the amplitude of the maximum value point of the current cycle's current data, the first quantity of the maximum value points of the current cycle's current data, and the abnormality degree of the previous cycle adjacent to the current cycle; Clustering the load data and the dissolved gas concentration based on the correlation relationship between the load data and the dissolved gas concentration to obtain multiple clustering clusters; Correcting the abnormality degree of each cycle according to the load data and the dissolved gas concentration in each clustering cluster to obtain the final abnormality degree after correction of each cycle. Select the cycle with the final abnormality degree greater than the first threshold as the abnormal fluctuation cycle. The final abnormality degree is to correct the abnormality degree of each cycle according to the continuity of the clustering cluster in time and the time distribution degree of the clustering cluster. The continuity is determined according to the second quantity of multiple data segments in the clustering cluster, the third quantity of data points in the clustering cluster, the fourth quantity of data points in the data segment, and the change amount between every two adjacent data points in the data segment. The data segment is obtained by segmenting the load data and the dissolved gas concentration in each clustering cluster after sorting them in time sequence. The time distribution degree is determined according to the time interval between every two adjacent data segments in the clustering cluster and the second quantity; Determining the possibility of a fault occurring in the abnormal fluctuation cycle according to the amplitude of each reference cycle in the abnormal fluctuation cycle and the time interval between adjacent reference cycles; Predicting whether there is a potential fault in the substation equipment according to the possibility of a fault occurring in the abnormal fluctuation cycle.

2. The method for predicting faults of power transformation equipment based on big data according to claim 1, wherein, The determining the abnormality degree of the current cycle according to the amplitude of the maximum value point of the current cycle's current data, the first quantity of the maximum value points of the current cycle's current data, and the abnormality degree of the previous cycle adjacent to the current cycle includes: Calculating the first difference between the first quantity and a predetermined value, and the second difference between the predetermined value and the abnormality degree of the adjacent previous cycle; Calculating the third difference between the amplitude of each maximum value point of the current cycle's current data and the average value of the amplitudes of all maximum value points of the previous cycle's current data, and superimposing each third difference to obtain a first superimposed value; Calculating the first product between the first difference and the second difference and the second product between the first superimposed value and the abnormality degree of the adjacent previous cycle; Determining the first sum value between the first product and the second product as the abnormality degree of the current cycle.

3. The method for predicting faults of power transformation equipment based on big data according to claim 1, wherein, The segmenting the load data and the dissolved gas concentration in each clustering cluster after sorting them in time sequence to obtain multiple data segments includes: Determining the time interval between the load data and the dissolved gas concentration at the current time point in the clustering cluster and the load data and the dissolved gas concentration at the adjacent previous time point; When the time interval is greater than the second threshold, the load data and the dissolved gas concentration at the current time point are used as the starting point of a new segment of data points. When the time interval is not greater than the second threshold, the load data and the dissolved gas concentration at the current time point are classified into the data segment where the previous adjacent load data is located.

4. The method for predicting the faults of power transformation equipment based on big data according to claim 1, wherein The determination of the temporal continuity of the cluster according to the second quantity of the data segments in the cluster, the third quantity of the data points in the cluster, the fourth quantity of the data points in the data segment, and the change amount between every two adjacent data points in the data segment includes: Calculating a first ratio between the second quantity and the third quantity and a second ratio between the fourth quantity and the third quantity; Calculating the average change amount of the change amounts between every two adjacent data points in the data segment and the variance of the change amounts between every two adjacent data points in the data segment; Calculating a third product between the second ratio, the reciprocal of the average change amount, and the reciprocal of the variance, and superimposing the third products to obtain a second superimposed value; Performing a normalization process on the fourth product between the second superimposed value and the first ratio to obtain the temporal continuity of the cluster.

5. The method for predicting faults of power transformation equipment based on big data according to claim 1, wherein The determination of the temporal distribution degree of the cluster according to the time interval between every two adjacent data segments in the cluster and the second quantity includes: Superimposing the time intervals between every two adjacent data segments in the cluster to obtain a third superimposed value, and calculating the reciprocal of the third difference between the second quantity and a predetermined value; Determining that the fifth product between the third superimposed value and the reciprocal of the third difference is the temporal distribution degree of the cluster.

6. The method for predicting the faults of power transformation equipment based on big data according to claim 1, wherein, The correction of the abnormality degree of each cycle according to the temporal continuity of the cluster and the temporal distribution degree to obtain the final corrected abnormality degree of each cycle includes: Calculating a third ratio between the temporal distribution degree of the cluster where each cycle is located and the temporal continuity of the cluster where each cycle is located, and a sixth product between the third ratio and the abnormality degree of each cycle; Performing a normalization process on the sixth products to obtain the final corrected abnormality degree of each cycle.

7. The method for predicting faults of power transformation equipment based on big data according to any one of claims 1-6, characterized in that, The determination of the probability of a fault occurring in the abnormal fluctuation cycle according to the amplitudes of the reference cycles of the abnormal fluctuation cycle and the time intervals between adjacent reference cycles includes: Taking the cycles in the nearest adjacent time period of the abnormal fluctuation cycle as the reference cycles of the abnormal fluctuation cycle; Calculating a fourth difference between the fifth quantity of the reference cycles and a predetermined value, and a fifth difference between the amplitudes of two adjacent reference cycles; Calculating a fourth ratio between the fifth difference and the amplitude of the previous reference cycle among the two adjacent reference cycles, and a second sum value between the fourth ratio and the reciprocal of the time interval between the adjacent reference cycles; Superimposing the second sum values to obtain a fourth superimposed value; Determine that the reciprocal of the fourth superposition value and the fourth difference value is the probability of a fault occurring in the abnormal fluctuation period.

8. The method for predicting faults of power transformation equipment based on big data according to claim 1, characterized in that Predicting whether there is a potential fault in the substation equipment according to the probability of a fault occurring in the abnormal fluctuation period includes: When the probability of a fault occurring in the abnormal fluctuation period of the substation equipment in the most recent time is greater than a third threshold, determine that there is a potential fault in the substation equipment.

9. The method for predicting faults of power transformation equipment based on big data according to claim 8, characterized in that, The third threshold is the third difference between the mean value of the probabilities of faults occurring in the abnormal fluctuation periods of multiple substation equipment in historical data and a preset value.

Images