A model training and abnormal data identification method, device, equipment and medium

Abstract

A model training and abnormal data identification method, device, equipment and medium are disclosed. The method comprises: determining an extended characteristic value corresponding to each existing characteristic value of a sample according to at least two groups of existing characteristic values in sample oil chromatogram data; wherein the existing characteristic values of different groups of samples are content values of at least two gases in transformer oil collected at different time points; training an abnormal data identification model according to the at least two groups of existing characteristic values, the extended characteristic values corresponding to the at least two groups of existing characteristic values, and an abnormal data label associated with the sample oil chromatogram data, for identifying abnormal data in the oil chromatogram data. Through the technical scheme provided in the embodiments of the present application, the accuracy of the trained abnormal data identification model is higher, and more accurate identification of abnormal data in the oil chromatogram data based on the model can be realized.

Description

TECHNICAL FIELD

[0001] Embodiments of the present application relate to the technical field of data processing, and in particular to a model training and abnormal data identification method, device, equipment and medium. BACKGROUND

[0002] With the development of transformer technology, oil-immersed transformers have been widely used. The body of the oil-immersed transformer is often installed in an oil tank filled with transformer oil, and the oil tank is welded with steel plates. In the application scenario of the oil-immersed transformer, an oil color acquisition device is needed to acquire the oil color spectrum data of the oil-immersed transformer. However, when the oil color acquisition device fails, the acquired oil color spectrum data of the oil-immersed transformer will be abnormal.

[0003] Therefore, how to detect the abnormality of the oil color spectrum data of the oil-immersed transformer and provide higher quality oil color spectrum data is a problem to be solved at present. SUMMARY

[0004] Embodiments of the present application provide a model training and abnormal data identification method, device, equipment and medium. The trained abnormal data identification model has higher accuracy, and can further realize more accurate identification of abnormal data in the oil color spectrum data based on the model.

[0005] In a first aspect, embodiments of the present application provide a model training method, comprising:

[0006] According to the at least two groups of sample existing characteristic values in the sample oil color spectrum data, determine the expansion characteristic values corresponding to each group of sample existing characteristic values; wherein the existing characteristic values of different groups of samples are the content values of at least two gases in the transformer oil collected at different time points;

[0007] According to the at least two groups of sample existing characteristic values, the expansion characteristic values corresponding to the at least two groups of sample existing characteristic values, and the abnormal data labels associated with the sample oil color spectrum data, train an abnormal data identification model for identifying abnormal data in the oil color spectrum data.

[0008] In a second aspect, embodiments of the present application also provide an abnormal data identification method based on oil color spectrum data, comprising:

[0009] Obtain at least two groups of target existing characteristic values in the target oil color spectrum data, wherein the existing characteristic values of different groups of targets are the content values of at least two gases in the transformer oil collected at different time points;

[0010] Input the at least two groups of target existing characteristic values in the target oil color spectrum data into the trained abnormal data identification model to obtain the abnormal data in the at least two groups of target existing characteristic values;

[0011] The abnormal data identification model is trained according to the model training method provided in any of the embodiments of the present application.

[0012] In a third aspect, the embodiments of the present application further provide a model training device, comprising:

[0013] The determination module is configured to determine an extended characteristic value corresponding to each of the existing characteristic values of the samples according to the existing characteristic values of the at least two groups of samples in the sample oil chromatographic data, wherein the existing characteristic values of different groups of samples are content values of at least two gases in the transformer oil collected at different time points.

[0014] The training module is configured to train an abnormal data identification model according to the existing characteristic values of the at least two groups of samples, the extended characteristic values corresponding to the existing characteristic values of the at least two groups of samples, and the abnormal data labels associated with the sample oil chromatographic data, so as to identify abnormal data in the oil chromatographic data.

[0015] In a fourth aspect, the embodiments of the present application further provide an abnormal data identification device based on oil chromatographic data, comprising:

[0016] The acquisition module is configured to acquire at least two groups of target existing characteristic values in target oil chromatographic data, wherein the existing characteristic values of different groups of target existing characteristic values are content values of at least two gases in the transformer oil collected at different time points.

[0017] The obtaining module is configured to input the at least two groups of target existing characteristic values in the target oil chromatographic data into the trained abnormal data identification model, and obtain abnormal data in the at least two groups of target existing characteristic values, wherein the abnormal data identification model is trained according to the model training method provided in any of the embodiments of the present application.

[0018] In a fifth aspect, the embodiments of the present application further provide an electronic device, which comprises:

[0019] one or more processors;

[0020] a memory configured to store one or more programs;

[0021] When the one or more programs are executed by the one or more processors, the one or more processors implement the model training method and / or the abnormal data identification method based on oil chromatographic data provided in any of the embodiments of the present application.

[0022] In a sixth aspect, the embodiments of the present application further provide a computer readable storage medium having a computer program stored thereon. When the program is executed by a processor, the model training method and / or the abnormal data identification method based on oil chromatographic data provided in any of the embodiments of the present application are implemented.

[0023] The scheme provided by the embodiment of the present application comprises the following steps: determining an extended characteristic value according to at least two groups of existing characteristic values of sample oil chromatographic data; training an abnormal data identification model according to the at least two groups of existing characteristic values, the corresponding extended characteristic values, and an abnormal data label associated with the sample oil chromatographic data, so that a model for identifying abnormal data in the oil chromatographic data can be obtained. Further, at least two groups of target existing characteristic values in target oil chromatographic data are obtained and input into the trained abnormal data identification model, so that abnormal data in the existing characteristic values can be obtained, and the abnormal data can be identified. In this way, the accuracy of the trained abnormal data identification model is higher, so that the model can be used to more accurately identify abnormal data in the oil chromatographic data and obtain higher-quality oil chromatographic data. BRIEF DESCRIPTION OF DRAWINGS

[0024] Figure 1A A flowchart of a model training method provided for the first embodiment of the present application;

[0025] Figure 1B A model structure diagram provided for the first embodiment of the present application;

[0026] Figure 2 A flowchart of a model training method provided for the second embodiment of the present application;

[0027] Figure 3 A flowchart of a model training method provided for the third embodiment of the present application;

[0028] Figure 4 A flowchart of an abnormal data identification method based on oil chromatographic data provided for the fourth embodiment of the present application;

[0029] Figure 5 A structural block diagram of a model training device provided for the fifth embodiment of the present application;

[0030] Figure 6 A structural block diagram of an abnormal data identification device based on oil chromatographic data provided for the sixth embodiment of the present application;

[0031] Figure 7 A structural diagram of an electronic device provided for the seventh embodiment of the present application. DETAILED DESCRIPTION

[0032] The present application will be further described below in conjunction with the drawings and embodiments. It can be understood that the specific embodiments described herein are only used to explain the present application, but not to limit the present application. In addition, it should be noted that, for the convenience of description, only the parts related to the present application are shown in the drawings, but not all the structures.

[0033] Embodiment one

[0034] Figure 1A This is a flowchart of a model training method provided in Embodiment 1 of the present invention. Figure 1B This is a schematic diagram of the model structure provided in Embodiment 1 of the present invention. This embodiment is applicable to the training of an anomaly data identification model, wherein the anomaly data in the oil chromatography data is caused by a malfunction of the oil color acquisition device. This method can be executed by a model training device, which can be implemented in software and / or hardware. Figure 1A As shown, the method specifically includes:

[0035] S101. Based on the existing characteristic values ​​of at least two groups of samples in the sample oil chromatographic data, determine the extended characteristic value corresponding to the existing characteristic value of each group of samples.

[0036] The sample oil chromatographic data refers to data containing at least two sets of existing characteristic values ​​and corresponding extended characteristic values. The labeling information for abnormal data in the sample oil chromatographic data is known. Existing characteristic values ​​refer to the content data of at least two dissolved gases in transformer oil collected by the oil color acquisition device. Extended characteristic values ​​refer to data characteristic values ​​determined by extending existing characteristic values. A set of existing characteristic values ​​refers to the content values ​​of at least two gases in transformer oil collected simultaneously. Different sets of existing characteristic values ​​refer to the content values ​​of at least two gases in transformer oil collected at different time points. The unit of gas content is expressed as concentration (ppm, parts per million). The gases contained in the transformer oil can be at least two of the following: hydrogen (H2), methane (CH4), ethylene (C2H4), acetylene (C2H2), carbon monoxide (CO), carbon dioxide (CO2), oxygen (O2), hydrocarbons (THC, total hydrocarbons), and ethane (C2H6).

[0037] Optionally, existing characteristic values ​​in the sample oil chromatographic data can be collected periodically. That is, at least two gases in the transformer oil can be collected at preset intervals to determine their content values. Alternatively, when relevant personnel specify or preset conditions are met, a gas collection in the transformer oil can be triggered to determine its content value.

[0038] Optionally, after determining the existing feature values ​​of at least two groups of samples in the sample oil chromatographic data, the existing feature values ​​can be calculated based on preset rules to determine the extended feature values; alternatively, the existing feature values ​​can be input into a pre-configured extended value determination model to output the extended feature values ​​corresponding to each group of existing feature values, that is, the extended feature values ​​corresponding to each group of existing feature values ​​can be determined based on the existing feature values ​​of at least two groups of samples in the sample oil chromatographic data.

[0039] For example, referring to Table 1, if three sets of sample existing characteristic values are stored in the sample oil chromatogram data, and the transformer oil gases contained in the different sets of sample existing characteristic values include hydrogen (h2), methane (ch4), ethylene (c2h4), acetylene (c2h2), carbon monoxide (co), carbon dioxide (co2), oxygen (o2), total hydrocarbons (thc), and ethane (c2h6), the sample existing characteristic values can be displayed in the form of the following table:

[0040] Table 1, sample existing characteristic values

[0041]

[0042] S102, according to at least two sets of sample existing characteristic values, at least two sets of sample existing characteristic values corresponding to the expanded characteristic values, and the abnormal data label associated with the sample oil chromatogram data, training an abnormal data identification model for identifying abnormal data in the oil chromatogram data.

[0043] Wherein, the abnormal data refers to the data whose value of characteristic data suddenly changes. The abnormal data label refers to the label that is labeled in advance for each data in the oil chromatogram data, indicating whether each set of characteristic value data is abnormal data or normal data.

[0044] Optionally, the existing characteristic value collected at each time and the corresponding expanded characteristic value can be used as a set of characteristic value data, for example, the corresponding expanded characteristic value can be added after each set of sample existing characteristic values (for example, a new row of supplementary characteristic value values can be added in Table 1), forming new sample oil chromatogram data. Thus, at least two sets of characteristic value data containing existing characteristic values and expanded characteristic values can be obtained, and further, each time, the preset number of groups (at least two groups) of characteristic value data are input into the abnormal data identification model to obtain the abnormal characteristic value data in the preset number of groups and the corresponding abnormal probability. Specifically, the abnormality of the last group of data can be determined according to the previous groups of data, or all the characteristic value data of the preset number of groups can be analyzed to determine the abnormal characteristic value data. For example, when the preset number of groups is 5, the abnormality of the fifth group of characteristic value data can be determined according to the first four groups of characteristic value data, or the abnormal data identification model can be directly used to analyze the trend of the values of the five groups of characteristic value data to determine the abnormal characteristic value data, and the data in the groups other than the abnormal data group is normal data.

[0045] Optionally, after determining the abnormal feature value data according to the abnormal data identification model, for each group of feature value data, a prediction result of whether the group of data is abnormal can be obtained, the prediction result is compared with the abnormal data label associated with the sample oil chromatogram data, the loss of the prediction result and the actual result is determined, and finally based on the loss, the parameters of the trained abnormal data identification model are adjusted, that is, the abnormal data identification model is trained, and the trained abnormal data identification model can be used to identify the abnormal data in the oil chromatogram data.

[0046] Optionally, the abnormal data identification model is constructed based on a long short-term memory (LSTM) network. Figure 1B For example, the input layer of the LSTM network can be set to 8 time series connected networks (containing 8 sigma), the output layer can be set to 2 nodes, the hidden layer can be set to 32 nodes, and the dropout parameter (Dropout) can be set to 0.25.

[0047] The advantage of such setting is that by using a classic LSTM neural network model to construct the abnormal data identification model, a more optimized abnormal data identification model can be trained, so that the identification accuracy of abnormal data is higher.

[0048] The scheme provided by the embodiment of the application determines the expanded feature values according to at least two groups of sample existing feature values in the sample oil chromatogram data, and then trains an abnormal data identification model according to the at least two groups of sample existing feature values, the corresponding expanded feature values, and the abnormal data label associated with the sample oil chromatogram data, so that a model for identifying abnormal data in the oil chromatogram data can be obtained. In this way, the trained abnormal data identification model has higher accuracy, which can facilitate subsequent more accurate identification of abnormal data in the oil chromatogram data based on the model, and obtain higher quality oil chromatogram data.

[0049] Optionally, training the abnormal data identification model according to the at least two groups of sample existing feature values, the expanded feature values corresponding to the at least two groups of sample existing feature values, and the abnormal data label associated with the sample oil chromatogram data includes: taking each group of sample existing feature values and the expanded feature values corresponding to the group of sample existing feature values as a group of data set; merging different groups of data sets according to the collection time of each group of sample existing feature values to obtain a training set, a validation set and a test set; and training the abnormal data identification model based on the training set, the validation set and the test set, and the abnormal data label associated with the sample oil chromatogram data.

[0050] Specifically, after the existing feature values collected at each time and the corresponding extended feature values are taken as a set of feature value data, the feature value data at a preset interval or a preset number (such as 5) of adjacent collection times can be taken as a set of data sets according to a preset rule, and then the multiple sets of data sets are sorted based on the collection time of the last set of data in each set of data, and then the sorted multiple sets of data sets are merged according to a preset proportion relationship, such as a proportion relationship of 7:2:1 of the training set, the validation set and the test set, to obtain the training set, the validation set and the test set. Finally, the related parameters of the abnormal data recognition model are adjusted by using the training set and the abnormal data label, and after the training is completed, the recognition effect of the abnormal data recognition model is verified by using the validation set and the abnormal data label, and after the verification is passed, the recognition accuracy of the abnormal data recognition model is tested by using the test set and the abnormal data label.

[0051] The advantage of such setting is that at least two sets of data sets are determined by merging at least two sets of existing feature values and extended feature values, and the data sets are further divided into a training set, a validation set and a test set to train the abnormal data recognition model, so that the abnormal data recognition model obtained by training has higher recognition accuracy for abnormal data in oil chromatographic data.

[0052] Embodiment Two

[0053] Figure 2 A flowchart of a model training method provided for Embodiment Two of the present application, which further explains and describes “determining the extended feature value corresponding to each set of sample existing feature values from at least two sets of sample existing feature values in sample oil chromatographic data” in detail based on the above-mentioned embodiment, such as shown in the following figure. Figure 2 The method specifically includes the following steps.

[0054] S201, determining the first extended feature value corresponding to each set of sample existing feature values according to the correlation between the sample existing feature values in the sample oil chromatographic data.

[0055] The correlation between the existing feature values refers to the numerical relationship between at least two existing feature values. The first extended feature value refers to the extended feature value corresponding to the existing feature value of the gas collected at the same time.

[0056] Optionally, the first extended feature value of the sample oil chromatographic data can be determined by the correlation between the content values of different gases collected at each time point. Specifically, the numerical relationship between the content values of different gases collected at each time point can be calculated according to a preset calculation rule to determine the first extended feature value; or a classical test method can be used to analyze the correlation between the content values of different gases collected at each time point to determine the first extended feature value.

[0057] Optionally, the numerical relationship between the content values of different gases collected at each time point is calculated according to a preset calculation rule, and there are many ways to determine the first extended characteristic value. Specifically, the first extended characteristic value of the sample oil chromatographic data can be determined according to the ratio of the content value of one kind of gas to the content value of one or more other kinds of gas collected at each time point, or the ratio of the content value of one or more other kinds of gas to the content value of one kind of gas collected at each time point. The first extended characteristic value of the sample oil chromatographic data can also be determined according to the difference between the content values of any two kinds of gas collected at each time point.

[0058] For example, the ratio of the content values of acetylene (C2H2) and ethylene (C2H4) collected at each time point can be used as the first extended value corresponding to the existing characteristic value of the group at this time point, i.e., the first extended value of each existing characteristic value is determined according to the formula C2H2 / C2H4. The ratio of the content values of methane (CH4) and hydrocarbon (THC) collected at each time point can also be used as the first extended value corresponding to the existing characteristic value of the group at this time point, i.e., the first extended value of each existing characteristic value is determined according to the formula CH4 / THC. The first extended value of each existing characteristic value can also be determined according to the formula (CH4+C2H4) / THC, CH4 / (H2+THC), CH4 / (CH4+C2H4+C2H2), or CH4-C2H4, where CH4, C2H4, THC, H2, and C2H2 represent the content values of methane, ethylene, hydrocarbon, hydrogen, and acetylene, respectively.

[0059] Optionally, the content values of different gases collected at each time point can be analyzed for correlation using a classical non-parametric rank sum test method to determine the first extended characteristic value. Accordingly, the first extended characteristic value corresponding to each existing characteristic value of each group of sample oil chromatographic data is determined based on the correlation between the existing characteristic values of each group of sample oil chromatographic data, including: using the rank sum test method to analyze the correlation between the existing characteristic values of each group of sample oil chromatographic data to obtain at least two gas correlation values corresponding to each existing characteristic value of each group; and determining the first extended characteristic value corresponding to each existing characteristic value of each group based on the at least two gas correlation values corresponding to each existing characteristic value of each group.

[0060] Optionally, for the existing characteristic values collected at the same time point in the sample oil chromatographic data, the rank sum test method can be used to analyze the correlation between the existing characteristic values of each gas and the existing characteristic values of other gases to obtain the correlation value of the gas. Specifically, the deviation of each gas from the existing characteristic values of other gases can be determined first, and then the sum of all deviations is obtained to obtain the total score of the gas, and the score of the total score is used as the correlation value of the gas. In this way, the correlation values of at least two gases are obtained.

[0061] Optionally, after obtaining the correlation values of the at least two gases, at least one correlation value can be further selected from the obtained correlation values of each gas as the first extended characteristic value of the sample oil chromatogram data. Specifically, the correlation values of the at least two gases can be sorted according to the score value size, and the correlation value of one or more gases with a high score is selected as the first extended characteristic value of the sample oil chromatogram data.

[0062] The advantage of such an arrangement is that by using a classical verification algorithm, the at least two gas correlation values corresponding to the existing characteristic values of each group of samples are first determined, and then the first extended characteristic value is determined, so that the determined sample oil chromatogram data can better represent the state of the transformer oil, and thus the abnormal data recognition model trained from the data is more accurate and reasonable.

[0063] S202, determining a second extended characteristic value corresponding to the existing characteristic values of each group of samples according to the content values of the same gas corresponding to the existing characteristic values of different groups of samples.

[0064] The second extended characteristic value refers to the extended characteristic value corresponding to the existing characteristic values of the gas collected at different time points.

[0065] Optionally, after determining the content values of the same gas corresponding to the existing characteristic values of each group of samples, the content values of the same gas corresponding to the existing characteristic values of each group of samples can be calculated according to a predetermined calculation rule to determine the corresponding second extended characteristic value. Alternatively, for each type of gas, the content values of the gas corresponding to the existing characteristic values of each group of samples can be input into a pre-trained supplementary characteristic determination model to output the corresponding supplementary characteristic, i.e., to determine the second supplementary characteristic.

[0066] S203, training an abnormal data recognition model according to the existing characteristic values of at least two groups of samples, the extended characteristic values corresponding to the existing characteristic values of at least two groups of samples, and the abnormal data labels associated with the sample oil chromatogram data, for identifying abnormal data in the oil chromatogram data.

[0067] The scheme provided by the embodiments of the present application determines the first extended characteristic value according to the correlation between the existing characteristic values of each group of samples in the sample oil chromatogram data, determines the second extended characteristic value according to the content values of the same gas corresponding to the existing characteristic values of different groups of samples, and further combines the abnormal data labels associated with the sample oil chromatogram data to train an abnormal data recognition model, thereby providing an implementable way to determine a supplementary characteristic value. Through such a way, the accuracy of the trained abnormal data recognition model is higher, which can facilitate more accurate identification of abnormal data in the oil chromatogram data based on the model, and obtain higher quality oil chromatogram data.

[0068] Embodiment three

[0069] Figure 3A flowchart of a model training method provided for Embodiment Three of the present application is shown in FIG. 3. The present embodiment is based on the above-mentioned embodiments and further explains and describes in detail the step of "determining the second extended characteristic value corresponding to the existing characteristic value of each group of samples according to the content value corresponding to the existing characteristic value of each group of samples of the same gas". As shown in FIG. 3, the method specifically includes the following steps. Figure 3

[0070] S301, determining the first extended characteristic value corresponding to the existing characteristic value of each group of samples according to the correlation between the existing characteristic values of each group of samples in the sample oil chromatographic data.

[0071] S302, determining the absolute growth amount and / or the relative growth rate of the target gas according to the content value corresponding to the existing characteristic value of the first group of samples and the existing characteristic value of the second group of samples, and taking the absolute growth amount and / or the relative growth rate as the second extended characteristic value corresponding to the existing characteristic value of the second group of samples.

[0072] The target gas refers to a gas whose content value needs to be determined to determine the supplementary characteristic value. The collection time point of the existing characteristic value of the first group of samples is earlier than the collection time point of the existing characteristic value of the second group of samples. The absolute growth amount refers to the growth amount obtained by subtracting the existing characteristic value of the same gas collected at different time points. The relative growth rate refers to the ratio of the absolute growth amount to the time difference between the collection of the existing characteristic values of at least two groups of samples.

[0073] Optionally, all the gases contained in the transformer oil can be sequentially taken as the target gas, or part of the gases contained in the transformer oil can be selected as the target gas according to a preset rule, i.e., the target gas is determined.

[0074] Optionally, after the target gas is determined, for each target gas, the content value of the target gas in the existing characteristic value of the first group and the second group of samples can be determined, the two determined content values can be subtracted to determine the absolute growth amount of the target gas as the second extended characteristic value corresponding to the existing characteristic value of the second group of samples; or after the absolute growth amount of the target gas is determined, the time difference between the collection of the existing characteristic values of the first group and the second group of samples can be further determined, and the ratio of the absolute growth amount to the time difference, i.e., the relative growth rate, can be taken as the second extended characteristic value corresponding to the existing characteristic value of the second group of samples; or both the absolute growth amount and the relative growth rate can be taken as the second extended characteristic value corresponding to the existing characteristic value of the second group of samples, which is not limited in the present embodiment.

[0075] ​Optionally, after determining the target gas, the relative growth amount of the target gas can also be determined as the second extended characteristic value corresponding to the existing characteristic value of the second group of samples. Specifically, if the collection time of the existing characteristic values of the first group and the second group of samples is p moment and q moment respectively, the difference between the existing characteristic values of the same gas collected at p moment and q moment can be calculated first, and then the ratio of the difference to the existing characteristic value of the gas at q moment is calculated as the relative growth amount. Further, the relative growth amount is also taken as the second extended characteristic value corresponding to the existing characteristic value of the second group of samples (i.e. the existing characteristic value of the gas collected at q moment).

[0076] S303, according to at least two groups of sample existing characteristic values, at least two groups of sample existing characteristic values corresponding to the extended characteristic values, and the abnormal data label associated with the sample oil chromatogram data, training the abnormal data recognition model for identifying the abnormal data in the oil chromatogram data.

[0077] The scheme provided by the embodiment of the application further determines the absolute growth amount and / or relative growth rate of the target gas according to the content value of the target gas corresponding to the existing characteristic values of the first group and the second group of samples after determining the first extended characteristic value, and takes the absolute growth amount and / or relative growth rate as the second extended characteristic value. Finally, the abnormal data recognition model is trained in combination with the abnormal data label associated with the sample oil chromatogram data, which gives an implementable way to determine the second extended characteristic value. Through such a way, the accuracy of the trained abnormal data recognition model is higher, which can facilitate more accurate identification of abnormal data in the oil chromatogram data based on the model subsequently, and obtain higher quality oil chromatogram data.

[0078] Embodiment four

[0079] Figure 4 A flowchart of an abnormal data recognition method based on oil chromatogram data provided by the fourth embodiment of the application. The present embodiment can be applicable to the case of identifying abnormal data caused by the failure of the oil color collection device, and is particularly suitable for the case of implementing abnormal data recognition by using the abnormal data recognition model trained by the above-mentioned embodiments. The method can be executed by an abnormal data recognition device based on oil chromatogram data. The device can be realized by software and / or hardware, and can be integrated into an electronic device with the function of abnormal data recognition based on oil chromatogram data. As shown in the figure, the abnormal data recognition method based on oil chromatogram data provided by the present embodiment specifically includes: Figure 4

[0080] S401, obtaining at least two groups of target existing characteristic values in target oil chromatogram data.

[0081] ​Wherein, different groups of target existing feature values are content values of at least two gases in transformer oil collected at different time points. The target oil chromatogram data refers to oil chromatogram data that needs to be detected for abnormal data. The label information of the abnormal data in the target oil chromatogram data is unknown. The target existing feature value refers to an existing feature value determined from the existing feature values and needing to be input into the model for prediction.

[0082] Optionally, the target oil chromatogram data can be selected by an artificial selection, or the oil chromatogram data meeting certain preset conditions can be determined as the target oil chromatogram data when the oil chromatogram data is detected to meet certain preset conditions.

[0083] Optionally, after the target oil chromatogram data is determined, all existing feature values of the groups contained in the target oil chromatogram data can be determined as the target existing feature values, or at least two groups of existing feature values can be selected as the target existing feature values according to a preset rule, that is, at least two groups of target existing feature values in the target oil chromatogram data are obtained.

[0084] S402, input at least two groups of target existing feature values in the target oil chromatogram data into the trained abnormal data identification model to obtain abnormal data in the at least two groups of target existing feature values.

[0085] Wherein, the abnormal data identification model can be trained by the model training method provided in any embodiment of the present application.

[0086] Optionally, the determined at least two groups of target existing feature values can be input into the trained abnormal data identification model together, or the determined at least two groups of target existing feature values can be divided into multiple groups according to a certain division rule, and then input into the trained abnormal data identification model, that is, at least two groups of target existing feature values in the target oil chromatogram data are input into the trained abnormal data identification model.

[0087] Optionally, after the at least two groups of target existing feature values are input into the abnormal data identification model, the abnormal existing feature value data and the corresponding abnormal probability in the at least two groups of target existing feature values can be output, that is, the abnormal data in the at least two groups of target existing feature values are obtained, and the identification of the abnormal data in the oil chromatogram data is completed.

[0088] Optionally, after the abnormal existing feature value data and the corresponding abnormal probability are output, the feature value data with a satisfactory abnormal probability value can be directly discarded to ensure the effectiveness of the oil chromatogram data.

[0089] The scheme provided by the embodiment of the present application can obtain at least two groups of target existing characteristic values in target oil chromatographic data, input the at least two groups of target existing characteristic values in the target oil chromatographic data into the trained abnormal data identification model, and obtain abnormal data in the at least two groups of target existing characteristic values. The trained abnormal data identification model can effectively identify abnormal data in oil chromatographic data, diagnose the fault condition of the oil-immersed transformer oil chromatographic acquisition device in advance, help to ensure the safe and stable operation of the power system, and improve the power supply reliability.

[0090] Embodiment five

[0091] Figure 5 The structural block diagram of the model training device provided by the fifth embodiment of the present application, the model training device provided by the embodiment of the present application can execute the model training method provided by any embodiment of the present application, and has the corresponding function modules and beneficial effects for executing the method.

[0092] The model training device can include a determination module 501 and a training module 502.

[0093] The determination module 501 is configured to determine an expansion characteristic value corresponding to each group of sample existing characteristic values according to at least two groups of sample existing characteristic values in sample oil chromatographic data; different groups of sample existing characteristic values are content values of at least two gases in transformer oil collected at different time points.

[0094] The training module 502 is configured to train an abnormal data identification model according to the at least two groups of sample existing characteristic values, the expansion characteristic values corresponding to the at least two groups of sample existing characteristic values, and an abnormal data label associated with the sample oil chromatographic data, so as to identify abnormal data in oil chromatographic data.

[0095] The scheme provided by the embodiment of the present application can determine an expansion characteristic value according to at least two groups of sample existing characteristic values in sample oil chromatographic data, and then train an abnormal data identification model according to the at least two groups of sample existing characteristic values, the corresponding expansion characteristic values, and an abnormal data label associated with the sample oil chromatographic data, so as to obtain a model for identifying abnormal data in oil chromatographic data. In this way, the trained abnormal data identification model has higher accuracy, and can facilitate more accurate identification of abnormal data in oil chromatographic data based on the model in the future, and obtain higher-quality oil chromatographic data.

[0096] Further, the determination module 501 can include:

[0097] The first determination unit is configured to determine a first expansion characteristic value corresponding to each group of sample existing characteristic values according to the correlation between each group of sample existing characteristic values in sample oil chromatographic data;

[0098] The second determining unit is used to determine the second extended feature value corresponding to the existing feature value of each group of samples based on the content value of the same gas in the existing feature values ​​of different groups of samples.

[0099] Furthermore, the first determining unit is specifically used for:

[0100] The rank-sum check method was used to analyze the correlation between the existing feature values ​​of each group of samples in the sample oil chromatographic data, and to obtain at least two gas correlation values ​​corresponding to the existing feature values ​​of each group of samples.

[0101] Based on at least two gas correlation values ​​corresponding to the existing feature values ​​of each group of samples, determine the first extended feature value corresponding to the existing feature values ​​of each group of samples.

[0102] Furthermore, the second determining unit is specifically used for:

[0103] Based on the content values ​​of the target gas in the existing characteristic values ​​of the first group of samples and the existing characteristic values ​​of the second group of samples, the absolute growth amount and / or relative growth rate of the target gas are determined, and the absolute growth amount and / or relative growth rate are used as the second extended characteristic value corresponding to the existing characteristic value of the second group of samples.

[0104] The time point at which the feature values ​​of the first group of samples were collected was earlier than the time point at which the feature values ​​of the second group of samples were collected.

[0105] Furthermore, training module 502 is specifically used for:

[0106] Each set of samples is divided into two datasets: the existing feature values ​​of each set of samples and the corresponding extended feature values ​​of that set of samples.

[0107] Based on the collection time of the existing feature values ​​of each group of samples, the datasets of different groups are merged to obtain the training set, validation set and test set;

[0108] An abnormal data identification model is trained based on the training set, validation set, and test set, as well as the abnormal data labels associated with the sample oil chromatographic data.

[0109] Furthermore, the abnormal data identification model is constructed based on a Long Short-Term Memory (LSTM) network.

[0110] Example 6

[0111] Figure 6 This is a structural block diagram of an abnormal data identification device based on oil chromatography data provided in Embodiment 6 of the present invention. The abnormal data identification device based on oil chromatography data provided in this embodiment of the present invention can execute the abnormal data identification method based on oil chromatography data provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of executing the method.

[0112] The abnormal data recognition device based on oil chromatographic data can include an acquisition module 601 and an obtaining module 602.

[0113] The acquisition module 601 is configured to acquire at least two groups of target existing characteristic values in target oil chromatographic data, wherein different groups of target existing characteristic values are content values of at least two gases in transformer oil collected at different time points.

[0114] The obtaining module 602 is configured to input the at least two groups of target existing characteristic values in the target oil chromatographic data into a trained abnormal data recognition model to obtain abnormal data in the at least two groups of target existing characteristic values.

[0115] The abnormal data recognition model is trained according to the model training method provided in any of the embodiments of the present application.

[0116] The scheme provided in the embodiments of the present application acquires at least two groups of target existing characteristic values in target oil chromatographic data, inputs the at least two groups of target existing characteristic values in the target oil chromatographic data into a trained abnormal data recognition model, and obtains abnormal data in the at least two groups of target existing characteristic values. The trained abnormal data recognition model can be used to effectively identify abnormal data in oil chromatographic data, diagnose fault conditions of an oil-immersed transformer oil chromatographic acquisition device in advance, help ensure safe and stable operation of a power system, and improve power supply reliability.

[0117] Embodiment Seven

[0118] Figure 7 A structural schematic diagram of an electronic device provided in the embodiment seven of the present application is shown. Figure 7 A block diagram of an exemplary device suitable to implement embodiments of the present application is shown. Figure 7 The device shown is merely an example, and should not impose any limitation on the functions and use range of the embodiments of the present application.

[0119] As shown in Figure 7 The electronic device 12 is in the form of a general computing device. The components of the electronic device 12 can include, but are not limited to, one or more processors or processing units 16, system memory 28, and a bus 18 connecting different system components, including the system memory 28 and the processing unit 16.

[0120] Bus 18 represents one or more of several types of bus structures, including a memory bus or memory controller, a peripheral bus, a graphics acceleration bus, a processor or local bus using any of a variety of bus architectures. By way of example, these architectures include Industry Standard Architecture (ISA) bus, Micro Channel Architecture (MCA) bus, Enhanced ISA (EISA) bus, Video Electronics Standards Association (VESA) local bus, and Peripheral Component Interconnect (PCI) bus.

[0121] Electronic device 12 typically includes a variety of computer system readable media. These media can be any available media that is accessible by electronic device 12 and includes both volatile and non-volatile media, removable and non-removable media.

[0122] System memory 28 can include computer system readable media in the form of volatile memory, such as random access memory (RAM) 30 and / or cache memory (cache 32). Electronic device 12 can further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, storage system 34 can be provided for reading from and writing to a non-removable, non-volatile magnetic media (e.g., a "hard drive"). Figure 7 not shown, a magnetic hard disk drive for reading from and writing to non-removable, non-volatile magnetic media (e.g., a "hard drive"). Although not specifically shown, such Figure 7 In alternative embodiments, a magnetic hard disk drive, a solid state drive (SSD) which is a non- volatile storage medium, a floppy disk drive for reading from and writing to a removable, non-volatile magnetic disk (e.g., a "floppy disk"), and an optical disk drive for reading from or writing to a removable, non-volatile optical disk (e.g., a CD-ROM, DVD-ROM, etc.) can be provided. In such instances, each can be connected to bus 18 by one or more data media interfaces. The system memory 28 can include at least one program product having a set (e.g., at least one) of program modules that are configured to carry out the functions of embodiments of the application.

[0123] Program / utility 40, having a set (at least one) of program modules 42, can be stored in system memory 28 by way of example, such as an operating system, one or more application programs, other program modules, and program data, each or some combination thereof, can include implementation of a network environment. Program modules 42 generally carry out the functions and / or methodologies of embodiments of the application described herein.

[0124] Electronic device 12 can also communicate with one or more external devices 14 (e.g., keyboard, pointing device, display 24, etc.), and with one or more devices that enable a user to interact with electronic device 12, and / or with any device that enables electronic device 12 to communicate with one or more other computing devices (e.g., network card, modem, etc.). This communication can be performed via input / output (I / O) interface 22. Furthermore, electronic device 12 can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), and / or public networks, such as the Internet) via network adapter 20. As shown, network adapter 20 communicates with other modules of electronic device 12 via bus 18. It should be understood that, although... Figure 7 As not shown, other hardware and / or software modules may be used in conjunction with electronic device 12, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0125] The processing unit 16 executes various functional applications and data processing by running programs stored in the system memory 28, such as implementing the model training method provided in the embodiments of the present invention, and / or the abnormal data identification method based on oil chromatography data.

[0126] Example 8

[0127] Embodiment 8 of the present invention also provides a computer-readable storage medium storing a computer program (or computer-executable instructions) thereon, which, when executed by a processor, is used to perform the model training method and / or the abnormal data identification method based on oil chromatography data provided in the embodiments of the present invention.

[0128] The computer storage medium of this invention can be any combination of one or more computer-readable media. A computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. A computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of computer-readable storage media (a non-exhaustive list) include: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this document, a computer-readable storage medium can be any tangible medium that contains or stores a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0129] A computer readable signal medium can include a propagated data signal with computer executable code embodied therein, for example, in baseband or as part of a carrier wave. Such a propagated signal can take any of a variety of forms, including, but not limited to, electro-magnetic, optical, or any suitable combination thereof. A computer readable signal medium can be any computer readable medium that is not a computer readable storage medium and that can communicate, propagate or transport programming code.

[0130] Program code embodied on a computer readable medium can be transmitted using any appropriate medium, including but not limited to wireless, wire line, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

[0131] Computer program code for carrying out operations for aspects of the present application can be written in any combination of one or more programming languages, including an object oriented programming language such as Python, Java, Smalltalk, C++ or the like, and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code can execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection can be made to an external computer (for example, through the Internet using an Internet Service Provider).

[0132] Note that the foregoing are merely examples of the preferred embodiments of the present application and the principles of the technology applied. It will be understood by those skilled in the art that the present application is not limited to the specific embodiments described herein, and that various obvious changes, modifications and substitutions can be made to the present application without departing from the scope of the present application. Therefore, although the embodiments of the present application have been described in detail through the above embodiments, the embodiments of the present application are not limited to the above embodiments, and can include more other equivalent embodiments without departing from the concept of the present application, and the scope of the present application is determined by the scope of the claims.

Claims

1. A model training method, characterized in that, include: Based on the existing characteristic values ​​of at least two groups of samples in the sample oil chromatography data, determine the extended characteristic value corresponding to the existing characteristic value of each group of samples; wherein, the existing characteristic values ​​of different groups of samples are the content values ​​of at least two gases in transformer oil collected at different time points; Based on the existing feature values ​​of the at least two sets of samples, the extended feature values ​​corresponding to the existing feature values ​​of the at least two sets of samples, and the abnormal data labels associated with the sample oil chromatographic data, an abnormal data identification model is trained to identify abnormal data in the oil chromatographic data. The step of determining the extended feature value corresponding to the existing feature value of each group of samples based on at least two groups of existing feature values ​​in the sample oil chromatographic data includes: determining the first extended feature value corresponding to the existing feature value of each group of samples based on the correlation between the existing feature values ​​of each group of samples in the sample oil chromatographic data; and determining the second extended feature value corresponding to the existing feature value of each group of samples based on the content value of the same gas in the existing feature values ​​of different groups of samples. The first extended feature value refers to the extended feature value corresponding to the existing feature value of the gas collected at the same time; the second extended feature value refers to the extended feature value corresponding to the existing feature value of the gas collected at different times. Specifically, determining the second extended feature value corresponding to the existing feature value of each group of samples based on the content value of the same gas in the existing feature values ​​of different groups of samples includes: determining the absolute growth rate and / or relative growth rate of the target gas based on the content value of the target gas in the existing feature values ​​of the first group of samples and the existing feature values ​​of the second group of samples, and using the absolute growth rate and / or relative growth rate as the second extended feature value corresponding to the existing feature value of the second group of samples; wherein the collection time point of the existing feature value of the first group of samples is earlier than the collection time point of the existing feature value of the second group of samples. The step of determining the first extended feature value corresponding to the existing feature value of each group of samples based on the correlation between the existing feature values ​​of each group of samples in the sample oil chromatographic data includes: The rank-sum check method is used. For each gas, the deviation of each gas from the existing feature values ​​of other gases is first determined. Then, all deviations are summed to obtain the total score of the gas. The total score is used as the correlation value of the gas, thereby obtaining the correlation values ​​of at least two gases corresponding to the existing feature values ​​of each sample. For each group of samples, the correlation values ​​of at least two gases corresponding to the existing feature values ​​are sorted according to the score value. One or more gases with high scores are selected as the first extended feature values ​​of the sample oil chromatographic data to determine the first extended feature value corresponding to the existing feature values ​​of each group of samples.

2. The method according to claim 1, characterized in that, The step of training an anomaly data identification model based on the existing feature values ​​of the at least two sets of samples, the expanded feature values ​​corresponding to the existing feature values ​​of the at least two sets of samples, and the anomaly data labels associated with the sample oil chromatographic data includes: Each set of samples is divided into two datasets: the existing feature values ​​of each set of samples and the corresponding extended feature values ​​of that set of samples. Based on the collection time of the existing feature values ​​of each group of samples, the datasets of different groups are merged to obtain the training set, validation set and test set; An abnormal data identification model is trained based on the training set, validation set, and test set, as well as the abnormal data labels associated with the sample oil chromatographic data.

3. The method according to any one of claims 1-2, characterized in that, The abnormal data identification model is built on a Long Short-Term Memory (LSTM) network.

4. A method for identifying abnormal data based on oil chromatography data, characterized in that, include: Obtain at least two sets of target existing feature values ​​from the target oil chromatographic data, wherein different sets of target existing feature values ​​are the content values ​​of at least two gases in transformer oil collected at different time points; Input at least two sets of target existing feature values ​​from the target oil chromatography data into the trained abnormal data identification model to obtain the abnormal data in the at least two sets of target existing feature values; The abnormal data identification model is trained using the method described in any one of claims 1-3.

5. A model training device, characterized in that, include: The determination module is used to determine the extended feature value corresponding to the existing feature value of each group of samples based on the existing feature values ​​of at least two groups of samples in the sample oil chromatography data; wherein the existing feature values ​​of different groups of samples are the content values ​​of at least two gases in transformer oil collected at different time points; The training module is used to train an abnormal data recognition model based on the existing feature values ​​of the at least two sets of samples, the expanded feature values ​​corresponding to the existing feature values ​​of the at least two sets of samples, and the abnormal data labels associated with the sample oil chromatographic data, so as to identify abnormal data in the oil chromatographic data. The determining module includes: a first determining unit, used to determine the first expanded feature value corresponding to the existing feature value of each group of samples based on the correlation between the existing feature values ​​of each group of samples in the sample oil chromatography data; and a second determining unit, used to determine the second expanded feature value corresponding to the existing feature value of each group of samples based on the content value of the same gas in the existing feature values ​​of different groups of samples; the first expanded feature value refers to the expanded feature value corresponding to the existing feature value of the gas collected at the same time; and the second expanded feature value refers to the expanded feature value corresponding to the existing feature value of the gas collected at different times. The second determining unit is specifically used to: determine the absolute growth rate and / or relative growth rate of the target gas based on the content value of the target gas in the existing characteristic values ​​of the first group of samples and the existing characteristic values ​​of the second group of samples, and use the absolute growth rate and / or relative growth rate as the second extended characteristic value corresponding to the existing characteristic value of the second group of samples; wherein the collection time point of the existing characteristic value of the first group of samples is earlier than the collection time point of the existing characteristic value of the second group of samples. The above-mentioned device is also used for: The rank-sum check method is used. For each gas, the deviation of each gas from the existing feature values ​​of other gases is first determined. Then, all deviations are summed to obtain the total score of the gas. The total score is used as the correlation value of the gas, thereby obtaining the correlation values ​​of at least two gases corresponding to the existing feature values ​​of each sample. For each group of samples, the correlation values ​​of at least two gases corresponding to the existing feature values ​​are sorted according to the score value. One or more gases with high scores are selected as the first extended feature values ​​of the sample oil chromatographic data to determine the first extended feature value corresponding to the existing feature values ​​of each group of samples.

6. An anomaly data identification device based on oil chromatography data, characterized in that, include: The acquisition module is used to acquire at least two sets of target existing feature values ​​in the target oil chromatographic data, wherein different sets of target existing feature values ​​are the content values ​​of at least two gases in transformer oil collected at different time points; The module is used to input at least two sets of existing target feature values ​​from the target oil chromatography data into a trained abnormal data identification model to obtain abnormal data from the at least two sets of existing target feature values. The abnormal data identification model is trained using the method described in any one of claims 1-3.

7. An electronic device, characterized in that, include: One or more processors; Memory, used to store one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the model training method as described in any one of claims 1-3, and / or the abnormal data identification method based on oil chromatography data as described in claim 4.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When executed by the processor, the program implements the model training method as described in any one of claims 1-3, and / or the abnormal data identification method based on oil chromatography data as described in claim 4.

Images