Distribution transformer weight and overload early warning method based on self-adaptive fuzzy neural algorithm

Abstract

The invention discloses a distribution transformer heavy overload early warning method based on an adaptive fuzzy neural algorithm, and relates to the technical field of electrical equipment heavy overload early warning, and the method comprises the steps: obtaining the related data of a distribution transformer, and analyzing the characteristic factors of the data through a Pearson's correlation coefficient; carrying out data dimension reduction based on a principal component analysis method, screening characteristic factors, carrying out improved clustering on the screened characteristic factors, and decomposing a training sample into IMF components by using variational mode decomposition; performing adaptive fuzzy neural prediction on the decomposed IMF components to obtain predicted values of all the IMF components, and reconstructing the predicted values by using a sample entropy to obtain a final predicted load; and carrying out weight and overload early warning based on the final predicted load. According to the method, the influence of the heavy overload problem of the transformer on equipment is solved, and when the sample number is large enough, the time span is long, and the randomness is very high, the effectiveness of the method is more obvious.

Description

Technical Field

[0001] The invention relates to the technical field of heavy overload warning for electrical equipment, and in particular to a heavy overload warning method for a distribution transformer based on an adaptive fuzzy neural algorithm. Background Art

[0002] With the development of my country's economic strength, the scale of the power grid continues to increase. In the energy consumption of the terminal, the proportion of electric energy is increasing, and the growth rate of power load has generally remained relatively rapid. my country's peak load level continues to rise, and the imbalance between supply and demand during the peak load period continues to worsen. If a distribution transformer is overloaded, it will cause problems such as accelerated aging and burning of the transformer, which will have an impact on the safety of the distribution network. In recent years, domestic power transmission and transformation equipment has made significant progress in information monitoring, storage and sharing, and realized online monitoring and diagnosis of equipment such as transformers. However, the current online monitoring equipment still has many defects such as short life, instability, and poor effect, and there are problems such as low efficiency in analyzing the operating status of the equipment. Therefore, accurate peak load prediction can help power grid-related staff to evaluate the load peak response capability of distribution transformers and other equipment in advance, ensure the reliability of power supply in heavy overload dangerous areas, and is of great significance to the safe and reliable operation of the distribution network.

[0003] Traditional measures to control heavy overload of distribution transformers are mainly based on the assessment of current problems, which makes it difficult to detect and eliminate the hidden dangers of heavy overload in advance. How to effectively carry out heavy overload warning of distribution transformers to help operators take preventive measures to avoid heavy overload after receiving the alarm information is a problem that power companies are more concerned about in the process of building a strong smart grid. The heavy overload warning of distribution transformers is mainly based on the historical load level of the distribution transformer to predict whether there will be heavy overload accidents in the future. A heavy overload warning method and device for distribution transformers based on variational mode decomposition-adaptive fuzzy neural algorithm are proposed. After comprehensive analysis of historical data, a risk assessment is conducted based on the probability of power outages and power outage losses that may be caused by heavy overloads. It mainly focuses on the strong randomness of data in the warning model and short-term and ultra-short-term time predictions. Summary of the invention

[0004] In view of the problems existing in the existing distribution transformer heavy overload warning method based on adaptive fuzzy neural algorithm, the present invention is proposed.

[0005] Therefore, the problem to be solved by the present invention is how to effectively carry out heavy overload warning of distribution transformers, timely discover and eliminate heavy overload hazards in advance, and ensure the power supply reliability in heavy overload risk areas.

[0006] In order to solve the above technical problems, the present invention provides the following technical solutions:

[0007] In the first aspect, an embodiment of the present invention provides a distribution transformer heavy overload warning method based on an adaptive fuzzy neural algorithm, which includes obtaining distribution transformer related data, analyzing the characteristic factors of the data using the Pearson correlation coefficient; reducing the data dimension based on the principal component analysis method, screening the characteristic factors, performing improved clustering on the screened characteristic factors, and decomposing the training samples into IMF components using variational mode decomposition; performing adaptive fuzzy neural prediction on the decomposed IMF components to obtain predicted values ​​of all IMF components, and reconstructing the predicted values ​​using sample entropy to obtain the final predicted load; and performing a heavy overload warning based on the final predicted load.

[0008] As a preferred solution of the distribution transformer heavy overload warning method based on adaptive fuzzy neural algorithm described in the present invention, wherein: the distribution transformer related data includes load data, climate data and periodic data; the load data includes three-phase current and voltage, active power and reactive power data; the climate data includes rainfall, wind speed, wind direction, maximum temperature, dew point temperature, minimum temperature and air pressure; the periodic data includes monthly periodicity, daily periodicity and holiday characteristic data.

[0009] As a preferred solution of the transformer overload warning method based on the adaptive fuzzy neural algorithm of the present invention, the weather data and the correlation of holiday factors and periodic factors are analyzed based on the Pearson correlation coefficient, and the overall mean is calculated to obtain the average value of each characteristic factor. The overall mean is expressed as:

[0010]

[0011] Calculate the covariance of the characteristic factors, expressed as:

[0012]

[0013] Calculate the overall Pearson coefficient and standard deviation of the characteristic factors. The standard deviation is expressed as:

[0014]

[0015] The overall Pearson coefficient is expressed as:

[0016]

[0017] Among them, X and Y represent two characteristic factors, E(X) and E(Y) represent the means of characteristics X and Y, and n represents the number of sampling points.

[0018] As a preferred solution of the transformer heavy overload warning method based on the adaptive fuzzy neural algorithm of the present invention, the data dimension reduction includes: inputting the characteristic factors of the record and standardizing them; obtaining multiple eigenvalues ​​according to the covariance matrix and sorting them from large to small to obtain corresponding multiple eigenvectors, and the covariance moment is:

[0019]

[0020] Among them, X k is represented as a random variable, Expressed as the mean of the random variable, r nk It is represented as the covariance corresponding to the characteristic factor, and n is represented as the number of characteristic factors; multiple eigenvalues ​​are calculated according to the covariance matrix and the corresponding multiple eigenvectors are obtained by sorting them from large to small, which is represented as:

[0021]

[0022] The contribution rate is calculated based on the eigenvalue of the characteristic factor, and the principal component corresponding to the eigenvalue whose cumulative contribution rate exceeds the threshold is selected, which is expressed as:

[0023] F i =a i X 1 +a i X 2 +…+a i X n (i=1,2,…,k)

[0024] Among them, F i is the principal component of the characteristic factor after dimensionality reduction, a i Expressed as the proportion of each characteristic factor, X n Characteristic factors, k represents random variables.

[0025] As a preferred solution of the transformer overload warning method based on the adaptive fuzzy neural algorithm of the present invention, one day's load data is obtained as a test sample, the characteristic factor principal component after principal component analysis is improved and clustered, the load of the same category of days in the previous three years is taken as a training sample, and the corresponding day load is regarded as a group of load signals; the modal decomposition expression is:

[0026]

[0027] In the formula, {u k}={u 1 ,…,u K} is the eigenmode component, {ω k}={ω 1 ,…,ω K}Introduce the quadratic penalty factor and Lagrange multiplication factor into the formula, calculate the center frequency of each modal component, and transform the variational constraint problem into an unconstrained problem:

[0028]

[0029] The multiplication alternating direction method is used for iterative solution, and the iterative update expression of each variable is:

[0030]

[0031] The convergence criterion of the above update formula is shown as follows:

[0032]

[0033] Where ε is the convergence tolerance.

[0034] As a preferred solution of the transformer heavy overload warning method based on the adaptive fuzzy neural algorithm of the present invention, the adaptive fuzzy neural prediction includes establishing an input network based on two fuzzy f-then rules, inputting 9 component IMFs, performing ANFIS intensity release on the input fuzzy set and generating fuzzy rules, which is expressed as:

[0035] Based on two fuzzy f-then rules, the input network is established, and the 9-component IMF is input. The input fuzzy set is subjected to the intensity release of ANFIS and the fuzzy rules are generated, which is expressed as:

[0036]

[0037]

[0038] Among them, x and y are the inputs of node i, usually selected and It is a bell-shaped function with a maximum value of 1 and a minimum value of 0. is the normalized fuzzy set of each layer;

[0039] Calculate the output of each rule, expressed as:

[0040]

[0041] in, is the proportion of each weight in the total weight, p, q and r are conclusion parameters, is the output of each layer;

[0042] Sum each output to get the total output:

[0043]

[0044] Among them, wi is the weight, O 3 is the forecast result for each IMF component.

[0045] As a preferred solution of the transformer heavy overload warning method based on the adaptive fuzzy neural algorithm of the present invention, the prediction reconstruction includes setting the 9 predicted IMF components as

[0046] {x(i), i=1,2,…,N}, reconstruct an m-dimensional vector, expressed as y(i):

[0047] y(i)={x(i),x(i+1),x(i+2),…,x(i+m-1)}

[0048] Calculate the Euclidean distance d{y(i),y(j)} between any components of y(i) and y(j), and define the maximum distance between the components as the maximum contribution component distance D{y(i),y(j)}, expressed as:

[0049] D{y(i),y(j)}=max{|y(i+k)-x(j+k)|}

[0050] Given a threshold r and an embedding dimension m, calculate the probability measure representing the regularity of the sequence {y(i)} It is expressed as:

[0051]

[0052] Setting the embedding dimension to m+1, the resulting sample entropy is:

[0053]

[0054] Determine the first threshold, set the samples whose sample entropy values ​​are less than or equal to the first threshold as the first category, determine the second threshold, set the samples whose sample entropy values ​​are greater than the first threshold and less than the second threshold as the second category, and set the samples whose sample entropy values ​​are greater than or equal to the second threshold as the third category. The reconstructed signal is the final predicted load, and the predicted load is compared with the test load to obtain the prediction accuracy.

[0055] In the second aspect, an embodiment of the present invention provides a distribution transformer heavy overload warning system based on an adaptive fuzzy neural algorithm, which includes: an analysis module, used to obtain distribution transformer related data, and use the Pearson correlation coefficient to analyze the characteristic factors of the data; a decomposition module, used to reduce the data dimension according to the principal component analysis method, screen the characteristic factors, improve the clustering of the screened characteristic factors, and use variational mode decomposition to decompose the training samples into IMF components; a prediction module, used to perform adaptive fuzzy neural prediction on the decomposed IMF components, obtain the predicted values ​​of all IMF components, and use sample entropy to reconstruct the predicted values ​​to obtain the final predicted load; a warning module, used to issue a heavy overload warning based on the final predicted load.

[0056] In a third aspect, an embodiment of the present invention provides a computer device, comprising a memory and a processor, wherein the memory stores a computer program, wherein: when the processor executes the computer program, it implements any step of the above-mentioned distribution transformer heavy overload warning method based on the adaptive fuzzy neural algorithm.

[0057] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein: when the computer program is executed by a processor, any step of the above-mentioned distribution transformer heavy overload warning method based on the adaptive fuzzy neural algorithm is implemented.

[0058] The beneficial effect of the present invention is that the present invention solves the impact of transformer heavy overload problem on equipment. The current literature on distribution transformer heavy overload prediction is mainly based on short-term prediction, and is based on the small premise that the model training sample size is sufficient. When the number of samples is large enough, the time span is long, and the randomness is very strong, the effectiveness of the present invention will be more obvious. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for describing the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. Among them:

[0060] Figure 1 It is a flow chart of the distribution transformer heavy overload warning method based on adaptive fuzzy neural algorithm;

[0061] Figure 2 This is the original data statistical chart;

[0062] Figure 3 This is the result of principal component analysis;

[0063] Figure 4 is the IMF component diagram;

[0064] Figure 5 This is the prediction result diagram. DETAILED DESCRIPTION

[0065] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0066] In the following description, many specific details are set forth to facilitate a full understanding of the present invention, but the present invention may also be implemented in other ways different from those described herein, and those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0067] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The term "in one embodiment" that appears in different places in this specification does not necessarily refer to the same embodiment, nor does it refer to a separate or selective embodiment that is mutually exclusive with other embodiments.

[0068] Example 1

[0069] Reference Figure 1 , which is the first embodiment of the present invention, and provides a distribution transformer heavy overload warning method based on an adaptive fuzzy neural algorithm, comprising:

[0070] S1: Obtain distribution transformer related data and use the Pearson correlation coefficient to analyze the characteristic factors of the data.

[0071] Collect the load data of distribution transformers in the past three years, including three-phase current and voltage, active power and reactive power.

[0072] Collect climate data for the past three years, including rainfall, wind speed, wind direction, maximum temperature, dew point temperature, minimum temperature, air pressure, etc.

[0073] The annual, monthly, daily and holiday characteristics of the three-year load are analyzed. The characteristic factors are shown in Table 1:

[0074] Table 1 Original characteristic factors

[0075]

[0076]

[0077] S2: Data dimensionality reduction based on principal component analysis, characteristic factors are screened, improved clustering is performed on the screened characteristic factors, and variational mode decomposition is used to decompose the training samples into IMF components.

[0078] The Pearson correlation coefficient is used to analyze the correlation between various weather data, holiday factors, and periodic factors, calculate the overall mean, and obtain the average value of each characteristic factor.

[0079] The overall mean is expressed as:

[0080]

[0081] Calculate the covariance of the characteristic factors, expressed as:

[0082]

[0083] Among them, X and Y represent two characteristic factors, E(X) and E(Y) represent the means of characteristics X and Y, and n represents the number of sampling points.

[0084] Calculate the overall Pearson coefficient and standard deviation of the characteristic factors. The standard deviation is expressed as:

[0085]

[0086] The overall Pearson coefficient is expressed as:

[0087]

[0088] Input the recorded characteristic factors and standardize them;

[0089] According to the covariance matrix, multiple eigenvalues ​​are obtained and the corresponding eigenvectors are obtained by sorting them from large to small. The covariance matrix is ​​expressed as:

[0090]

[0091] Among them, X k is represented as a random variable, Expressed as the mean of the random variable, r nk It is represented as the covariance corresponding to the characteristic factors, and n is represented as the number of characteristic factors;

[0092] According to the covariance matrix, multiple eigenvalues ​​are calculated and sorted from large to small to obtain the corresponding multiple eigenvectors, which are expressed as:

[0093]

[0094] The contribution rate is calculated based on the eigenvalue of the characteristic factor, and the principal component corresponding to the eigenvalue whose cumulative contribution rate exceeds the threshold is selected, which is expressed as:

[0095] F i =a i X 1 +a i X 2 +…+a i X n (i=1,2,…,k)

[0096] Among them, F i is the principal component of the characteristic factor after dimensionality reduction, a i Expressed as the proportion of each characteristic factor, X n Characteristic factors, k represents random variables.

[0097] S3: Perform adaptive fuzzy neural prediction on the decomposed IMF components to obtain predicted values ​​of all IMF components, and use sample entropy to reconstruct the predicted values ​​to obtain the final predicted load.

[0098] VMD is a completely non-recursive signal decomposition method that can adaptively decompose the input sequence into several intrinsic modal components. The decomposition process of VMD is essentially the process of summing the bandwidth of all modal components, and the sum of the modal components is equal to the original signal. The modal decomposition expression is:

[0099]

[0100] Among them, {u k}={u 1 ,…,u K} is the eigenmode component, {ω k}={ω 1 ,…,ω K}Introduce the quadratic penalty factor and Lagrange multiplication factor into the formula, calculate the center frequency of each modal component, and transform the variational constraint problem into an unconstrained problem:

[0101]

[0102] The multiplication alternating direction method is used for iterative solution, and the iterative update expression of each variable is:

[0103]

[0104] The convergence criterion of the above update formula is shown as follows:

[0105]

[0106] Here, ε is the convergence tolerance.

[0107] Based on two fuzzy f-then rules, the input network is established, and the 9-component IMF is input. The input fuzzy set is subjected to the intensity release of ANFIS and the fuzzy rules are generated, which is expressed as:

[0108]

[0109] Among them, x and y are the inputs of node i, usually selected and It is a bell-shaped function with a maximum value of 1 and a minimum value of 0. is the normalized fuzzy set of each layer;

[0110] Calculate the output of each rule, expressed as:

[0111]

[0112] in, is the proportion of each weight in the total weight, p, q and r are conclusion parameters, is the output of each layer;

[0113] Sum each output to get the total output:

[0114]

[0115] Among them, w i is the weight, O 3 is the forecast result for each IMF component.

[0116] S4: issuing a heavy overload warning based on the final predicted load.

[0117] Set the 9 predicted IMF components as {x(i), i = 1, 2, ..., N}, reconstruct the m-dimensional vector, expressed as y(i):

[0118] y(i)={x(i),x(i+1),x(i+2),…,x(i+m-1)}

[0119] Calculate the Euclidean distance d{y(i),y(j)} between any components of y(i) and y(j), and define the maximum distance between the components as the maximum contribution component distance D{y(i),y(j)}, expressed as:

[0120] D{y(i),y(j)}=max{|y(i+k)-x(j+k)|}

[0121] Given a threshold r and an embedding dimension m, calculate the probability measure representing the regularity of the sequence {y(i)} It is expressed as:

[0122]

[0123] Setting the embedding dimension to m+1, the resulting sample entropy is:

[0124]

[0125] The distribution transformer operation can be evaluated according to the distribution transformer load rate, and the distribution transformer operation risk can be divided into three situations: normal, heavy load and overload. In actual production, the situation where the load reaches more than 80% of the distribution transformer rated capacity and lasts for two hours or more is judged as heavy load; the situation where the load reaches more than 100% of the distribution transformer rated capacity and lasts for two hours or more is judged as overload.

[0126] In summary, the present invention solves the impact of transformer overload problems on equipment. The current literature on distribution transformer overload prediction is mainly based on short-term prediction, and is based on the premise that the model training sample size is relatively sufficient. When the number of samples is large enough, the time span is long, and the randomness is very strong, the effectiveness of the present invention will be more obvious. And in terms of short-term and ultra-short-term time prediction, the present invention can also greatly improve the prediction accuracy. VMD divides the input signal into 9 IMF components, and obtains the final prediction signal after weighting the components, which is more conducive to data extraction and training. The ANFIS algorithm makes better predictions for samples with strong randomness and uncertain probability, improves the accuracy of the prediction model for heavy overload prediction, reduces errors, and improves the stability of distribution transformers.

[0127] Example 2

[0128] Reference Figure 1 Based on the first embodiment, this embodiment further provides a distribution transformer heavy overload warning system based on an adaptive fuzzy neural algorithm, including:

[0129] An analysis module is used to obtain distribution transformer related data and analyze the characteristic factors of the data using the Pearson correlation coefficient;

[0130] The decomposition module is used to reduce the data dimension according to the principal component analysis method, screen the characteristic factors, perform improved clustering on the screened characteristic factors, and decompose the training samples into IMF components using variational mode decomposition;

[0131] A prediction module is used to perform adaptive fuzzy neural prediction on the decomposed IMF components to obtain the predicted values ​​of all IMF components, and reconstruct the predicted values ​​using sample entropy to obtain the final predicted load;

[0132] The early warning module is used to issue a heavy overload early warning according to the final predicted load.

[0133] This embodiment also provides a computer device, which is suitable for the distribution transformer heavy overload warning method based on the adaptive fuzzy neural algorithm, including a memory and a processor; the memory is used to store computer executable instructions, and the processor is used to execute computer executable instructions to implement the distribution transformer heavy overload warning method based on the adaptive fuzzy neural algorithm as proposed in the above embodiment.

[0134] The computer device may be a terminal, and the computer device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer device may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device may be a touch layer covering the display screen, or a key, trackball or touchpad provided on the housing of the computer device, or an external keyboard, touchpad or mouse, etc.

[0135] This embodiment further provides a storage medium on which a computer program is stored. When the program is executed by a processor, the method for early warning of heavy overload of a distribution transformer based on an adaptive fuzzy neural algorithm as proposed in the above embodiment is implemented.

[0136] The storage medium proposed in this embodiment and the data storage method proposed in the above embodiment belong to the same inventive concept. The technical details not fully described in this embodiment can be found in the above embodiment, and this embodiment has the same beneficial effects as the above embodiment.

[0137] Example 3

[0138] Reference Figure 2 to Figure 5 This is the third embodiment of the present invention. Based on the previous two embodiments, this embodiment provides a distribution transformer heavy overload warning method based on an adaptive fuzzy neural algorithm. In order to verify the beneficial effects of the present invention, scientific demonstration is carried out through simulation experiments.

[0139] Taking the actual power system in Anshun area as an example, the load data set of three transformers under the typical distribution network line from January 1, 2022 to December 31, 2023 is used to verify the effectiveness of the method proposed in this paper. Each data structure of the sample set is two-dimensional load data, collected once every 15 minutes, and 96 data points per day. Due to the large number of samples and the long time span, only one month of data from all transformers is selected for display. The load data statistics of the three transformers under this line in August 2022 are as follows Figure 2 shown.

[0140] The correlation analysis of the characteristic factors is carried out. The correlation coefficients of each factor are shown in Table 2:

[0141] Table 2 Correlation of characteristic factors

[0142] Characteristic factors Correlation coefficient r Characteristic factors Correlation coefficient r A2 0.9 A9 0.88 A3 0.84 A10 0.72 A4 0.72 A11 0.53 A5 0.62 A12 0.71 A6 0.49 A13 0.44 A7 0.58

[0143] After the data dimension reduction, the characteristic factors are ranked by contribution, and finally the remaining temperature, rainfall, minimum temperature, maximum temperature, dew point temperature, holiday factors, and periodic factors are used as clustering indicators. Principal component analysis is performed, and the analysis diagram is shown in Table 3.

[0144] The VMD decomposition of the corresponding day load obtained by improved clustering yields 9 IMF components, such as Figure 4 shown.

[0145] Each decomposed IMF component is input into the ANFIS prediction training model to obtain the prediction result, and then the prediction result is reconstructed by sample entropy to obtain the final load, which is then compared with the true value to verify the effectiveness. The comparison results are shown in Figure 5 .

[0146] Depend on Figure 5 It can be seen that the prediction accuracy of the training data set reaches 0.99434, and the prediction accuracy of the test set reaches 0.9658, which is higher than the prediction accuracy of the traditional algorithm, verifying the reliability of this method.

[0147] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical solutions of the present invention, which should all be included in the scope of the claims of the present invention.

Claims

1. A distribution transformer heavy overload warning method based on adaptive fuzzy neural algorithm, characterized by: include, Obtain distribution transformer related data and use Pearson correlation coefficient to analyze the characteristic factors of the data; Data dimension reduction based on principal component analysis, characteristic factors are screened, improved clustering is performed on the screened characteristic factors, and the training samples are decomposed into IMF components using variational mode decomposition; Perform adaptive fuzzy neural prediction on the decomposed IMF components to obtain predicted values ​​of all IMF components, and use sample entropy to reconstruct the predicted values ​​to obtain the final predicted load; A heavy overload warning is performed based on the final predicted load.

2. The distribution transformer heavy overload early warning method based on adaptive fuzzy neural algorithm according to claim 1 is characterized in that: The distribution transformer related data includes load data, climate data and periodic data; The load data includes three-phase current and voltage, active power and reactive power data; The climate data includes rainfall, wind speed, wind direction, maximum temperature, dew point temperature, minimum temperature and air pressure; The periodic data include monthly periodicity, daily periodicity and holiday characteristic data.

3. The distribution transformer heavy overload early warning method based on adaptive fuzzy neural algorithm according to claim 2 is characterized in that: Based on the Pearson correlation coefficient, the correlation between weather data, holiday factors, and periodic factors is analyzed, and the overall mean is calculated to obtain the average value of each characteristic factor. The overall mean is expressed as: Calculate the covariance of the characteristic factors, expressed as: Calculate the overall Pearson coefficient and standard deviation of the characteristic factors. The standard deviation is expressed as: The overall Pearson coefficient is expressed as: Among them, X and Y represent two characteristic factors, E(X) and E(Y) represent the means of characteristics X and Y, and n represents the number of sampling points.

4. The distribution transformer heavy overload early warning method based on adaptive fuzzy neural algorithm as claimed in claim 3 is characterized by: The data dimension reduction includes: Input the recorded characteristic factors and standardize them; According to the covariance matrix, multiple eigenvalues ​​are obtained and the corresponding eigenvectors are obtained by sorting them from large to small. The covariance matrix is ​​expressed as: Among them, X k is represented as a random variable, Expressed as the mean of the random variable, r nk It is represented as the covariance corresponding to the characteristic factors, and n is represented as the number of characteristic factors; According to the covariance matrix, multiple eigenvalues ​​are calculated and sorted from large to small to obtain corresponding multiple eigenvectors, which are expressed as: The contribution rate is calculated based on the eigenvalue of the characteristic factor, and the principal component corresponding to the eigenvalue whose cumulative contribution rate exceeds the threshold is selected, which is expressed as: F i =a i X1+a i X2+…+a i X n (i=1,2,…,k) Among them, F i is the principal component of the characteristic factor after dimensionality reduction, a i Expressed as the proportion of each characteristic factor, X n is the characteristic factor, and k represents a random variable.

5. The distribution transformer heavy overload early warning method based on adaptive fuzzy neural algorithm according to claim 4 is characterized in that: The variational mode decomposition includes: One day's load data is obtained as a test sample, and the characteristic factors after principal component analysis are clustered. The loads of the same type of days in the previous three years are taken as training samples, and the corresponding day loads are regarded as a group of load signals. The expression of the modal decomposition is: Among them, {u k }={u1,…,u K } is the eigenmode component, {ω k }={ω1,…,ω K }Introduce the quadratic penalty factor and Lagrange multiplication factor into the formula, calculate the center frequency of each modal component, and transform the variational constraint problem into an unconstrained problem: The multiplication alternating direction method is used for iterative solution, and the iterative update expression of each variable is: The convergence criterion of the above update formula is shown as follows: Here, ε is the convergence tolerance.

6. The distribution transformer heavy overload warning method based on adaptive fuzzy neural algorithm according to claim 5 is characterized by: The adaptive fuzzy neural prediction includes: Based on two fuzzy f-then rules, the input network is established, and the 9-component IMF is input. The input fuzzy set is subjected to the intensity release of ANFIS and the fuzzy rules are generated, which is expressed as: Among them, x and y are the inputs of node i, and we select and is a bell-shaped function with a maximum value of 1 and a minimum value of 0. is the normalized fuzzy set of each layer; Calculate the output of each rule, expressed as: in, is the proportion of each weight in the total weight, p, q and r are conclusion parameters, is the output of each layer; Sum each output to get the total output: Among them, w i is the weight, and O3 is the forecast result of each IMF component.

7. The distribution transformer heavy overload warning method based on adaptive fuzzy neural algorithm according to claim 6 is characterized in that: The prediction reconstruction includes, Set the 9 predicted IMF components as {x(i), i = 1, 2, ..., N}, reconstruct the m-dimensional vector, expressed as y(i): y(i)={x(i),x(i+1),x(i+2),…,x(i+m-1)} Calculate the Euclidean distance d{y(i),y(j)} between any components of y(i) and y(j), and define the maximum distance between the components as the maximum contribution component distance D{y(i),y(j)}, expressed as: D{y(i),y(j)}=max{|y(i+k)-x(j+k)|} Given a threshold r and an embedding dimension m, calculate the probability measure representing the regularity of the sequence {y(i)} It is expressed as: Setting the embedding dimension to m+1, the resulting sample entropy is: Determine the first threshold, set the samples whose sample entropy values ​​are less than or equal to the first threshold as the first category, determine the second threshold, set the samples whose sample entropy values ​​are greater than the first threshold and less than the second threshold as the second category, and set the samples whose sample entropy values ​​are greater than or equal to the second threshold as the third category. The reconstructed signal is the final predicted load, and the predicted load is compared with the test load to obtain the prediction accuracy.

8. A distribution transformer overload warning system based on an adaptive fuzzy neural algorithm, based on the distribution transformer overload warning method based on an adaptive fuzzy neural algorithm according to any one of claims 1 to 7, characterized in that: include, An analysis module is used to obtain distribution transformer related data and analyze the characteristic factors of the data using the Pearson correlation coefficient; The decomposition module is used to reduce the data dimension according to the principal component analysis method, screen the characteristic factors, perform improved clustering on the screened characteristic factors, and decompose the training samples into IMF components using variational mode decomposition; A prediction module is used to perform adaptive fuzzy neural prediction on the decomposed IMF components to obtain the predicted values ​​of all IMF components, and reconstruct the predicted values ​​using sample entropy to obtain the final predicted load; The early warning module is used to issue a heavy overload early warning according to the final predicted load.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the distribution transformer heavy overload warning method based on the adaptive fuzzy neural algorithm described in any one of claims 1 to 7 are implemented.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the distribution transformer heavy overload warning method based on an adaptive fuzzy neural algorithm as described in any one of claims 1 to 7 are implemented.

Images