Transformer fault diagnosis method, device, equipment and medium

Abstract

The invention discloses a transformer fault diagnosis method, device, equipment and medium, and relates to the technical field of transformer fault diagnos.The transformer fault diagnosis method comprises the steps that terahertz spectrum information of transformer fault insulating oil is obtained through the terahertz time-domain spectroscopy technology, and CEEMD is used for decomposing terahertz spectrum sequence data; the fuzzy entropy of each sub-sequence after decomposition is solved to form a characteristic quantity reflecting the fault state of the transformer, only terahertz spectrum information of transformer insulating oil is extracted and processed in the process, influences of the insulating oil, internal and external environments and transformer equipment are neglected, and the fault state of the transformer is obtained. And then a brand new transformer fault diagnosis model is established according to a regularization extreme learning machine RELM and a hiking optimization method HOA, the fault type of the transformer is diagnosed according to the transformer fault diagnosis model, and the accuracy of transformer fault diagnosis is greatly improved.

Description

Technical Field

[0001] The present invention relates to the technical field of transformer fault diagnosis, and in particular to a transformer fault diagnosis method, device, equipment and medium. Background Art

[0002] As one of the important electrical equipment in the power system, the transformer undertakes the important tasks of electric energy transmission and voltage conversion. It is a static electrical equipment, which is used to convert a certain value of AC voltage into another or several different values ​​of voltage with the same frequency. It usually has two or more windings. In order to transmit electric energy, at the same frequency, the AC voltage and current of one system are converted into the voltage and current of another system through electromagnetic induction. Usually, the values ​​of these currents and voltages are different.

[0003] The role of the transformer is multifaceted. It can not only increase the voltage to deliver electricity to the power consumption area, but also reduce the voltage to various levels of usage voltage to meet the power consumption needs. Once it fails, it will cause local or even large-scale power outages, which will inevitably cause huge economic losses. The main faults of the transformer usually include oil leakage at the welding point, oil leakage at the seal, oil leakage at the flange connection, oil leakage at the bolt or pipe thread, oil leakage at the cast iron part, oil leakage at the radiator, and oil leakage at the porcelain bottle and glass oil meter. Therefore, establishing an efficient and accurate transformer fault diagnosis method and mastering its movement state are of great significance to improving the transformer operation level and ensuring the safe and reliable operation of the power system.

[0004] At present, the commonly used transformer fault diagnosis methods include IEC ratio method, Rogers ratio method and David triangle method, as well as intelligent diagnosis methods based on the characteristics of dissolved gas in oil. These methods are based on chromatography to obtain the gas data dissolved in oil, and use the gas data to determine the type of transformer fault. However, the gas content data is easily affected by the insulating oil in the transformer and the internal and external environment, as well as the transformer equipment itself, making it difficult for the measured gas content data to reflect the actual transformer status, and thus the diagnostic accuracy of transformer faults is low. Summary of the invention

[0005] The embodiments of the present invention provide a transformer fault diagnosis method, device, equipment and medium, which can solve the problem in the prior art that the current method is easily affected by the insulating oil in the transformer and the internal and external environment, as well as the transformer equipment itself, making it difficult for the measured gas content data to reflect the actual transformer state, and thus the diagnosis accuracy of the transformer fault is low.

[0006] An embodiment of the present invention provides a transformer fault diagnosis method, comprising the following steps:

[0007] Collect terahertz time-domain spectrum signal data of transformer insulating oil and form time-domain sequence data;

[0008] The complementary set empirical mode decomposition (CEEMD) is used to decompose the time domain series data to obtain the intrinsic mode component IMFs subsequences with different center frequencies; the fuzzy entropy FE method is used to obtain the fuzzy entropy value of each intrinsic mode component IMFs subsequence, and the fuzzy entropy value of each intrinsic mode component IMFs subsequence is combined into multiple feature column vectors;

[0009] A transformer fault diagnosis model based on regularized extreme learning machine RELM is constructed; multiple feature column vectors are used to train the transformer fault diagnosis model, and during the training process, the walking optimization method HOA is used to simulate the constraints on the transformer fault diagnosis model to update the model parameters of the transformer fault diagnosis model and obtain the trained transformer fault diagnosis model;

[0010] The characteristic column vector of the transformer to be tested is input into the trained transformer fault diagnosis model to obtain the fault type of the transformer after diagnosis.

[0011] Preferably, the obtaining of intrinsic mode component IMFs subsequences with different center frequencies comprises:

[0012] Add different noise instances to the original time domain series data s(t) to obtain a new set of time domain series data, the acquisition equation is:

[0013] s i (t) = s(t) + noise i

[0014] Where: s(t) represents the original signal; s i (t) represents a signal with noise added; noise i represents a random noise instance;

[0015] For each signal s with added noise i (t) Perform empirical mode decomposition (EMD) to obtain a set of intrinsic mode components, whose decomposition equation is:

[0016]

[0017] Of which: IMF i，k (t) represents the kth intrinsic mode function of the ith signal; N represents the number of intrinsic mode functions; residue i (t) represents the residual of the i-th signal;

[0018] After performing empirical mode decomposition (EMD), a set of intrinsic mode components is obtained for ensemble averaging, and the ensemble average equation is:

[0019]

[0020] in: represents the final k-th intrinsic mode function; M represents the number of noise instances; represents the k-th mode of the ith noise instance;

[0021] The intrinsic mode components IMFs with high consistency in different noise instances are selected from the set after the set average, and the intrinsic mode components IMFs are used as the intrinsic mode components IMFs subsequences with different center frequencies.

[0022] Preferably, the method of obtaining the fuzzy entropy value of each intrinsic mode component IMFs subsequence by using the fuzzy entropy FE method includes:

[0023] Get the intrinsic mode component IMFs subsequence and The similarity of is, and the similarity acquisition equation is:

[0024]

[0025] Where: n represents the exponent of the fuzzy function; r represents the similarity tolerance threshold; Representation sequence and Chebyshev distance; represents a sequence of length m;

[0026] According to the similarity between each intrinsic mode component IMFs subsequence, the average similarity of each intrinsic mode component IMFs subsequence is obtained, and the acquisition equation is:

[0027]

[0028] Where: N represents the total length of the sequence;

[0029] According to each intrinsic mode component IMFs subsequence, the fuzzy entropy value of each intrinsic mode component IMFs subsequence is obtained, and the acquisition equation is:

[0030] FuzzyEn(m,n,r)=lnφ m,n,r -lnφ m+1,r,n

[0031] Among them: m+1 represents the dimension.

[0032] Preferably, the updating of the model parameters of the transformer fault diagnosis model comprises:

[0033] When training the transformer fault diagnosis model, firstly, the parameters of the regularized extreme learning machine RELM and the hiking optimization method HOA are set, and the parameters include the population size, the maximum number of iterations, the number of hidden layer nodes and the regularization parameter;

[0034] According to the step size and fitness threshold formed by the walking optimization method HOA when continuously iteratively optimizing the speed and position of the hiker in the Tobler walking function, the parameters of the regularized extreme learning machine RELM are continuously updated; and the input layer weights and hidden layer biases obtained when the iteration is completed are set as the input parameters of the regularized extreme learning machine RELM to complete the optimization update of the model parameters of the transformer fault diagnosis model.

[0035] Preferably, the iterative optimization of the speed and position of the hiker within the Tobler hiking function comprises:

[0036] Tobler's walking function is expressed as:

[0037]

[0038] Where: V i，t represents the speed of hiker i at iteration or time t; S i，t Indicates the slope of the terrain;

[0039] The speed of the hiker in the Tobler hiking function is expressed as:

[0040] V i,t =V i,t-1 +γ i,t (β best -α i,t β i,t )

[0041] Where: V i，t Indicates the current speed; V i，t-1 represents the initial velocity; γ i,t represents a random number in the range [0,1]; β best represents the position of the leading hiker; α i，t It indicates that the sweep factor SF of the hiker is in the range [1,3];

[0042] The position of the hiker in the Tobler hiking function is expressed as:

[0043] β i,t+1 =β i,t +V i,t

[0044] Where: β i,t+1 Indicates the current location;

[0045] Initialize the positions of the hikers. Each hiker corresponds to a result, which contains the input layer weights and hidden layer biases.

[0046] Get the walking speed of each hiker according to the Tobler hiking function, update the real-time position of each hiker, and update the current fitness and terrain to adjust the step length;

[0047] The walking speed and real-time location of each hiker are continuously optimized iteratively to obtain the optimal fitness threshold and step length.

[0048] The embodiment of the present invention further provides a transformer fault diagnosis device, comprising:

[0049] A data acquisition module is used to collect terahertz time-domain spectrum signal data of transformer insulating oil and form time-domain sequence data;

[0050] The data processing module is used to decompose the time domain sequence data by using complementary set empirical mode decomposition (CEEMD) to obtain intrinsic mode component IMFs subsequences with different center frequencies; the fuzzy entropy FE method is used to obtain the fuzzy entropy value of each intrinsic mode component IMFs subsequence, and the fuzzy entropy value of each intrinsic mode component IMFs subsequence is combined into multiple feature column vectors;

[0051] A model module is used to construct a transformer fault diagnosis model based on a regularized extreme learning machine RELM; multiple feature column vectors are used to train the transformer fault diagnosis model, and during the training process, a walking optimization method HOA is used to simulate constraints on the transformer fault diagnosis model to update the model parameters of the transformer fault diagnosis model, thereby obtaining a trained transformer fault diagnosis model;

[0052] The fault identification module is used to input the characteristic column vector of the transformer to be tested into the trained transformer fault diagnosis model to obtain the fault type of the transformer after diagnosis.

[0053] An embodiment of the present invention further provides an electronic device, including a memory and a processor;

[0054] The memory is used to store computer programs;

[0055] The processor is used to implement the steps of the transformer fault diagnosis method as described above when executing the computer program stored in the memory.

[0056] An embodiment of the present invention further provides a computer-readable storage medium for storing a computer program, wherein when the computer program is executed by a processor, the steps of the transformer fault diagnosis method described above are implemented.

[0057] The embodiments of the present invention provide a transformer fault diagnosis method, device, equipment and medium. Compared with the prior art, the advantages thereof are as follows:

[0058] The present invention adopts terahertz time-domain spectroscopy technology to obtain terahertz spectrum information of transformer fault insulating oil, and uses complementary set empirical mode decomposition (CEEMD) to decompose terahertz spectrum sequence data, and obtains the fuzzy entropy of each subsequence after decomposition to form a feature quantity reflecting the fault state of the transformer. This process only extracts and processes the terahertz spectrum information of the transformer insulating oil, that is, only extracts the feature quantity reflecting the fault state of the transformer, and ignores the influence of the insulating oil, internal and external environment and the transformer equipment itself on the terahertz spectrum information. Then, a new transformer fault diagnosis model is established according to the regularized extreme learning machine RELM and the hiking optimization method HOA. The established model can efficiently search for more accurate fault information in the feature quantity. Finally, the fault type of the transformer is diagnosed according to the transformer fault diagnosis model, and the accuracy of transformer fault diagnosis is greatly improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 A schematic diagram of the overall process of a transformer fault diagnosis method provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0060] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the specific embodiments of the present invention are described in detail below in conjunction with the accompanying drawings. In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention can be implemented in many other ways different from those described herein, and those skilled in the art can make similar improvements without violating the connotation of the present invention, so the present invention is not limited by the specific embodiments disclosed below.

[0061] See also Figure 1 The embodiment of the present invention provides a transformer fault diagnosis method, comprising the following steps:

[0062] Step 1: Use a terahertz spectrometer to collect the terahertz time-domain spectrum signal of the insulating oil under transformer fault.

[0063] Step 2: Process the collected terahertz time-domain spectrum signal of transformer fault insulating oil.

[0064] 1. First, the collected terahertz time-domain spectrum signal of the transformer fault insulating oil is decomposed by complementary set empirical mode decomposition to obtain multiple modal function (IMFs) components after decomposition; the specific steps of complementary set empirical mode decomposition include:

[0065] ①Add noise.

[0066] Add different noise instances to the original signal s(t) to generate a new set of signals, whose formula is:

[0067] s i (t) = s(t) + noise i

[0068] Where: s(t) represents the original signal; s i (t) represents a signal with noise added; noise i represents a random noise instance.

[0069] ②EMD decomposition.

[0070] For each signal s with added noise i (t) Perform empirical mode decomposition (EMD) to obtain a set of intrinsic mode functions, whose formula is:

[0071]

[0072] Of which: IMF i，k (t) represents the kth intrinsic mode function of the ith signal; N represents the number of intrinsic mode functions; residue i (t) represents the residual of the i-th signal.

[0073] ③ Ensemble average, the formula is:

[0074]

[0075] in: represents the final k-th intrinsic mode function; M represents the number of noise instances; represents the k-th mode of the ith noise instance.

[0076] ④Mode selection.

[0077] The stable IMFs components with high consistency in different noise instances are selected from the set.

[0078] 2. Then calculate the fuzzy entropy for each IMF component. Fuzzy entropy is an indicator to measure the complexity and uncertainty of the signal, which can effectively characterize the randomness and regularity of the signal. The specific calculation steps include:

[0079] ①Sequence definition.

[0080] The time series data is divided into sequences of length m and recorded as in and subtract its mean from Get the mean-removed sequence.

[0081] ②Definition of distance.

[0082] Define two sequences and The distance between is the Chebyshev distance, that is, the absolute value of the maximum difference between the corresponding elements of the two sequences.

[0083] ③Similarity definition.

[0084] Introducing fuzzy membership To measure and The similarity is defined using an exponential function, and the formula is:

[0085]

[0086] Where: n represents the exponent of the fuzzy function; r represents the similarity tolerance threshold, which is usually taken as a certain proportion of the sequence standard deviation (such as 0.2 times the standard deviation).

[0087] ④Average similarity calculation.

[0088] Calculate the average similarity φ for each sequence m，r，n , and its calculation formula is:

[0089]

[0090] Where: N represents the total length of the sequence.

[0091] ⑤Fuzzy entropy calculation.

[0092] Increase the dimension to m+1 and repeat the above steps to calculate φ m+1，r，n , and then calculate the fuzzy entropy, the formula is:

[0093] FuzzyEn(m,n,r)=lnφ m,n,r -lnφ m+1,r,n .

[0094] 3. Finally, the fuzzy entropy of each IMF component is combined into a column vector to form the characteristic sample of the terahertz spectrum signal of the transformer fault insulating oil.

[0095] Step 3: Sample division of terahertz time-domain spectrum characteristics of transformer fault insulating oil.

[0096] The random sampling method is used to divide the terahertz time-domain spectrum feature samples of the pre-processed transformer fault insulating oil into a training set and a test set in the ratio of 70% and 30%.

[0097] Step 4: Establish a transformer fault diagnosis model based on terahertz time-domain spectral characteristics.

[0098] The training set is used to train the transformer fault diagnosis model, and a transformer fault diagnosis model based on terahertz time-domain spectral features is established. Then the test set is used to test the diagnostic effect of the model to verify the model performance.

[0099] The present invention selects a regularized extreme learning machine (RELM) to establish a transformer fault diagnosis model, and adopts a hiking optimization algorithm (HOA) to determine RELM model parameters.

[0100] 1. Regularized extreme learning machine.

[0101] Regularized extreme learning machine (RELM) is based on structural risk minimization. It introduces regularization into extreme learning machine (ELM) and considers the terms that control the complexity of machine learning in the optimization objective function. It improves the problem of ELM's easy overfitting and improves the generalization and robustness of the algorithm.

[0102] ELM is a single hidden layer feedforward neural network. The input layer weights and hidden layer biases are randomly generated. During the execution process, only the number of hidden layer nodes needs to be set, and the weight matrix between the hidden layer and the output layer is obtained through learning or training.

[0103] Any given N samples where x i =[x1,x2,…,x n ] T ∈R n ,y i =[y1,y2,…,y m ] T ∈R m , then the number of input and output neurons of the model are n and m respectively, and the ELM with S hidden layer nodes should try to satisfy the following formula:

[0104]

[0105] Where: β i =[β1,β2,···,β n ] T represents the output weight vector between the output node and the i-th hidden layer node; g(·) represents the activation function; w i =[w1,w2,…,w n ] T represents the input weight vector between the input node and the i-th hidden layer node; b i represents the bias of the i-th hidden layer node.

[0106] set up Then the above formula can be rewritten as: Hβ=Y, where Therefore, the learning objective function of ELM is:

[0107] min||Hβ-Y||

[0108] When g(·) is infinitely differentiable, the output parameters of the extreme learning machine do not need to be adjusted and can be directly obtained by calculation; therefore, the minimum and only solution β that satisfies the above equation can be calculated as:

[0109] β=H + Y

[0110] Where: H + represents the Moore-Penrose generalized inverse of the hidden layer output matrix H.

[0111] RELM introduces a regular term to control the norm of the output weight matrix (β), and its learning objective function is:

[0112]

[0113] Where: λ>0 means and Tradeoff parameters.

[0114] 2. Walking optimization algorithm.

[0115] Since the hyperparameters of the RELM model have a significant impact on its training and learning effects, HOA is introduced to optimize the input layer weights and hidden layer biases of RELM to enhance the generalization of the RELM model. HOA is a meta-heuristic optimization algorithm inspired by hiking experience. The mathematical model of HOA is based on the famous Tobler's Hiking Function (THF), which determines the speed of the hiker by considering the altitude and walking distance of the terrain. The specific implementation process of the HOA algorithm includes:

[0116] ①Tobler walking function (THF), its formula is:

[0117]

[0118] Where: V i，t represents the speed of hiker i at iteration or time t (unit: km / h); S i，t Indicates the slope of the terrain.

[0119] ② Slope calculation, the formula is:

[0120]

[0121] Where: dh is the change in altitude; dt is the distance covered by the hiker; θ i，t represents the hiker's inclination angle at time t.

[0122] ③ The walker speed is updated, and the formula is:

[0123] V i,t =V i,t-1 +γ i,t (β best -α i,t β i,t )

[0124] Where: V i，t Indicates the current speed; V i，t-1 represents the initial velocity; γ i,t represents a random number in the range [0,1]; β best represents the position of the leading hiker; α i，t It indicates that the sweep factor (SF) of the hiker is in the range [1,3].

[0125] ④The hiker’s position is updated, and the formula is:

[0126] β i,t+1 =β i,t +V i,t

[0127] Where: β i,t+1 Indicates the updated position.

[0128] 3. The specific steps of optimizing RELM based on HOA are:

[0129] ① Input data and initialization operation, input pre-processed transformer fault insulating oil terahertz time domain spectrum data.

[0130] ② Set the parameters of HOA and RELM, including: population size, maximum number of iterations, number of hidden layer nodes, regularization parameters, etc.

[0131] ③ Initialize the positions of the hikers, each hiker corresponds to a potential solution (a combination of the input layer weights and the hidden layer bias).

[0132] ④ Calculate the corresponding fitness size for each hiker, that is, the value of the objective function.

[0133] ⑤ Calculate the walking speed of each hiker according to the Tobler hiking function, update the position, and adjust the step length according to the current fitness and terrain (search space).

[0134] ⑥Continuously iterate to find the optimal hiker position until the termination condition is met (the maximum number of iterations or the fitness threshold is reached).

[0135] ⑦ The input layer weights and hidden layer biases obtained after the iteration are used as the input parameters of RELM, and the optimal HOA-RELM model is established for transformer fault diagnosis.

[0136] As a new type of fast, reliable and non-destructive testing technology, terahertz time-domain spectroscopy technology has excellent coherence, extremely high signal-to-noise ratio and good penetration, and has been widely used in physical and chemical testing, non-destructive testing, materials science, biomedicine and other fields. The present invention uses terahertz time-domain spectroscopy technology to extract new feature quantities reflecting the fault state of the transformer as the input of the transformer fault diagnosis model, so as to realize accurate diagnosis of the transformer fault.

[0137] The present invention adopts terahertz time-domain spectroscopy technology to obtain terahertz spectrum information of transformer fault insulating oil, and proposes a new HOA-RELM transformer fault diagnosis method to solve the shortcomings of the existing transformer fault diagnosis method based on chromatography to obtain dissolved gas in oil. The method can realize rapid and non-contact detection of transformer faults, greatly improving the accuracy and efficiency of fault diagnosis.

[0138] The embodiment of the present invention further provides a transformer fault diagnosis device, comprising:

[0139] The data acquisition module is used to collect terahertz time-domain spectrum signal data of transformer insulating oil and form time-domain sequence data.

[0140] The data processing module is used to decompose the time domain series data by using complementary set empirical mode decomposition (CEEMD) to obtain intrinsic mode component IMFs subsequences with different center frequencies; the fuzzy entropy FE method is used to obtain the fuzzy entropy value of each intrinsic mode component IMFs subsequence, and the fuzzy entropy value of each intrinsic mode component IMFs subsequence is combined into multiple feature column vectors.

[0141] The model module is used to construct a transformer fault diagnosis model based on the regularized extreme learning machine RELM; multiple feature column vectors are used to train the transformer fault diagnosis model, and during the training process, the hiking optimization method HOA is used to simulate the constraints on the transformer fault diagnosis model to update the model parameters of the transformer fault diagnosis model, and obtain the trained transformer fault diagnosis model.

[0142] The fault identification module is used to input the characteristic column vector of the transformer to be tested into the trained transformer fault diagnosis model to obtain the fault type of the transformer after diagnosis.

[0143] An embodiment of the present invention further provides an electronic device, including a memory and a processor.

[0144] The memory is used to store computer programs.

[0145] When the processor is used to execute the computer program stored in the memory, the steps of the above transformer fault diagnosis method are implemented.

[0146] An embodiment of the present invention further provides a computer-readable storage medium for storing a computer program, and when the computer program is executed by a processor, the steps of the above transformer fault diagnosis method are implemented.

[0147] The above-mentioned embodiments only express several implementation methods of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for ordinary technicians in this field, several variations and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

Claims

1. A transformer fault diagnosis method, characterized in that: The following steps are involved: Collect terahertz time-domain spectrum signal data of transformer insulating oil and form time-domain sequence data; The complementary set empirical mode decomposition (CEEMD) is used to decompose the time domain series data to obtain the intrinsic mode component IMFs subsequences with different center frequencies; the fuzzy entropy FE method is used to obtain the fuzzy entropy value of each intrinsic mode component IMFs subsequence, and the fuzzy entropy value of each intrinsic mode component IMFs subsequence is combined into multiple feature column vectors; A transformer fault diagnosis model based on regularized extreme learning machine RELM is constructed; multiple feature column vectors are used to train the transformer fault diagnosis model, and during the training process, the walking optimization method HOA is used to simulate the constraints on the transformer fault diagnosis model to update the model parameters of the transformer fault diagnosis model and obtain the trained transformer fault diagnosis model; The characteristic column vector of the transformer to be tested is input into the trained transformer fault diagnosis model to obtain the fault type of the transformer after diagnosis.

2. A transformer fault diagnosis method according to claim 1, characterized in that: The obtaining of the intrinsic mode component IMFs subsequences with different center frequencies comprises: Add different noise instances to the original time domain series data s(t) to obtain a new set of time domain series data, the acquisition equation is: s i (t)=s(t)+noise i Where: s(t) represents the original signal; s i (t) represents a signal with noise added; noise i represents a random noise instance; For each signal s with added noise i (t) Perform empirical mode decomposition (EMD) to obtain a set of intrinsic mode components, whose decomposition equation is: Of which: IMF i，k (t) represents the kth intrinsic mode function of the ith signal; N represents the number of intrinsic mode functions; residue i (t) represents the residual of the i-th signal; After performing empirical mode decomposition (EMD), a set of intrinsic mode components is obtained for ensemble averaging, and the ensemble average equation is: in: represents the final k-th intrinsic mode function; M represents the number of noise instances; represents the k-th mode of the ith noise instance; The intrinsic mode components IMFs with high consistency in different noise instances are selected from the set after the set average, and the intrinsic mode components IMFs are used as the intrinsic mode components IMFs subsequences with different center frequencies.

3. A transformer fault diagnosis method according to claim 1, characterized in that: The fuzzy entropy FE method is used to obtain the fuzzy entropy value of each intrinsic mode component IMFs subsequence, including: Get the intrinsic mode component IMFs subsequence and The similarity of is, and the similarity acquisition equation is: Where: n represents the exponent of the fuzzy function; r represents the similarity tolerance threshold; Representation sequence and Chebyshev distance; represents a sequence of length m; According to the similarity between each intrinsic mode component IMFs subsequence, the average similarity of each intrinsic mode component IMFs subsequence is obtained, and the acquisition equation is: Where: N represents the total length of the sequence; According to each intrinsic mode component IMFs subsequence, the fuzzy entropy value of each intrinsic mode component IMFs subsequence is obtained, and the acquisition equation is: FuzzyEn(m,n,r)=lnφ m,n,r -lnφ m+1,r,n Among them: m+1 represents the dimension.

4. A transformer fault diagnosis method according to claim 1, characterized in that: The updating of the model parameters of the transformer fault diagnosis model includes: When training the transformer fault diagnosis model, firstly, the parameters of the regularized extreme learning machine RELM and the hiking optimization method HOA are set, and the parameters include the population size, the maximum number of iterations, the number of hidden layer nodes and the regularization parameter; According to the step size and fitness threshold formed by the walking optimization method HOA when continuously iteratively optimizing the speed and position of the hiker in the Tobler walking function, the parameters of the regularized extreme learning machine RELM are continuously updated; and the input layer weights and hidden layer biases obtained when the iteration is completed are set as the input parameters of the regularized extreme learning machine RELM to complete the optimization update of the model parameters of the transformer fault diagnosis model.

5. A transformer fault diagnosis method according to claim 4, characterized in that: The iterative optimization of the speed and position of the hiker in the Tobler hiking function includes: Tobler's walking function is expressed as: Where: V i，t represents the speed of hiker i at iteration or time t; S i，t Indicates the slope of the terrain; The speed of the hiker in the Tobler hiking function is expressed as: V i,t =V i,t-1 +g i,t (b best -a i,t b i,t ) Where: V i，t Indicates the current speed; V i，t-1 represents the initial velocity; γ i,t represents a random number in the range [0,1]; β best represents the position of the leading hiker; α i，t It indicates that the sweep factor SF of the hiker is in the range [1,3]; The position of the hiker in the Tobler hiking function is expressed as: β i,t+1 ＝β i,t +V i,t Where: β i,t+1 Indicates the current location; Initialize the positions of the hikers. Each hiker corresponds to a result, which contains the input layer weights and hidden layer biases. Get the walking speed of each hiker according to the Tobler hiking function, update the real-time position of each hiker, and update the current fitness and terrain to adjust the step length; The walking speed and real-time location of each hiker are continuously optimized iteratively to obtain the optimal fitness threshold and step length.

6. A transformer fault diagnosis device, characterized in that: include: A data acquisition module is used to collect terahertz time-domain spectrum signal data of transformer insulating oil and form time-domain sequence data; The data processing module is used to decompose the time domain sequence data by using complementary set empirical mode decomposition (CEEMD) to obtain intrinsic mode component IMFs subsequences with different center frequencies; the fuzzy entropy FE method is used to obtain the fuzzy entropy value of each intrinsic mode component IMFs subsequence, and the fuzzy entropy value of each intrinsic mode component IMFs subsequence is combined into multiple feature column vectors; A model module is used to construct a transformer fault diagnosis model based on a regularized extreme learning machine RELM; multiple feature column vectors are used to train the transformer fault diagnosis model, and during the training process, a walking optimization method HOA is used to simulate constraints on the transformer fault diagnosis model to update the model parameters of the transformer fault diagnosis model, thereby obtaining a trained transformer fault diagnosis model; The fault identification module is used to input the characteristic column vector of the transformer to be tested into the trained transformer fault diagnosis model to obtain the fault type of the transformer after diagnosis.

7. An electronic device, characterized in that: include: Memory and processor; The memory is used to store computer programs; The processor is used to implement the steps of a transformer fault diagnosis method according to any one of claims 1 to 5 when executing the computer program stored in the memory.

8. A computer-readable storage medium, characterized in that: Used to store a computer program, which, when executed by a processor, implements the steps of a transformer fault diagnosis method as claimed in any one of claims 1 to 5.

Images