A method, device and storage medium for insulator condition detection

Abstract

This invention discloses an insulator condition detection method, device, and storage medium. The method includes: acquiring insulator operating condition data and insulator image data; generating a first feature data matrix based on the insulator operating condition data and a second feature data matrix based on the insulator image data; determining a weight matrix of the first feature data matrix using a first neural network model, and generating a third feature data matrix based on the first feature data matrix and the weight matrix; using the third feature data matrix as input and generating a fourth feature data matrix through a second neural network model; concatenating the fourth feature data matrix and the second feature data matrix to generate a fifth feature data matrix; using the fifth feature data matrix as input and generating a sixth feature data matrix through a third neural network model; and using the sixth feature data matrix as input and generating an insulator condition detection result through a fourth neural network model.

Description

Technical Field

[0001] The present invention relates to the field of detection technology, and in particular to an insulator condition detection method, device and storage medium. Background Technology

[0002] Insulators are widely used in power transmission lines, but when insulators experience pollution flashover, it can seriously threaten the safe and stable operation of the power system. To prevent pollution flashover accidents, it is necessary to conduct online detection of the degree of pollution on the insulator surface. This allows for timely detection and handling of problems with the insulators, ensuring the reliable operation of the power system.

[0003] To date, existing methods for quantifying the factors influencing insulator surface salt density based on correlation coefficients are ill-suited for adaptive quantification under different environments, resulting in low accuracy in predicting insulator surface salt density. Furthermore, leakage current data measured by sensors during condition monitoring cannot fully reflect the insulator surface condition. Summary of the Invention

[0004] This invention provides an insulator condition detection method, device, and storage medium to achieve accurate insulator condition detection.

[0005] In a first aspect, embodiments of the present invention provide an insulator condition detection method, comprising:

[0006] Acquire insulator operating condition data and insulator image data;

[0007] A first feature data matrix is ​​generated based on the insulator operating condition data, and a second feature data matrix is ​​generated based on the insulator image data.

[0008] The weight matrix of the first feature data matrix is ​​determined using a first neural network model, and a third feature data matrix is ​​generated based on the first feature data matrix and the weight matrix.

[0009] The third feature data matrix is ​​used as input to generate the fourth feature data matrix through the second neural network model.

[0010] The fourth feature data matrix and the second feature data matrix are concatenated to generate the fifth feature data matrix;

[0011] The fifth feature data matrix is ​​used as input to generate the sixth feature data matrix through the third neural network model.

[0012] The sixth feature data matrix is ​​used as input to generate insulator state detection results through a fourth neural network model. Optionally, the weight matrix of the first feature data matrix is ​​determined using a first neural network model, including:

[0013] Y1=σ1(W1X)

[0014] Y2=σ2(W2Y1)

[0015] In the formula, σ1 represents the first nonlinear activation function, σ2 represents the second nonlinear activation function, X represents the insulator operating data, W1 represents the first parameter matrix, W2 represents the second parameter matrix, and Y2 represents the weight matrix.

[0016] Optionally, the second neural network model uses a first LSTM model layer, the third neural network model uses a second LSTM model layer, and the fourth neural network model uses a shallow neural network layer;

[0017] The second neural network model, the third neural network model, and the fourth neural network model constitute an ML-LSTM model;

[0018] Obtain the hyperparameter matrix of the ML-LSTM model;

[0019] The parameters in the hyperparameter matrix are determined using a cross-hatching algorithm, wherein chaotic mapping is used to generate the population used in the cross-hatching algorithm.

[0020] Optionally, the first LSTM model layer and the second LSTM model layer have the same structure;

[0021] The function expressions for the first LSTM model layer and the second LSTM model layer are as follows:

[0022] f t =σ(W f ·[h t-1 ,x t ]+b f )

[0023] i t =σ(W i ·[h t-1 ,x t ]+b i )

[0024] o t =σ(W o ·[h t-1 ,x t ]+b o )

[0025]

[0026]

[0027] h t =o t ·tanh(Ct )

[0028] In the formula, σ represents the first activation function, x t W represents the input features at the current time. f Let W represent the first parameter training matrix. i W represents the training matrix with the second parameter. o W represents the training matrix with third parameter. c Let b represent the fourth parameter training matrix. f b represents the first training bias term. i b represents the second training bias term. o b represents the third training bias term. c h represents the fourth training bias term. t-1 f represents the output of the hidden layer at the previous moment. t Represents the forget gate, i t Indicates the input gate value. Represents candidate values, o t Indicates the output gate value, h t Indicates the hidden layer state;

[0029] The function expression for the shallow neural network layer is:

[0030] P Fi =f[Wf g (wX+b)+B]

[0031] In the formula, f g Let w represent the second activation function, b represent the first weight parameter, b represent the first bias parameter, W represent the second weight parameter, B represent the second bias parameter, and P represent the second activation function. Fi This indicates the insulator condition detection results;

[0032] The hyperparameter matrix is:

[0033] θ=[W f W i W o W c ,b f ,b i ,b o ,b c ,w,b,W,B).

[0034] Optionally, the insulator operating condition data includes:

[0035] Ambient temperature, ambient humidity, equivalent salt density of insulator, effective value of leakage current, maximum value of leakage current, ratio of third harmonic to fundamental frequency of leakage current, and standard deviation of leakage current.

[0036] Optionally, the insulator image data can be used as input to generate the second feature data matrix using a convolutional neural network model.

[0037] Optionally, the insulator operating condition data is normalized to obtain the first feature data matrix.

[0038] Secondly, embodiments of the present invention also provide an insulator condition detection method, comprising:

[0039] Acquire insulator operating condition data and insulator image data;

[0040] A first feature data matrix is ​​generated based on the insulator operating condition data, and a second feature data matrix is ​​generated based on the insulator image data.

[0041] The weight matrix of the first feature data matrix is ​​determined using a first neural network model, and a third feature data matrix is ​​generated based on the first feature data matrix and the weight matrix.

[0042] The third feature data matrix is ​​used as input to generate the first weights through the second neural network model.

[0043] The second feature data matrix is ​​used as input, and the second weights are generated through the third neural network model.

[0044] The first penalty function of the fourth neural network is generated based on the first weight, the second weight, the third feature data matrix, and the second feature data matrix.

[0045] The fourth neural network model is configured to concatenate the third feature data matrix and the second feature data matrix to generate the fourth feature data matrix;

[0046] A fifth neural network model is configured to take the fourth feature data matrix as input and generate insulator state detection results;

[0047] An insulator state prediction model is constructed by the second neural network model, the third neural network model, and the fifth neural network model.

[0048] Obtain the initial hyperparameter matrix of the insulator state prediction model, update the initial parameter matrix based on the second penalty function, and complete the training of the insulator state prediction model;

[0049] Using insulator operating condition data and insulator image data as input, the trained insulator condition prediction model is used to generate insulator condition detection results.

[0050] Optionally, the first weight includes the loss function of the second neural network model, expressed as:

[0051] a1 = 1 - L1

[0052]

[0053] In the formula, a1 represents the first weight, X1 represents the third feature data matrix, and Y1 represents the first output;

[0054] The second weight includes the loss function of the third neural network model, expressed as:

[0055] a1 = 1 - L1

[0056]

[0057] In the formula, a2 represents the second weight, X2 represents the second feature data matrix, and Y2 represents the second output;

[0058] The expression for the first penalty function is:

[0059]

[0060] In the formula, L represents the first penalty function.

[0061] Optionally, the second penalty function is:

[0062]

[0063] In the formula, f obj Let N represent the fitness function, and N represent the number of samples. Represents the actual value. This represents the predicted value.

[0064] Optionally, based on the second penalty function, the parameters in the hyperparameter matrix are determined using a cross-tab algorithm.

[0065] Thirdly, embodiments of the present invention also provide an electronic device, including at least one processor and a memory communicatively connected to the at least one processor;

[0066] The memory stores a computer program that can be executed by the at least one processor, which enables the at least one processor to perform any of the insulator state detection methods described in the embodiments of the present invention.

[0067] Fourthly, embodiments of the present invention also provide a computer-readable storage medium storing computer instructions, which are used to cause a processor to execute any of the insulator state detection methods described in the embodiments of the present invention.

[0068] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention proposes an insulator condition detection method, which includes acquiring insulator operating condition data and insulator image data. Based on these two types of data, the insulator filling is detected. Because the insulator image data is combined, when using the fusion of these two types of data for detection, relatively independent multi-dimensional features of the insulator can be determined. This improves the accuracy of the detection results while avoiding the introduction of redundant feature information. Specifically, a first feature data matrix is ​​generated based on the insulator operating condition data, and a weight matrix of this data matrix is ​​generated. The influence of each data parameter in the first feature data matrix on the detection result is determined through the weight matrix, thereby obtaining the third feature data. The system uses a third feature data matrix to generate a fourth feature data matrix using a second neural network model. The fourth and second feature data matrices are then concatenated to generate a fifth feature data matrix. This fifth feature data matrix is ​​used as input to generate a sixth feature data matrix using the third neural network model. Finally, the sixth feature data matrix is ​​used as input to generate the insulator condition detection result using a fourth neural network model. By processing the third, fifth, and sixth feature data matrices separately using a single neural network model, the data features obtained at each level can be refined, and the correlation between data features at each level can be fully combined, thereby achieving effective data feature fusion and ultimately improving the accuracy of the final insulator condition detection. Attached Figure Description

[0069] Figure 1 This is a flowchart of the insulator condition detection method in the embodiment;

[0070] Figure 2 This is a flowchart of another insulator condition detection method in the embodiment;

[0071] Figure 3 This is a flowchart of another insulator condition detection method in the embodiments;

[0072] Figure 4 This is a schematic diagram of the electronic device structure in the embodiment. Detailed Implementation

[0073] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, it should be noted that, for ease of description, the accompanying drawings show only the parts relevant to the present invention, and not all of the structures.

[0074] Example 1

[0075] Figure 1 This is a flowchart of the insulator condition detection method in the embodiment, for reference. Figure 1 Insulator condition detection methods include:

[0076] S101. Obtain insulator operating condition data and insulator image data.

[0077] In this embodiment, the insulator operating condition data is set to non-image data. The insulator operating condition data can be current, voltage, environmental parameters, leakage current, preset insulator parameters (for leakage current threshold, leakage current standard deviation, etc.).

[0078] Insulator operating condition data can be obtained through corresponding sensors or measuring devices, or by reading from a designated storage space in the power grid system.

[0079] In this embodiment, the insulator image data is defined to include images of the insulator. The method of acquiring the insulator image is not limited; for example, the insulator image can be acquired in the following way:

[0080] The system automatically captures inspection images along the power transmission line path using equipment such as drones and unmanned vehicles.

[0081] Feature extraction is performed on the inspection image region to determine the insulators contained in the inspection image region. The method for extracting insulators is not limited. For example, feature extraction can be achieved through edge detection, corner point extraction, etc.

[0082] S102. Generate a first feature data matrix based on the insulator operating condition data, and generate a second feature data matrix based on the insulator image data.

[0083] For example, in this embodiment, generating the first feature data matrix based on the insulator operating condition data can be as follows:

[0084] Data cleaning and other operations are performed on the data to remove obviously abnormal data from the insulator operating condition data, and the normal data are used to form the first feature data matrix according to the specified format.

[0085] For example, the first feature data matrix can be in the form of:

[0086] X1 = [D 1_1 D 1_2 …D 1_i ]

[0087] In the formula, X1 represents the first feature data matrix, and D 1_1 ~D 1_i This represents the preset operating data (parameters) for each type of insulator.

[0088] For example, in this embodiment, generating the second feature data matrix based on the insulator image data can be as follows:

[0089] The second feature data matrix is ​​set to include specified feature parameters obtained from the insulator image data. The feature parameters can be: the maximum temperature of the insulator, the minimum temperature, whether there is breakage or damage, etc.

[0090] The second feature data matrix can take the form of:

[0091] X2 = [D 2_1 D 2_2 …D 2_i ]

[0092] In the formula, X2 represents the first feature data matrix, and D 2_1 ~D 2_i This represents the preset operating data (parameters) for each type of insulator.

[0093] For example, in this embodiment, the method of determining the above-mentioned feature parameters from the insulator image data is not limited. For example, the above-mentioned feature parameters can be determined based on neural networks or other methods.

[0094] For example, the neural network described above for determining feature parameters from insulator image data can be a multi-scale image feature extraction network, which includes multiple multi-scale hybrid convolutional blocks.

[0095] Each multi-scale hybrid convolutional block includes horizontal and vertical gradients, and the update process for each convolutional block is as follows:

[0096] H(x,y)=Z(x+1,y)-Z(x-1,y)

[0097] V(x,y)=Z(x,y+1)-Z(x,y-1)

[0098] F c =f(H(x,y),V(x,y))

[0099] In the formula, H(x,y) represents the horizontal gradient calculation, Z represents the gradient, V(x,y) represents the vertical gradient calculation, x represents the horizontal direction of a pixel, y represents the vertical direction of a pixel, and F... c Represents a multi-scale hybrid convolutional block;

[0100] The multi-scale image feature extraction network consists of multiple multi-scale hybrid convolutional blocks connected in series. The update process of the multi-scale image feature extraction network is as follows:

[0101] F = f(F) c1 ,F c2 ,…F cn )

[0102] In the formula, F ciLet F represent the i-th multi-scale hybrid convolutional block, and let F represent the multi-scale image feature extraction network.

[0103] For example, after grayscale conversion, denoising, and enhancement processing of the insulator image, insulator image data is obtained. The insulator image data is then input into a multi-scale image feature extraction network to obtain a second feature data matrix X2.

[0104] S103. Use the first neural network model to determine the weight matrix of the first feature data matrix, and generate the third feature data matrix based on the first feature data matrix and the weight matrix.

[0105] In this embodiment, each weight (parameter) in the weight matrix is ​​set to represent the degree of influence of an insulator condition data on the insulator condition detection result.

[0106] In this scheme, the specific structure of the first neural network model is not limited, and it can be freely selected according to actual needs.

[0107] S104. Using the third feature data matrix as input, generate the fourth feature data matrix through the second neural network model.

[0108] In this embodiment, a second feature data matrix is ​​set to represent the hidden layer state of the second neural network model.

[0109] S105. Concatenate the fourth feature data matrix and the second feature data matrix to generate the fifth feature data matrix.

[0110] In this embodiment, there is no limitation on the way the fourth feature data matrix and the second feature data matrix are spliced. For example, the second feature data matrix can be expanded in dimension, and then the fourth feature data matrix and the second feature data matrix can be directly spliced ​​to finally form the fifth feature data matrix.

[0111] Alternatively, the parameters in the fourth and second feature data matrices can be rearranged and combined to generate a fifth feature data matrix of a specified dimension.

[0112] S106. Using the fifth feature data matrix as input, generate the sixth feature data matrix through the third neural network model.

[0113] In this embodiment, a sixth feature data matrix is ​​set to represent the hidden layer state of the third neural network model.

[0114] S107. Using the sixth feature data matrix as input, generate the insulator state detection result through the fourth neural network model.

[0115] In conjunction with steps S102 to S107, in this embodiment, specifically, the fourth neural network model is configured to take the sixth feature data matrix as input to generate the insulator state detection result.

[0116] An insulator state prediction model is constructed by using a second neural network model, a third neural network model, and a fourth neural network model.

[0117] The initial hyperparameter matrix of the insulator state prediction model is obtained, and the parameters in the hyperparameter matrix are determined based on the cross-parameter algorithm. Chaotic mapping is used to generate the population used in the cross-parameter algorithm to complete the training of the insulator state prediction model.

[0118] Using insulator operating condition data and insulator image data as input, the trained insulator condition prediction model is used to generate insulator condition detection results.

[0119] In this embodiment, the output of the fourth neural network model, namely the insulator state detection result, is used as the basis for determining the insulator state.

[0120] This embodiment proposes an insulator condition detection method. The method includes acquiring insulator operating condition data and insulator image data. Based on these two types of data, the insulator filling is detected. Because the insulator image data is combined, the detection using the fusion of these two types of data can determine the relatively independent multi-dimensional features of the insulator, improving the accuracy of the detection results while avoiding the introduction of redundant feature information. Specifically, a first feature data matrix is ​​generated based on the insulator operating condition data, and a weight matrix of this data matrix is ​​generated. The weight matrix determines the magnitude of the influence of each data parameter in the first feature data matrix on the detection result, thus obtaining a third feature data matrix. The third feature... The data matrix is ​​used to generate a fourth feature data matrix through a second neural network model. The fourth feature data matrix and the second feature data matrix are then concatenated to generate a fifth feature data matrix. The fifth feature data matrix is ​​used as input to generate a sixth feature data matrix through a third neural network model. The sixth feature data matrix is ​​used as input to generate the insulator condition detection result through a fourth neural network model. By using a neural network model to process the third, fifth, and sixth feature data matrices respectively, the data features obtained at each level can be refined, and the correlation between the data features at each level can be fully combined to achieve effective fusion of data features, thereby improving the accuracy of the final insulator condition detection.

[0121] exist Figure 1 Based on the scheme shown, in one possible implementation, the weight matrix for determining the first feature data matrix using the first neural network model includes:

[0122] Y1=σ1(W1X1)

[0123] Y2=σ2(W2Y1)

[0124] In the formula, σ1 represents the first nonlinear activation function, σ2 represents the second nonlinear activation function, X1 represents the first feature data matrix, W1 represents the first parameter matrix, W2 represents the second parameter matrix, and Y2 represents the weight matrix.

[0125] For example, in this scheme, the first neural network model is configured to include two fully connected layers, wherein the specific method for generating the weight matrix through the first neural network model is as follows:

[0126] The first feature data matrix is ​​input into the first fully connected layer to obtain Y1, where Y1 is:

[0127] Y1=σ1(W1X1)

[0128] Here, σ1 is specifically defined as the nonlinear activation function ReLU, and W1 specifically represents the layer parameters of the first fully connected layer.

[0129] Inputting Y1 into the second fully connected layer yields Y2, which is:

[0130] Y2=σ2(W2Y1)

[0131] Here, σ2 is specifically defined as the nonlinear activation function Sigmoid, and W2 is specifically defined as the layer parameter of the second fully connected layer.

[0132] In this scheme, the third feature data matrix is ​​specifically the product of Y2 and the first feature data matrix.

[0133] In this scheme, the third feature data matrix X3 can be equivalently represented by the following formula:

[0134] X3=X1sigmiod(W2Relu(W1·Pool(X1)))

[0135] In the above formula, Pool represents the global average pooling operation, ReLU represents the non-linear activation function ReLU, and sigmoid represents the non-linear activation function Sigmoid.

[0136] Based on any of the aforementioned schemes, in one possible implementation scheme, the second neural network model adopts the first LSTM model layer, the third neural network model adopts the second LSTM model layer, and the fourth neural network model adopts the shallow neural network layer.

[0137] The second, third, and fourth neural network models constitute the ML-LSTM model;

[0138] Obtain the hyperparameter matrix of the ML-LSTM model;

[0139] The parameters in the hyperparameter matrix are determined using the cross-hatching algorithm, where chaotic mapping is used to generate the population used in the cross-hatching algorithm.

[0140] In this scheme, the specific form of the ML-LSTM model is not limited. The hyperparameter matrix of the ML-LSTM model is set as θ, which can be determined as follows:

[0141] The initial parameters of the hyperparameter matrix θ are set based on experience;

[0142] The fitness function of the cross-hatching algorithm is defined as follows:

[0143]

[0144] In the formula, f obj Let N represent the fitness function, and N represent the number of samples. Represents the actual value. Indicates the predicted value;

[0145] In the population initialization phase of the cross-cutting algorithm, a (Chebyshev) chaotic map is used to initialize the population to ensure a uniform initial population distribution. The expression for the chaotic map is:

[0146] x(i+1)=cos(4×cos -1 (x(i)))

[0147] In the formula, x(i) represents the randomly generated population, and x(i+1) represents the population generated after applying chaotic mapping;

[0148] Let parent particles θ(i) and θ(j) randomly choose their nth dimension to cross each other. During the crossing process, random mutation is introduced. The expression for random mutation is:

[0149] M hc (i,n)=r1·θ(i,n)+(1-r1)·θ(j,n)+c1·θ(i,n)-θ(j,n)+r3(θ a -θ(i,n))-r4(θ′(i,n)-θ(i,n)

[0150] M hc (j,n)=r2·θ(j,n)+(1-r2)·θ(i,n)+c2·θ(j,n)-θ(i,n)+r3(θ a -θ(j,n))-r4(θ′(j,n)-θ(j,n)

[0151] In the formula, r1, r2, r3, and r4 represent random numbers in the range of 0 to 1, c1 and c2 represent random numbers in the range of -1 to 1, θ(i,n) is the nth dimension of particle θ(i), and θ(j,n) is the nth dimension of particle θ(j). a Let θ' represent the globally optimal particle, and θ'(i,n) and θ'(j,n) represent random particles.

[0152] The parent particle θ(q) randomly selects the v-th and k-th dimensions to intersect, and its expression is:

[0153] MS vc (q,v)=rθ(q,v)+(1-r)θ(q,k)

[0154] In the formula, MS vc (q,v) is the v-th dimension offspring generated by the vertical intersection of θ(q,v) and θ(q,k), where r is a random number in the range of 0 to 1, θ(q,v) represents the v-th dimension of particle θ(q), and θ(q,k) is the k-th dimension of particle θ(q).

[0155] Based on the set number of iterations, repeat the random mutation and the mutual crossing of the vth and kth dimensions in the previous two steps, and stop iterating when the preset number of iterations is reached.

[0156] In this scheme, the parameters in the hyperparameter matrix after iterative calculation are substituted into the ML-LSTM model to obtain an ML-LSTM model that meets the preset conditions.

[0157] In this scheme, the cross-hatching algorithm is used to determine the model parameters in the ML-LSTM model. Simultaneously, chaotic mapping is employed in the cross-hatching algorithm to generate the required population. Based on chaotic mapping, the initial population is uniformly distributed, which can solve the potential local optima problem and effectively improve the intelligent detection of insulator states.

[0158] Based on the scheme of constructing an ML-LSTM model using a second neural network model, a third neural network model, and a fourth neural network model, in one possible implementation, the first LSTM model layer and the second LSTM model layer are designed to have the same structure.

[0159] In this scheme, the function expressions for the first LSTM model layer and the second LSTM model layer are defined as follows:

[0160] f t =σ(W f ·[h t-1 ,x t ]+b f )

[0161] i t =σ(W i ·[ht-1 ,x t ]+b i )

[0162] o t =σ(W o ·[h t-1 ,x t ]+b o )

[0163]

[0164]

[0165] h t =o t ·tanh(C t )

[0166] In the formula, σ represents the first activation function, x t W represents the input features at the current time. f Let W represent the first parameter training matrix. i W represents the training matrix with the second parameter. o W represents the training matrix with third parameter. c Let b represent the fourth parameter training matrix. f b represents the first training bias term. i b represents the second training bias term. o b represents the third training bias term. c h represents the fourth training bias term. t-1 f represents the output of the hidden layer at the previous moment. t Represents the forget gate, i t Indicates the input gate value. Represents candidate values, o t Indicates the output gate value, h t This represents the hidden layer state.

[0167] Furthermore, in this scheme, the function expression of the shallow neural network layer is defined as follows:

[0168] P Fi =f[Wf g (wX+b)+B]

[0169] In the formula, f g Let w represent the second activation function, b represent the first weight parameter, b represent the first bias parameter, W represent the second weight parameter, B represent the second bias parameter, and P represent the second activation function. Fi This indicates the insulator condition detection result.

[0170] In this scheme, based on the above function expression, the corresponding hyperparameter matrix is:

[0171] θ=[W f W i W o W c ,b f ,b i ,b o ,b c ,w,b,W,B).

[0172] In this scheme, the first LSTM model layer and the second LSTM model layer are designed to have the same structure. Based on this, it is convenient to obtain the model parameters in the ML-LSTM model through the cross-multiplication algorithm, thereby reducing the design and training costs of the ML-LSTM model and improving the efficiency of model training.

[0173] Based on any of the aforementioned schemes, in one possible implementation scheme, the insulator operating condition data includes:

[0174] Ambient temperature, ambient humidity, equivalent salt density of insulator, effective value of leakage current, maximum value of leakage current, ratio of third harmonic to fundamental frequency of leakage current, and standard deviation of leakage current.

[0175] Based on any of the aforementioned schemes, in one possible implementation, insulator image data is used as input, and a convolutional neural network model is used to generate a second feature data matrix.

[0176] For example, in this solution, the structure of the convolutional neural network model is not limited. For instance, the convolutional neural network may include 20 convolutional layers and 10 pooling layers.

[0177] Based on any of the aforementioned schemes, in one possible implementation scheme, the insulator operating condition data is normalized to obtain a first feature data matrix.

[0178] For example, in this solution, the min-max normalization method can be used to normalize the insulator operating condition data.

[0179] Figure 2 This is a flowchart of another insulator condition detection method in the embodiment, for reference. Figure 2 Based on any of the aforementioned solutions, in one possible implementation, the method includes:

[0180] S201. Obtain insulator operating condition data and insulator image data.

[0181] In this scheme, the insulator operating condition data includes ambient temperature, ambient humidity, historical insulator equivalent salt density, effective value of leakage current at the corresponding salt density, maximum value of leakage current, ratio of third harmonic to fundamental frequency of leakage current, and standard deviation of leakage current.

[0182] In this scheme, the insulator image data is set to insulator images that do not contain background image information.

[0183] S202. Normalize the insulator operating condition data to generate the first feature data matrix.

[0184] In this scheme, the insulator operating condition data is specifically processed by min-max normalization to generate the first feature data matrix.

[0185] S203. Using the insulator image data as input, a convolutional neural network model is used to generate the second feature data matrix.

[0186] In this scheme, the convolutional neural network is set to include 20 convolutional layers and 10 pooling layers. The output of the pooling layers is set to construct a second feature data matrix. The second feature data matrix includes several insulator feature parameters, including insulator contour information, insulator temperature information and insulator fracture information.

[0187] S204. Use the first neural network model to determine the weight matrix of the first feature data matrix, and generate the third feature data matrix based on the first feature data matrix and the weight matrix.

[0188] In this scheme, the first neural network model is defined as consisting of two fully connected layers. The specific method for generating the weight matrix using the first neural network model is as follows:

[0189] The first feature data matrix is ​​input into the first fully connected layer to obtain Y1, where Y1 is:

[0190] Y1=σ1(W1X1)

[0191] Here, σ1 is specifically defined as the nonlinear activation function ReLU, and W1 specifically represents the layer parameters of the first fully connected layer.

[0192] Inputting Y1 into the second fully connected layer yields Y2, which is:

[0193] Y2=σ2(W2Y1)

[0194] Here, σ2 is specifically defined as the nonlinear activation function Sigmoid, and W2 is specifically defined as the layer parameter of the second fully connected layer.

[0195] In this scheme, the third feature data matrix is ​​specifically the product of Y2 and the first feature data matrix.

[0196] S205. Using the third feature data matrix as input, generate the fourth feature data matrix through the first LSTM model layer.

[0197] In this scheme, a fourth feature data matrix is ​​set to represent the hidden layer state of the first LSTM model layer.

[0198] S206. Concatenate the fourth feature data matrix and the second feature data matrix to generate the fifth feature data matrix.

[0199] S207. Using the fifth feature data matrix as input, generate the sixth feature data matrix through the second LSTM model layer.

[0200] In this scheme, a sixth feature data matrix is ​​set to represent the hidden layer state of the second LSTM model layer.

[0201] S208. Using the sixth feature data matrix as input, generate insulator state detection results through a shallow neural network layer.

[0202] Combining steps S205 to S208, in this scheme, the first LSTM model layer, the second LSTM model layer, and the shallow neural network layer constitute the ML-LSTM model.

[0203] The first and second LSTM model layers are assumed to have the same structure. Both layers are defined as Long Short-Term Memory (LSTM) networks. The specific function expression for each LSTM layer is defined as follows:

[0204] f t =σ(W f ·[h t-1 ,x t ]+b f )

[0205] i t =σ(W i ·[h t-1 ,x t ]+b i )

[0206] o t =σ(W o ·[h t-1 ,x t ]+b o )

[0207]

[0208]

[0209] h t =o t ·tanh(C t )

[0210] The function expression for the shallow neural network layer is defined as follows:

[0211] P Fi =f[Wf g (wX+b)+B]

[0212] In this scheme, the hyperparameter matrix of the ML-LSTM model is set as follows:

[0213] θ=[W f W i W o W c ,b f ,b i ,b o ,b c ,w,b,W,B]

[0214] In this scheme, the hyperparameter matrix θ is determined by combining the cross-hatching algorithm and chaotic mapping, specifically including:

[0215] The fitness function of the cross-hatching algorithm is defined as follows:

[0216]

[0217] In the formula, f obj Let N represent the fitness function, and N represent the number of samples. Represents the actual value. Indicates the predicted value;

[0218] In the population initialization phase of the cross-cutting algorithm, a (Chebyshev) chaotic map is used to initialize the population to ensure a uniform initial population distribution. The expression for the chaotic map is:

[0219] x(i+1)=cos(4×cos -1 (x(i)))

[0220] In the formula, x(i) represents the randomly generated population, and x(i+1) represents the population generated after applying chaotic mapping;

[0221] Let parent particles θ(i) and θ(j) randomly choose their nth dimension to cross each other. During the crossing process, random mutation is introduced. The expression for random mutation is:

[0222] M hc (i,n)=r1·θ(i,n)+(1-r1)·θ(j,n)+c1·θ(i,n)-θ(j,n)+r3(θ a -θ(i,n))-r4(θ′(i,n)-θ(i,n)

[0223] M hc(j,n)=r2·θ(j,n)+(1-r2)·θ(i,n)+c2·θ(j,n)-θ(i,n)+r3(θ a -θ(j,n))-r4(θ′(j,n)-θ(j,n)

[0224] In the formula, r1, r2, r3, and r4 represent random numbers in the range of 0 to 1, c1 and c2 represent random numbers in the range of -1 to 1, θ(i,n) is the nth dimension of particle θ(i), and θ(j,n) is the nth dimension of particle θ(j). a Let θ' represent the globally optimal particle, and θ'(i,n) and θ'(j,n) represent random particles.

[0225] The parent particle θ(q) randomly selects the v-th and k-th dimensions to intersect, and its expression is:

[0226] MS vc (q,v)=rθ(q,v)+(1-r)θ(q,k)

[0227] In the formula, MS vc (q,v) is the v-th dimension offspring generated by the vertical intersection of θ(q,v) and θ(q,k), where r is a random number in the range of 0 to 1, θ(q,v) represents the v-th dimension of particle θ(q), and θ(q,k) is the k-th dimension of particle θ(q).

[0228] Based on the set number of iterations, repeat the random mutation and the mutual crossing of the vth and kth dimensions in the previous two steps. Stop iterating when the preset number of iterations is reached, and finally obtain the required hyperparameter matrix θ.

[0229] Example 2

[0230] Figure 3 This is a flowchart of another insulator condition detection method in the embodiments, see reference. Figure 3 In this embodiment, the detection method includes:

[0231] S301. Acquire insulator operating condition data and insulator image data.

[0232] S302. Normalize the insulator operating condition data to generate the first feature data matrix.

[0233] S303. Using the insulator image data as input, a convolutional neural network model is used to generate the second feature data matrix.

[0234] S304. Use a first neural network model to determine the weight matrix of the first feature data matrix, and generate a third feature data matrix based on the first feature data matrix and the weight matrix.

[0235] In conjunction with steps S301 to S304, the content of the above steps in this embodiment is the same as the corresponding content recorded in steps S101 to S104 in the embodiment, and the specific content will not be described in detail.

[0236] S305. Using the third feature data matrix as input, the first weight is generated through the second neural network model.

[0237] S306. Using the second feature data matrix as input, the second weights are generated through the third neural network model.

[0238] S307. Generate the first penalty function of the fourth neural network based on the first weight, the second weight, the third feature data matrix, and the second feature data matrix.

[0239] In conjunction with steps S305 to S307, in this embodiment, a first weight is set to represent the influence factor of the second neural network model on the output result in the insulator state prediction model;

[0240] A second weight is set to represent the influence factor of the third neural network model on the output result in the insulator state prediction model.

[0241] In this scheme, the loss function of the second neural network model is specifically set as the calculation formula for the first weight, and the loss function of the third neural network model is set as the calculation formula for the second weight.

[0242] S308. Configure a fourth neural network model to concatenate the third feature data matrix and the second feature data matrix to generate the fourth feature data matrix.

[0243] For example, in this embodiment, the method of concatenating the third feature data matrix and the second feature data matrix into the fourth neural network model is not limited. For example, the concatenation of the feature data matrix can be achieved by summation concatenation, product concatenation, correlation analysis fusion, etc.

[0244] S309. Configure the fifth neural network model to take the fourth feature data matrix as input and generate insulator state detection results.

[0245] S310. An insulator state prediction model is constructed by using a second neural network model, a third neural network model, and a fifth neural network model.

[0246] S311. Obtain the initial hyperparameter matrix of the insulator state prediction model, update the initial parameter matrix based on the second penalty function, and complete the training of the insulator state prediction model.

[0247] In this embodiment, the second penalty function can be set according to requirements. For example, the second penalty function can be used to represent the minimum root mean square error of the insulator state model.

[0248] In this embodiment, the method for updating the initial parameter matrix based on the second penalty function can be as follows:

[0249] Based on the second penalty function, the cross-parameter algorithm is used to determine the parameters in the hyperparameter matrix, where the second penalty function is used as the fitness function of the cross-parameter algorithm.

[0250] S312. Using insulator operating condition data and insulator image data as input, generate insulator condition detection results using the trained insulator condition prediction model.

[0251] This embodiment proposes an insulator condition detection method. The method includes acquiring insulator operating condition data and insulator image data. Based on these two types of data, the insulator filling is detected. Because the image data is combined, the fusion of these two types of data allows for the determination of relatively independent multi-dimensional features of the insulator, improving the accuracy of the detection results while avoiding the introduction of redundant feature information. Specifically, a first feature data matrix is ​​generated based on the insulator operating condition data, and a weight matrix is ​​generated for this data matrix. The weight matrix determines the magnitude of the influence of each data parameter in the first feature data matrix on the detection results, thereby obtaining a third feature data matrix. The first feature data matrix is ​​used as input to generate the first weight through the second neural network model; the second feature data matrix is ​​used as input to generate the second weight through the third neural network model; the first penalty function of the fourth neural network is generated based on the first weight, the second weight, the third feature data matrix, and the second feature data matrix; the fourth neural network model is configured to concatenate the third feature data matrix and the second feature data matrix to generate the fourth feature data matrix; the fifth neural network model is configured to use the fourth feature data matrix as input to generate the insulator state detection result. This process can refine the data features obtained at each level and fully combine the correlation between the data features at each level, thereby achieving effective fusion of data features and ultimately improving the accuracy of the final insulator state detection.

[0252] exist Figure 3 Based on the scheme shown, in one possible implementation, the first weight includes the loss function of the second neural network model, expressed as:

[0253] a1 = 1 - L1

[0254]

[0255] In the formula, a1 represents the first weight, X1 represents the third feature data matrix, and Y1 represents the first output;

[0256] The second weight includes the loss function of the third neural network model, expressed as:

[0257] a1 = 1 - L1

[0258]

[0259] In the formula, a2 represents the second weight, X2 represents the second feature data matrix, and Y2 represents the second output;

[0260] The expression for the first penalty function is:

[0261]

[0262] In the formula, L represents the first penalty function.

[0263] Optionally, based on any of the aforementioned schemes, in one possible implementation, the fourth neural network model can be a deep reinforcement learning model.

[0264] Furthermore, in this scheme, when the fourth neural network model is a deep reinforcement learning model, the concatenation method of the third feature data matrix and the second feature data matrix can be linear weighted fusion.

[0265] In this scheme, the specific forms of the second and third neural network models are not limited. For example, the second and third neural network models can each be an LSTM model layer.

[0266] In this scheme, the specific form of the fifth neural network model is not limited. For example, the fifth neural network model can be a shallow neural network layer.

[0267] exist Figure 3 Based on the scheme shown, in one feasible implementation, the second penalty function is:

[0268]

[0269] In the formula, f obj Let N represent the fitness function, and N represent the number of samples. Represents the actual value. This represents the predicted value.

[0270] For example, in this embodiment, based on any of the aforementioned solutions, in one possible implementation, the second neural network model and the third neural network model are set as the first LSTM model layer and the second LSTM model layer, respectively.

[0271] The fifth neural network model is set as a shallow neural network layer.

[0272] The function expressions for the first LSTM model layer and the second LSTM model layer are defined as follows:

[0273] f t =σ(W f ·[h t-1 ,x t ]+b f )

[0274] i t =σ(W i ·[h t-1 ,x t ]+b i )

[0275] o t =σ(W o ·[h t-1 ,x t ]+b o )

[0276]

[0277]

[0278] h t =o t ·tanh(C t )

[0279] In the formula, σ represents the first activation function, x t W represents the input features at the current time. f Let W represent the first parameter training matrix. i W represents the training matrix with the second parameter. o W represents the training matrix with third parameter. c Let b represent the fourth parameter training matrix. f b represents the first training bias term. i b represents the second training bias term. o b represents the third training bias term. c h represents the fourth training bias term. t-1 f represents the output of the hidden layer at the previous moment. t Represents the forget gate, i t Indicates the input gate value. Represents candidate values, o t Indicates the output gate value, h t This represents the hidden layer state.

[0280] Furthermore, in this scheme, the function expression of the shallow neural network layer is defined as follows:

[0281] P Fi =f[Wf g (wX+b)+B]

[0282] In the formula, fg Let w represent the second activation function, b represent the first weight parameter, b represent the first bias parameter, W represent the second weight parameter, B represent the second bias parameter, and P represent the second activation function. Fi This indicates the insulator condition detection result.

[0283] In this scheme, based on the above function expression, the corresponding hyperparameter matrix is:

[0284] θ=[W f W i W o W c ,b f ,b i ,b o ,b c ,w,b,W,B).

[0285] The hyperparameter matrix is ​​determined using the cross-tab algorithm, and the fitness function of the cross-tab algorithm, i.e., the second penalty function, is defined as follows:

[0286]

[0287] In the formula, f obj Let N represent the fitness function, and N represent the number of samples. Represents the actual value. Indicates the predicted value;

[0288] In the cross-linking algorithm, parent particles θ(i) and θ(j) are randomly selected to cross each other in the nth dimension. During the cross-linking process, random mutation is introduced, and the expression for random mutation is:

[0289] M hc (i,n)=r1·θ(i,n)+(1-r1)·θ(j,n)+c1·θ(i,n)-θ(j,n)+r3(θ a -θ(i,n))-r4(θ′(i,n)-θ(i,n)

[0290] M hc (j,n)=r2·θ(j,n)+(1-r2)·θ(i,n)+c2·θ(j,n)-θ(i,n)+r3(θ a -θ(j,n))-r4(θ′(j,n)-θ(j,n)

[0291] In the formula, r1, r2, r3, and r4 represent random numbers in the range of 0 to 1, c1 and c2 represent random numbers in the range of -1 to 1, θ(i,n) is the nth dimension of particle θ(i), and θ(j,n) is the nth dimension of particle θ(j). a Let θ' represent the globally optimal particle, and θ'(i,n) and θ'(j,n) represent random particles.

[0292] During particle iteration, a multi-penalty function update mechanism is adopted. If the offspring particle is inferior to the updated particle, a penalty is imposed; if the offspring particle is superior to the updated particle, a reward is given. The specific representation process is as follows:

[0293] f(MS hc (i,n))>f(X(i,n)),reward-=1

[0294] f(MS hc (i,n)) <f(X(i,n)),reward+=1

[0295] In the formula, f(MS) hc (i,n)) represents the value of the second penalty function using the update particle as a parameter, and f(X(i,n)) represents the value of the second penalty function using the offspring particle as a parameter;

[0296] Based on the set number of iterations, the particle swarm iterates using a penalty mechanism. The iteration stops when the preset number of iterations is reached, and the required hyperparameter matrix θ is finally obtained.

[0297] Example 3

[0298] Figure 4 A schematic diagram of an electronic device 10 that can be used to implement embodiments of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (e.g., helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0299] like Figure 4As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12 or a random access memory (RAM) 13, communicatively connected to the at least one processor 11. The memory stores computer programs executable by the at least one processor. The processor 11 can perform various appropriate actions and processes based on the computer program stored in the ROM 12 or loaded from storage unit 18 into the RAM 13. The RAM 13 may also store various programs and data required for the operation of the electronic device 10. The processor 11, ROM 12, and RAM 13 are interconnected via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0300] Multiple components in electronic device 10 are connected to I / O interface 15, including: input unit 16, such as keyboard, mouse, etc.; output unit 17, such as various types of displays, speakers, etc.; storage unit 18, such as disk, optical disk, etc.; and communication unit 19, such as network card, modem, wireless transceiver, etc. Communication unit 19 allows electronic device 10 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0301] Processor 11 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. Processor 11 performs the various methods and processes described above, such as insulator state detection methods.

[0302] In some embodiments, the insulator condition detection method may be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or installed on electronic device 10 via ROM 12 and / or communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of the insulator condition detection method described above may be performed. Alternatively, in other embodiments, processor 11 may be configured to perform the insulator condition detection method by any other suitable means (e.g., by means of firmware).

[0303] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0304] Computer programs used to implement the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0305] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0306] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0307] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or computing systems that include middleware components (e.g., application servers), or computing systems that include frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0308] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.

[0309] Note that the above description is merely a preferred embodiment of the present invention and the technical principles employed. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and various obvious changes, readjustments, and substitutions can be made without departing from the scope of protection of the present invention. Therefore, although the present invention has been described in detail through the above embodiments, the present invention is not limited to the above embodiments, and may include many other equivalent embodiments without departing from the concept of the present invention, the scope of which is determined by the scope of the appended claims.

Claims

1. A method for detecting the condition of an insulator, characterized in that, include: Acquire insulator operating condition data and insulator image data; A first feature data matrix is ​​generated based on the insulator operating condition data, and a second feature data matrix is ​​generated based on the insulator image data. The weight matrix of the first feature data matrix is ​​determined using a first neural network model, and a third feature data matrix is ​​generated based on the first feature data matrix and the weight matrix. The third feature data matrix is ​​used as input to generate the fourth feature data matrix through the second neural network model. The fourth feature data matrix and the second feature data matrix are concatenated to generate the fifth feature data matrix; The fifth feature data matrix is ​​used as input to generate the sixth feature data matrix through the third neural network model. A fourth neural network model is configured to take the sixth feature data matrix as input to generate insulator state detection results; An insulator state prediction model is constructed using the second neural network model, the third neural network model, and the fourth neural network model. The initial hyperparameter matrix of the insulator state prediction model is obtained, and the parameters in the hyperparameter matrix are determined based on the cross-hatching algorithm. Chaotic mapping is used to generate the population used in the cross-hatching algorithm to complete the training of the insulator state prediction model. Using insulator operating condition data and insulator image data as input, the trained insulator condition prediction model is used to generate insulator condition detection results.

2. The insulator condition detection method as described in claim 1, characterized in that, The weight matrix for determining the first feature data matrix using the first neural network model includes: In the formula, Denotes the first nonlinear activation function. This represents the second nonlinear activation function. This represents the first feature data matrix. Denotes the first parameter matrix. Denotes the second parameter matrix. This represents the weight matrix.

3. The insulator condition detection method as described in claim 2, characterized in that, The second neural network model uses a first LSTM model layer, the third neural network model uses a second LSTM model layer, and the fourth neural network model uses a shallow neural network layer; The second neural network model, the third neural network model, and the fourth neural network model constitute an ML-LSTM model; Obtain the hyperparameter matrix of the ML-LSTM model; The parameters in the hyperparameter matrix are determined using a cross-hatching algorithm, wherein chaotic mapping is used to generate the population used in the cross-hatching algorithm.

4. The insulator condition detection method as described in claim 3, characterized in that, The first LSTM model layer and the second LSTM model layer have the same structure; The function expressions for the first LSTM model layer and the second LSTM model layer are as follows: In the formula, This represents the first activation function. This represents the input features at the current moment. Denotes the first parameter training matrix. This represents the training matrix for the second parameter. This represents the training matrix with the third parameter. This represents the fourth parameter, the training matrix. This represents the first training bias term. This indicates the second training bias term. This indicates the third training bias term. This indicates the fourth training bias term. This represents the output of the hidden layer at the previous moment. Represents the Gate of Oblivion Indicates the input gate value. Indicates candidate values, Indicates the output gate value. Indicates the hidden layer state; The function expression for the shallow neural network layer is: In the formula, This represents the second activation function. This represents the first weight parameter. Indicates the first bias parameter. This represents the second weighting parameter. This represents the second bias parameter. This indicates the insulator condition detection results; The hyperparameter matrix is: 。 5. An electronic device, characterized in that, It includes at least one processor and a memory communicatively connected to the at least one processor; The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the insulator condition detection method according to any one of claims 1-4.

6. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that cause a processor to execute the insulator condition detection method according to any one of claims 1-4.

Images