Transformer state evaluation method and device, terminal equipment and storage medium

Abstract

The invention discloses a transformer state evaluation method and device, terminal equipment and a storage medium. The method comprises the steps of obtaining equipment data of a to-be-evaluated transformer and transformer fault gas data; constructing a transformer digital twinborn model according to the equipment data, and obtaining a running state gas data ratio of the to-be-evaluated transformer; inputting the running state gas data ratio into a preset transformer state evaluation model for evaluation to obtain a state evaluation result of the to-be-evaluated transformer; wherein the training of the transformer state evaluation model comprises the following steps: calculating a ratio between every two transformer fault gas data according to the transformer fault gas data, and taking the ratio as a to-be-trained sample set; and taking the to-be-trained sample set as input, taking the transformer state type as output, training the to-be-trained transformer state evaluation model until a preset number of iterations is reached, and obtaining a trained transformer state evaluation model. According to the invention, the transformer state evaluation precision is improved.

Description

Technical Field

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

[0002] The working principle of the transformer is complex, and there are many physical fields inside the equipment. Due to the high voltage and large size of the equipment, the maintenance and operation of the transformer lacks timeliness. Digital twin is a technology that connects physical entities and digital models through a platform, and simulates the entire life cycle of physical entities with the help of historical data, real-time data and related algorithms. It has the characteristics of bidirectional, continuous, open and interconnected. Digital twin technology can comprehensively integrate a variety of physical information inside and outside the equipment, establish a comprehensive mapping of the physical world in the digital world, and quickly reflect the current operating status of the physical entity in real time. According to the recorded sensor data and historical data, the digital model is analyzed and corrected in real time to achieve the full life cycle management of the physical entity. The digital twin technology of the transformer is to use the super timeliness and equipment management and maintenance functions of the digital twin to establish a one-to-one digital twin for the transformer, which fully guarantees that the transformer can work normally under high voltage.

[0003] At present, the evaluation of transformer status often uses DGA data as feature parameter mining and uses machine learning models to evaluate the transformer status. However, the use of machine learning models requires model training first. However, transformer status monitoring data usually includes a large amount of complex data such as equipment operation data and environmental meteorological data. Machine learning models are prone to overfitting on training data. Although this can be solved by methods such as regularization and cross-validation, these methods increase the time and complexity of model tuning, resulting in insufficient model generalization ability, and thus the accuracy of the model in transformer status evaluation is not high. Summary of the invention

[0004] The embodiments of the present invention provide a transformer state assessment method, an apparatus, a terminal device and a storage medium, which can effectively solve the problem of low accuracy of transformer state assessment in the prior art.

[0005] An embodiment of the present invention provides a transformer state assessment method, comprising:

[0006] Obtain equipment data of the transformer to be evaluated and transformer fault gas data;

[0007] constructing a transformer digital twin model of the transformer to be evaluated based on the equipment data;

[0008] According to the transformer digital twin model, determining the operating state gas data of the transformer to be evaluated in the transformer digital twin model; and calculating the operating state gas data ratio according to the operating state gas data;

[0009] Inputting the operating status gas data ratio into a preset transformer status assessment model for assessment to obtain a status assessment result of the transformer to be assessed;

[0010] The training of the transformer state assessment model includes:

[0011] According to the transformer fault gas data, a ratio between every two transformer fault gas data is calculated, and the ratio is used as a sample set to be trained;

[0012] Taking the sample set to be trained as input and the transformer state type as output, the transformer state assessment model to be trained is trained until a preset number of iterations is reached to obtain a trained transformer state assessment model.

[0013] Further, according to the transformer fault gas data, the ratio between every two transformer fault gas data is calculated, and the ratio is used as a sample set to be trained, including:

[0014] Determining the transformer fault gas content according to the transformer fault gas data;

[0015] According to the transformer fault gas content, calculate the gas content ratio of every two transformer fault gas data;

[0016] According to the gas content ratio, calculate the ratio mean and ratio standard deviation corresponding to the gas content ratio;

[0017] According to the ratio mean, ratio standard deviation and gas content ratio, the correlation coefficient of the gas content ratio is calculated;

[0018] The gas content ratios corresponding to the correlation coefficients being greater than a preset threshold are eliminated to obtain a final gas content ratio, and the final gas content ratio is used as a sample set to be trained.

[0019] Further, taking the sample set to be trained as input and the transformer state type as output, the transformer state assessment model to be trained is trained until a preset number of iterations is reached to obtain a trained transformer state assessment model, including:

[0020] Inputting the sample set to be trained into the transformer state assessment model to be trained for random optimization to obtain an initial training sample group;

[0021] According to the initial training sample group, obtaining initial sample group parameters of the initial training sample group;

[0022] Performing chaotic mapping on the initial training sample group to obtain the initial sample group position;

[0023] According to the initial sample group position and the sample set to be trained, the iterative optimization operation is repeatedly performed until a preset number of iterations is reached, and the final sample group position is output as the transformer state type to obtain a trained transformer state assessment model;

[0024] The iterative optimization operation includes:

[0025] According to the current sample group position, the current sample group parameters and the sample set to be trained, a fitness objective function is constructed; the current sample group position of the first iterative optimization operation is the initial sample group position, and the current sample group parameters are the initial sample group parameters;

[0026] The current fitness threshold is calculated according to the current number of iterative optimization operations, the preset number of iterations and the preset threshold coefficient;

[0027] Obtaining an updated sample group position according to the fitness objective function, the current fitness threshold and a preset sample group update weight;

[0028] The current sample group position is updated according to the updated sample group position; and the current sample group parameters are updated according to the updated current sample group position.

[0029] Furthermore, according to the current sample group position and the sample set to be trained, a fitness objective function is constructed, including:

[0030] Randomly divide the sample set to be trained into several data sets;

[0031] Calculate the ratio of the number of positions to the number of data sets according to the number of positions corresponding to the current sample group position and the number of data sets corresponding to the plurality of data sets;

[0032] The ratio of the number of positions to the number of data sets is used as the fitness value;

[0033] A fitness objective function is constructed according to the fitness value and the number of data sets.

[0034] Furthermore, according to the fitness objective function, the current fitness threshold and the preset sample group update weight, the updated sample group position is obtained, including:

[0035] Determining a current fitness value according to the fitness objective function;

[0036] Compare the current fitness value and the current fitness threshold to determine the current sample group position update type;

[0037] When the current fitness value is less than the current fitness threshold, the updated sample group position is obtained according to the current sample group position, the preset number of iterations, the preset sample group update weight, the distance between the current sample group position and the sample group position optimized in the previous iteration, and the current decreasing coefficient;

[0038] When the current fitness value is greater than or equal to the current fitness threshold, the updated sample group position is obtained according to the current sample group position, the preset number of iterations, the distance between the current sample group position and the sample group position optimized in the previous iteration, and the preset sample group update weight;

[0039] Among them, during the first iterative optimization operation, the position of the sample group optimized in the previous iterative optimization is the current sample group position; the current decreasing coefficient is calculated according to the preset number of iterations and the current number of iterative optimization operations.

[0040] Furthermore, the transformer digital twin model includes a data analysis layer; the transformer state assessment model is deployed in the data analysis layer.

[0041] As an improvement of the above solution, another embodiment of the present invention provides a transformer state assessment device, including:

[0042] A transformer data acquisition module is used to acquire equipment data of the transformer to be evaluated and transformer fault gas data;

[0043] A digital twin model building module, used to build a transformer digital twin model of the transformer to be evaluated based on the device data;

[0044] A gas ratio determination module is used to determine the operating state gas data of the transformer to be evaluated in the transformer digital twin model according to the transformer digital twin model; and calculate the operating state gas data ratio according to the operating state gas data;

[0045] A transformer state evaluation module, used for inputting the operating state gas data ratio into a preset transformer state evaluation model for evaluation, and obtaining a state evaluation result of the transformer to be evaluated;

[0046] Among them, the transformer status assessment module includes:

[0047] A training sample determination unit, used to calculate the ratio between every two transformer fault gas data according to the transformer fault gas data, and use the ratio as a sample set to be trained;

[0048] The model training unit is used to take the sample set to be trained as input and the transformer state type as output, and train the transformer state assessment model to be trained until a preset number of iterations is reached to obtain a trained transformer state assessment model.

[0049] Furthermore, the training sample determination unit further includes:

[0050] A fault gas content determination subunit, used for determining the transformer fault gas content according to the transformer fault gas data;

[0051] The content ratio calculation subunit is used to calculate the gas content ratio of every two transformer fault gas data according to the transformer fault gas content;

[0052] A mean and standard deviation calculation subunit, used to calculate a ratio mean and a ratio standard deviation corresponding to the gas content ratio according to the gas content ratio;

[0053] A correlation coefficient calculation subunit, used for calculating the correlation coefficient of the gas content ratio according to the ratio mean, the ratio standard deviation and the gas content ratio;

[0054] The sample processing subunit is used to remove the gas content ratios corresponding to the correlation coefficients being greater than a preset threshold value, obtain a final gas content ratio, and use the final gas content ratio as a sample set to be trained.

[0055] Another embodiment of the present invention provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, and when the processor executes the computer program, a transformer state assessment method as described in the above embodiment is implemented.

[0056] Another embodiment of the present invention provides a computer-readable storage medium, which includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute a transformer state assessment method described in the above embodiment.

[0057] By implementing the present invention, at least the following beneficial effects are achieved:

[0058] The present invention provides a transformer state assessment method, device, terminal equipment and storage medium, wherein the method can obtain equipment data of a transformer to be assessed and transformer fault gas data; construct a transformer digital twin model of the transformer to be assessed according to the equipment data; determine the operating state gas data of the transformer to be assessed in the transformer digital twin model according to the transformer digital twin model; and calculate the operating state gas data ratio according to the operating state gas data; input the operating state gas data ratio into a preset transformer state assessment model for assessment to obtain a state assessment result of the transformer to be assessed; wherein the training of the transformer state assessment model comprises: calculating the ratio between every two transformer fault gas data according to the transformer fault gas data, and using the ratio as a sample set to be trained; using the sample set to be trained as input and the transformer state type as output, training the transformer state assessment model to be trained until a preset number of iterations is reached to obtain a trained transformer state assessment model. By constructing a digital twin model of the transformer, the actual operating status of the transformer can be simulated more accurately, so that the operating status gas data obtained is more accurate, and then the status assessment result of the transformer to be evaluated obtained based on the ratio of the operating status gas data is more accurate, thereby improving the accuracy of the transformer status assessment; the ratio of the transformer fault gas data generated by the historical faults of the transformer is used as the training data set of the transformer status assessment model, which simplifies the feature space of the model training data and reduces redundant information, thereby enhancing the generalization ability of the model for different operating statuses of different transformers, reducing the model tuning time and training complexity, making the transformer status assessment model training process more efficient, and thus the model has higher accuracy in transformer status assessment. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] Figure 1 It is a flowchart of a transformer state assessment method provided by an embodiment of the present invention;

[0060] Figure 2 It is a schematic diagram of a transformer digital twin model of a transformer state assessment method provided by an embodiment of the present invention;

[0061] Figure 3 is a relationship diagram between the number of iterations and the evaluation accuracy of a transformer state evaluation method provided by an embodiment of the present invention;

[0062] Figure 4 It is a structural schematic diagram of a transformer state assessment device provided by an embodiment of the present invention;

[0063] Figure 5 It is another structural schematic diagram of a transformer state assessment device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0064] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0065] See also Figure 1 , is a flow chart of a transformer state assessment method provided by an embodiment of the present invention, comprising:

[0066] S1. Obtain equipment data of the transformer to be evaluated and transformer fault gas data;

[0067] S2. constructing a transformer digital twin model of the transformer to be evaluated according to the equipment data;

[0068] S3. According to the transformer digital twin model, determine the operating state gas data of the transformer to be evaluated in the transformer digital twin model; and calculate the operating state gas data ratio according to the operating state gas data;

[0069] S4, inputting the operating status gas data ratio into a preset transformer status assessment model for assessment, and obtaining a status assessment result of the transformer to be assessed;

[0070] The training of the transformer state assessment model includes:

[0071] According to the transformer fault gas data, a ratio between every two transformer fault gas data is calculated, and the ratio is used as a sample set to be trained;

[0072] Taking the sample set to be trained as input and the transformer state type as output, the transformer state assessment model to be trained is trained until a preset number of iterations is reached to obtain a trained transformer state assessment model.

[0073] Specifically, the equipment data of the transformer to be evaluated includes the electrical performance parameters of the transformer during operation, such as voltage, current power and power factor; environmental data, such as temperature and humidity; historical operation and maintenance data, such as maintenance records and parts replacement records, and structural data. Transformer fault gas data refers to the data of gases generated when various faults occur in the transformer, including gas types and contents such as hydrogen (H2), methane (CH4), ethane (C2H6), ethylene (C2H4) and acetylene (C2H2). The operating state gas data of the transformer to be evaluated refers to the gas data generated by the transformer to be evaluated during operation. The status of the transformer can be evaluated based on the operating state gas data. Dissolved Gas Analysis (DGA) is one of the most commonly used fault diagnosis methods for oil-immersed transformers. During the operation of the transformer, the oil-paper composite insulation material will undergo a chemical reaction to release low-molecular hydrogen, hydrocarbons and carbon-containing gas compounds, which are dissolved in the insulating oil. Transformers in different states will cause different gases to be generated.

[0074] Preferably, the transformer fault gas data includes H2 (hydrogen), CH4 (methane), C2H6 (ethane), C2H4 (ethylene) and C2H2 (acetylene), and the selected transformer fault gas data is processed using an uncoded ratio, and the processed ratio is used as a training set and input into the transformer state assessment model to be trained for training; the transformer state assessment model to be trained includes a radial basis function neural network input layer. In the oil dissolved gas analysis method, the uncoded ratio processing of the characteristic gas is to help identify and diagnose the fault type of the transformer by calculating the ratio between different gas concentrations.

[0075] In a preferred embodiment of the present invention, the equipment data and transformer fault gas data of the transformer to be evaluated are first obtained; then the transformer digital twin model of the transformer to be evaluated is constructed according to the equipment data; then the operating state gas data of the transformer to be evaluated in the transformer digital twin model is determined according to the transformer digital twin model; and the operating state gas data ratio is calculated according to the operating state gas data; finally, the operating state gas data ratio is input into the preset transformer state evaluation model for evaluation to obtain the state evaluation result of the transformer to be evaluated; wherein, the training of the transformer state evaluation model includes: according to the transformer fault gas data, the ratio between every two transformer fault gas data is calculated, and the ratio is used as the sample set to be trained; with the sample set to be trained as input and the transformer state type as output, the transformer state evaluation model to be trained is trained until the preset number of iterations is reached to obtain the trained transformer state evaluation model. The state evaluation results of the transformer to be evaluated include normal, medium and low temperature overheating, high temperature overheating, low energy discharge, high energy discharge and partial discharge. Normal: Indicates that the transformer is operating normally and no abnormalities are detected; Medium and low temperature overheating: Indicates that the transformer is overheated to a certain extent, but has not yet reached a serious state; High temperature overheating: Indicates that the transformer is seriously overheated and may need to be dealt with immediately; Low-energy discharge: Indicates that there may be low-energy discharge inside the transformer; High-energy discharge: Indicates that there is high-energy discharge inside the transformer, which is a serious fault signal; Partial discharge: Indicates that there is discharge in certain parts inside the transformer, which may indicate degradation of the insulating material.

[0076] Specifically, according to the transformer fault gas data, the ratio between every two transformer fault gas data is calculated, and the ratio is used as a sample set to be trained, including:

[0077] Determining the transformer fault gas content according to the transformer fault gas data;

[0078] According to the transformer fault gas content, calculate the gas content ratio of every two transformer fault gas data;

[0079] According to the gas content ratio, calculate the ratio mean and ratio standard deviation corresponding to the gas content ratio;

[0080] According to the ratio mean, ratio standard deviation and gas content ratio, the correlation coefficient of the gas content ratio is calculated;

[0081] The gas content ratios corresponding to the correlation coefficients being greater than a preset threshold are eliminated to obtain a final gas content ratio, and the final gas content ratio is used as a sample set to be trained.

[0082] Preferably, the ratio between the gas contents of transformer faults is used as a training sample to construct a transformer fault feature set, and Pearson correlation analysis is used to identify training samples with strong linear relationships in the transformer fault feature set, and the training samples with strong linear relationships are removed from the training sample set, and the remaining training samples are used as inputs of the transformer state assessment model (radial basis function neural network) to be trained. If all gas content ratios are used as inputs, the dimension may be too high, increasing the complexity of the model. Therefore, in the present invention, in the step of constructing a transformer training sample set using the ratio between the transformer fault gases as a training sample, for any two training samples X i and X j , the Pearson correlation coefficient between the two is expressed as:

[0083]

[0084] In the formula, Represents X i The mean of Represents X i The standard deviation of X k represents the ith training sample in the kth training sample set, r represents the Pearson correlation coefficient, the value range of r is [-1, 1], and the value of |r| is greater than 0.75 (preset threshold), indicating a strong linear relationship. The present invention uses Pearson correlation analysis to identify training samples with strong linear relationships in the transformer training sample set, removes the training samples with strong linear relationships from the training sample set, and uses the remaining training samples as inputs of the radial basis function neural network. By reducing the training samples as model inputs, the gas content ratios with linear relationships are removed, thereby avoiding the introduction of redundant information, improving the generalization ability of the model, and reducing the training time and computational cost of the model.

[0085] In a preferred embodiment of the present invention, the transformer state assessment model to be trained, i.e., the radial basis function neural network, includes an input layer, a hidden layer and an output layer, and the expression of the radial basis function of the hidden layer is:

[0086]

[0087] In the formula, x represents the input variable of the input layer, that is, the input training sample (gas content ratio); m represents the number of neurons in the radial basis function layer, represents the unknown coefficient of the radial basis function, G represents the basis function of the radial basis function; the basis function of the radial basis function is a Gaussian function, which is expressed as:

[0088]

[0089] Where C represents the center of the Gaussian basis function, and σ represents the width parameter of the Gaussian basis function.

[0090] Specifically, taking the sample set to be trained as input and the transformer state type as output, the transformer state assessment model to be trained is trained until a preset number of iterations is reached to obtain a trained transformer state assessment model, including:

[0091] Inputting the sample set to be trained into the transformer state assessment model to be trained for random optimization to obtain an initial training sample group;

[0092] According to the initial training sample group, obtaining initial sample group parameters of the initial training sample group;

[0093] Performing chaotic mapping on the initial training sample group to obtain the initial sample group position;

[0094] According to the initial sample group position and the sample set to be trained, the iterative optimization operation is repeatedly performed until a preset number of iterations is reached, and the final sample group position is output as the transformer state type to obtain a trained transformer state assessment model;

[0095] The iterative optimization operation includes:

[0096] According to the current sample group position, the current sample group parameters and the sample set to be trained, a fitness objective function is constructed; the current sample group position of the first iterative optimization operation is the initial sample group position, and the current sample group parameters are the initial sample group parameters;

[0097] The current fitness threshold is calculated according to the current number of iterative optimization operations, the preset number of iterations and the preset threshold coefficient;

[0098] Obtaining an updated sample group position according to the fitness objective function, the current fitness threshold and a preset sample group update weight;

[0099] The current sample group position is updated according to the updated sample group position; and the current sample group parameters are updated according to the updated current sample group position.

[0100] Preferably, the hyperparameters of the transformer state assessment model to be trained are optimized to obtain a trained transformer state assessment model. The sample set to be trained is input into the transformer state assessment model to be trained for random optimization to obtain an initial training sample group. Based on the initial training sample group, the initial sample group parameters of the initial training sample group are obtained. The initial sample group parameters include: population size, penalty coefficient, population position upper limit, population position lower limit, etc. The initial training sample group is subjected to chaotic mapping to obtain the initial sample group position. The initial sample group position is initialized using Tent chaotic mapping, which is expressed as:

[0101]

[0102] In the formula, Z i represents the function value of the i-th mapping, and i represents the number of mappings. Traditionally, when using the whale algorithm to optimize the model, randomly generated data is generally used as the initial whale population position (initial sample group position) of the whale algorithm, so it is difficult to ensure the diversity of the whale population, which easily leads to poor optimization results of the algorithm; based on this, the present invention uses Tent chaotic mapping to initialize the initial sample group position. From the advantages of randomness, regularity and ergodicity of chaotic mapping, the diversity of the whale population (initial training sample group) can be guaranteed. The quality of the initial training sample group affects the solution accuracy and convergence speed of the algorithm. Therefore, the present invention uses these characteristics of chaotic mapping to enable the algorithm to effectively escape from the local optimum, thereby maintaining the diversity of the initial training sample group and improving the global search capability.

[0103] According to the initial sample group position and the sample set to be trained, the iterative optimization operation is repeatedly performed until a preset number of iterations is reached, and the final sample group position is output as the transformer state type to obtain a trained transformer state assessment model; wherein, the iterative optimization operation includes: constructing a fitness objective function according to the current sample group position, the current sample group parameters and the sample set to be trained; in the optimization algorithm, the fitness objective function plays an important role in the final optimization effect. Therefore, in the process of establishing the fitness objective function, in order to prevent the final result from being biased due to problems such as uneven division of the sample set to be trained or overfitting, the present invention uses K-fold cross validation to train and evaluate the radial basis function neural network model. The current sample group position of the first iterative optimization operation is the initial sample group position, and the current sample group parameters are the initial sample group parameters; the current fitness threshold is calculated based on the current iterative optimization operation number, the preset number of iterations, and the preset threshold coefficient; the updated sample group position is obtained based on the fitness objective function, the current fitness threshold, and the preset sample group update weight; the current sample group position is updated based on the updated sample group position; and the current sample group parameters are updated based on the updated current sample group position. The formula for the current sample group position is:

[0104]

[0105] In the formula, p * represents the current fitness threshold, t represents the number of current iterative optimization operations for the current sample group position update, T maxrepresents the maximum number of iterations, k represents the adjustment coefficient, which is used to adjust the size of the preset sample group update weight w; D represents the distance between the current position of the individual in the current sample group and the position of the sample group optimized in the previous iteration, A represents the decreasing coefficient, and X * (t) represents the current position of the individual in the current sample group.

[0106] Preferably, a fitness objective function is constructed according to the current sample group position and the sample set to be trained, including:

[0107] Randomly divide the sample set to be trained into several data sets;

[0108] Calculate the ratio of the number of positions to the number of data sets according to the number of positions corresponding to the current sample group position and the number of data sets corresponding to the plurality of data sets;

[0109] The ratio of the number of positions to the number of data sets is used as the fitness value;

[0110] A fitness objective function is constructed according to the fitness value and the number of data sets.

[0111] In a preferred embodiment of the present invention, the sample set to be trained is divided into K parts to form K data sets; the i-th (i=1, 2, ..., K) part is used as a test set, and the remaining K-1 parts are used as training sets; a radial basis function neural network is created, and the ratio of the number of positions to the number of data sets is calculated according to the number of positions corresponding to the current sample group position and the number of data sets corresponding to a number of data sets, and the ratio of the number of positions to the number of data sets is used as a fitness value, and a fitness objective function is constructed according to the fitness value and the number of data sets. The fitness objective function is constructed:

[0112]

[0113] Where f(i) represents the fitness value obtained each time cross validation is performed, and K represents the number of data sets.

[0114] Schematically, according to the fitness objective function, the current fitness threshold and the preset sample group update weight, the updated sample group position is obtained, including:

[0115] Determining a current fitness value according to the fitness objective function;

[0116] Compare the current fitness value and the current fitness threshold to determine the current sample group position update type;

[0117] When the current fitness value is less than the current fitness threshold, the updated sample group position is obtained according to the current sample group position, the preset number of iterations, the preset sample group update weight, the distance between the current sample group position and the sample group position optimized in the previous iteration, and the current decreasing coefficient;

[0118] When the current fitness value is greater than or equal to the current fitness threshold, the updated sample group position is obtained according to the current sample group position, the preset number of iterations, the distance between the current sample group position and the sample group position optimized in the previous iteration, and the preset sample group update weight;

[0119] Among them, during the first iterative optimization operation, the position of the sample group optimized in the previous iterative optimization is the current sample group position; the current decreasing coefficient is calculated according to the preset number of iterations and the current number of iterative optimization operations.

[0120] Specifically, in the general whale optimization algorithm, the probability threshold is often set to 0.5 to coordinate the synchronization of the spiral and encirclement processes and to randomly select the two motion modes. However, this method may cause the model to converge too slowly, especially as the number of iterations gradually increases. Therefore, in order to improve the convergence speed, the present invention creatively introduces the fitness threshold p * Instead of the original probability threshold, the fitness threshold p of the present invention is * The fitness threshold p in the embodiment of the present invention will change with the change of the number of iterations. * The value range is [0, 1], and the fitness threshold p * The expression is expressed as:

[0121]

[0122] In the formula, t represents the current iteration number, T max Represents the maximum number of iterations; λ and μ are both constants, λ is 0.5, and μ is 0.3.

[0123] The current decreasing coefficient A is controlled by the convergence factor α, and the expression of the convergence factor α is:

[0124]

[0125] Where, T maxRepresents the maximum number of iterations (the preset number of iterative optimization operations), t represents the current number of iterative optimization operations, and the coefficient A is controlled by the convergence factor α. The change of the convergence factor α leads to certain changes in the random search and shrinkage encirclement mechanisms. Specifically, a convergence factor α with a larger change has a better global search capability, which prevents the algorithm from falling into the local optimum; a smaller convergence factor α has a stronger local search capability, which can speed up the convergence of the algorithm. However, the convergence factor α in the whale optimization algorithm decreases linearly from 2 to 0 with the number of iterations, and cannot fully reflect the exploration and development process of the whale algorithm. The nonlinear decreasing convergence factor α involved in the present invention is a nonlinear decreasing convergence factor α that changes rapidly in the early stage and relatively slowly in the later stage, which is used to balance the exploration and development process of the whale algorithm.

[0126] In another preferred embodiment of the present invention, Gaussian mutation is performed on the current sample group position after the position update, and the positions before and after the mutation are compared and judged, and the optimal item is selected as the current optimal position. The main purpose of Gaussian mutation is to improve the ability of the algorithm to jump out of the local optimum and enhance the algorithm's optimization ability. The expression formula of the present invention using Gaussian mutation for the current sample group position is as follows:

[0127] X(N)=X(t)+X(t)·N(0,1)

[0128]

[0129] Where N(0, 1) generates a Gaussian distributed random number between 0 and 1, and X(N) represents the position vector generated after Gaussian mutation.

[0130] Preferably, the transformer digital twin model includes a data analysis layer; the transformer state assessment model is deployed in the data analysis layer. The transformer digital twin model also includes a device layer, a perception layer and a service layer. Figure 2As shown, the transformer equipment model is assembled and integrated from a multi-level perspective, and a virtual-real mapping relationship between the equipment entity and the model is established. The equipment entity and the model together constitute the transformer digital twin model. The digital twin model includes 5 layers, namely, the equipment layer, the perception layer, the data transmission layer, the data analysis layer and the service layer from bottom to top. The perception layer in the transformer digital twin model of this embodiment adopts a sensor group, which is used to detect various data in the corresponding transformer equipment; for example, the sensor group detects the characteristic gas dissolved in the transformer oil in the transformer equipment. Specifically, the present invention evaluates the operating status of the transformer by detecting the gas content dissolved in the transformer oil. The equipment layer is the transformer equipment of the substation, which is the main object and basis for the evaluation of the equipment operating status, and provides the perception layer with information such as the gas content dissolved in the transformer oil. The perception layer includes a series of sensors. The data transmission layer is located between the perception layer and the data analysis layer, and plays the role of a connecting link to realize the input, spatial storage and output of data. This layer uses digital twin technology to grasp the operating status of the transformer equipment in real time. The data analysis layer uses the transformer status assessment model to achieve in-depth analysis and mining of transformer equipment data, and then evaluates the operating status and health of the transformer equipment. This layer combines digital twin technology with the transformer status assessment model built based on the radial basis function neural network to ultimately achieve quantitative analysis of transformer equipment operating status data. The service layer visualizes the operating status of the transformer equipment through the human-computer interaction system and implements reasonable management and control of the transformer equipment.

[0131] By implementing this embodiment, it is possible to obtain equipment data and transformer fault gas data of the transformer to be evaluated; construct a transformer digital twin model of the transformer to be evaluated based on the equipment data; determine the operating status gas data of the transformer to be evaluated in the transformer digital twin model based on the transformer digital twin model; and calculate the operating status gas data ratio based on the operating status gas data; input the operating status gas data ratio into a preset transformer state evaluation model for evaluation to obtain a state evaluation result of the transformer to be evaluated; wherein, the training of the transformer state evaluation model includes: calculating the ratio between every two transformer fault gas data based on the transformer fault gas data, and using the ratio as a sample set to be trained; using the sample set to be trained as input and the transformer state type as output, training the transformer state evaluation model to be trained until a preset number of iterations is reached to obtain a trained transformer state evaluation model. By constructing a digital twin model of the transformer, the actual operating status of the transformer can be simulated more accurately, so that the operating status gas data obtained is more accurate, and then the status assessment result of the transformer to be evaluated obtained based on the ratio of the operating status gas data is more accurate, thereby improving the accuracy of the transformer status assessment; the ratio of the transformer fault gas data generated by the historical faults of the transformer is used as the training data set of the transformer status assessment model, which simplifies the feature space of the model training data and reduces redundant information, thereby enhancing the generalization ability of the model for different operating statuses of different transformers, reducing the model tuning time and training complexity, making the transformer status assessment model training process more efficient, and thus the model has higher accuracy in transformer status assessment.

[0132] This embodiment uses Pearson correlation analysis to identify training samples with strong linear relationships in the transformer training sample set, removes the training samples with strong linear relationships from the training sample set, and uses the remaining training samples as inputs of the radial basis function neural network. By reducing the training samples as model inputs and removing data with linear relationships, the introduction of redundant information is avoided, the generalization ability of the model is improved, and the training time and computational cost of the model are reduced.

[0133] This embodiment optimizes the hyperparameters of the radial basis function neural network based on the parameter-adaptive whale optimization algorithm, which solves the problems of long training time, low evaluation accuracy and low learning efficiency of the traditional machine learning model; the whale optimization algorithm of this embodiment uses Tent chaotic mapping to initialize the position of the sample group, so that the population distribution is more uniform and the convergence speed of the algorithm is accelerated; the whale optimization algorithm of this embodiment introduces the sample group update weight, so that the whale continuously updates the position in the process of surrounding the prey in a spiral motion, that is, continuously updates the position of the sample group. As the number of iterations increases, the weight will decrease linearly to avoid falling into the local optimum and achieve an improvement in evaluation accuracy. The relationship between the number of iterations and the evaluation accuracy of this embodiment is as follows: Figure 3 As shown; at the end of each iteration of the whale optimization algorithm of this embodiment, the Gaussian detection mechanism is used to update and judge the fitness of the whale position, so that the whale algorithm has a better exploration position and accelerates the convergence speed of the model.

[0134] See also Figure 4 and Figure 5 , is a schematic diagram of a transformer state assessment device provided by an embodiment of the present invention, comprising:

[0135] A transformer data acquisition module is used to acquire equipment data of the transformer to be evaluated and transformer fault gas data;

[0136] A digital twin model building module, used to build a transformer digital twin model of the transformer to be evaluated based on the device data;

[0137] A gas ratio determination module is used to determine the operating state gas data of the transformer to be evaluated in the transformer digital twin model according to the transformer digital twin model; and calculate the operating state gas data ratio according to the operating state gas data;

[0138] A transformer state evaluation module, used for inputting the operating state gas data ratio into a preset transformer state evaluation model for evaluation, and obtaining a state evaluation result of the transformer to be evaluated;

[0139] Among them, the transformer status assessment module includes:

[0140] A training sample determination unit, used to calculate the ratio between every two transformer fault gas data according to the transformer fault gas data, and use the ratio as a sample set to be trained;

[0141] The model training unit is used to take the sample set to be trained as input and the transformer state type as output, and train the transformer state assessment model to be trained until a preset number of iterations is reached to obtain a trained transformer state assessment model.

[0142] Specifically, the training sample determination unit further includes:

[0143] A fault gas content determination subunit, used for determining the transformer fault gas content according to the transformer fault gas data;

[0144] The content ratio calculation subunit is used to calculate the gas content ratio of every two transformer fault gas data according to the transformer fault gas content;

[0145] A mean and standard deviation calculation subunit, used to calculate a ratio mean and a ratio standard deviation corresponding to the gas content ratio according to the gas content ratio;

[0146] A correlation coefficient calculation subunit, used for calculating the correlation coefficient of the gas content ratio according to the ratio mean, the ratio standard deviation and the gas content ratio;

[0147] The sample processing subunit is used to remove the gas content ratios corresponding to the correlation coefficients being greater than a preset threshold value, obtain a final gas content ratio, and use the final gas content ratio as a sample set to be trained. The present invention provides a transformer state evaluation device, which acquires equipment data and transformer fault gas data of a transformer to be evaluated according to a transformer data acquisition module; in a digital twin model construction module, constructs a transformer digital twin model of the transformer to be evaluated according to the equipment data; in a gas ratio determination module, determines the operating state gas data of the transformer to be evaluated in the transformer digital twin model according to the transformer digital twin model; and calculates the operating state gas data ratio according to the operating state gas data; through the transformer state evaluation module, inputs the operating state gas data ratio into a preset transformer state evaluation model for evaluation, and obtains a state evaluation result of the transformer to be evaluated; wherein the transformer state evaluation module comprises: a training sample determination unit, which is used to calculate the ratio between every two transformer fault gas data according to the transformer fault gas data, and uses the ratio as a sample set to be trained; a model training unit, which is used to train the transformer state evaluation model to be trained with the sample set to be trained as input and the transformer state type as output, until a preset number of iterations is reached, and a trained transformer state evaluation model is obtained. By constructing a digital twin model of the transformer, the actual operating status of the transformer can be simulated more accurately, so that the operating status gas data obtained is more accurate, and then the status assessment result of the transformer to be evaluated obtained based on the ratio of the operating status gas data is more accurate, thereby improving the accuracy of the transformer status assessment; the ratio of the transformer fault gas data generated by the historical faults of the transformer is used as the training data set of the transformer status assessment model, which simplifies the feature space of the model training data and reduces redundant information, thereby enhancing the generalization ability of the model for different operating statuses of different transformers, reducing the model tuning time and training complexity, making the transformer status assessment model training process more efficient, and thus the model has higher accuracy in transformer status assessment.

[0148] It should be noted that the device embodiments described above are merely schematic, wherein the units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they may be located in one place, or they may be distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the scheme of this embodiment. In addition, in the accompanying drawings of the device embodiments provided by the present invention, the connection relationship between the modules indicates that there is a communication connection between them, which may be specifically implemented as one or more communication buses or signal lines. A person of ordinary skill in the art may understand and implement it without paying any creative effort.

[0149] Those skilled in the art can clearly understand that, for the sake of convenience and brevity, the specific working process of the device described above can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0150] Another embodiment of the present invention further provides a terminal device, including a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor implements a transformer state assessment method as described in the above embodiment when executing the computer program. The terminal device may be a computing device such as a desktop computer, a notebook, a palm computer, and a cloud server. The terminal device may include, but is not limited to, a processor and a memory.

[0151] The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or any conventional processor, etc. The processor is the control center of the terminal device, and uses various interfaces and lines to connect various parts of the entire terminal device.

[0152] The memory can be used to store the computer program, and the processor realizes various functions of the terminal device by running or executing the computer program stored in the memory and calling the data stored in the memory. The memory can mainly include a program storage area and a data storage area, wherein the program storage area can store an operating system, at least one application required for a function, etc.; the data storage area can store data created according to the use of the mobile phone, etc. In addition, the memory can include a high-speed random access memory, and can also include a non-volatile memory, such as a hard disk, a memory, a plug-in hard disk, a smart memory card (Smart Media Card, SMC), a secure digital (Secure Digital, SD) card, a flash card (Flash Card), at least one disk storage device, a flash memory device or other volatile solid-state storage device.

[0153] Another embodiment of the present invention provides a computer-readable storage medium, which includes a stored computer program, wherein when the computer program is running, the device where the computer-readable storage medium is located is controlled to execute a transformer state assessment method described in the above embodiment.

[0154] The storage medium is a computer-readable storage medium, and the computer program is stored in the computer-readable storage medium. When the computer program is executed by the processor, the steps of each of the above-mentioned method embodiments can be implemented. Among them, the computer program includes computer program code, and the computer program code can be in source code form, object code form, executable file or some intermediate form. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium.

[0155] The above is a preferred embodiment of the present invention. It should be pointed out that a person skilled in the art can make several improvements and modifications without departing from the principle of the present invention. These improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A transformer status assessment method, characterized in that: include: Obtain equipment data of the transformer to be evaluated and transformer fault gas data; constructing a transformer digital twin model of the transformer to be evaluated based on the equipment data; According to the transformer digital twin model, determining the operating state gas data of the transformer to be evaluated in the transformer digital twin model; and calculating the operating state gas data ratio according to the operating state gas data; Inputting the operating status gas data ratio into a preset transformer status assessment model for assessment to obtain a status assessment result of the transformer to be assessed; The training of the transformer state assessment model includes: According to the transformer fault gas data, a ratio between every two transformer fault gas data is calculated, and the ratio is used as a sample set to be trained; Taking the sample set to be trained as input and the transformer state type as output, the transformer state assessment model to be trained is trained until a preset number of iterations is reached to obtain a trained transformer state assessment model.

2. A transformer status assessment method according to claim 1, characterized in that: According to the transformer fault gas data, the ratio between every two transformer fault gas data is calculated, and the ratio is used as a sample set to be trained, including: Determining the transformer fault gas content according to the transformer fault gas data; According to the transformer fault gas content, calculate the gas content ratio of every two transformer fault gas data; According to the gas content ratio, calculate the ratio mean and ratio standard deviation corresponding to the gas content ratio; According to the ratio mean, ratio standard deviation and gas content ratio, the correlation coefficient of the gas content ratio is calculated; The gas content ratios corresponding to the correlation coefficients being greater than a preset threshold are eliminated to obtain a final gas content ratio, and the final gas content ratio is used as a sample set to be trained.

3. A transformer status assessment method according to claim 1, characterized in that: Taking the sample set to be trained as input and the transformer state type as output, training is performed in the transformer state assessment model to be trained until a preset number of iterations is reached to obtain a trained transformer state assessment model, including: Inputting the sample set to be trained into the transformer state assessment model to be trained for random optimization to obtain an initial training sample group; According to the initial training sample group, obtaining initial sample group parameters of the initial training sample group; Performing chaotic mapping on the initial training sample group to obtain the initial sample group position; According to the initial sample group position and the sample set to be trained, the iterative optimization operation is repeatedly performed until a preset number of iterations is reached, and the final sample group position is output as the transformer state type to obtain a trained transformer state assessment model; The iterative optimization operation includes: According to the current sample group position, the current sample group parameters and the sample set to be trained, a fitness objective function is constructed; the current sample group position of the first iterative optimization operation is the initial sample group position, and the current sample group parameters are the initial sample group parameters; The current fitness threshold is calculated according to the current number of iterative optimization operations, the preset number of iterations and the preset threshold coefficient; Obtaining an updated sample group position according to the fitness objective function, the current fitness threshold and a preset sample group update weight; The current sample group position is updated according to the updated sample group position; and the current sample group parameters are updated according to the updated current sample group position.

4. A transformer status assessment method according to claim 3, characterized in that: According to the current sample group position and the sample set to be trained, a fitness objective function is constructed, including: Randomly divide the sample set to be trained into several data sets; Calculate the ratio of the number of positions to the number of data sets according to the number of positions corresponding to the current sample group position and the number of data sets corresponding to the plurality of data sets; The ratio of the number of positions to the number of data sets is used as the fitness value; A fitness objective function is constructed according to the fitness value and the number of data sets.

5. A transformer status assessment method according to claim 4, characterized in that: According to the fitness objective function, the current fitness threshold and the preset sample group update weight, the updated sample group position is obtained, including: Determining a current fitness value according to the fitness objective function; Compare the current fitness value and the current fitness threshold to determine the current sample group position update type; When the current fitness value is less than the current fitness threshold, the updated sample group position is obtained according to the current sample group position, the preset number of iterations, the preset sample group update weight, the distance between the current sample group position and the sample group position optimized in the previous iteration, and the current decreasing coefficient; When the current fitness value is greater than or equal to the current fitness threshold, the updated sample group position is obtained according to the current sample group position, the preset number of iterations, the distance between the current sample group position and the sample group position optimized in the previous iteration, and the preset sample group update weight; Among them, during the first iterative optimization operation, the position of the sample group optimized in the previous iterative optimization is the current sample group position; the current decreasing coefficient is calculated according to the preset number of iterations and the current number of iterative optimization operations.

6. A transformer status assessment method according to claim 1, characterized in that: The transformer digital twin model includes a data analysis layer; the transformer state assessment model is deployed in the data analysis layer.

7. A transformer status assessment device, characterized in that: include: A transformer data acquisition module is used to acquire equipment data of the transformer to be evaluated and transformer fault gas data; A digital twin model building module, used to build a transformer digital twin model of the transformer to be evaluated based on the device data; A gas ratio determination module is used to determine the operating state gas data of the transformer to be evaluated in the transformer digital twin model according to the transformer digital twin model; and calculate the operating state gas data ratio according to the operating state gas data; A transformer state evaluation module, used for inputting the operating state gas data ratio into a preset transformer state evaluation model for evaluation, and obtaining a state evaluation result of the transformer to be evaluated; Among them, the transformer status assessment module includes: A training sample determination unit, used to calculate the ratio between every two transformer fault gas data according to the transformer fault gas data, and use the ratio as a sample set to be trained; The model training unit is used to take the sample set to be trained as input and the transformer state type as output, and train the transformer state assessment model to be trained until a preset number of iterations is reached to obtain a trained transformer state assessment model.

8. A transformer status assessment device according to claim 7, characterized in that: The training sample determination unit further includes: A fault gas content determination subunit, used for determining the transformer fault gas content according to the transformer fault gas data; The content ratio calculation subunit is used to calculate the gas content ratio of every two transformer fault gas data according to the transformer fault gas content; A mean and standard deviation calculation subunit, used to calculate a ratio mean and a ratio standard deviation corresponding to the gas content ratio according to the gas content ratio; A correlation coefficient calculation subunit, used for calculating the correlation coefficient of the gas content ratio according to the ratio mean, the ratio standard deviation and the gas content ratio; The sample processing subunit is used to remove the gas content ratios corresponding to the correlation coefficients being greater than a preset threshold value, obtain a final gas content ratio, and use the final gas content ratio as a sample set to be trained.

9. A terminal device, characterized in that: The method comprises a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein when the processor executes the computer program, a transformer state assessment method according to any one of claims 1 to 6 is implemented.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a stored computer program, wherein when the computer program is executed, the device where the computer-readable storage medium is located is controlled to execute a transformer state assessment method according to any one of claims 1 to 6.

Images