A method for early warning of operating status of a transformer

Abstract

The present application relates to the field of state monitoring technology, and in particular to a method for early warning of the operation state of a transformer, the method comprising: collecting the electrical values ​​of the transformer at each moment in a preset time period to obtain a first power sequence, wherein the electrical values ​​are secondary voltage or secondary current; correcting the electrical values ​​at each moment according to the change in environmental parameters to obtain a second power sequence; inputting the second power sequence and the control parameter sequence of the controlled device into a classification model to obtain the operation state, and issuing an early warning according to the operation state. Through the technical solution of the present application, the operation state of the transformer can be accurately judged, and an early warning signal can be issued in a timely manner.

Description

Technical Field

[0001] The present application relates to the technical field of status monitoring, and in particular to a method for early warning of the operating status of a transformer. Background Art

[0002] A transformer is an electrical device used in power systems. The transformer includes a voltage transformer and a current transformer. The voltage transformer is used to convert the high voltage (primary voltage) on the high voltage side into a small voltage (secondary voltage) on the low voltage side, and the current transformer is used to convert the high current (primary current) on the high voltage side into a small current (primary current) on the low voltage side, so that the measuring equipment and relays on the low voltage side of the power system can work under safe conditions. The high voltage side is the primary side and the low voltage side is the secondary side. Therefore, the operating status of the transformer directly affects the safety of the power system.

[0003] At present, a patent application document with application publication number CN117972460A discloses a method for distinguishing operating faults of high-voltage current transformers, the method comprising: collecting fault types and secondary current data of all current transformers at several measuring points; decomposing the secondary current data to obtain a first residual; calculating a second residual based on the secondary current data; performing residual decomposition, correlation analysis and cluster analysis on both the first residual and the second residual to obtain fault components corresponding to each fault type; splicing the fault components and extracting features through a model to obtain features corresponding to each fault type; obtaining a feature vector of the faulty current transformer through the secondary current data A and the features corresponding to each fault type; constructing a current transformer fault classification model through a graph convolutional neural network; and performing fault judgment on the transformer to be evaluated through the current transformer fault classification model.

[0004] The above method can perform fault identification based on the secondary current data of the current transformer. However, the operating environment of the current transformer is complex and changeable, and the complex operating environment will affect the collection accuracy of the operating data (voltage data and current data), resulting in inaccurate judgment of the operating status. Summary of the invention

[0005] In order to solve the technical problem of inaccurate operation status judgment, the present application provides a transformer operation status early warning method, which can accurately judge the operation status of the transformer and issue an early warning signal in time.

[0006] In a first aspect of the present application, a method for warning the operating status of a transformer is provided, the warning method comprising: collecting the electrical values ​​of the transformer at each moment in a preset time period to obtain a first power sequence, wherein the electrical value is a secondary voltage or a secondary current; correcting the electrical value at each moment according to the change in environmental parameters to obtain a second power sequence, wherein the environmental parameters include temperature and humidity, comprising: calculating the measured increment of the electrical value at any moment in the first power sequence; inputting the sequence of changes in each environmental parameter in an adjacent time period at any moment and the average value of each environmental parameter into a target model, and outputting the environmental increment of the electrical value at the moment; subtracting the environmental increment of the electrical value from the measured increment of the electrical value to obtain the true increment of the electrical value at each moment; starting from the start moment of the preset time period, taking the sum of the true increment of the electrical value at each moment and the electrical value at the previous adjacent moment as the corrected electrical value at the corresponding moment; inputting the second power sequence and the control parameter sequence of the controlled device into the classification model to obtain the operating status, and issuing a warning according to the operating status, wherein the operating status is normal or abnormal, and the controlled device is an electrical device connected to the secondary side of the transformer.

[0007] The secondary voltage or secondary current of the transformer at each moment in a preset time period is collected as an electrical value to obtain a first power sequence; considering the complexity of the transformer's operating environment, changes in environmental parameters will also cause changes in electrical values, making the first power sequence unable to accurately reflect the working state of the transformer, and the electrical values ​​at each moment in the first power sequence are corrected using the changes in environmental parameters to obtain a second power sequence; further, considering that the control parameters of the electrical equipment on the secondary side of the transformer will also affect the electrical values, the second power sequence and the control parameter sequence of the controlled equipment are input into a classification model to obtain an operating state, which is normal or abnormal, and then an early warning is issued based on the operating state to accurately judge the operating state of the transformer and realize a status early warning of the transformer.

[0008] Preferably, the time Electrical Value The measured increase for: , is the moment in the first power sequence Electrical Value The value of is the moment in the first power sequence Electrical Value The value of .

[0009] Preferably, the method for determining the preset time period includes: calculating the degree of environmental change at each historical moment before the current moment, the degree of environmental change being equal to the sum of the changes in each environmental parameter; taking the historical moment whose degree of environmental change is less than the change threshold and is closest to the current moment as the starting moment of the preset time period, and taking the current moment as the ending moment of the preset time period.

[0010] Preferably, the target model is a polynomial model, and the target model is:

[0011] ; For the moment Electrical value Environmental increment, is the number of environmental parameters, is the number of moments in the adjacent time period, is the time in the adjacent time period Environmental parameters The amount of change, Electrical value Environmental parameters The influence coefficient of Environmental parameters The coefficient of fluctuation, Environmental parameters The average value of .

[0012] The average value of each environmental parameter can reflect the average state of the environment in which the transformer is located. The average state of the environment and the change of environmental parameters at each moment can be combined to accurately predict the environmental increment of the electrical value caused by environmental changes.

[0013] Preferably, the target model is a BP neural network.

[0014] Preferably, the method for constructing the target model includes: constructing an initial model; collecting basic data and experimental data, both of which include the measured increments of electrical values ​​at any time, as well as the change sequence of each environmental parameter in adjacent time periods and the average value of each environmental parameter, and the change sequence of each environmental parameter in the basic data is 0; using the difference between the measured increments of electrical values ​​in the experimental data and the basic data as the environmental increment label to calculate the mean square error loss, and updating the initial model with the goal of minimizing the mean square error loss to obtain the target model.

[0015] In the experimental environment, only the environmental parameters are changed and the corresponding measured increments of electrical values ​​are collected. The collected measured increments of electrical values ​​are compared with the measured increments of electrical values ​​in the basic data (all changes are 0). The changes in electrical values ​​caused by environmental changes in the experimental data can be obtained to ensure the accuracy of the experimental data labels.

[0016] Preferably, the mean square error loss satisfy:

[0017] , For experimental data The initial model output results are: For experimental data The environment increment tag, For experimental data The maximum mean value of the changes in each environmental parameter, For experimental data The maximum variance of each environmental parameter change.

[0018] The variance of the change of each environmental parameter and the mean value of the change of the environmental parameter can reflect the complexity of the environment. The larger the mean value of the change of the environmental parameter, the more obvious the environmental change is. The larger the variance of the change of the environmental parameter, the more unstable the environmental change is and the greater the complexity of the environment is. In order to ensure that the target model can accurately predict the environmental increment of the electrical value in a complex environment, a larger weight is assigned to the experimental data with a higher degree of complexity.

[0019] Preferably, the classification model includes multiple timing feature extraction sub-models and classification sub-models. The timing feature extraction sub-model is used to extract the timing features of the second power sequence and the control parameter sequence of the controlled device, and input the timing features into the classification sub-model to obtain the operating status.

[0020] When the working state of the transformer is normal, the change of the control parameters of the controlled device on the secondary side of the transformer will also cause the electrical value of the transformer to fluctuate. If the change of the control parameters of the controlled device is ignored, the working state of the transformer cannot be accurately judged. Therefore, the control parameter sequence of the controlled device on the secondary side of the transformer is used as the input of the classification model. The classification model combines the second power sequence and the control parameter sequence of the controlled device to accurately judge the working state of the transformer.

[0021] Preferably, the training method of the classification model includes: taking the second power sequence and the control parameter sequence of the controlled device collected in the historical time as training samples, inputting the training samples into the classification model to obtain output results; calculating the cross entropy function between the output results and the actual state of the training samples to train the classification model until the cross entropy function is less than a preset function value.

[0022] The technical solution of this application has the following beneficial technical effects:

[0023] The secondary voltage or secondary current of the transformer at each moment in a preset time period is collected as an electrical value to obtain a first power sequence; considering the complexity of the transformer's operating environment, changes in environmental parameters will also cause changes in electrical values, making the first power sequence unable to accurately reflect the working state of the transformer, and the electrical values ​​at each moment in the first power sequence are corrected using the changes in environmental parameters to obtain a second power sequence; further, considering that the control parameters of the electrical equipment on the secondary side of the transformer will also affect the electrical values, the second power sequence and the control parameter sequence of the controlled equipment are input into a classification model to obtain an operating state, which is normal or abnormal, and then an early warning is issued based on the operating state to accurately judge the operating state of the transformer and realize a status early warning of the transformer. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] By reading the detailed description below with reference to the accompanying drawings, the above and other purposes, features and advantages of the exemplary embodiments of the present application will become easy to understand. In the accompanying drawings, several embodiments of the present application are shown in an exemplary and non-limiting manner, and the same or corresponding reference numerals represent the same or corresponding parts, wherein:

[0025] Figure 1 is a flow chart of a transformer operating status early warning method according to an embodiment of the present application;

[0026] Figure 2 is a flowchart of a second power sequence acquisition method according to an embodiment of the present application. DETAILED DESCRIPTION

[0027] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present application.

[0028] It should be understood that when the terms "first", "second", etc. are used in the claims, specification and drawings of the present application, they are only used to distinguish different objects, rather than to describe a specific order. The terms "include" and "comprise" used in the specification and claims of the present application indicate the presence of the described features, wholes, steps, operations, elements and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components and / or their collections.

[0029] According to a first aspect of the present application, the present application provides a method for early warning of operating status of a transformer. Figure 1 1 is a flow chart of a method for early warning of a transformer operation status according to an embodiment of the present application. Figure 1As shown, the transformer operation status early warning method includes steps S101 to S103, which are described in detail below.

[0030] S101, collecting electrical values ​​of a mutual inductor at each moment in a preset time period to obtain a first power sequence, wherein the electrical value is a secondary voltage or a secondary current.

[0031] In one embodiment, a first power sequence of a transformer is collected, wherein the first power sequence includes electrical values ​​at each moment in a preset time period, and the electrical value is a secondary voltage or a secondary current. The preset time period includes a current moment and multiple moments before the current moment. The preset time period includes a total of 10 moments.

[0032] It can be understood that when the transformer is a current transformer, the electrical value is the secondary current; when the transformer is a voltage transformer, the electrical value is the secondary voltage.

[0033] The first power sequence can reflect the volatility of the electrical value of the transformer within a preset time period, and further can reflect the operating state of the transformer.

[0034] S102, correcting the electrical value at each moment according to the change amount of the environmental parameters to obtain a second power sequence, wherein the environmental parameters include temperature and humidity.

[0035] In one embodiment, the operating status of a mining transformer can be determined based on changes in electrical values ​​at adjacent moments in a first power sequence. However, the use environments of transformers are diverse. They can be used in mine power systems as well as in building or rail transit power systems. Changes in environmental parameters such as temperature and humidity in the use environment will cause changes in electrical values, making it impossible for the first power sequence to accurately reflect the operating status of the transformer. Therefore, it is necessary to correct the electrical values ​​at each moment based on the change in environmental parameters to obtain a second power sequence so that the second power sequence can accurately reflect the operating status of the transformer.

[0036] Among them, the time The change of environmental parameters is The value of the environmental parameter minus the time The numerical value of the environmental parameter. The environmental parameter may also include dust concentration and amplitude.

[0037] Specifically, the second power sequence includes the corrected electrical values ​​at each moment, Figure 2 is a flow chart of a second power sequence acquisition method according to an embodiment of the present application. Figure 2 As shown, the second power sequence acquisition method includes steps S201 to S204.

[0038] S201, calculating the measured increment of the electrical value at any time in the first power sequence.

[0039] Among them, the time Electrical Value The measured increase for: , is the moment in the first power sequence Electrical Value The value of is the moment in the first power sequence Electrical Value The value of .

[0040] It can be understood that when the electrical value is a secondary voltage, the measured increment of the electrical value is used to reflect the first power sequence moment Voltage fluctuation; when the electrical value is secondary current, the measured increment of the electrical value is used to reflect the moment of the first power sequence current fluctuations.

[0041] S202, inputting the variation sequence of each environmental parameter in the adjacent time period at any moment and the average value of each environmental parameter into the target model, and outputting the electrical value environmental increment at the moment.

[0042] Among them, the change of each environmental parameter can reflect the environmental changes in the transformer usage scenario, and the environmental changes will cause changes in the electrical value. The increase in the electrical value caused by the change of each environmental parameter is recorded as the electrical value environmental increment, and the measured electrical value increment can be regarded as the sum of the electrical value environmental increment and the electrical value real increment. In other words, when the change of each environmental parameter does not cause the electrical value to change, the measured electrical value increment is equal to the electrical value real increment.

[0043] Specifically, the target model can predict the electrical value environmental increment at any time; the input of the target model is the change sequence of each environmental parameter in the adjacent time period at any time and the average value of each environmental parameter, and the output is the electrical value environmental increment at the time.

[0044] The adjacent time period of a moment includes multiple moments before the moment, and optionally, the adjacent time period includes 3 moments in total. The average value of each environmental parameter can reflect the average state of the environment where the transformer is located. For example, when the environmental parameter is humidity, the average value of the environmental parameter can reflect the average humidity of the environment where the transformer is located.

[0045] In one embodiment, the target model is a polynomial model, specifically satisfying the following relationship:

[0046] ; For the moment Electrical value Environmental increment, is the number of environmental parameters, is the number of moments in the adjacent time period, is the time in the adjacent time period Environmental parameters The amount of change, Electrical value Environmental parameters The influence coefficient of Environmental parameters The coefficient of fluctuation, Environmental parameters The average value of .

[0047] In another embodiment, the target model is a BP neural network; the input data of the BP neural network is a sequence of changes in each environmental parameter in a time period adjacent to a moment, as well as the average value of each environmental parameter, and the output is the electrical value environmental increment at that moment. For example, the environmental parameters include two types, temperature and humidity, and the adjacent time period includes a total of three moments, then the input layer of the BP neural network includes 2×3+2, a total of 8 neurons, for receiving input data, and the output layer of the BP neural network includes 1 neuron, for outputting the electrical value environmental increment.

[0048] It can be understood that one electrical value corresponds to one target model.

[0049] In one embodiment, a method for constructing a target model includes: constructing an initial model; collecting basic data and experimental data, wherein both the basic data and the experimental data include the measured increments of electrical values ​​at any time, as well as the change sequence of each environmental parameter in adjacent time periods and the average value of each environmental parameter, and the change sequence of each environmental parameter in the basic data is 0; using the difference between the measured increments of electrical values ​​in the experimental data and the basic data as an environmental increment label to calculate the mean square error loss, and updating the initial model with the goal of minimizing the mean square error loss to obtain a target model.

[0050] Among them, the basic data is the measured increment of electrical values ​​collected when there is no environmental change. Multiple groups of experimental data can be collected by applying different changes to each environmental parameter, and the difference between the measured increment of electrical values ​​in each group of experimental data and the measured increment of electrical values ​​in the basic data can be regarded as the increment of electrical values ​​caused by environmental changes. That is to say, the difference between the measured increment of electrical values ​​in each group of experimental data and the measured increment of electrical values ​​in the basic data is the environmental increment of electrical values ​​actually collected for each group of experimental data. Therefore, the difference between the measured increment of electrical values ​​in the experimental data and the basic data can be used as a label.

[0051] Among them, the mean square error loss satisfy: , For experimental data The corresponding initial model output results are: For experimental data The environment increment tag.

[0052] In other embodiments, in a set of experimental data, the variance of each environmental parameter change and the mean of the environmental parameter change can reflect the complexity of the environment. The larger the mean of the environmental parameter change, the more obvious the environmental change. The larger the variance of the environmental parameter change, the more unstable the environmental change and the greater the complexity of the environment. In order to ensure that the target model can accurately predict the electrical value environmental increment in a complex environment, the mean square error loss is improved. satisfy:

[0053] , For experimental data The initial model output results are: For experimental data The environment increment tag, For experimental data The maximum mean value of the changes in each environmental parameter, For experimental data The maximum variance of each environmental parameter change.

[0054] In this way, the target model is used to accurately predict the electrical value environment increment at the time.

[0055] S203, subtracting the environmental increment of the electrical value from the actual increment of the electrical value to obtain the real increment of the electrical value at each moment.

[0056] Among them, the measured increment of the electrical value can be regarded as the sum of the environmental increment of the electrical value and the real increment of the electrical value. At this point, the measured increment of the electrical value and the environmental increment of the electrical value have been obtained. The real increment of the electrical value at each moment in the preset time period can be obtained by subtracting the environmental increment of the electrical value from the measured increment of the electrical value. The real increment of the electrical value can reflect the change of the electrical value of the transformer.

[0057] S204, starting from the start time of the preset time period, the sum of the actual increment of the electrical value at each time and the electrical value at the last adjacent time is sequentially taken as the corrected electrical value at the corresponding time.

[0058] Among them, after obtaining the true increment of the electrical value at each moment, starting from the starting moment of the preset time period, the true increment of the electrical value at the second moment is added to the electrical value at the starting moment to obtain the corrected electrical value at the second moment, and the true increment of the electrical value at the third moment is added to the corrected electrical value at the second moment to obtain the corrected electrical value at the third moment; and so on, the corrected electrical value at each moment is obtained, thereby obtaining the second power sequence.

[0059] In an optional embodiment, the accuracy of the electrical value at the start of the preset time period directly affects the accuracy of the second power sequence, and further affects the accuracy of the subsequent operating state. Therefore, to ensure that the electrical value at the start of the preset time period is a true increment of the electrical value, the method for determining the preset time period includes: calculating the degree of environmental change at each historical moment before the current moment, wherein the degree of environmental change is equal to the sum of the changes in each environmental parameter; taking the historical moment with an environmental change degree less than a change threshold and closest to the current moment as the start moment of the preset time period, and taking the current moment as the end moment of the preset time period.

[0060] S103, inputting the second power sequence and the control parameter sequence of the controlled device into the classification model to obtain the operating state, and issuing an early warning according to the operating state, wherein the operating state is normal or abnormal, and the controlled device is an electrical device connected to the secondary side of the transformer.

[0061] In one embodiment, the second power sequence eliminates the influence of complex environment on the electrical value, and when the control parameters of the controlled device do not change, the second power sequence can accurately reflect the working state of the transformer.

[0062] The controlled device is an electrical device connected to the secondary side (low-voltage side) of the transformer. When the working state of the transformer is normal, changes in the control parameters of the controlled device on the secondary side of the transformer will also cause fluctuations in the electrical values ​​of the transformer. If the changes in the control parameters of the controlled device are ignored, the working state of the transformer cannot be accurately judged. Therefore, the control parameter sequence of the controlled device on the secondary side of the transformer is used as the input of the classification model. The classification model combines the second power sequence and the control parameter sequence of the controlled device to accurately judge the working state of the transformer.

[0063] The classification model includes multiple time series feature extraction sub-models and classification sub-models. The time series feature extraction sub-model is used to extract the time series features of the second power sequence and the control parameter sequence of the controlled device, and input the time series features into the classification sub-model to obtain the operating status.

[0064] The time series feature extraction submodel is LSTM, RNN or GRU, and the classification submodel is Softmax, logistic regression or SVM. The number of time series feature extraction submodels is the same as the total number of the second power sequence and the controlled device control parameter sequence.

[0065] In one embodiment, the training method of the classification model includes: taking the second power sequence and the control parameter sequence of the controlled device collected in the historical time as training samples, inputting the training samples into the classification model, and obtaining output results; calculating the cross entropy function between the output results and the real state of the training samples to train the classification model until the cross entropy function is less than a preset function value. The preset function value is 0.001; the real state is manually marked, and the real state is normal or abnormal.

[0066] In this way, the operating status is output according to the classification model, and an early warning is issued according to the operating status. When the operating status of the transformer is abnormal, an early warning signal is issued.

[0067] The technical principle and implementation details of a transformer operating status warning method of the present application are introduced above through specific embodiments. The secondary voltage or secondary current of the transformer at each moment in a preset time period is collected as an electrical value to obtain a first power sequence; considering the complexity of the transformer use environment, changes in environmental parameters will also cause changes in electrical values, making the first power sequence unable to accurately reflect the working state of the transformer, and the electrical values ​​at each moment in the first power sequence are corrected using the changes in environmental parameters to obtain a second power sequence; further, considering that the control parameters of the electrical equipment on the secondary side of the transformer will also affect the electrical values, therefore, the second power sequence and the control parameter sequence of the controlled equipment are input into the classification model to obtain the operating state, the operating state is normal or abnormal, and then a warning is issued according to the operating state, the operating state of the transformer is accurately judged, and the status warning of the transformer is realized.

[0068] The technical features of the above-described embodiments may be arbitrarily combined. To make the description concise, not all possible combinations of the technical features in the above-described embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0069] The above-described embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be construed as limiting the scope of the patent application. It should be pointed out that, for a person of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. Therefore, the protection scope of the patent application shall be subject to the attached claims.

Claims

1. A method for early warning of transformer operation status, characterized in that: The early warning method comprises: Collecting electrical values ​​of the mutual inductor at each moment in a preset time period to obtain a first power sequence, wherein the electrical values ​​are secondary voltages or secondary currents; The electrical value at each moment is corrected according to the variation of the environmental parameters to obtain a second power sequence, wherein the environmental parameters include temperature and humidity, including: calculating the measured increment of the electrical value at any moment in the first power sequence; inputting the variation sequence of each environmental parameter in the adjacent time period at any moment and the average value of each environmental parameter into the target model, and outputting the environmental increment of the electrical value at the moment; subtracting the environmental increment of the electrical value from the measured increment of the electrical value to obtain the real increment of the electrical value at each moment; starting from the start moment of the preset time period, sequentially taking the sum of the real increment of the electrical value at each moment and the electrical value at the last adjacent moment as the corrected electrical value at the corresponding moment; The second power sequence and the control parameter sequence of the controlled device are input into the classification model to obtain the operating state, and an early warning is issued according to the operating state, wherein the operating state is normal or abnormal, and the controlled device is an electrical device connected to the secondary side of the transformer.

2. A transformer operating status early warning method according to claim 1, characterized in that: time Electrical Value The measured increase for: , is the moment in the first power sequence Electrical Value The value of is the moment in the first power sequence Electrical Value The value of .

3. A transformer operating status early warning method according to claim 1, characterized in that: The method for determining the preset time period includes: Calculate the degree of environmental change at each historical moment before the current moment, where the degree of environmental change is equal to the sum of the changes in each environmental parameter; The historical moment whose environmental change degree is less than the change threshold and is closest to the current moment is taken as the starting moment of the preset time period, and the current moment is taken as the ending moment of the preset time period.

4. A transformer operating status early warning method according to claim 1, characterized in that: The target model is a polynomial model, and the target model is: ; For the moment Electrical value Environmental increment, is the number of environmental parameters, is the number of moments in the adjacent time period, is the time in the adjacent time period Environmental parameters The amount of change, Electrical value Environmental parameters The influence coefficient of Environmental parameters The coefficient of fluctuation, Environmental parameters The average value of .

5. A transformer operating status early warning method according to claim 1, characterized in that: The target model is a BP neural network.

6. A transformer operating status early warning method according to claim 1, characterized in that: The target model construction method includes: Build an initial model; Collect basic data and experimental data, both of which include the measured increment of electrical values ​​at any time, as well as the variation sequence of each environmental parameter in adjacent time periods and the average value of each environmental parameter, and the variation sequence of each environmental parameter in the basic data is 0; The difference between the experimental data and the measured increment of the electrical value in the basic data is used as the environmental increment label to calculate the mean square error loss. The initial model is updated with the goal of minimizing the mean square error loss to obtain the target model.

7. A transformer operating status early warning method according to claim 6, characterized in that: The mean square error loss satisfy: , For experimental data The initial model output results are: For experimental data The environment increment tag, For experimental data The maximum mean value of the changes in each environmental parameter, For experimental data The maximum variance of each environmental parameter change.

8. A transformer operating status early warning method according to claim 1, characterized in that: The classification model includes multiple time series feature extraction sub-models and classification sub-models. The time series feature extraction sub-model is used to extract the time series features of the second power sequence and the control parameter sequence of the controlled device, and input the time series features into the classification sub-model to obtain the operating status.

9. A transformer operating status early warning method according to claim 8, characterized in that: The training method of the classification model includes: The second power sequence and the control parameter sequence of the controlled device collected in the historical time are used as training samples, and the training samples are input into the classification model to obtain output results; The cross entropy function between the output result and the true state of the training sample is calculated to train the classification model until the cross entropy function is less than the preset function value.

Images