A photovoltaic array fault detection method based on improved artificial neural network

Abstract

The application discloses a photovoltaic array fault detection method based on an improved artificial neural network, and the photovoltaic array fault diagnosis model constructed by the photovoltaic array fault detection method comprises an input layer, a gated convolution layer, a flattening layer, an output layer and one or a plurality of fully connected layers connected in sequence; by adding the gated convolution layer and the plurality of fully connected layers, the gradient disappearance problem existing in the current photovoltaic array fault diagnosis is solved; the adaptive Softmax activation function is used as the activation function of the output layer, so that the accuracy and stability of the model are improved; in addition, the photovoltaic array is constructed through simulation simulation, a large number of working state parameters of the photovoltaic array can be acquired according to the module parameters of the photovoltaic array under the condition that the data is less; the application can process complex tasks with large data sets and provide fast prediction after training, and has good generalization ability and extremely high feasibility and application prospect.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of solar photovoltaic fault detection, and particularly relates to a photovoltaic array fault detection method based on an improved artificial neural network. BACKGROUND

[0002] In recent years, the number of installed solar photovoltaic systems has experienced rapid growth among all major energy types. According to the latest data from the International Energy Agency (IEA), the global installed capacity of photovoltaic systems reached 420 GW in 2023, a significant increase of 85% compared to 228 GW in 2022. However, solar photovoltaic systems are installed in complex outdoor environments. As an important component of photovoltaic power stations, the array part has a higher failure rate in the entire photovoltaic system. Adverse external factors make the array more susceptible to various faults such as shadowing, aging, short circuit, and open circuit. These faults have a significant impact on the reliability and performance of solar photovoltaic systems. The annual power loss caused by faults in solar photovoltaic systems is about 18.9%. Therefore, it is of great significance to monitor and supervise solar photovoltaic arrays to detect, classify, and diagnose possible faults to reduce losses and ensure the normal operation of the system.

[0003] In recent years, a large number of scholars and experts have proposed various advanced artificial intelligence algorithms such as decision trees, extreme learning machines, and neural networks for solar photovoltaic array fault diagnosis, improving the stability of photovoltaic systems. Although different machine learning methods and deep learning methods are gradually replacing traditional fault diagnosis methods such as manual detection, each method has both advantages and disadvantages. Siva Ramakrishna Madeti et al. proposed a photovoltaic fault diagnosis model based on the k-nearest neighbor (kNN) algorithm, which has high accuracy. However, as the dimensionality of the data set increases, the increasing computational load of the kNN algorithm will result in slower speed, and the algorithm accuracy is greatly affected by noise. Liu Xingxing et al. used a fault diagnosis model based on extreme gradient boosting (XGBoost) to diagnose photovoltaic array faults, and used grid search and cross-validation to optimize the hyperparameters of the diagnosis model. The resulting model has good generalization and high accuracy. However, XGBoost has high spatial complexity in the training process, requiring large computational resources and potentially causing memory shortages and other issues. Appiah A Y et al. proposed a photovoltaic array fault diagnosis technology based on long short-term memory networks (LSTM), which can automatically extract features from photovoltaic array fault classification raw data for classification. Experiments show that the proposed method has high accuracy, but LSTM has high computational complexity and may have insufficient generalization ability when there is little data.

[0004] As the mainstream method of machine learning, artificial neural network (ANN) is widely used in photovoltaic array DC side fault diagnosis field, which can capture feature information in large data and establish an effective fault diagnosis model. Hussain verified that ANN has more advantages than other methods in photovoltaic fault diagnosis field through t-test technology, because ANN model can determine the difference of power and irradiance level of photovoltaic array, and has the possibility to distinguish the change of these parameters, including shading and non-shading conditions. Cherifa proposed to use ANN for photovoltaic fault diagnosis research on small grid-connected photovoltaic generators, and the experiment showed that this method has high accuracy. Adel compared the fault detection and classification of photovoltaic array by applying ANN, KNN, XGBoost and other machine learning methods, and concluded that ANN is more accurate and has high feasibility in classification. Similarly, Baojie Li et al. used eight fault diagnosis methods for photovoltaic fault diagnosis comparison experiment, and concluded that ANN is the best fault diagnosis model. Compared with other machine learning methods, ANN can be applied to solve nonlinear models, can handle complex tasks with large data sets and provide fast prediction after training, and also has good generalization ability. However, the neurons in the hidden layer of ANN will be compressed when propagating forward, which will cause the weights close to the output layer to approach the optimal solution, while the weights close to the input are still in a random solution state, resulting in the gradient disappearance phenomenon. When the gradient disappearance phenomenon is slight, it will affect the training efficiency and cause the accuracy to decrease, and when it is serious, it will cause the model to be unable to learn, so that the fault diagnosis result is almost random guess. SUMMARY

[0005] In view of the gradient disappearance problem existing in the ANN model in the existing photovoltaic array fault diagnosis field, the application provides a photovoltaic array fault detection method based on an improved artificial neural network, which realizes accurate prediction and diagnosis of solar photovoltaic faults.

[0006] The application provides a photovoltaic array fault detection method based on an improved artificial neural network, which comprises the following steps:

[0007] Step one, obtain electrical parameter data of photovoltaic array in different working states as a sample set, and divide the sample set into training samples and test samples; the working state of the photovoltaic array includes normal state and fault state.

[0008] Step two, constructing a photovoltaic array fault diagnosis model; the photovoltaic array fault diagnosis model comprises an input layer, a gated convolution layer, a flattening layer, an output layer, and one or a plurality of fully connected layers connected in sequence; the gated convolution layer comprises a feature extraction unit and a gated selection unit; after the electrical parameters are input into the input layer, the feature extraction unit is used to extract features, and the gated selection unit is used to filter noise; the flattening layer is used to flatten the two-dimensional array output by the gated convolution layer into a one-dimensional array, which is then input into the fully connected layer to complete the classification task, and is input into the output layer.

[0009] Step three, training the photovoltaic array fault diagnosis model using training samples, and taking the model with the highest accuracy during the training process as the optimal photovoltaic array fault diagnosis model.

[0010] Step four, testing the optimal photovoltaic array fault diagnosis model using test samples; if the accuracy of the optimal photovoltaic array fault diagnosis model is lower than the preset value, reconstructing the photovoltaic array fault diagnosis model, training the reconstructed photovoltaic array fault diagnosis model using training samples, and updating the optimal photovoltaic array fault diagnosis model; repeating the above process until the accuracy of the optimal photovoltaic array fault diagnosis model is higher than the preset value.

[0011] Step five, using the optimal photovoltaic array fault diagnosis model to diagnose the fault of the measured solar photovoltaic array according to the electrical parameters of the measured solar photovoltaic array, obtaining the diagnosis result, and processing the photovoltaic array according to the diagnosis result.

[0012] As a preferred, in the step one, the method for obtaining electrical parameter data is as follows: a photovoltaic array simulation model is established, and the electrical parameters of the photovoltaic array under normal state and fault state are simulated, and the electrical parameters under the same working state are labeled with the same state label.

[0013] As a preferred, in the photovoltaic array simulation model, the irradiance range is set to 100 W / m 2 ~1000 W / m 2 , and the change step is 50 W / m 2 ; the temperature range is set to 10℃~48℃, and the change step is 2℃.

[0014] As a preferred, in the step two, the number of fully connected layers in the photovoltaic array fault diagnosis model is two.

[0015] As a preferred, the fault state includes single-row fault and parallel fault; the single-row fault includes short circuit, partial open circuit, aging, shadow shielding and dirt shielding; the parallel fault includes open circuit shadow fault and open circuit dirt fault.

[0016] As a preferred, the electrical parameters include irradiance, temperature, open circuit voltage, short circuit current, working voltage and working current.

[0017] As preferred, the activation function in the output layer adopts an adaptive Softmax activation function; and the activation function in the full connection layer adopts a Relu activation function.

[0018] As preferred, the method for reconstructing the photovoltaic array fault diagnosis model is to reset the number of filters and neurons and the size of the convolution kernel.

[0019] As preferred, the output of each neuron in the full connection layer is

[0020]

[0021] wherein, bl is the bias of the i-th neuron in the l-th layer; wl is the weight learned between the j-th neuron in the (l-1)-th layer and the i-th neuron in the l-th layer; Nl is the number of neurons in the l-th layer; Nl is the number of neurons in the l-th layer; wl is the weight learned between the j-th neuron in the (l-1)-th layer and the i-th neuron in the l-th layer; Nl is the number of neurons in the l-th layer; Nl is the number of neurons in the l-th layer; Nl is the number of neurons in the l-th layer.

[0022] As preferred, the definition of the gated convolution layer is as follows:

[0023]

[0024]

[0025]

[0026] wherein, Conv1 and Conv2 are respectively the output of the gate selection unit and the feature extraction unit; W g and W f are respectively the convolution filters for calculating the gating value and the feature value; I is the channel value obtained by the convolution kernel at the feature position; and O is the output of the gated convolution layer.

[0027] As preferred, in step three, the training process is as follows:

[0028] a. setting the cross-entropy function as the loss function Loss;

[0029] b. training the initial photovoltaic array fault diagnosis model by using the training samples, until the training times Epochs are greater than the maximum training times Max_Epochs, to obtain the trained photovoltaic array fault diagnosis model; in this embodiment, the maximum training times Epochs is 100.

[0030] c. using the training samples ω train ​The error error and the accuracy accuracy of the verification set calculation model are calculated, and the network parameters of each level are saved.

[0031] d. Repeat the above process until the iteration times Iterator_Times is greater than the set maximum iteration times T.

[0032] e. Compare the photovoltaic array fault diagnosis models obtained in multiple iterations, and select the model with the highest accuracy as the optimal photovoltaic array fault diagnosis model.

[0033] The present application has the beneficial effects of:

[0034] 1. The present application adds a gating convolution layer and multiple fully connected layers in the photovoltaic array fault diagnosis model, solving the gradient disappearance problem existing in the existing photovoltaic array fault diagnosis; at the same time, the present application uses an adaptive Softmax activation function as the activation function of the output layer, improving the accuracy and stability of the model.

[0035] 2. The present application can obtain a large number of working state parameters of the photovoltaic array according to the module parameters of the photovoltaic array under the condition of less data through simulation; at the same time, the present application can process complex tasks with large data sets and provide fast prediction after training, and has good generalization ability and high feasibility and application prospect. BRIEF DESCRIPTION OF DRAWINGS

[0036] Figure 1 The present application is a whole flow chart.

[0037] Figure 2 The present application is a photovoltaic array simulation model schematic diagram.

[0038] Figure 3 The present application is a photovoltaic array fault diagnosis model structure schematic diagram.

[0039] Figure 4 The present application is a confusion matrix schematic diagram of XGBoost, KNN, ANN and LSTM; wherein, (a) is a confusion matrix schematic diagram of the present application; (b) is a confusion matrix schematic diagram of XGBoost; (c) is a confusion matrix schematic diagram of KNN; (d) is a confusion matrix schematic diagram of ANN; (e) is a confusion matrix schematic diagram of LSTM. DETAILED DESCRIPTION

[0040] The present application will be further described below in conjunction with the drawings.

[0041] As shown in the drawings, a photovoltaic array fault detection method based on an improved artificial neural network includes the following steps: Figure 1

[0042] ​Step one, as shown in Figure 2 , a 4x3 array of photovoltaic array simulation model is established in MATLAB / Simulink software by selecting a photovoltaic module of model lSoltech lSTH-350-WH; the module parameters of the photovoltaic array are set as: the maximum power point voltage is 43V, the maximum power point current is 8.13A, the open circuit voltage is 51.5V, and the short circuit current is 9.4A. In order to cover a variety of operating environments, the input irradiance range is set to 100W / m 2 ~1000W / m 2 by the constant module in MATLAB / Simulink software, with a change step of 50W / m 2 ; the temperature range is set to 10℃~48℃, with a change step of 2℃. The actual single-row fault and parallel fault of the photovoltaic array are simulated by the established photovoltaic array simulation model; the single-row fault includes short circuit, partial open circuit, aging, shadow shading and dirt shading; the parallel fault includes open circuit shadow fault and open circuit dirt fault. In the simulation experiment, multiple groups of data are obtained through Goto, From and Scope modules, each group of data including irradiance, temperature, open circuit voltage, short circuit current, working voltage and working current, and each group of data is labeled, and a complete data set is formed by multiple groups of different experiments. The multiple groups of labeled data are used as a sample set ω, and the data in the sample set ω is shown in Table 1.

[0043] Table 1: Running state, number and label of simulation data set

[0044] Running state Description Number Label Normal No fault operation 400 0 Short circuit Two levels, respectively set one block, two blocks of photovoltaic module short circuit, each 400 groups of data 800 1 Part open circuit Two levels, respectively set one branch open circuit and two branch open circuit, each 400 groups of data 800 2 Aging The same branch is sequentially connected to three different levels of resistance, and 400 groups of data are obtained respectively 1200 3 Shadow shielding Reduce the irradiance of the entire photovoltaic array 400 4 Dirt shielding Two levels of dirt shielding, reduce the irradiance of one block and two blocks of photovoltaic modules of one branch, each 400 groups of data 800 5 Open circuit shadow Open circuit and shadow fault 400 6 Open circuit dirt Open circuit and dirt fault 400 7

[0045] Due to the large difference between different types of data values in the sample set ω, the model fault diagnosis accuracy and generalization ability are reduced, etc. In order to better train the improved artificial neural network model, the above data set is normalized, and the normalization formula is as follows:

[0046]

[0047] Where X is the normalized data; x is the original data of the sample set ω; x max and x min are the maximum and minimum values of the original data, respectively.

[0048] The sample set ω is divided into categories with relatively independent features using one-hot encoding, which completes the fault classification. The processed sample set ω is divided into training sample ω train and test sample ω test in the proportion of 7:3, and the number of iterations of cross-validation is set to 10.

[0049] Step two, constructing a photovoltaic array fault diagnosis model

[0050] As shown in Figure 3 , the photovoltaic array fault diagnosis model includes an input layer, a gated convolution layer, a Flatten layer, two fully connected layers and an output layer connected in turn. The gated convolution layer includes a feature extraction unit and a gated selection unit, combining the advantages of gated recurrent unit (GRU) and convolutional neural network. The feature extraction unit generates new features from the input data, improving the feature extraction capability; the gated selection unit retains the update gate and reset gate in GRU in order to simplify the model structure and adapt to the characteristics of convolution operation. As a variant of RNN, GRU is simpler in structure than LSTM, making it easier to capture the timing information of the input signal and not prone to gradient disappearance over time. The update gate and reset gate in the gated selection unit can filter the features, selectively retaining useful information while filtering out unimportant or noisy information, and selecting the main features affecting the result for backward transmission, thereby improving the expression ability and generalization performance of the model. The reset process, update process and output propagation process of GRU are as follows:

[0051] r t = σ( W r · [ ℎ t−1 , x t ]+ b r )

[0052] z t = σ( W z · [ ℎ t−1 , x t ]+ b z )

[0053]

[0054] wherein, r t and z t are the outputs of the reset gate and the update gate, respectively, 0 < r t , z t < 1; σ is the sigmoid function; W r and Wz are weight matrices corresponding to the reset gate and the update gate, respectively; h t-1 is the hidden state of the previous time step; x t is the input of the current time step; b r and b z are bias terms of the reset gate and the update gate, respectively. is the candidate hidden state.

[0055] The definition of the gated convolution layer is as follows:

[0056]

[0057]

[0058]

[0059] wherein, Conv1 and Conv2 are the outputs of the gated selection unit and the feature extraction unit, respectively; W g and W f are convolution filters for calculating the gating value and the feature value, respectively; I is the channel value obtained by the convolution kernel at the feature position; O is the output of the gated convolution layer. is the pixel-level product.

[0060] In this embodiment, the size of the convolution kernel in the gated convolution layer is 3, and the number of filters is 16.

[0061] Since the output of the gated convolution layer is a two-dimensional array, but the fully connected layer usually requires a one-dimensional input array, directly using the fully connected layer to process high-dimensional input data will result in a very large number of parameters, causing waste of computing resources and increasing the risk of overfitting, so the dimension reduction process from two-dimensional to one-dimensional is particularly important.

[0062] The two-dimensional array output by the gated convolution layer is mapped to a one-dimensional array through the flattening layer, which greatly reduces the redundant parameters of the network structure. The calculation formula of the flattening layer is as follows:

[0063]

[0064] wherein, x ij is the data in the i-th row and the j-th column of the two-dimensional array; and array is an array.

[0065] The one-dimensional array output by the flattening layer is input into a full connection layer; two full connection layers play a classification role and have stronger classification ability than one full connection layer, and more full connection layers can further improve the classification ability of the model, but also increase the complexity and calculation cost of the model. In order to improve the performance of the model as much as possible, two full connection layers are used in the present application. Each layer of the full connection layer contains a large number of neurons, each neuron is connected to the neurons of the previous layer, receives the output of the neurons of the previous layer, and outputs to the next layer. The full connection layer is essentially a linear conversion from one feature space to another feature vector. For the i-th neuron of the l-th layer, its output z i The expression of (l) is:

[0066]

[0067] wherein, is the bias of the i-th neuron of the l-th layer; is the number of neurons in the l-th layer; is the weight learned between the j-th neuron of the l-1-th layer and the i-th neuron of the l-th layer; is the output of the j-th neuron of the l-1-th layer after the activation function. Each full connection layer is connected with an activation function; since the Relu activation function is nonlinear, there is no gradient explosion problem compared with other activation functions. And using a nonlinear activation function, the neural network can not only perform linear transformation, but also process and represent the nonlinear relationship in the data, so the Relu activation function is selected as the activation function of the full connection layer.

[0068] The activation function of the output layer uses Adaptive Softmax; Adaptive Softmax has sufficient universality, and selecting Adaptive Softmax as the classifier of the improved ANN fault diagnosis model can effectively reduce the increase of calculation time caused by the large size of the data set without affecting the accuracy

[0069] Step three, using the training sample ω train The photovoltaic array fault diagnosis model is trained; during the training process, the error between the expected utility distribution and the target distribution evaluated by the cross-entropy loss function is used, and then the Adam optimization algorithm is used to optimize the loss function value. The process of optimizing the loss function value by the Adam optimization algorithm is as follows:

[0070] 3-1. Set the cross-entropy function as the loss function Loss;

[0071]

[0072] ​​​3-2.Using the training sample ω train The initial photovoltaic array fault diagnosis model is trained until the training times Epochs are greater than the maximum training times Max_Epochs, and the trained photovoltaic array fault diagnosis model is obtained; in the embodiment, the maximum training times Epochs is 100.

[0073] 3-3.Using the training sample ω train The error error and the accuracy accuracy of the model are calculated by the verification set, and the network parameters at each level are saved.

[0074] 3-4.The above process is repeated until the iteration times Iterator_Times are greater than the set maximum iteration times T.

[0075] 3-5.The photovoltaic array fault diagnosis models obtained in multiple iterations are compared, and the model with the highest accuracy is selected as the optimal photovoltaic array fault diagnosis model.

[0076] Step four, input the test sample ω test into the optimal photovoltaic array fault diagnosis model, output the fault diagnosis result of the test sample, and verify the effectiveness of the model established in step two; if the accuracy is lower than the preset value, the number of filters and neurons and the size of the convolution kernel are reset, and the updated photovoltaic array fault diagnosis model is retrained, and the optimal photovoltaic array fault diagnosis model is updated; the above process is repeated until the accuracy of the optimal photovoltaic array fault diagnosis model is higher than the preset value.

[0077] Step five, model evaluation

[0078] The model of the application and XGBoost, KNN, ANN and LSTM are tested respectively, and the results output by the expected utility module are taken as the diagnosis basis for fault diagnosis. The effect of the final model training confusion matrix is shown in Figure 4 ; the accuracy of the five models is shown in Table 2. From Table 2 and Figure 4 , it can be seen that the single fault diagnosis accuracy and the overall fault diagnosis accuracy of the application are better than those of the other three models, and the diagnosis accuracy of the model can be as high as 98.6%, and the model is established.

[0079] Table 2 Accuracy of five diagnostic models of data set

[0080] Model name The present application XGBoost KNN ANN LSTM Accuracy rate 98.6% 88.4% 88.0% 96.0% 92.0%

[0081] Step six, input the parameters of the measured solar photovoltaic array into the photovoltaic array fault diagnosis model, and diagnose the measured solar photovoltaic fault.

Claims

1. A photovoltaic array fault detection method based on an improved artificial neural network, characterized in that: Includes the following steps: Step 1: Obtain electrical parameter data of the photovoltaic array under different operating conditions as a sample set, and divide the sample set into training samples and test samples; the operating conditions of the photovoltaic array include normal state and fault state; In step one, the method for obtaining electrical parameter data is as follows: establish a photovoltaic array simulation model, simulate the electrical parameters of the photovoltaic array under normal and fault conditions, and assign the same status label to the electrical parameters under the same working conditions; Step 2: Construct a photovoltaic array fault diagnosis model. The photovoltaic array fault diagnosis model includes an input layer, a gated convolutional layer, a flattening layer, an output layer, and two fully connected layers. The gated convolutional layer includes a feature extraction unit and a gate selection unit. After the electrical parameters are input into the input layer, features are extracted by the feature extraction unit, and noise is filtered by the gate selection unit. The flattening layer flattens the two-dimensional array output by the gated convolutional layer into a one-dimensional array, which is then input into the fully connected layer to complete the classification task and then into the output layer. The gated convolutional layer is defined as follows: ； ； ； Wherein, Conv1 and Conv2 are the outputs of the gating selection unit and the feature extraction unit, respectively; W g and W f These are the convolutional filters that calculate the gate value and the feature value, respectively; I is the channel value obtained by the convolutional kernel at the feature location; O is the output of the gated convolutional layer; The product is at the pixel level; The gated convolutional layer has a kernel size of 3 and a filter count of 16. The flattening layer is used to map the two-dimensional array output by the gated convolutional layer to a one-dimensional array. The formula for calculating the flattening layer is: ； Where, x ij This represents the data in the i-th row and j-th column of a two-dimensional array; array is an array. The output z of each neuron in the fully connected layer i (l) is: ； in, This is the bias of the i-th neuron in the l-th layer; For the first The number of neurons in the layer; For the first The weights learned between the j-th neuron in layer j and the i-th neuron in layer l; For the first The output of the j-th neuron in the layer after passing through the activation function; Step 3: Train the photovoltaic array fault diagnosis model using training samples, and use the model with the highest accuracy during training as the optimal photovoltaic array fault diagnosis model. Step 4: Test the optimal photovoltaic array fault diagnosis model using test samples. If the accuracy of the optimal photovoltaic array fault diagnosis model is lower than the preset value, reconstruct the photovoltaic array fault diagnosis model and train the reconstructed photovoltaic array fault diagnosis model using training samples to update the optimal photovoltaic array fault diagnosis model. Repeat the above process until the accuracy of the optimal photovoltaic array fault diagnosis model is higher than the preset value. Step 5: Use the optimal photovoltaic array fault diagnosis model to perform fault diagnosis based on the electrical parameters of the tested solar photovoltaic array and obtain the diagnosis results.

2. The photovoltaic array fault detection method based on an improved artificial neural network according to claim 1, characterized in that: In the photovoltaic array simulation model, the irradiance range is set to 100W / m. 2 ~1000W / m 2 The change step size is 50W / m 2 The temperature range is set to 10℃~48℃, with a variation step of 2℃.

3. The photovoltaic array fault detection method based on an improved artificial neural network according to claim 1, characterized in that: In step two, the photovoltaic array fault diagnosis model has two fully connected layers.

4. The photovoltaic array fault detection method based on an improved artificial neural network according to claim 1, characterized in that: The fault states include single-line faults and parallel faults; single-line faults include short circuit, partial open circuit, aging, shadowing, and dirt occlusion; parallel faults include open-circuit shadowing faults and open-circuit dirt faults.

5. The photovoltaic array fault detection method based on an improved artificial neural network according to claim 1, characterized in that: The electrical parameters include irradiance, temperature, open-circuit voltage, short-circuit current, operating voltage, and operating current.

6. The photovoltaic array fault detection method based on an improved artificial neural network according to claim 1, characterized in that: The activation function in the output layer is the adaptive Softmax activation function; the activation function in the fully connected layer is the ReLU activation function.

7. The photovoltaic array fault detection method based on an improved artificial neural network according to claim 1, characterized in that: The method for reconstructing the photovoltaic array fault diagnosis model is as follows: reset the number of filters and neurons, as well as the size of the convolution kernel.

8. The photovoltaic array fault detection method based on an improved artificial neural network according to claim 1, characterized in that: In step three, the training process is as follows: a. Set the cross-entropy function as the loss function (Loss); b. Train the initial photovoltaic array fault diagnosis model using training samples until the number of training epochs exceeds the maximum number of training epochs, thus obtaining the trained photovoltaic array fault diagnosis model. c. Calculate the model's error and accuracy using the validation set extracted from the training samples, and save the network parameters at each level; d. Repeat the above process until the number of iterations Iterator_Times is greater than the set maximum number of iterations T; e. Compare the photovoltaic array fault diagnosis models obtained in multiple iterations, and select the model with the highest accuracy as the optimal photovoltaic array fault diagnosis model.

Images