A capacitor fault identification method, system and device based on improved ANFIS

Abstract

A capacitor fault identification method based on improved ANFIS, the method first constructs a capacitor fault identification model; takes dielectric loss angle tangent and capacitance value as input, takes capacitor insulation aging degree as output, trains capacitor fault identification model, obtains capacitor insulation aging identification model, used to identify insulation aging degree of capacitor to be detected; takes active current component and leakage current as input, takes capacitor internal leakage degree as output, trains capacitor fault identification model, obtains capacitor internal leakage identification model, used to identify leakage degree of capacitor to be detected. This design reasonably selects dielectric loss angle and capacitance value as characteristic state quantities for judging insulation aging state, and reasonably selects active current component and leakage current as characteristic state quantities for judging capacitor leakage state, thus realizing accurate monitoring of capacitor insulation aging degree and leakage degree.

Description

Technical Field

[0001] The present invention belongs to the technical field of capacitor state monitoring, and in particular relates to a capacitor fault identification method, system and equipment based on an improved ANFIS. Background Art

[0002] At present, DC link capacitors (DC-Link) are widely used as power conversion equipment in wind power, photovoltaic, aerospace and electric vehicles, but component failures are frequent and the failure rate is high, which seriously affects the operating performance of DC-Link. In order to ensure the reliability of capacitor operation, it is necessary to closely monitor the status of capacitor banks and issue early warnings to make maintenance and diagnosis more convenient.

[0003] Common capacitor faults include insulation aging and leakage faults. The invention patent with application number 202010925125.0 provides an online identification system and method for the degree of aging of electrical lines based on ANFIS. The method establishes an adaptive neural network fuzzy inference system ANFIS, selects residual current, load voltage, ambient temperature and ambient humidity as characteristic quantities of line aging, and performs real-time online monitoring of the degree of aging of the line; although the system can monitor the degree of aging in real time online, it cannot realize online monitoring of leakage faults. The present invention studies insulation aging and leakage faults, two common capacitor faults; and because the electrical properties of electrical lines and capacitors are different, how to reasonably select the fault characteristic quantities of capacitors and improve the accuracy of capacitor fault identification is also the research focus of the present invention. Summary of the invention

[0004] The purpose of the present invention is to provide a capacitor fault identification method, system and device based on an improved ANFIS, which can accurately identify the insulation aging and internal leakage degree of the capacitor, in view of the above-mentioned problems existing in the prior art.

[0005] To achieve the above objectives, the technical solution of the present invention is as follows:

[0006] A capacitor fault identification method based on improved ANFIS, the capacitor fault identification method comprising the following steps:

[0007] S1. Construct capacitor fault identification model based on improved ANFIS network;

[0008] S2. Using dielectric loss tangent and capacitance as input and capacitor insulation aging degree as output, the capacitor fault identification model is trained to obtain a capacitor insulation aging identification model; using active current component and leakage current as input and capacitor internal leakage degree as output, the capacitor fault identification model is trained to obtain a capacitor internal leakage identification model;

[0009] S3. Based on the capacitor insulation aging identification model and the capacitor internal leakage identification model, the insulation aging degree and the leakage degree of the capacitor to be tested are respectively identified.

[0010] In step S1, the capacitor fault identification model constructed has five layers, including an input layer, a fuzzification layer, a batch normalization layer, a defuzzification layer, and an output layer; wherein,

[0011] The input layer is used to obtain the membership value corresponding to the input data according to the following formula:

[0012]

[0013]

[0014] In the above formula, represents the output of node i in the first layer; f i (x) represents the membership value corresponding to the input data x output by the first layer node i; f i (y) represents the membership value corresponding to the input data y output by the first layer node i; f represents the membership function of the fuzzy set;

[0015] The fuzzification layer is used to fuzzify the input data according to the following formula:

[0016] ω i =f i (x)·f i (y);

[0017] In the above formula, ω i represents the output of the second layer node i; represents the fuzzy operator;

[0018] The batch normalization layer is used to introduce batch normalization to normalize the input data to fix the mean and variance of the input data:

[0019]

[0020]

[0021]

[0022]

[0023] In the above formula, is the output of node i in the third layer; Represents the result after batch normalization of the input data; represents the variance of the input signal; β and γ represent the translation factor and scale factor, respectively, both of which are learnable parameters; ω1 and ω2 represent the outputs of node 1 and node 2 in the second layer, respectively; μ βRepresents the mean of the input signal; e represents a decimal close to 0;

[0024] The defuzzification layer is used to obtain the actual output value by weighted averaging according to the following formula:

[0025]

[0026] In the above formula, represents the actual output value of node i in the fourth layer; p i ,q i 、r i They are all parameter sets of node i automatically generated by the ANFIS network;

[0027] The output layer is used to calculate the total output according to the following formula:

[0028]

[0029] In the above formula, O 5 Represents the output of the fifth layer, which is the total output of the model.

[0030] The step S2 also includes optimizing the learnable parameters of the capacitor insulation aging identification model and the capacitor internal leakage identification model using the BP neural network, specifically: calculating the loss function of the BP neural network based on the model prediction value and the actual observation value, updating the network parameters of the BP neural network by back propagation, and iterating the calculation until the BP neural network converges; the loss function is:

[0031]

[0032] In the above formula, Q i Represents the true label value obtained from the actual dataset; Represents the output value of node i in the third layer of the capacitor fault identification model.

[0033] The calculation formula of the dielectric loss tangent is:

[0034]

[0035]

[0036] In the above formula, tanδ is the dielectric loss tangent, δ is the dielectric loss angle; f is the power frequency; ε is the dielectric constant; ρ is the volume resistivity of the insulating medium; S is the plate area; d is the plate spacing; k is the electrostatic force constant;

[0037] The calculation formula of the leakage current is:

[0038]

[0039] In the above formula, Ileak is the leakage current; C is the capacitance of the capacitor; V is the voltage of the capacitor, R is the resistance of the capacitor; t is the time;

[0040] The calculation formula of the active current component is:

[0041] I R =I·sinδ;

[0042] In the above formula, I R Active current component; I is the measured current.

[0043] A capacitor fault identification system based on improved ANFIS, the capacitor fault identification system includes a model building module, a model training module, and a fault identification module:

[0044] The model building module is used to build a capacitor fault identification model based on the improved ANFIS network;

[0045] The model training module is used to train the capacitor fault identification model with the dielectric loss tangent and the capacitance value as input and the capacitor insulation aging degree as output to obtain the capacitor insulation aging identification model; and to train the capacitor fault identification model with the active current component and the leakage current as input and the capacitor internal leakage degree as output to obtain the capacitor internal leakage identification model;

[0046] The fault identification module is used to identify the insulation aging degree and the leakage degree of the capacitor to be detected based on the capacitor insulation aging identification model and the capacitor internal leakage identification model.

[0047] The capacitor fault identification model constructed by the model building module has five layers, including an input layer, a fuzzification layer, a batch normalization layer, a defuzzification layer, and an output layer; wherein,

[0048] The input layer is used to obtain the membership value corresponding to the input data according to the following formula:

[0049]

[0050]

[0051] In the above formula, represents the output of node i in the first layer; f i (x) represents the membership value corresponding to the input data x output by the first layer node i; f i (y) represents the membership value corresponding to the input data y output by the first layer node i; f represents the membership function of the fuzzy set;

[0052] The fuzzification layer is used to fuzzify the input data according to the following formula:

[0053] ω i =f i (x)·f i (y);

[0054] In the above formula, ω i represents the output of the second layer node i; represents the fuzzy operator;

[0055] The batch normalization layer is used to introduce batch normalization to normalize the input data to fix the mean and variance of the input data:

[0056]

[0057]

[0058]

[0059]

[0060] In the above formula, is the output of node i in the third layer; Represents the result after batch normalization of the input data; represents the variance of the input signal; β and γ represent the translation factor and scale factor, respectively, both of which are learnable parameters; ω1 and ω2 represent the outputs of node 1 and node 2 in the second layer, respectively; μ β Represents the mean of the input signal; e represents a decimal close to 0;

[0061] The defuzzification layer is used to obtain the actual output value by weighted averaging according to the following formula:

[0062]

[0063] In the above formula, represents the actual output value of node i in the fourth layer; p i ,q i 、r i They are all parameter sets of node i automatically generated by the ANFIS network;

[0064] The output layer is used to calculate the total output according to the following formula:

[0065]

[0066] In the above formula, O 5 Represents the output of the fifth layer, which is the total output of the model.

[0067] The model training module is also used to optimize the learnable parameters of the capacitor insulation aging identification model and the capacitor internal leakage identification model using the BP neural network, specifically: the loss function of the BP neural network is calculated based on the model prediction value and the actual observation value, the network parameters of the BP neural network are updated by back propagation, and the calculation is iterated until the BP neural network converges; the loss function is:

[0068]

[0069] In the above formula, Q i Represents the true label value obtained from the actual dataset; Represents the output value of node i in the third layer of the capacitor fault identification model.

[0070] The calculation formula of the dielectric loss tangent is:

[0071]

[0072]

[0073] In the above formula, tanδ is the dielectric loss tangent, δ is the dielectric loss angle; f is the power frequency; ε is the dielectric constant; ρ is the volume resistivity of the insulating medium; S is the plate area; d is the plate spacing; k is the electrostatic force constant;

[0074] The calculation formula of the leakage current is:

[0075]

[0076] In the above formula, I leak is the leakage current; C is the capacitance of the capacitor; V is the voltage of the capacitor, R is the resistance of the capacitor; t is the time;

[0077] The calculation formula of the active current component is:

[0078] I R =I·sinδ;

[0079] In the above formula, I R Active current component; I is the measured current.

[0080] A capacitor fault identification device based on an improved ANFIS, the capacitor fault identification device comprising a memory and a processor;

[0081] The memory is used to store computer program code and transmit the computer program code to the processor;

[0082] The processor is used to execute the aforementioned method according to the instructions in the computer program code.

[0083] A computer-readable storage medium stores a computer program, and the computer program implements the aforementioned method when executed by a processor.

[0084] Compared with the prior art, the present invention has the following beneficial effects:

[0085] 1. The present invention is a capacitor fault identification method based on improved ANFIS. First, a capacitor fault identification model is constructed. Secondly, the dielectric loss tangent and capacitance value are reasonably selected as characteristic state quantities for judging the insulation aging state, thereby realizing accurate monitoring of the insulation aging degree of the capacitor; the active current component and leakage current are reasonably selected as characteristic state quantities for judging the leakage state of the capacitor, thereby realizing accurate monitoring of the leakage degree of the capacitor. Therefore, the present invention realizes accurate monitoring of the insulation aging state and leakage state of the capacitor.

[0086] 2. The present invention is a capacitor fault identification method based on improved ANFIS, which combines batch normalization with the ANFIS adaptive neural fuzzy inference system, processes input data through batch normalization, thereby avoiding the problem of internal covariate offset of ANFIS, and can also improve the stability and training efficiency of the model, and more accurately classify the insulation aging degree and leakage degree of the capacitor. Therefore, the present invention can improve the stability, training efficiency and recognition accuracy of the model.

[0087] 3. The present invention is a capacitor fault identification method based on improved ANFIS. On the basis of batch normalization, the introduction of scale factor γ helps to restore the original scale difference of the data, making the data closer to the true value, and the introduction of translation factor β helps to restore the distribution deviation of the data, so as to more accurately capture and analyze the aging of the fault, and further improve the recognition accuracy; the BP neural network optimizes the scale factor γ and the translation factor β, which can further improve the network training efficiency. Therefore, the present invention can further improve the recognition accuracy and network training efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0088] Figure 1 The present invention is a flowchart of the method.

[0089] Figure 2 The flowchart is for fault identification using the capacitor insulation aging identification model and the capacitor internal leakage identification model obtained in Example 1.

[0090] Figure 3 This is the recognition result of the capacitor insulation aging recognition model in Example 1 identifying the insulation aging degree of the capacitor to be detected.

[0091] Figure 4 This is the result of identifying the leakage degree of the capacitor to be detected by the capacitor internal leakage identification model in Example 1.

[0092] Figure 5 It is a schematic diagram of the structure of the system of the present invention.

[0093] Figure 6 It is a structural schematic diagram of the device described in the present invention. DETAILED DESCRIPTION

[0094] The present invention is further described in detail below in conjunction with specific implementations and drawings.

[0095] Embodiment 1:

[0096] See also Figure 1 , a capacitor fault identification method based on improved ANFIS is specifically carried out in the following steps:

[0097] 1. Construct a capacitor fault identification model based on the improved ANFIS1 network; the capacitor fault identification model has five layers, including input layer, fuzzification layer, batch normalization layer, defuzzification layer, and output layer; among them,

[0098] The input layer is used to obtain the membership value corresponding to the input data according to the following formula:

[0099]

[0100]

[0101] In the above formula, represents the output of node i in the first layer; f i (x) represents the membership value corresponding to the input data x output by the first layer node i; f i (y) represents the membership value corresponding to the input data y output by the first layer node i; f represents the membership function of the fuzzy set;

[0102] The fuzzification layer is used to fuzzify the input data according to the following formula:

[0103] ω i =f i (x)·f i (y);

[0104] In the above formula, ω i represents the output of the second layer node i; represents the fuzzy operator;

[0105] The batch normalization layer is used to introduce batch normalization to normalize the input data to fix the mean and variance of the input data:

[0106]

[0107]

[0108]

[0109]

[0110] In the above formula, is the output of node i in the third layer; Represents the result after batch normalization of the input data; represents the variance of the input signal; β and γ represent the translation factor and scale factor, respectively, both of which are learnable parameters; ω1 and ω2 represent the outputs of node 1 and node 2 in the second layer, respectively; μ β Represents the mean of the input signal; e represents a decimal close to 0;

[0111] The defuzzification layer is used to obtain the actual output value by weighted averaging according to the following formula:

[0112]

[0113] In the above formula, represents the actual output value of node i in the fourth layer; p i ,q i 、r i They are all parameter sets of node i automatically generated by the ANFIS network;

[0114] The output layer is used to calculate the total output according to the following formula:

[0115]

[0116] In the above formula, O 5 Represents the output of the fifth layer, which is the total output of the model;

[0117] 2. (1) When the capacitor insulation ages, the capacitance of the capacitor will decrease. In addition, since the insulation material performance deteriorates or is damaged after insulation aging, the conductive path of the capacitor will increase, resulting in increased dielectric loss. The dielectric loss tangent reflects this loss in the capacitor circuit. Therefore, the dielectric loss tangent and the capacitance are selected as characteristic state quantities as the input of the model, and the degree of capacitor insulation aging is used as the output. 0, 1, 2, and 3 correspond to severe aging, moderate aging, mild aging, and normal state, respectively. After training the capacitor fault identification model, the capacitor insulation aging identification model is obtained. The calculation formula of the dielectric loss tangent is:

[0118]

[0119]

[0120] In the above formula, tanδ is the dielectric loss tangent, δ is the dielectric loss angle; f is the power frequency; ε is the dielectric constant; ρ is the volume resistivity of the insulating medium; S is the plate area; d is the plate spacing; k is the electrostatic force constant;

[0121] (2) When leakage occurs inside the capacitor, the insulation performance inside the capacitor decreases, causing abnormal current to pass through the capacitor, increasing the DC component of the current and the leakage current. Therefore, the active current component and the leakage current are selected as the characteristic state quantities as the input of the model, and the internal leakage degree of the capacitor is used as the output. 0, 1, 2, and 3 correspond to severe leakage, moderate leakage, mild leakage, and normal state, respectively. After training the capacitor fault identification model, the internal leakage identification model of the capacitor is obtained.

[0122] The calculation formula of the leakage current is:

[0123]

[0124] In the above formula, I leak is the leakage current; C is the capacitance of the capacitor; V is the voltage of the capacitor, R is the resistance of the capacitor; t is the time;

[0125] The calculation formula of the active current component is:

[0126] I R =I·sinδ;

[0127] In the above formula, I R Active current component; I is the measured current;

[0128] (3) Using the BP neural network to optimize the learnable parameters of the capacitor insulation aging identification model and the capacitor internal leakage identification model, specifically: calculating the loss function of the BP neural network based on the model prediction value and the actual observation value, updating the network parameters of the BP neural network through back propagation, and iterating the calculation until the BP neural network converges; the loss function is:

[0129] The loss function is:

[0130]

[0131] In the above formula, Q i Represents the true label value obtained from the actual dataset; represents the actual output value of node i in the third layer of the capacitor fault identification model;

[0132] The calculation process of the back propagation is:

[0133]

[0134]

[0135]

[0136]

[0137]

[0138]

[0139]

[0140]

[0141] In the above formula, γ′ and β′ represent the updated γ and β respectively;

[0142] 3. If Figure 2 As shown, the monitored capacitor dielectric loss tangent and capacitance value are input into the capacitor insulation aging identification model to identify the insulation aging degree of the capacitor to be detected; the monitored capacitor active current component and leakage current are input into the capacitor internal leakage identification model to identify the leakage degree of the capacitor to be detected, and the identification result is output; according to the identification result, the capacitor is warned, and the data is stored in an external data storage device for re-inspection.

[0143] The identification results of insulation aging degree and leakage degree of 8 capacitors are as follows: Figure 3 , 4 As shown, it can be seen that the predicted value is close to the true value, which indicates that the capacitor fault identification method of the present invention has a high recognition accuracy.

[0144] Embodiment 2:

[0145] See also Figure 5 A capacitor fault identification system based on improved ANFIS includes a model building module, a model training module, and a fault identification module. The model building module is used to build a capacitor fault identification model. The capacitor fault identification model has five layers, including an input layer, a fuzzification layer, a batch normalization layer, a defuzzification layer, and an output layer; wherein,

[0146] The input layer is used to obtain the membership value corresponding to the input data according to the following formula:

[0147]

[0148]

[0149] In the above formula, represents the output of node i in the first layer; f i(x) represents the membership value corresponding to the input data x output by the first layer node i; f i (y) represents the membership value corresponding to the input data y output by the first layer node i; f represents the membership function of the fuzzy set;

[0150] The fuzzification layer is used to fuzzify the input data according to the following formula:

[0151] ω i =f i (x)·f i (y);

[0152] In the above formula, ω i represents the output of the second layer node i; represents the fuzzy operator;

[0153] The batch normalization layer is used to introduce batch normalization to normalize the input data to fix the mean and variance of the input data:

[0154]

[0155]

[0156]

[0157]

[0158] In the above formula, is the output of node i in the third layer; Represents the result after batch normalization of the input data; represents the variance of the input signal; β and γ represent the translation factor and scale factor, respectively, both of which are learnable parameters; ω1 and ω2 represent the outputs of node 1 and node 2 in the second layer, respectively; μ β Represents the mean of the input signal; e represents a decimal close to 0;

[0159] The defuzzification layer is used to obtain the actual output value by weighted averaging according to the following formula:

[0160]

[0161] In the above formula, represents the actual output value of node i in the fourth layer; p i ,q i 、r i They are all parameter sets of node i automatically generated by the ANFIS network;

[0162] The output layer is used to calculate the total output according to the following formula:

[0163]

[0164] In the above formula, O 5 Represents the output of the fifth layer, which is the total output of the model;

[0165] The model training module is used to train the capacitor fault identification model with the dielectric loss tangent and the capacitance value as input and the capacitor insulation aging degree as output to obtain the capacitor insulation aging identification model; the active current component and the leakage current are used as input and the capacitor internal leakage degree is used as output to train the capacitor fault identification model to obtain the capacitor internal leakage identification model; the calculation formula of the dielectric loss tangent is:

[0166]

[0167]

[0168] In the above formula, tanδ is the dielectric loss tangent, δ is the dielectric loss angle; f is the power frequency; ε is the dielectric constant; ρ is the volume resistivity of the insulating medium; S is the plate area; d is the plate spacing; k is the electrostatic force constant;

[0169] The calculation formula of the leakage current is:

[0170]

[0171] In the above formula, I leak is the leakage current; C is the capacitance of the capacitor; V is the voltage of the capacitor, R is the resistance of the capacitor; t is the time;

[0172] The calculation formula of the active current component is:

[0173] I R =I·sinδ;

[0174] In the above formula, I R Active current component; I is the measured current;

[0175] After obtaining the capacitor insulation aging identification model and the capacitor internal leakage identification model, the BP neural network is used to optimize the learnable parameters of the capacitor insulation aging identification model and the capacitor internal leakage identification model, specifically: the loss function of the BP neural network is calculated based on the model prediction value and the actual observation value, the network parameters of the BP neural network are updated by back propagation, and the calculation is iterated until the BP neural network converges; the loss function is:

[0176] The loss function is:

[0177]

[0178] In the above formula, Q iRepresents the true label value obtained from the actual dataset; represents the output value of node i in the third layer of the capacitor fault identification model;

[0179] The fault identification module is used to identify the insulation aging degree and the leakage degree of the capacitor to be detected based on the capacitor insulation aging identification model and the capacitor internal leakage identification model.

[0180] Embodiment 3:

[0181] See also Figure 6 A capacitor fault identification device based on an improved ANFIS comprises a memory and a processor; the memory is used to store computer program code and transmit the computer program code to the processor; the processor is used to execute the method described in Example 1 according to the instructions in the computer program code.

[0182] A computer-readable storage medium having a computer program stored thereon, wherein the computer program implements the method described in Embodiment 1 when executed by a processor.

[0183] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0184] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0185] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0186] Although the preferred embodiments of the present application have been described, those skilled in the art may make other changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications falling within the scope of the present application.

[0187] Obviously, those skilled in the art can make various changes and modifications to the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalents, the present application is also intended to include these modifications and variations.

Claims

1. A capacitor fault identification method based on improved ANFIS, characterized by: The capacitor fault identification method comprises the following steps: S1. Construct capacitor fault identification model based on improved ANFIS network; S2. Using dielectric loss tangent and capacitance as input and capacitor insulation aging degree as output, the capacitor fault identification model is trained to obtain a capacitor insulation aging identification model; using active current component and leakage current as input and capacitor internal leakage degree as output, the capacitor fault identification model is trained to obtain a capacitor internal leakage identification model; S3, identifying the insulation aging degree and leakage degree of the capacitor to be tested respectively based on the capacitor insulation aging identification model and the capacitor internal leakage identification model; In step S1, the capacitor fault identification model constructed has five layers, including an input layer, a fuzzification layer, a batch normalization layer, a defuzzification layer, and an output layer; wherein, The input layer is used to obtain the membership value corresponding to the input data according to the following formula: In the above formula, represents the output of node i in the first layer; f i (x) represents the membership value corresponding to the input data x output by the first layer node i; f i (y) represents the membership value corresponding to the input data y output by the first layer node i; f represents the membership function of the fuzzy set; The fuzzification layer is used to fuzzify the input data according to the following formula: ω i =f i (x)·f i (y); In the above formula, ω i represents the output of the second layer node i; represents the fuzzy operator; The batch normalization layer is used to introduce batch normalization to normalize the input data to fix the mean and variance of the input data: In the above formula, is the output of node i in the third layer; Represents the result after batch normalization of the input data; represents the variance of the input signal; β and γ represent the translation factor and scale factor, respectively, both of which are learnable parameters; ω1 and ω2 represent the outputs of node 1 and node 2 in the second layer, respectively; μ β Represents the mean of the input signal; e represents a decimal close to 0; The defuzzification layer is used to obtain the actual output value by weighted averaging according to the following formula: In the above formula, represents the actual output value of node i in the fourth layer; p i ,q i 、r i They are all parameter sets of node i automatically generated by the ANFIS network; The output layer is used to calculate the total output according to the following formula: In the above formula, O 5 Represents the output of the fifth layer, which is the total output of the model.

2. The capacitor fault identification method based on improved ANFIS according to claim 1 is characterized in that: The step S2 also includes optimizing the learnable parameters of the capacitor insulation aging identification model and the capacitor internal leakage identification model using the BP neural network, specifically: calculating the loss function of the BP neural network based on the output value of the third layer of the capacitor fault identification model and the real label value obtained from the actual data set, updating the network parameters of the BP neural network by back propagation, and iterating the calculation until the BP neural network converges; the loss function is: In the above formula, Q i Represents the true label value obtained from the actual dataset; Represents the output value of node i in the third layer of the capacitor fault identification model.

3. The capacitor fault identification method based on improved ANFIS according to claim 1 is characterized in that: The calculation formula of the dielectric loss tangent is: In the above formula, tanδ is the dielectric loss tangent, δ is the dielectric loss angle; f is the power frequency; ε is the dielectric constant; ρ is the volume resistivity of the insulating medium; S is the plate area; d is the plate spacing; k is the electrostatic force constant; The calculation formula of the leakage current is: In the above formula, I leak is the leakage current; C is the capacitance of the capacitor; V is the voltage of the capacitor, R is the resistance of the capacitor; t is the time; The calculation formula of the active current component is: I R =I·sinδ; In the above formula, I R Active current component; I is the measured current.

4. A capacitor fault identification system based on improved ANFIS, characterized by: The capacitor fault identification system includes a model building module, a model training module, and a fault identification module: The model building module is used to build a capacitor fault identification model based on the improved ANFIS network; The model training module is used to train the capacitor fault identification model with the dielectric loss tangent and the capacitance value as input and the capacitor insulation aging degree as output to obtain the capacitor insulation aging identification model; Taking the active current component and the leakage current as input and the internal leakage degree of the capacitor as output, the capacitor fault identification model is trained to obtain the capacitor internal leakage identification model; The fault identification module is used to identify the insulation aging degree and leakage degree of the capacitor to be detected based on the capacitor insulation aging identification model and the capacitor internal leakage identification model respectively; The capacitor fault identification model constructed by the model building module has five layers, including an input layer, a fuzzification layer, a batch normalization layer, a defuzzification layer, and an output layer; wherein, The input layer is used to obtain the membership value corresponding to the input data according to the following formula: In the above formula, represents the output of node i in the first layer; f i (x) represents the membership value corresponding to the input data x output by the first layer node i; f i (y) represents the membership value corresponding to the input data y output by the first layer node i; f represents the membership function of the fuzzy set; The fuzzification layer is used to fuzzify the input data according to the following formula: ω i =f i (x)·f i (y); In the above formula, ω i represents the output of the second layer node i; represents the fuzzy operator; The batch normalization layer is used to introduce batch normalization to normalize the input data to fix the mean and variance of the input data: In the above formula, is the output of node i in the third layer; Represents the result after batch normalization of the input data; represents the variance of the input signal; β and γ represent the translation factor and scale factor, respectively, both of which are learnable parameters; ω1 and ω2 represent the outputs of node 1 and node 2 in the second layer, respectively; μ β Represents the mean of the input signal; E represents a decimal close to 0; The defuzzification layer is used to obtain the actual output value by weighted averaging according to the following formula: In the above formula, Represents the actual output value of node i in the fourth layer; p i ,q i 、r i They are all parameter sets of node i automatically generated by the ANFIS network; The output layer is used to calculate the total output according to the following formula: In the above formula, O 5 Represents the output of the fifth layer, which is the total output of the model.

5. The capacitor fault identification system based on improved ANFIS according to claim 4 is characterized in that: The model training module is also used to optimize the learnable parameters of the capacitor insulation aging identification model and the capacitor internal leakage identification model using the BP neural network, specifically: the loss function of the BP neural network is calculated based on the model prediction value and the actual observation value, the network parameters of the BP neural network are updated by back propagation, and the calculation is iterated until the BP neural network converges; the loss function is: In the above formula, Q i Represents the true label value obtained from the actual dataset; Represents the output value of node i in the third layer of the capacitor fault identification model.

6. The capacitor fault identification system based on improved ANFIS according to claim 4 is characterized in that: The calculation formula of the dielectric loss tangent is: In the above formula, tanδ is the dielectric loss tangent, δ is the dielectric loss angle; f is the power frequency; ε is the dielectric constant; ρ is the volume resistivity of the insulating medium; S is the plate area; d is the plate spacing; k is the electrostatic force constant; The calculation formula of the leakage current is: In the above formula, I leak is the leakage current; C is the capacitance of the capacitor; V is the voltage of the capacitor, R is the resistance of the capacitor; t is the time; The calculation formula of the active current component is: I R =I·sinδ; In the above formula, I R Active current component; I is the measured current.

7. A capacitor fault identification device based on improved ANFIS, characterized in that: The capacitor fault identification device includes a memory and a processor; The memory is used to store computer program code and transmit the computer program code to the processor; The processor is configured to execute the method according to any one of claims 1 to 3 according to instructions in the computer program code.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 3 is implemented.

Images