A dry-type transformer winding temperature rise abnormality diagnosis method

Abstract

The application provides a dry-type transformer winding temperature rise abnormality diagnosis method, comprising the following steps: S1. constructing a dry-type transformer winding temperature dynamic model; S2. obtaining each-phase theoretical temperature rise data of the dry-type transformer winding according to the dry-type transformer winding temperature dynamic model; S3. obtaining each-phase actual temperature rise data of the dry-type transformer winding in real time; S4. diagnosing whether the dry-type transformer winding temperature rise is abnormal according to the each-phase actual temperature rise data and the each-phase theoretical temperature rise data, and outputting an abnormality diagnosis result if yes. The application can accurately predict the theoretical winding temperature of the dry-type transformer through the winding temperature dynamic model, and accurately judge the specific reason of the dry-type transformer temperature rise abnormality in combination with the actual temperature rise data and a preset rule, so as to remind the staff to detect and maintain the dry-type transformer components based on the specific reason.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of power distribution fault detection, in particular to a dry-type transformer winding temperature rise abnormality diagnosis method. BACKGROUND

[0002] The dry-type transformer is a very important power supply and distribution equipment, which has a series of advantages such as no pollution, good fireproof performance, strong short-circuit resistance, and is widely used in various occasions. During the operation of the dry-type transformer, various losses will be generated, which will convert into heat energy and cause the temperature of the winding to rise. When the temperature of the winding reaches the set value, the dry-type transformer adopts forced air cooling to dissipate heat. If the heat dissipation system works abnormally, the internal overheating of the dry-type transformer winding will accelerate the aging of the insulation material, and even cause production accidents. As can be seen, the accurate judgment of the working state of the heat dissipation system of the dry-type transformer has an important influence on the safe and efficient operation of the dry-type transformer.

[0003] The existing dry-type transformer heat dissipation system mainly measures the temperature signal through the temperature sensor embedded in the winding. When the temperature of the winding reaches the preset value, the system automatically starts the fan to cool. However, this method cannot determine the specific diagnosis results such as whether the winding fan is dusty or operates normally, and whether the transformer cabinet fan operates normally.

[0004] Therefore, it is an urgent problem for those skilled in the art to provide a dry-type transformer winding temperature rise abnormality diagnosis method that can accurately predict the temperature of the dry-type transformer winding, diagnose whether the winding temperature rise is normal, and give a comprehensive diagnosis result. SUMMARY

[0005] The purpose of the present application is to provide a dry-type transformer winding temperature rise abnormality diagnosis method, which has clear logic, is safe, effective, reliable and easy to operate, can accurately predict the temperature of the dry-type transformer winding, diagnose whether the winding temperature rise is normal, and give a comprehensive diagnosis result when the winding temperature rise is abnormal.

[0006] Based on the above purpose, the technical scheme provided by the present application is as follows:

[0007] A dry-type transformer winding temperature rise abnormality diagnosis method, comprising the following steps:

[0008] S1. Constructing a dry-type transformer winding temperature dynamic model;

[0009] S2. Obtaining the theoretical temperature rise data of each phase of the dry-type transformer winding according to the dry-type transformer winding temperature dynamic model;

[0010] S3. Real-time obtaining the actual temperature rise data of each phase of the dry-type transformer winding;

[0011] S4. Diagnosing whether the winding temperature rise of the dry-type transformer is abnormal according to the actual temperature rise data of each phase and the theoretical temperature rise data of each phase, and outputting an abnormal diagnosis result if yes.

[0012] Preferably, the step S1 comprises the following steps:

[0013] A1. Obtaining original identification data of a dry-type transformer in a normal working state;

[0014] A2. Preprocessing the original identification data;

[0015] A3. Constructing an initial dry-type transformer winding temperature model;

[0016] A4. Selecting and initializing model parameters;

[0017] A5. Inputting the preprocessed original identification data into the initial dry-type transformer winding temperature model to obtain a predicted output sequence;

[0018] A6. Constructing a loss function, and updating the model parameters to minimize the loss function;

[0019] A7. Repeating steps A4 to A6 to obtain the dry-type transformer winding temperature dynamic model.

[0020] Preferably, the original identification data specifically includes average low-voltage side winding phase current, average winding phase current harmonic, average winding temperature, and average ambient temperature within a preset time.

[0021] Preferably, the step A2 comprises the following steps:

[0022] Dividing the original identification data into a training set and a test set;

[0023] Normalizing the original identification data.

[0024] Preferably, the initial dry-type transformer winding temperature model comprises:

[0025] An ARX model, an LSTM model, and a fully connected layer of a dry-type transformer winding temperature;

[0026] The LSTM model is configured to fit nonlinear regression coefficients of the ARX model;

[0027] The fully connected layer is configured to convert a feature vector.

[0028] Preferably, the step A5 comprises the following steps:

[0029] B1. Inputting a first state vector into the LSTM model to obtain a feature vector;

[0030] B2. transforming the feature vector into a state-dependent vector;

[0031] B3. obtaining a predicted output sequence according to the state-dependent vector and the second state vector;

[0032] B4. repeating steps B1 to B3 to obtain a plurality of predicted output sequences according to the training set and the test set;

[0033] wherein the second state vector comprises the low-voltage-side winding phase current mean value, the phase current harmonic mean value, and the ambient temperature mean value.

[0034] Preferably, the step A7 is specifically:

[0035] C1. obtaining a plurality of loss functions according to the plurality of predicted output sequences and the model parameters;

[0036] C2. obtaining the dry-type transformer winding temperature dynamic model according to the minimum value of the plurality of loss functions and the corresponding model parameters.

[0037] Preferably,

[0038] The each-phase theoretical temperature rise data is specifically: the low-voltage-side difference absolute value maximum value, the difference change rate maximum value, and the theoretical maximum difference value of each-phase winding temperature;

[0039] The each-phase actual temperature rise data is specifically: the low-voltage-side model difference value in the first cycle, the variance value in the first cycle, and the difference change rate in the first cycle of each-phase winding temperature;

[0040] and the low-voltage-side difference absolute value maximum value in the second cycle and the difference change rate maximum value in the second cycle of each-phase winding temperature;

[0041] wherein the first cycle is N unit times, and the second cycle is N+1 unit times.

[0042] Preferably, the step S4 further comprises: if no, updating the low-voltage-side difference absolute value maximum value and the difference change rate maximum value of each-phase winding temperature.

[0043] Preferably, the step S4 is specifically:

[0044] D11. if the each-phase actual temperature rise data is greater than the positive value of the each-phase theoretical temperature rise data, outputting each-phase winding temperature sensor failure according to a preset rule;

[0045] D12. if the each-phase actual temperature rise data is less than the negative value of the each-phase theoretical temperature rise data, outputting each-phase winding fan operation abnormality or cabinet fan abnormality or each-phase winding fan operation abnormality and cabinet fan abnormality according to the preset rule.

[0046] The dry-type transformer winding temperature rise abnormality diagnosis method disclosed by the application is a kind of winding temperature dynamic model constructed according to the key influencing factors of the dry-type transformer winding temperature, and the theoretical temperature rise data of the dry-type transformer winding temperature is more accurately obtained through the dynamic model; at the same time, the actual temperature rise data of the dry-type transformer winding is obtained in real time, the theoretical temperature rise data is combined to judge whether the temperature rise of the dry-type transformer winding is abnormal, if it is abnormal, the specific abnormal reason is calculated through the theoretical temperature rise data and the actual temperature rise data, and the abnormal diagnosis result is output. Compared with the prior art, the application can accurately predict the theoretical winding temperature of the dry-type transformer through the winding temperature dynamic model, and accurately judge the specific reason of the dry-type transformer temperature rise abnormality by combining the actual temperature rise data with the preset rule, so as to remind the staff to detect and maintain the dry-type transformer components based on this. BRIEF DESCRIPTION OF DRAWINGS

[0047] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or the prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0048] Figure 1 A dry-type transformer winding temperature rise abnormality diagnosis method flow chart is provided for the embodiments of the application.

[0049] Figure 2 A flow chart of step S1 is provided for the embodiments of the application.

[0050] Figure 3 A structure diagram of the dry-type transformer winding temperature dynamic characteristic model is provided for the embodiments of the application.

[0051] Figure 4 A flow chart of step A5 is provided for the embodiments of the application.

[0052] Figure 5 A flow chart of step A7 is provided for the embodiments of the application.

[0053] Figure 6 A flow chart of step S4 is provided for the embodiments of the application. DETAILED DESCRIPTION

[0054] With reference to the drawings and embodiments of the present application, the technical solutions in the embodiments of the present application will be described clearly and completely. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments of the present application, all other embodiments obtained by those skilled in the art without creative efforts belong to the scope of the present application.

[0055] The embodiments of the present application are written in a progressive manner.

[0056] The embodiments of the present application provide a dry-type transformer winding temperature rise abnormality diagnosis method. The technical problem that the predicted winding temperature of the dry-type transformer is inaccurate in the prior art, resulting in inaccurate diagnosis results of the temperature rise abnormality is solved.

[0057] A dry-type transformer winding temperature rise abnormality diagnosis method comprises the following steps:

[0058] S1. Constructing a dry-type transformer winding temperature dynamic model;

[0059] S2. Obtaining each-phase theoretical temperature rise data of the dry-type transformer winding according to the dry-type transformer winding temperature dynamic model;

[0060] S3. Real-time obtaining each-phase actual temperature rise data of the dry-type transformer winding;

[0061] S4. Diagnosing whether the dry-type transformer winding temperature rise is abnormal according to the each-phase actual temperature rise data and the each-phase theoretical temperature rise data, and outputting an abnormality diagnosis result if the dry-type transformer winding temperature rise is abnormal.

[0062] It should be noted that, in the present embodiment, the dry-type transformer is a three-phase winding type, and the steps of constructing the winding temperature dynamic model for each-phase winding are consistent.

[0063] In step S1, the winding temperature dynamic model is constructed according to the key influencing factors of the dry-type transformer winding temperature;

[0064] In step S2, each-phase theoretical temperature rise data of the winding is obtained through the constructed winding temperature dynamic model;

[0065] In step S3, each-phase actual temperature rise data of the winding is obtained in real time;

[0066] In step S4, whether the dry-type transformer winding temperature rise is abnormal is judged according to the each-phase theoretical temperature rise data and the each-phase actual temperature rise data of the winding, and if the dry-type transformer winding temperature rise is abnormal, the abnormality reason is further diagnosed, and an abnormality diagnosis result is outputted.

[0067] Preferably, step S1 comprises the following steps:

[0068] A1. obtaining original identification data of the dry-type transformer in a normal working state;

[0069] A2. preprocessing the original identification data;

[0070] A3. constructing an initial dry-type transformer winding temperature model;

[0071] A4. selecting and initializing model parameters;

[0072] A5. inputting the preprocessed original identification data into the initial dry-type transformer winding temperature model to obtain a predicted output sequence;

[0073] A6. constructing a loss function and updating the model parameters to minimize the loss function;

[0074] A7. repeating steps A4 to A6 to obtain a dry-type transformer winding temperature dynamic model.

[0075] It should be noted that, taking any phase winding as an example, the construction of the dry-type transformer winding temperature dynamic model can be roughly divided into: data collection and processing, model structure design, model parameter selection and initialization, forward operation, model parameter updating, and optimization to select the optimal winding temperature dynamic model.

[0076] In step A1, the original identification data of the dry-type transformer in a normal state is obtained by a collection device; in this embodiment, the data sampling period is set to 1 second, and the original identification data of the dry-type transformer in a normal working state is collected for one week, and after the large amount of data is simplified, the original identification data is obtained;

[0077] In step A2, the original identification data is preprocessed to convert the original identification data into a format that can be input into the model;

[0078] In step A3, an initial dry-type transformer winding temperature model is constructed, and since the model is an initial model, the original identification data needs to be input for optimization;

[0079] In step A4, the model parameters are selected and initialized;

[0080] In step A5, the preprocessed original identification data is input into the initial winding temperature dynamic model to obtain a predicted output sequence;

[0081] In step A6, a loss function is constructed, and the model parameters in step A4 are updated to minimize the loss function value;

[0082] In step A7, the process of steps A4 to A6 is repeated, that is, the initial winding is continuously optimized, and finally the dry-type transformer winding temperature dynamic model is obtained.

[0083] Preferably, the original identification data specifically includes: the average phase current of the low-voltage side winding, the average phase current harmonics, the average winding temperature, and the average ambient temperature within a preset time period.

[0084] In practical applications, the original identification data specifically includes the average phase current, phase current harmonics, winding temperature, and ambient temperature of the low-voltage side winding within a preset time period. In order to reduce the impact of large load fluctuations on modeling, the statistical characteristics of the sampled data were extracted, and the corresponding data for 5 consecutive minutes were used as the original identification data for the winding temperature dynamic model.

[0085] Preferably, step A2 includes the following steps:

[0086] The original identification data is divided into a training set and a test set;

[0087] Normalized raw identification data.

[0088] In practical applications, the original identification data is divided into a training set and a test set in an 8:2 ratio. The training set is used to identify model parameters, and the test set is used to test the model.

[0089] The original identification data is then normalized. In this embodiment, to ensure the reliability of the results, the present invention uses the maximum-minimum method to standardize the identification data, transforming all data to the range [0,1]. The transformation formula is as follows:

[0090]

[0091] Where y represents the normalized data, and x represents the original identification data of the model. max x min These are the maximum and minimum values ​​of phase current, phase current harmonics, winding temperature, and ambient temperature, respectively.

[0092] Preferably, the initial dry-type transformer winding temperature model includes:

[0093] ARX ​​model, LSTM model, and fully connected layer for dry transformer winding temperature;

[0094] LSTM model, used to fit the nonlinear regression coefficients of ARX model;

[0095] Fully connected layers are used to transform feature vectors.

[0096] In practical applications, such as Figure 3 As shown, an ARX model and an LSTM model for winding temperature are established, along with a fully connected layer for transforming eigenvectors. The LSTM model is used to fit the nonlinear regression coefficients of the ARX model.

[0097] It should be noted that the Auto-Regressive with Extra Inputs (ARX) model is a time series analysis method, and the model parameters thereof condense important information of a system state, and an accurate ARX model can deeply and centrally express a running law of the system.

[0098] The long-short term memory (LSTM) model is a special RNN model, and is proposed to solve the gradient dispersion problem of the RNN model; in the traditional RNN, the training algorithm uses the BPTT, and when the time is relatively long, the residual error that needs to be back propagated is exponentially decreased, which leads to slow network weight updating and cannot reflect the long-term memory effect of the RNN, and therefore a storage unit is needed to store the memory, and therefore the LSTM model is proposed.

[0099] Preferably, the step A5 comprises the following steps:

[0100] B1. inputting the first state vector into the LSTM model to obtain a feature vector;

[0101] B2. converting the feature vector into a state-dependent vector;

[0102] B3. obtaining a predicted output sequence according to the state-dependent vector and the second state vector;

[0103] B4. repeating the steps B1 to B3 to obtain a plurality of predicted output sequences according to the training set and the test set;

[0104] The second state vector comprises a low-voltage side winding phase current mean value, a phase current harmonic mean value and an ambient temperature mean value.

[0105] In the steps B1 to B3, in the embodiment, the state vector x(t-1) is first input into the m-layer LSTM network, the feature vector is calculated through the LSTM network, the feature vector is converted into the state-dependent vector by the full connection layer, the state-dependent vector is taken as the parameter of the ARX model (that is, the state-dependent vector is actually the ARX coefficient), the state-dependent vector and the input a(t-1) of the ARX model are input into the ARX model, and finally the theoretical predicted value of the winding temperature is obtained. In the model, the winding temperature is taken as the output of the ARX model, the low-voltage side winding phase current, the phase current harmonic and the ambient temperature are taken as the input of the ARX model, and the model structure can be represented as:

[0106]

[0107] Wherein, t is a sampling period of the identification data; u(t) = [u1(t) u2(t) u3(t)] TThe system identification data sampling time input is the phase current mean value, the phase current harmonic mean value, and the environment temperature mean value; y(t) = y1(t) is the system identification data sampling time output, that is, the winding temperature mean value; E(t) is a Gaussian white noise signal; p and q are the orders of the model output and input, respectively; Φ 0 j j is a coefficient matrix, which is calculated by the state vector x(t-1) through the LSTM network; θ is a state-dependent coefficient vector output by the LSTM end of the model, is the feature vector of the mth layer LSTM network at the Tth time step, W fc fc are the weights and biases of the full connection layer, is the activation function of the full connection layer.

[0108] It should be noted that after determining the model, step A4 in this embodiment is specifically: selecting the number of series-connected layers in the LSTM, the output dimension of each layer, the output variable order of the ARX model, the input variable order, the state vector order T, and initializing the above model parameters.

[0109] Step B4 in this embodiment is specifically: inputting the training set and the test set into the model to obtain a plurality of predicted output sequences; in this embodiment, the training set and the test set are used to construct the input of the LSTM, the input of the ARX, and the model output according to the determined structure, and the output at the LSTM end, that is, the coefficients of the ARX model, are calculated forwardly. Continuing to multiply the output vector with the input at the ARX end, a plurality of predicted output sequences can be obtained.

[0110] Step A6 in this embodiment is specifically: constructing a loss function E, and constantly updating the parameters of the model until the loss function is minimized through the back propagation algorithm to obtain the final model. The E formula is as follows:

[0111]

[0112] wherein E represents the mean square error of the predicted value and the actual value, and N represents the number of training samples.

[0113] Preferably, step A7 is specifically:

[0114] C1. According to a plurality of predicted output sequences and changed model parameters, a plurality of loss functions are obtained.

[0115] C2. According to the minimum value of the plurality of loss functions and the corresponding model parameters, a dry-type transformer winding temperature dynamic model is obtained.

[0116] ​​​In actual application, steps A4 to A6 are repeated, after all model structures are traversed, the loss function values under different structures are compared, the model structure and model parameters with the minimum loss function value are selected as the final model considering the model structure complexity and real-time calculation requirements.

[0117] Preferably,

[0118] The phase-by-phase theoretical temperature rise data specifically includes: the maximum value of the absolute value of the low-voltage side difference of the phase-by-phase winding temperature, the maximum value of the difference change rate, and the theoretical maximum difference;

[0119] The phase-by-phase actual temperature rise data specifically includes: the low-voltage side model difference of the phase-by-phase winding temperature in the first cycle, the variance value in the first cycle, and the difference change rate in the first cycle;

[0120] and the maximum value of the absolute value of the low-voltage side difference of the phase-by-phase winding temperature in the second cycle and the maximum value of the difference change rate in the second cycle;

[0121] The first cycle is N unit times, and the second cycle is N+1 unit times.

[0122] In actual application, the method for obtaining the above data specifically includes: in the modeling data collection and processing stage of the dry-type transformer winding temperature dynamic characteristic model, the maximum value of the absolute value of the low-voltage side phase-by-phase winding temperature difference TAB max-s , TAC max-s , TBC max-s , the maximum value of the difference change rate TABV max-s , TACV max-s , TBCV max-s , the theoretical maximum difference of the phase-by-phase winding temperature TDA max-s , TDB max-s , TDC max-s is obtained by using the final model structure and parameters to obtain the optimal output sequence, and the difference between the model actual output sequence is calculated to obtain the theoretical maximum difference of the phase-by-phase winding temperature TDA max-s , TDB max-s , TDC max-s , which is used as the initial judgment threshold in the subsequent winding temperature rise abnormality diagnosis process.

[0123] The system collects the low-voltage side phase-by-phase winding current, phase current harmonic, winding temperature, and ambient temperature of the dry-type transformer in real time, and sets the data sampling frequency to 1 second each time. The data is continuously collected, and every 5 minutes, the average of the phase-by-phase winding current, the average of the phase current harmonic, the average of the winding temperature, and the average of the ambient temperature in the time period are calculated. A 1-hour sliding moving window is established, and the data sampling frequency in the window is 1 minute each time.

[0124] When the required data amount of the model structure is reached, when the Nth 5-minute (i.e., the first cycle) collection is completed, the model prediction is entered, and the difference TDAmax-c , TDB max-c , TDC max-c At the same time, the variance S of the temperature of each phase winding in the Nth 5 minutes is calculated A , S B , S C , and the absolute value of the difference TAB i , TAC i , TBC i of the temperature of each phase winding and the maximum value of the difference variation rate TABV i , TACV i , TBCV i are calculated, where i ranges from 1 to 5.

[0125] The sliding moving window is moved forward by 5 minutes to cover the data of the Nth 5 minutes (i.e. the second period), and the absolute value of the difference of the temperature of each phase winding TAB max-h , TAC max-h , TBC max-h , and the maximum value of the difference variation rate TABV max-h , TACV max-h , TBCV max-h in the sliding moving window are calculated.

[0126] Preferably, the step S4 further comprises: if no, updating the maximum value of the absolute value of the difference of the temperature of each phase winding on the low-voltage side and the maximum value of the difference variation rate.

[0127] In actual application, in the embodiment, when -TDA max-s ≤ TDA max-c ≤ TDA max-s , -TDB max-s ≤ TDB max-c ≤ TDB max-s , and -TDC max-s ≤ TDC max-c ≤ TDC max-s , it indicates that the temperature rise of each phase winding is normal, and at this time, the threshold value can be updated, and the threshold value updating method is as follows:

[0128] If TAB max-h > TAB max-s , then TAB max-s = TAB max-h , otherwise, TAB max-s remains unchanged;

[0129] If TAC max-h > TAC max-s , then TAC max-s = TAC max-h , otherwise, TAC max-s remains unchanged;

[0130] If TBC max-h > TBC max-s , then TBC max-s = TBC max-h , otherwise TBC max-s is unchanged.

[0131] If TABV max-h > TABV max-s , then TABV max-s = TABV max-h , otherwise TABV max-s is unchanged.

[0132] If TACV max-h > TACV max-s , then TACV max-s = TACV max-h , otherwise TACV max-s is unchanged.

[0133] If TBCV max-h > TBCV max-s , then TBCV max-s = TBCV max-h , otherwise TBCV max-s is unchanged.

[0134] Preferably, step S4 is specifically:

[0135] D11. If the actual temperature rise data of each phase is greater than the positive value of the theoretical temperature rise data of each phase, output the winding temperature sensor fault of each phase according to the preset rule;

[0136] D12. If the actual temperature rise data of each phase is less than the negative value of the theoretical temperature rise data of each phase, output the fan operation abnormality of each phase winding, the cabinet fan abnormality, or the fan operation abnormality of each phase winding and the cabinet fan abnormality according to the preset rule.

[0137] In actual application, step D11 is specifically: when TDA max-c > TDA max-s , it indicates that the actual temperature rise of the A-phase winding model is greater than the positive value of the theoretical temperature rise, at this time if S A ≤ A, A is a small enough empirical threshold value, and at least one of TAB i is greater than TAB max-s , at least one of TAC i is greater than TAC max-s , and TBC i is less than TBC max-s , it is judged that the temperature rise abnormality is caused by the A-phase winding temperature sensor fault or the installation position moving;

[0138] Step D12 is specifically: when TDAmax-c < TDA max-s , indicates that the actual temperature rise of the A-phase winding model is less than the theoretical temperature rise negative value, at this time if TAB i At least one is greater than TAB max-s , TAC ii At least one is greater than TAC max-s , TBC i All are less than TBC max-s , TABV i All are less than TBCV max-s , then determine the cause of the abnormal temperature rise A-phase winding fan is not normal or dust accumulation; if TAB i All are less than TAB max-s , TAC i All are less than TAC max-s , TBC i All are less than TBC max-s , TABV i At least one is greater than TABV max-s , TACV ii At least one is greater than TACV max-s , TBCV i At least one is greater than TBCV max-s , then determine the cause of the abnormal temperature rise cabinet fan is not running normally; if TAB i At least one is greater than TAB max-s , TAC i At least one is greater than TAC max-s , TBC i All are less than TBC max-s , TABV i At least one is greater than TABV max-s , TACV ii At least one is greater than TACV max-s , TBCV i At least one is greater than TBCV max-s , then determine the cause of the abnormal temperature rise A-phase winding fan is not running normally or dust accumulation, and the cabinet fan is not running normally.

[0139] It should be noted that the A-phase winding temperature rise abnormality determination method is as above, since the steps of establishing the winding temperature dynamic model of each phase winding are consistent, therefore, B-phase and C-phase same as A-phase are also determined in the same way, ultimately can more accurately and more comprehensively diagnose which phase of the dry-type transformer has a temperature rise fault and the specific reason of the fault, and output the diagnosis result.

[0140] Those skilled in the art can understand that all or part of the steps of the above-mentioned method embodiments can be completed by program instructions and related hardware. The aforementioned program instructions can be stored in a computer-readable storage medium, and the program instructions perform the steps of the above-mentioned method embodiments when executed. The aforementioned storage medium includes a mobile storage device, a read-only memory (ROM), a magnetic disc or an optical disc, and various media that can store program codes.

[0141] It should be understood that if "system", "device", "unit" and / or "module" are used in the present application, it is only a method for distinguishing different components, elements, parts, portions or assemblies at different levels. However, if other words can achieve the same purpose, the words can be replaced by other expressions.

[0142] As shown in the present application and claims, unless the context clearly indicates otherwise, "one", "a", "an", and / or "the" do not refer to the singular, but can include the plural as well. Generally, the terms "comprising" and "including" only indicate the inclusion of the steps and elements explicitly identified, and these steps and elements do not constitute an exclusive list of steps or elements. The element defined by the phrase "comprising a" does not exclude the presence of additional identical elements in the process, method, product or device comprising the element.

[0143] If flowcharts are used in the present application, the flowcharts are used to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or subsequent operations are not necessarily performed in sequence. Instead, each step can be processed in reverse order or simultaneously. Meanwhile, other operations can be added to these processes, or one or more steps can be removed from these processes.

[0144] The above has carried on the detailed introduction to the dry-type transformer winding temperature rise abnormal diagnosis method provided by the present application. The above description of the disclosed embodiments enables those skilled in the art to implement or use the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for diagnosing abnormal temperature rise in the windings of a dry-type transformer, characterized in that, Includes the following steps: S1. Construct a dynamic model of the winding temperature of a dry-type transformer; S2. Obtain the theoretical temperature rise data of each phase of the dry transformer winding based on the dynamic temperature model of the dry transformer winding; S3. Real-time acquisition of actual temperature rise data for each phase of the dry-type transformer winding; S4. Based on the actual temperature rise data of each phase and the theoretical temperature rise data of each phase, diagnose whether the temperature rise of the dry-type transformer winding is abnormal. If so, output the abnormal diagnosis result. The initial dry-type transformer winding temperature model includes: ARX ​​model, LSTM model, and fully connected layer for dry transformer winding temperature; The LSTM model is used to fit the nonlinear regression coefficients of the ARX model; The fully connected layer is used to transform the feature vector; B1. Input the first state vector into the LSTM model to obtain the feature vector; B2. The feature vectors are transformed into nonlinear regression coefficients of the ARX model by a fully connected layer; B3. Based on the nonlinear regression coefficients and the second state vector of the ARX model, obtain the predicted output sequence through the ARX model; B4. Repeat steps B1 to B3. Based on the training set and test set, obtain several predicted output sequences through the ARX model. The output is the ideal predicted value of the winding temperature. Specifically, the first state vector is: state vector x(t-1); The second state vector includes: the average phase current of the low-voltage side winding, the average phase current harmonics, and the average ambient temperature; The specific theoretical temperature rise data for each phase are: the maximum absolute value of the temperature difference on the low-voltage side of each phase winding, the maximum rate of change of the temperature difference, and the theoretical maximum temperature difference. The actual temperature rise data for each phase are specifically: the low-voltage side model difference of each phase winding temperature in the first cycle, the variance value in the first cycle, and the rate of change of the difference in the first cycle. The maximum absolute value of the temperature difference on the low-voltage side during the second cycle and the maximum rate of change of the temperature difference during the second cycle; The first period consists of N units of time, and the second period consists of N+1 units of time. Step S4 specifically involves: D11. If the difference in the low-voltage side model temperature of each phase winding is greater than the positive value of the theoretical maximum difference of each phase in the first cycle, output that the temperature sensor of that phase winding has failed according to the preset rule. D12. If the difference in low-voltage side model temperature of each phase winding within the first cycle is less than the negative value of the theoretical maximum difference of each phase, then according to the preset rule, output that the phase winding fan is abnormal or the cabinet fan is abnormal or the phase winding fan and cabinet fan are abnormal.

2. The method for diagnosing abnormal temperature rise in dry-type transformer windings as described in claim 1, characterized in that, Step S1 includes the following steps: A1. Obtain the raw identification data of the dry-type transformer under normal operating conditions; A2. Preprocess the original identification data; A3. Construct an initial dry-type transformer winding temperature model; A4. Select and initialize model parameters; A5. Input the pre-processed raw identification data into the initial dry transformer winding temperature model to obtain the predicted output sequence; A6. Construct a loss function and update the model parameters to minimize the loss function; A7. Repeat steps A4 to A6 to obtain the dynamic model of the dry transformer winding temperature; Specifically, the predicted output sequence is the theoretical predicted value of the winding temperature.

3. The method for diagnosing abnormal temperature rise in dry-type transformer windings as described in claim 2, characterized in that, The original identification data specifically includes: the average phase current of the low-voltage side winding, the average phase current harmonics, the average winding temperature, and the average ambient temperature within a preset time period.

4. The method for diagnosing abnormal temperature rise in dry-type transformer windings as described in claim 3, characterized in that, Step A2 includes the following steps: The original identification data is divided into a training set and a test set; The original identification data is normalized.

5. The method for diagnosing abnormal temperature rise in dry-type transformer windings as described in claim 1, characterized in that, Step A7 specifically involves: C1. Obtain several loss functions based on several predicted output sequences and by changing the model parameters; C2. Obtain the dynamic model of the dry-type transformer winding temperature based on the minimum value among several loss functions and the corresponding model parameters.

6. The method for diagnosing abnormal temperature rise in dry-type transformer windings as described in claim 1, characterized in that, Step S4 further includes: if not, updating the maximum absolute value of the low-voltage side temperature difference and the maximum rate of change of the difference for each phase winding.

Images