A transformer operation and maintenance method based on improved K-means

Abstract

The present invention proposes a transformer operation and maintenance method based on improved K-means, collects the operation time and operation time difference sequence of the transformer, calculates the reliability of the transformer using the three-parameter Weibull distribution, and evaluates the stability of the transformer by the ADF statistic; taking reliability and stability as key indicators, constructs a transformer data set, and uses kernel density estimation KDE and local outlier factor LOF to optimize the initial cluster center selection of the K-means algorithm, specifically by calculating the cluster center variation coefficient and the local reachable density, ensuring that the dispersion and density of the cluster center distribution are reasonable. After the transformer data set is clustered and analyzed by the improved K-means algorithm, the reliability and stability means of each cluster are calculated, and weights are assigned according to user needs, and the comprehensive score is calculated, and the comprehensive score is divided into five levels to evaluate the transformer status. Finally, different operation and maintenance strategies are formulated according to the evaluation results, which improves the accuracy and reliability of transformer operation and maintenance.

Description

Technical Field

[0001] The invention relates to a transformer operation and maintenance method based on improved K-means, belonging to the technical field of power metering online monitoring and operation and maintenance. Background Art

[0002] The instrument transformer is one of the key devices used for measurement and protection in the power system. In the operation and maintenance of the power system, the operation and maintenance strategy of the instrument transformer directly affects the stability and safety of the power system. Therefore, it is extremely important to implement a suitable operation and maintenance strategy for the instrument transformer.

[0003] For example, the patent document with publication number CN114626769A discloses a method for operation and maintenance of a capacitor voltage transformer, including: obtaining the operation and maintenance data of the target capacitor voltage transformer, and calculating the stability, importance and fault degree of the target capacitor voltage transformer based on the operation and maintenance data; the operation and maintenance data at least includes error assessment data, basic data and historical fault data; calculating the operation and maintenance coefficient of the target capacitor voltage transformer based on its stability, importance and fault degree; determining the operation and maintenance strategy of the target capacitor voltage transformer based on the current operation and maintenance coefficient and its preset threshold. This application integrates multiple indicators such as the stability, importance and fault degree of the capacitor voltage transformer, and calculates the operation and maintenance coefficient and determines the operation and maintenance strategy based on them, which effectively guides the operation and maintenance management of the capacitor voltage transformer, greatly reduces the blindness of operation and maintenance, and saves human resource costs. The above invention calculates the operation and maintenance coefficient based on stability, importance and degree of failure, and determines the operation and maintenance strategy based on the operation and maintenance coefficient. However, the multi-dimensional analysis method usually focuses on the overall trend and is difficult to capture the uniqueness of each device in the subtle state. This means that even if the devices perform similarly in macro indicators, their micro states may have significant differences, and these differences make it impossible for the invention to accurately and effectively identify the differences in state, limiting the personalization and pertinence of the operation and maintenance strategy. Therefore, there is an urgent need for an operation and maintenance method that can more finely identify the subtle differences in device states, subdivide device states, thereby improving the accuracy of state recognition, and formulating more targeted operation and maintenance strategies. Summary of the invention

[0004] In order to solve the above problems existing in the prior art, the present invention proposes a transformer operation and maintenance method based on improved K-means.

[0005] The technical solution of the present invention is as follows:

[0006] A transformer operation and maintenance method based on improved K-means, the method comprising:

[0007] Obtaining the operating time and operating time difference sequence of the transformer to be tested; calculating the reliability and stability of the transformer to be tested according to the operating time and operating time difference sequence, and constructing a transformer data set to be tested according to the reliability and stability;

[0008] The K-means clustering algorithm is improved by using kernel density estimation KDE and local outlier factor LOF. The data set of the transformer to be tested is clustered based on the improved K-means clustering algorithm to obtain clustering results. The status of the transformer is evaluated according to the clustering results to obtain evaluation results.

[0009] Different operation and maintenance strategies are formulated for the transformer to be tested according to the evaluation results, and the transformer to be tested is operated and maintained according to the operation and maintenance strategies.

[0010] As a preferred embodiment of the present invention, the method further comprises:

[0011] Obtain failure time data of a failed transformer of the same model as the transformer to be tested, wherein the failure time data includes Based on the failure time data, the likelihood function of the three-parameter Weibull distribution is obtained by using the maximum likelihood estimation MLE calculation , expressed as:

[0012] ;

[0013] ;

[0014] In the formula, is the total number of failure time data, is the probability density function of the three-parameter Weibull distribution; the three parameters include , and , is the size parameter, is the shape parameter, is the threshold parameter; For the failure time;

[0015] The logarithmic sum of the probability density function is used to obtain the log-likelihood function , expressed as:

[0016] ;

[0017] Compute the gradient of the negative log-likelihood function , the negative log-likelihood function Recorded as , the calculation gradient is expressed as:

[0018] ;

[0019] ;

[0020] ;

[0021] ;

[0022] Use the quasi-Newton method BFGS to update the three parameters, specifically:

[0023] Initial three parameters and the approximate Hessian matrix , where the identity matrix is ​​chosen as the approximate Hessian matrix;

[0024] In the three parameters In the iteration, the search direction is calculated , expressed as:

[0025] ;

[0026] in, For the current three parameters The gradient at For the current three parameters The approximate Hessian matrix at ;

[0027] Determining step size by line search , according to the search direction and step length Update three parameters , expressed as:

[0028] ;

[0029] Calculate the difference between step size and gradient respectively and , expressed as:

[0030] ;

[0031] ;

[0032] In the formula, For the first of the three parameters In the iteration, the three parameters The gradient at

[0033] According to the difference and Update the approximate Hessian matrix , expressed as:

[0034] ;

[0035] In the formula, Represents a transpose operation;

[0036] Check whether the iteration termination condition is met, that is, check whether the three parameter changes or the gradient norm are less than the preset threshold. If the iteration termination condition is met, stop the iteration; otherwise, return to the calculation search direction and continue iterating until the iteration termination condition is met;

[0037] Get the final three parameters , and .

[0038] As a preferred implementation mode of the present invention, the reliability of the transformer to be tested is calculated according to the running time, specifically:

[0039] According to the running time and the final three parameters , and Calculate the failure probability density for the three-parameter Weibull distribution , expressed as:

[0040] ;

[0041] In the formula, is the running time;

[0042] The failure probability density The corresponding distribution function The formula is:

[0043] ;

[0044] According to the above distribution function Calculate the reliability function , expressed as:

[0045] .

[0046] As a preferred implementation mode of the present invention, the stability of the transformer to be tested is calculated according to the running time ratio difference sequence, specifically:

[0047] Get the current detection time of the transformer under test and its first 50 moments The ratio difference value is used to construct the operating time ratio difference sequence of the transformer to be tested. , the running time difference sequence Recorded as ;

[0048] Calculate the ADF statistic based on the running time difference series , expressed as:

[0049] ;

[0050] The ADF statistic Perform the inversion operation to obtain the stability evaluation function , expressed as:

[0051] .

[0052] As a preferred embodiment of the present invention, according to the reliability and stability of the transformer to be tested, a transformer data set to be tested is constructed. ,in, , , where For the The reliability and stability of the transformer under test, is the total number of transformers to be tested.

[0053] As a preferred embodiment of the present invention, the initialization cluster center strategy of the K-means clustering algorithm is improved by using kernel density estimation KDE and local outlier factor LOF. The specific steps include:

[0054] Randomly select K data points from the transformer data set to be tested as the initial cluster centers, and record the coordinates of each cluster center;

[0055] Calculate the coefficient of variation of the cluster center, compare the coefficient of variation of the cluster center with a preset dispersion threshold, if the coefficient of variation of the cluster center is less than the dispersion threshold, repeat the steps of randomly selecting a data point as the initial cluster center and calculating the coefficient of variation of the cluster center until the coefficient of variation of the cluster center is greater than or equal to the dispersion threshold, wherein the preset dispersion threshold is the coefficient of variation of the transformer data set to be tested;

[0056] The local reachable density value of each cluster center is calculated by combining the local reachable density calculation in the local outlier factor LOF and the Gaussian kernel density estimation KDE;

[0057] The local reachable density value of each cluster center is compared with the preset density threshold. If the local reachable density value of the current cluster center is greater than or equal to the preset density threshold, the current cluster center is retained; if the local reachable density value of the current cluster center is less than the preset density threshold, the current cluster center is replaced by the data point that is closest to the current cluster center and whose local reachable density value is greater than the preset density threshold, wherein the preset density threshold is 10% of the maximum local reachable density of the transformer data set to be tested.

[0058] As a preferred embodiment of the present invention, the specific steps of calculating the coefficient of variation of cluster centers include:

[0059] Calculate the mean of the coordinate dimensions corresponding to all cluster centers ;

[0060] Calculate the standard deviation of each cluster center to the mean ;

[0061] According to the mean of all cluster centers and the standard deviation of each cluster center to the mean , calculate the coefficient of variation , expressed as:

[0062] .

[0063] As a preferred embodiment of the present invention, the local reachable density calculation in the local outlier factor LOF is combined with the Gaussian kernel density estimation KDE to calculate the local reachable density value of each cluster center. The specific steps include:

[0064] The Gaussian kernel function is used to calculate the global density estimate of the current cluster center, which is expressed as:

[0065] ;

[0066] In the formula, For the current cluster center The Gaussian kernel density estimate KDE at the current cluster center The global density estimate of is bandwidth; For the The reliability and stability of the transformer under test;

[0067] Calculate the local reachable density LRD of the current cluster center, expressed as:

[0068] ;

[0069] In the formula, To consider the distance from the current cluster center of The cluster center is calculated by the nearest neighbor The local reachable density LRD, The current cluster center No. Distance neighborhood, For the From any point in the neighborhood, The current cluster center To data point The reachable distance;

[0070] When calculating the local reachable density of the current cluster center, the distance from the current cluster center is considered. nearest neighbors, for the Each data point in the nearest neighbor is assigned a weight, which is calculated based on the global density estimation of the current cluster center and expressed as:

[0071] ;

[0072] In the formula, To assign data points The weight of for Any data point in the nearest neighbor Global density estimation of ;

[0073] In giving After assigning weights to each data point in the nearest neighbors, calculate the weighted local reachability density of the current cluster center , expressed as:

[0074] .

[0075] As a preferred implementation mode of the present invention, the status of the mutual inductor is evaluated according to the clustering result, specifically:

[0076] After the transformer data set to be tested is clustered by the improved K-means clustering algorithm, K clusters and their corresponding cluster centers are obtained;

[0077] Calculate the mean of reliability and stability for each cluster separately and ;

[0078] According to the user's requirements for reliability and stability, the mean and Assign weights and calculate the comprehensive score, expressed as follows:

[0079] ;

[0080] In the formula, For the comprehensive rating, is the reliability mean weight, is the stability mean weight;

[0081] The comprehensive scores are sorted in order from large to small, and the sorting results are equally divided into five levels, the higher the ranking, the lower the number, and the comprehensive scores and corresponding ranking levels are the evaluation results.

[0082] As a preferred embodiment of the present invention, different operation and maintenance strategies are formulated for the transformer under test according to the evaluation results, specifically, basic maintenance is performed at the first ranking level; preventive maintenance is performed at the second ranking level; corrective maintenance is performed at the third ranking level; corrective maintenance is performed at the fourth ranking level; asset management and planned maintenance is performed at the fifth ranking level, and the transformer under test is operated and maintained according to the ranking level and the corresponding operation and maintenance strategy.

[0083] The present invention has the following beneficial effects:

[0084] 1. The present invention is a transformer operation and maintenance method based on improved K-means. The K-means algorithm is improved by introducing kernel density estimation KDE and local outlier factor LOF, including calculating the coefficient of variation of cluster centers and combining local reachable density calculation, optimizing the initialization cluster center strategy of the K-means algorithm, reducing the uncertainty of initial cluster center selection, and improving the efficiency and stability of the clustering algorithm. Therefore, in practical applications, the transformer data set can be clustered more effectively, thereby accurately performing state evaluation and prediction, and enhancing the guidance of transformer operation and maintenance;

[0085] 2. The present invention is a transformer operation and maintenance method based on improved K-means, which combines the precise calculation based on the running time and the running time difference sequence, constructs a data set using the two key indicators of reliability and stability, and formulates different operation and maintenance strategies based on the evaluation results of these indicators, making the operation and maintenance strategies more accurate and scientific, and meeting the maintenance needs under different transformer states. BRIEF DESCRIPTION OF THE DRAWINGS

[0086] Figure 1 The figure is a flow chart of the method of the present invention. DETAILED DESCRIPTION

[0087] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0088] It should be understood that the step numbers used in this document are only for convenience of description and are not intended to limit the order in which the steps are executed.

[0089] It should be understood that the terms used in the present specification are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the present specification and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include plural forms.

[0090] The terms “include” and “comprising” indicate the presence of described features, integers, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, components and / or combinations thereof.

[0091] The term "and / or" means and includes any and all possible combinations of one or more of the associated listed items.

[0092] Embodiment 1:

[0093] See also Figure 1 , a transformer operation and maintenance method based on improved K-means, comprising the following steps:

[0094] S1. Obtaining the operating time and operating time difference sequence of the transformer to be tested; calculating the reliability and stability of the transformer to be tested according to the operating time and operating time difference sequence, and constructing a transformer data set to be tested according to the reliability and stability;

[0095] S11. Reliability of the transformer to be tested:

[0096] Transformer reliability usually refers to the ability of the transformer to operate without failure under its expected working conditions. As a continuous probability distribution, the three-parameter Weibull distribution has good flexibility and can adapt to different types of data distributions. , shape parameters and threshold parameters , can fit a variety of different life distribution patterns, including early failure period, random failure period and wear period. In reliability engineering, Weibull distribution is widely used to describe the failure time of electronic components, mechanical parts and electrical equipment. As a kind of electrical equipment, the failure mode and life characteristics of the transformer also meet the description ability of Weibull distribution. And the threshold parameter in the three-parameter Weibull distribution is The early failure of the product can be described, which is very important for electrical equipment such as transformers, because early failure will seriously affect the reliability and service life of the equipment. Therefore, in this embodiment, the three-parameter Weibull distribution is selected to calculate the reliability of the transformer to be tested.

[0097] S111. To calculate the reliability of a transformer based on the three-parameter Weibull distribution, we must first clarify the three parameters of the three-parameter Weibull distribution: size parameter , shape parameters and threshold parameters ,in:

[0098] , which is often negatively correlated with the transformer load. This parameter affects the distribution function The average and scale of ; To describe the dispersion of the measured values, The failure function is a decreasing function, indicating that the transformer is in the running-in period. and They respectively indicate that the failure function is a constant function and an increasing function, that is, the transformer is in a short failure period and a loss period; is the failure threshold and can also be considered as the translation parameter of the probability density curve.

[0099] Therefore, before calculating the reliability of the transformer to be tested, the method further includes:

[0100] Obtain failure time data of a failed transformer of the same model as the transformer to be tested, wherein the failure time data includes Based on the failure time data, the likelihood function of the three-parameter Weibull distribution is obtained by using the maximum likelihood estimation MLE calculation , expressed as:

[0101] ;

[0102] ;

[0103] In the formula, is the total number of failure time data, is the probability density function of the three-parameter Weibull distribution; the three parameters include , and , is the size parameter, is the shape parameter, is the threshold parameter; For the failure time;

[0104] The logarithmic sum of the probability density function is used to obtain the log-likelihood function , expressed as:

[0105] ;

[0106] In order to find the parameter value that best fits the observed data, this implementation maximizes the likelihood function of the data and uses the quasi-Newton method BFGS to solve the maximum likelihood estimation of the three-parameter Weibull distribution, specifically:

[0107] Compute the gradient of the negative log-likelihood function , the negative log-likelihood function Recorded as , the calculation gradient is expressed as:

[0108] ;

[0109] ;

[0110] ;

[0111] ;

[0112] The quasi-Newton method BFGS is used to update the three parameters. BFGS updates the parameters through iterations, including:

[0113] Initial three parameters and the approximate Hessian matrix , where the identity matrix is ​​chosen as the approximate Hessian matrix;

[0114] In the three parameters In the iteration, the search direction is calculated , expressed as:

[0115] ;

[0116] in, For the current three parameters The gradient at For the current three parameters The approximate Hessian matrix at ;

[0117] Determining step size by line search , so that the objective function minimum, wherein the line search in this implementation is an Armijo line search;

[0118] According to the search direction and step length Update three parameters , expressed as:

[0119] ;

[0120] Calculate the difference between step size and gradient respectively and , expressed as:

[0121] ;

[0122] ;

[0123] In the formula, For the first of the three parameters In the iteration, the three parameters The gradient at

[0124] According to the difference and Update the approximate Hessian matrix , expressed as:

[0125] ;

[0126] In the formula, Represents a transpose operation;

[0127] Check whether the iteration termination condition is met, that is, check whether the three parameter changes or the gradient norm are less than the preset threshold. If the iteration termination condition is met, stop the iteration; otherwise, return to the calculation search direction and continue iterating until the iteration termination condition is met;

[0128] Get the final parameters, that is, the three parameters of the three-parameter Weibull distribution , and .

[0129] S112, calculating the reliability of the transformer to be tested according to the operating time;

[0130] According to the running time and the three parameters of the three-parameter Weibull distribution obtained in step S111 , and , calculate the failure probability density of the three-parameter Weibull distribution , expressed as:

[0131] ;

[0132] In the formula, is the running time;

[0133] The failure probability density The corresponding distribution function The formula is:

[0134] ;

[0135] According to the above distribution function Calculate the reliability function , expressed as:

[0136] ;

[0137] S12. Stability of the transformer to be tested:

[0138] The stability of the transformer focuses on the system's responsiveness to external influences and the changes in data over a period of time. This embodiment measures the stability of the transformer based on the stability of the ratio difference time series data over a period of time. Specifically, this embodiment selects the ADF unit root test for the stability of the time series data. The principle is:

[0139] (1) Assume that the sequence has a unit root, which is the null hypothesis, that is, the time series is non-stationary;

[0140] (2) Calculate the significance test statistic, also known as the ADF statistic;

[0141] (3) Compare the significance test statistic with the critical value:

[0142] If the significance test statistic is less than the given critical value (1%, 5%, 10% confidence level), the null hypothesis is rejected and the series is considered to be stationary.

[0143] If the significance test statistic is greater than or equal to the given critical value, the null hypothesis cannot be rejected and the series is considered to be non-stationary.

[0144] According to the above principle, the operating time ratio difference sequence of the transformer to be tested is obtained, where the ratio difference is defined as the difference between the output value of the transformer at a certain moment and the ideal output value. Assume Indicates at time The difference in time, = Actual output value The ideal output value is worth noting that For example only, the running time in step S11 are not the same; the ratio difference sequence refers to a sequence of ratio difference values ​​at a series of time points. In this embodiment, the current detection time of the transformer to be tested is obtained. and its first 50 moments The ratio difference value is used to construct the operating time ratio difference sequence of the transformer to be tested. , the running time difference sequence Recorded as ;

[0145] Calculate the ADF statistic based on the running time difference series , expressed as:

[0146] ;

[0147] The ADF statistic Perform the inversion operation to obtain the stability evaluation function , expressed as:

[0148] ;

[0149] Evaluate the function based on stability The value of determines the stability of the transformer to be tested:

[0150] When the stability evaluation function The larger the value of , the smaller the ADF statistic, and the more likely it is to reject the null hypothesis (the existence of a unit root), that is, to believe that the sequence is stationary, and therefore the higher the stability;

[0151] When the stability evaluation function The smaller the value of , the larger the ADF statistic, and the lower the possibility of rejecting the null hypothesis (the existence of a unit root), that is, the sequence is considered to be non-stationary, and therefore the stability is lower.

[0152] S13, constructing a two-dimensional transformer data set to be tested according to the reliability and stability obtained in step S11 and step S12 ,in, , , where For the The reliability and stability of the transformer under test, is the total number of transformers to be tested;

[0153] S2, using kernel density estimation KDE and local outlier factor LOF to improve the K-means clustering algorithm;

[0154] The K-means clustering algorithm is an unsupervised method and an effective tool for processing unlabeled data. It automatically divides data points into different clusters through similarity measurement, thereby helping to discover potential patterns in the data. When processing the transformer data set to be tested, the K-means model can be selected to effectively group data objects according to the characteristics of reliability and stability without prior knowledge. The K-means clustering algorithm performs well when processing large-scale data. As the data set increases, the efficiency of the algorithm remains high, which is suitable for the evaluation requirements of the transformer data set to be tested described in this embodiment. Therefore, in this embodiment, the K-means clustering model is used to cluster the transformer data set to be tested. Perform clustering;

[0155] The goal of the K-means clustering algorithm is to divide the data set into K clusters so that the sum of the squares of the distances from each data point to the center of its cluster is minimized. Therefore, initializing the cluster centers is a very critical step. Reasonable initialization of the cluster centers can reduce the number of iterations required for the algorithm to reach a stable state and reduce computational costs. It can also reduce the uncertainty caused by random initialization, making the algorithm more robust to changes in input data. Generally, places with dense data points are more likely to be the actual centers of each cluster; the distance between each cluster center should cover more data space to avoid falling into a local optimum. To this end, this embodiment follows the principle of "the density of cluster centers should be relatively large, and the cluster centers should be relatively dispersed" to improve the strategy for initializing cluster centers. The specific steps include:

[0156] S21, randomly selecting K data points from the transformer data set to be tested as initial cluster centers, and recording the coordinates of each cluster center. Preferably, in this embodiment, the K value is set to 5;

[0157] S22. Calculate the coefficient of variation of cluster centers:

[0158] S221. Calculate the mean of all cluster centers , that is, the mean of each coordinate dimension;

[0159] S222, calculate the standard deviation from each cluster center to the mean ;

[0160] S223. The coefficient of variation is a statistical indicator used to measure the degree of dispersion of data in a data set. It is the ratio of the standard deviation to the mean, usually expressed as a percentage, based on the mean of all cluster centers. and the standard deviation of each cluster center to the mean , calculate the coefficient of variation , expressed as:

[0161] ;

[0162] S23, using the coefficient of variation to measure the dispersion between data points, specifically comparing the coefficient of variation of the cluster center with a preset dispersion threshold. If the coefficient of variation of the cluster center is less than the dispersion threshold, repeating steps S21 and S22 until the coefficient of variation of the cluster center is greater than or equal to the dispersion threshold, wherein the preset dispersion threshold is the coefficient of variation of the transformer data set to be tested, and the specific calculation method is the same as the principle of the calculation method of the coefficient of variation of the cluster center, which will not be repeated here;

[0163] S24. Combine the local reachable density calculation in the local outlier factor LOF with the Gaussian kernel density estimation KDE to calculate the local reachable density value of each cluster center:

[0164] S241. Kernel density estimation KDE is a non-parametric method. Its basic idea is to assume that data points are independently sampled from an unknown probability density function, and its goal is to estimate this unknown probability density function. Kernel density estimation does not rely on any specific probability distribution assumption, so it can be applied to any type of data. In this embodiment, the kernel function uses the Gaussian kernel function, which is expressed as follows:

[0165] ;

[0166] In the formula, For the current cluster center The Gaussian kernel density estimate KDE at the current cluster center The global density estimate of The bandwidth is a great influence on the density estimation result. The optimal bandwidth is usually selected by cross-validation and other methods, which will not be described in detail in this embodiment. For the The reliability and stability of the transformer under test;

[0167] S242. Local outlier factor LOF is an anomaly detection algorithm. The local reachable density LRD of a data point can be calculated internally. The size of the LRD value can measure the local density of a data point based on the distance between the data point and the points in its neighborhood. The higher the density between the point and the surrounding points, the larger the LRD value; the lower the density between the point and the surrounding points, the smaller the LRD value.

[0168] Calculate the local reachable density LRD of the current cluster center, expressed as:

[0169] ;

[0170] In the formula, To consider the distance from the current cluster center of The cluster center is calculated by the nearest neighbor The local reachable density LRD, The current cluster center No. Distance neighborhood, For the From any point in the neighborhood, The current cluster center To data point The reachable distance;

[0171] S243, combining the local reachable density calculation in the local outlier factor LOF with the Gaussian kernel density estimation KDE to calculate the local reachable density value of each cluster center, specifically:

[0172] When calculating the local reachable density of the current cluster center, the distance from the current cluster center is considered. nearest neighbors, for the Each data point in the nearest neighbor is assigned a weight, which is calculated based on the global density estimation of the current cluster center and expressed as:

[0173] ;

[0174] In the formula, To assign data points The weight of for Any data point in the nearest neighbor The calculation principle of the global density estimation of the current cluster center is the same as that of the global density estimation of the current cluster center in step S241, which will not be repeated here;

[0175] In giving After assigning weights to each data point in the nearest neighbors, calculate the weighted local reachability density of the current cluster center , the local reachable density The weights of the data points in the neighborhood are taken into account and expressed as:

[0176] ;

[0177] S25, compare the local reachable density value of each cluster center with the preset density threshold, if the local reachable density value of the current cluster center is greater than or equal to the preset density threshold, retain the current cluster center; if the local reachable density value of the current cluster center is less than the preset density threshold, replace the current cluster center with the data point closest to the current cluster center and whose local reachable density value is greater than the preset density threshold, wherein the preset density threshold is 10% of the maximum local reachable density of the transformer data set to be tested, and the calculation principle of the local reachable density of the transformer data set to be tested is the same as step S242, which will not be repeated here;

[0178] S3, clustering the transformer data set to be tested based on the improved K-means clustering algorithm to obtain clustering results, performing status evaluation of the transformer according to the clustering results to obtain evaluation results, formulating different operation and maintenance strategies for the transformer to be tested according to the evaluation results, and the transformer to be tested is operated and maintained according to the operation and maintenance strategies;

[0179] S31. After the transformer data set to be tested is clustered by the improved K-means clustering algorithm, K clusters and their corresponding cluster centers are obtained. In this embodiment, since the K value is set to 5, 5 clusters are obtained. Since the cluster center data of each cluster represents the current state of all transformers to be tested in the cluster, the mean of the reliability and stability of the cluster is calculated according to the reliability and stability of each transformer to be tested in the cluster, which is the mean of the reliability and stability of the cluster center. and ;

[0180] According to the user's requirements for reliability and stability, the mean and Assign weights and calculate the comprehensive score, expressed as follows:

[0181] ;

[0182] In the formula, For the comprehensive rating, is the reliability mean weight, is the stability mean weight;

[0183] The comprehensive scores are sorted from large to small, and the sorting results are equally divided into five levels. The higher the ranking, the lower the score. Each level corresponds to a different operation and maintenance strategy.

[0184] S32. The specific operation and maintenance strategy is:

[0185] The first level of sorting is basic maintenance, including appearance inspection, transformer cleaning, fastening inspection and operation status inspection. The appearance inspection includes checking whether the transformer housing, grounding wire, joints and insulation parts are damaged, aged or corroded; the transformer cleaning includes removing dust and dirt on the surface of the transformer to ensure that the equipment is clean and prevent flashover or breakdown caused by accumulated dirt; the fastening inspection includes checking whether the bolts of each connection part are loose and tightening them when necessary; the operation status inspection includes checking whether the current and voltage parameters of the transformer are within the normal range during operation;

[0186] The second level is sorted for preventive maintenance, including insulation resistance test, dielectric loss test, oil sample detection, partial discharge test and preventive test, wherein the insulation resistance test includes regularly measuring the insulation resistance of the primary and secondary sides of the transformer to determine whether the insulation condition is good; the dielectric loss test includes measuring the dielectric loss angle of the voltage transformer to evaluate the degree of insulation aging; the oil sample detection is for oil-immersed transformers, and oil sample detection is required to be carried out regularly to analyze whether the oil quality has deteriorated; the partial discharge test specifically determines whether there are potential insulation defects inside the transformer by testing the partial discharge situation; the preventive test includes regularly conducting preventive electrical tests such as withstand voltage test and ratio test to ensure the stability of the transformer electrical performance;

[0187] The three levels of corrective maintenance are carried out, including fault diagnosis, component replacement, oil treatment or replacement, and post-repair testing. The fault diagnosis is specifically to determine the fault type and cause, such as turn-to-turn short circuit, insulation breakdown, etc., through electrical testing and appearance inspection; the component replacement is specifically to replace or repair damaged windings, insulating materials, seals, etc.; the oil treatment or replacement is for oil-immersed transformers, to treat or replace deteriorated insulating oil; the post-repair testing is specifically to carry out a series of electrical tests after repair to ensure that the transformer has restored normal performance;

[0188] The four levels of corrective maintenance are sorted, including performance optimization, technical transformation, online monitoring, and technical consultation and support. The performance optimization is to adjust the operating parameters of the transformer according to its actual operating conditions to ensure the best performance; the technical transformation is to upgrade or transform old transformers or equipment with backward technology, such as replacing them with more efficient or environmentally friendly models; the online monitoring is to introduce an online monitoring system to monitor the operating status of the transformer in real time, such as partial discharge, temperature, vibration and other parameters; the technical consultation and support is to cooperate with manufacturers or professional third parties to provide technical consultation to ensure that the transformer operates in the best condition;

[0189] The asset management and planned maintenance are carried out in five levels, including asset ledger management, life cycle management, full life cycle cost analysis, regular maintenance plan and emergency plan. Among them, the asset ledger management includes establishing a detailed transformer ledger to record its model, specification, commissioning time, maintenance history and other information; the life cycle management is specifically to plan the replacement or upgrade time of the transformer according to its operating time, status and historical maintenance records; the full life cycle cost analysis is specifically to conduct a comprehensive analysis of the initial investment, operation and maintenance costs, failure losses, etc. of the transformer to optimize the efficiency of asset use; the regular maintenance plan is specifically to formulate an annual or quarterly maintenance plan to ensure that the transformer is fully maintained and maintained during operation; the emergency plan is specifically to formulate an emergency plan for possible serious failures to ensure rapid response and processing in the event of sudden failures;

[0190] The transformer to be tested is operated and maintained according to the sorting level and the corresponding operation and maintenance strategy. Each level corresponds to different goals and operation contents, covering the entire life cycle of the transformer from daily basic maintenance to long-term asset management. This embodiment can effectively improve the performance of the transformer, optimize its operating efficiency and extend its service life through a systematic maintenance plan.

[0191] In the embodiments of the present application, "at least one" refers to one or more, and "more than one" refers to two or more. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent the existence of A alone, the existence of A and B at the same time, and the existence of B alone. Among them, A and B can be singular or plural. The character " / " generally indicates that the previous and next associated objects are in an "or" relationship. "At least one of the following" and similar expressions refer to any combination of these items, including any combination of single or plural items. For example, at least one of a, b and c can be represented by: a, b, c, a and b, a and c, b and c, or a and b and c, where a, b, c can be single or multiple.

[0192] Those of ordinary skill in the art will appreciate that the various units and algorithm steps described in the embodiments disclosed herein can be implemented in a combination of electronic hardware, computer software, and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0193] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0194] In several embodiments provided in the present application, if any function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application can be essentially or partly embodied in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory; hereinafter referred to as: ROM), random access memory (Random Access Memory; hereinafter referred to as: RAM), disk or optical disk, and other media that can store program codes.

[0195] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A transformer operation and maintenance method based on improved K-means, characterized in that: The method comprises: Obtaining the operating time and operating time difference sequence of the transformer to be tested; calculating the reliability and stability of the transformer to be tested according to the operating time and operating time difference sequence, and constructing a transformer data set to be tested according to the reliability and stability; The K-means clustering algorithm is improved by using kernel density estimation KDE and local outlier factor LOF. The data set of the transformer to be tested is clustered based on the improved K-means clustering algorithm to obtain clustering results. The status of the transformer is evaluated according to the clustering results to obtain evaluation results. Formulate different operation and maintenance strategies for the transformer to be tested according to the evaluation results, and the transformer to be tested is operated and maintained according to the operation and maintenance strategies; Among them, the kernel density estimation KDE and local outlier factor LOF are used to improve the initialization cluster center strategy of the K-means clustering algorithm. The specific steps include: Randomly select K data points from the transformer data set to be tested as the initial cluster centers, and record the coordinates of each cluster center; Calculate the coefficient of variation of the cluster center, compare the coefficient of variation of the cluster center with a preset dispersion threshold, if the coefficient of variation of the cluster center is less than the dispersion threshold, repeat the steps of randomly selecting a data point as the initial cluster center and calculating the coefficient of variation of the cluster center until the coefficient of variation of the cluster center is greater than or equal to the dispersion threshold, wherein the preset dispersion threshold is the coefficient of variation of the transformer data set to be tested; The local reachable density value of each cluster center is calculated by combining the local reachable density in the local outlier factor LOF and the Gaussian kernel density estimation KDE. The specific steps include: The Gaussian kernel function is used to calculate the global density estimate of the current cluster center, which is expressed as: ； In the formula, For the current cluster center The Gaussian kernel density estimate KDE at the current cluster center The global density estimate of is the bandwidth, For the The reliability and stability of the transformer under test, For the The reliability and stability of the transformer under test, is the total number of transformers to be tested; Calculate the local reachable density LRD of the current cluster center, expressed as: ； In the formula, To consider the distance from the current cluster center of The cluster center is calculated by the nearest neighbor The local reachable density LRD, The current cluster center No. Distance neighborhood, For the From any point in the neighborhood, The current cluster center To data point The reachable distance; When calculating the local reachable density of the current cluster center, the distance from the current cluster center is considered. nearest neighbors, for the Each data point in the nearest neighbor is assigned a weight, which is calculated based on the global density estimation of the current cluster center and expressed as: ； In the formula, To assign data points The weight of for Any data point in the nearest neighbor Global density estimation of ; In giving After assigning weights to each data point in the nearest neighbors, calculate the weighted local reachability density of the current cluster center , expressed as: ； The local reachable density value of each cluster center is compared with the preset density threshold. If the local reachable density value of the current cluster center is greater than or equal to the preset density threshold, the current cluster center is retained; if the local reachable density value of the current cluster center is less than the preset density threshold, the current cluster center is replaced by the data point that is closest to the current cluster center and whose local reachable density value is greater than the preset density threshold, wherein the preset density threshold is 10% of the maximum local reachable density of the transformer data set to be tested.

2. According to claim 1, a transformer operation and maintenance method based on improved K-means is characterized in that: The method further comprises: Obtain failure time data of a failed transformer of the same model as the transformer to be tested, wherein the failure time data includes Based on the failure time data, the likelihood function of the three-parameter Weibull distribution is obtained by using the maximum likelihood estimation MLE calculation , expressed as: ； ； In the formula, is the total number of failure time data, is the probability density function of the three-parameter Weibull distribution; the three parameters include , and , is the size parameter, is the shape parameter, is the threshold parameter; For the failure time; The logarithmic sum of the probability density function is used to obtain the log-likelihood function , expressed as: ； Compute the gradient of the negative log-likelihood function , the negative log-likelihood function Recorded as , the calculation gradient is expressed as: ； ； ； ； Use the quasi-Newton method BFGS to update the three parameters, specifically: Initial three parameters and the approximate Hessian matrix , where the identity matrix is ​​chosen as the approximate Hessian matrix; In the three parameters In the iteration, the search direction is calculated , expressed as: ； in, For the current three parameters The gradient at For the current three parameters The approximate Hessian matrix at ; Determining step size by line search , according to the search direction and step length Update three parameters , expressed as: ； Calculate the difference between step size and gradient respectively and , expressed as: ； ； In the formula, For the first of the three parameters In the iteration, the three parameters The gradient at According to the difference and Update the approximate Hessian matrix , expressed as: ； In the formula, Represents a transpose operation; Check whether the iteration termination condition is met, that is, check whether the three parameter changes or the gradient norm are less than the preset threshold. If the iteration termination condition is met, stop the iteration; otherwise, return to the calculation search direction and continue iterating until the iteration termination condition is met; Get the final three parameters , and .

3. The transformer operation and maintenance method based on improved K-means according to claim 2 is characterized in that: The reliability of the transformer to be tested is calculated based on the running time, specifically: According to the running time and the final three parameters , and Calculates the failure probability density of a three-parameter Weibull distribution , expressed as: ； In the formula, is the running time; The failure probability density The corresponding distribution function The formula is: ； According to the above distribution function Calculate the reliability function , expressed as: 。 4. The method for operating and maintaining a mutual inductor based on improved K-means according to claim 3, characterized in that: The stability of the transformer to be tested is calculated according to the running time difference sequence, specifically: Get the current detection time of the transformer under test and its first 50 moments The ratio difference value is used to construct the operating time ratio difference sequence of the transformer to be tested. , the running time difference sequence Recorded as ; Calculate the ADF statistic based on the running time difference series , expressed as: ； For the ADF statistic Perform the inversion operation to obtain the stability evaluation function , expressed as: 。 5. The method for operating and maintaining a mutual inductor based on improved K-means according to claim 4, characterized in that: According to the reliability and stability of the transformer to be tested, the transformer data set to be tested is constructed ,in, , , where For the The reliability and stability of the transformer under test, is the total number of transformers to be tested.

6. The method for operating and maintaining a mutual inductor based on improved K-means according to claim 5, characterized in that: The specific steps for calculating the coefficient of variation of cluster centers include: Calculate the mean of the coordinate dimensions corresponding to all cluster centers ; Calculate the standard deviation of each cluster center to the mean ; According to the mean of all cluster centers and the standard deviation of each cluster center to the mean , calculate the coefficient of variation , expressed as: 。 7. The method for operating and maintaining a mutual inductor based on improved K-means according to claim 6, characterized in that: The status of the transformer is evaluated based on the clustering results, specifically: After the transformer data set to be tested is clustered by the improved K-means clustering algorithm, K clusters and their corresponding cluster centers are obtained; Calculate the mean of reliability and stability for each cluster separately and ; According to the user's requirements for reliability and stability, the mean and Assign weights and calculate the comprehensive score, expressed as follows: ； In the formula, For the comprehensive rating, is the reliability mean weight, is the stability mean weight; The comprehensive scores are sorted in order from large to small, and the sorting results are equally divided into five levels, the higher the ranking, the lower the number, and the comprehensive scores and corresponding ranking levels are the evaluation results.

8. The transformer operation and maintenance method based on improved K-means according to claim 7, characterized in that: Different operation and maintenance strategies are formulated for the transformer under test according to the evaluation results, specifically, basic maintenance is performed at the first ranking level; preventive maintenance is performed at the second ranking level; corrective maintenance is performed at the third ranking level; corrective maintenance is performed at the fourth ranking level; and asset management and planned maintenance is performed at the fifth ranking level. The transformer under test is operated and maintained according to its ranking level and the corresponding operation and maintenance strategy.

Images