Photovoltaic array fault diagnosis method based on composite model soft voting integration

Abstract

The invention provides a photovoltaic array fault diagnosis method based on composite model soft voting integration. The photovoltaic array fault diagnosis method comprises the following steps: acquiring a plurality of P-V curve data when a photovoltaic array normally operates and fails; preprocessing the plurality of P-V curve data, and formulating fault type labels and performing division processing to obtain a training data set and a test data set; constructing a plurality of composite sub-models, integrating the composite sub-models based on a soft voting integration algorithm, determining various hyper-parameters of the integrated diagnosis model based on a global optimization algorithm, and completing the construction of the integrated diagnosis model; inputting the training data set into the integrated diagnosis model for training to obtain a trained integrated diagnosis model; inputting the test data set into the trained integrated diagnosis model to test and evaluate the performance of the integrated diagnosis model; and the photovoltaic array is diagnosed through the integrated diagnosis model, and the fault type of the photovoltaic array is determined.

Description

Technical Field

[0001] The invention relates to the technical field of photovoltaic array fault diagnosis, and in particular to a photovoltaic array fault diagnosis method based on composite model soft voting integration. Background Art

[0002] Photovoltaic power station fault detection is a photovoltaic operation management technology. Solar photovoltaic power stations are green energy equipment that are being developed at home and abroad. Issues such as the optimization, improvement and operation cost of photovoltaic power stations have seriously restricted the development of photovoltaic power generation. Among them, photovoltaic arrays are prone to photovoltaic cell component "fragmentation", "line aging" and "hot spot phenomenon" due to their large footprint and wide distribution. Grid-connected inverters are prone to overvoltage, overcurrent, power tube short circuit and open circuit. These seriously affect the life of photovoltaic cell components and the safe and stable operation of photovoltaic power stations;

[0003] Artificial intelligence is generally used for fault retrieval of photovoltaic arrays. The existing artificial intelligence-based photovoltaic array diagnosis method is: first, collect fault data, process the data, and divide it into training sets and test sets; then, build a machine learning or artificial neural network diagnosis model and use the divided training data set to train the model, optimize the parameters, so that the adjusted hyperparameters of the diagnosis model reach the optimal value to improve the performance of the diagnosis model; finally, use the divided test data set to verify the diagnostic performance of the model; finally, put the best performing diagnosis model into use; however, in the prior art, most of the existing machine learning or artificial neural network diagnosis models use different modules to be connected together in a horizontal series manner. Although this can make up for the defects between models, it will also cause the shortcomings of the models to overlap, affecting the reliability of the diagnosis results. Summary of the invention

[0004] In view of the deficiencies in the prior art, the present invention provides a photovoltaic array fault diagnosis method based on composite model soft voting integration, which solves the problems in the prior art.

[0005] According to an embodiment of the present invention, a photovoltaic array fault diagnosis method based on composite model soft voting integration includes the following steps:

[0006] Obtain some PV curve data when the PV array is operating normally and when a fault occurs;

[0007] Preprocess several PV curve data, formulate fault type labels and divide them into different categories, and obtain training data sets and test data sets;

[0008] Construct multiple composite sub-models, integrate the composite sub-models based on the soft voting integration algorithm, and determine the hyperparameters of the integrated diagnosis model based on the global optimization algorithm to complete the construction of the integrated diagnosis model;

[0009] Inputting the training data set into the integrated diagnosis model for training to obtain a trained integrated diagnosis model;

[0010] Input the test data set into the trained integrated diagnostic model to test and evaluate its performance;

[0011] The photovoltaic array is diagnosed through an integrated diagnostic model to determine the fault type of the photovoltaic array.

[0012] As an embodiment of the present invention, several PV curve data when the photovoltaic array is operating normally and when a fault occurs are obtained, including:

[0013] Through Matlab or Simulink, three photovoltaic strings are connected in parallel. Each photovoltaic string is a photovoltaic array including three photovoltaic modules. Then simulation is performed to obtain several PV curve data when the photovoltaic array operates normally and fails.

[0014] As an embodiment of the present invention, a plurality of PV curve data are preprocessed, and fault type labels and division processes are formulated to obtain a training data set and a test data set, including:

[0015] Based on the linear normalization method, several PV curve data are normalized, and the calculation formula is as follows:

[0016]

[0017] Among them, X is the PV curve data, X min is the minimum value of the PV curve data, X max is the maximum value of the PV curve data, X norm is the normalized data;

[0018] Corresponding fault labels are set for the normalized data, and the data with fault labels are divided according to the preset division rules to obtain training data sets and test data sets; the fault labels include: normal operation, open circuit fault, short circuit fault, aging fault, occlusion fault, occlusion and open circuit concurrent fault, occlusion and aging concurrent fault, and occlusion and short circuit concurrent fault.

[0019] As an embodiment of the present invention, multiple composite sub-models are constructed, the composite sub-models are integrated based on a soft voting integration algorithm, and various hyperparameters of the integrated diagnosis model are determined based on a global optimization algorithm to complete the construction of the integrated diagnosis model, including:

[0020] Construct multiple composite sub-models; wherein the composite sub-models include: CNN-SVM model, CNN-BiGRU model and DBN-ELM model;

[0021] Based on the soft voting integration algorithm, multiple composite sub-models are integrated to obtain an integrated diagnosis model;

[0022] Based on the whale optimization algorithm, various parameters of multiple composite sub-models are optimized and adjusted to determine various parameters of multiple composite sub-models;

[0023] Determine the weight values ​​of multiple composite sub-models in the soft voting integration algorithm and complete the construction of the integrated diagnosis model.

[0024] As an embodiment of the present invention, a test data set is input into a trained integrated diagnostic model to test and evaluate its performance, including:

[0025] Input the test data set into the trained integrated diagnosis model for testing and obtain the test results;

[0026] Based on the test results, the accuracy, precision, recall and F1-Score are calculated respectively. The calculation formula is as follows:

[0027]

[0028] Among them, T P is the number of samples predicted as positive examples in the test results, T N is the number of samples predicted as negative examples in the test results, P is the number of samples in the test results where the true negative examples are predicted as positive examples, F N is the number of samples predicted as negative examples from the true positive examples in the test results; A is the accuracy, R is the recall, P is the precision, and F1 is the F1-Score.

[0029] As an embodiment of the present invention, the output layers of the plurality of composite sub-models all adopt a softmax function so as to process the probability of the fault type of the photovoltaic array being diagnosed.

[0030] As an embodiment of the present invention, in the process of diagnosis through the integrated diagnosis model, the probability of the fault type output by each composite sub-model is weighted averaged according to the determined weights through the soft voting algorithm to obtain the final diagnosis probability, and the one with the highest probability is determined as the sample fault type. The calculation formula is as follows:

[0031]

[0032] E(x))=maxP n (x)

[0033] Among them, K is the number of composite sub-models, k is the serial number of the composite sub-model, x is the data serial number of the entire photovoltaic array to be tested, n is the serial number of the fault type, ω k is the weight value assigned to the kth composite sub-model, P kn (x) is the probability that the kth composite sub-model x is judged as n, P n (x) is the probability that the integrated diagnosis model judges it to be n, and E(x) is the final judgment result of x.

[0034] Compared with the prior art, the present invention has the following beneficial effects: combining the post-integrated diagnosis model, adopting the soft voting integration algorithm, cleverly combining multiple composite sub-models with different characteristics, giving full play to the advantages of each model and using multiple composite sub-models to jointly determine the fault diagnosis results can improve the accuracy and stability of photovoltaic single fault and concurrent fault diagnosis, and provide strong technical support for the efficient operation and maintenance of photovoltaic arrays; at the same time, by optimizing the parameter settings of multiple composite sub-models and the weight distribution settings of soft voting, the model performance is improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 A schematic diagram of a flow chart of an embodiment of the present invention;

[0036] Figure 2 A schematic diagram of another embodiment of the present invention;

[0037] Figure 3 This is an operation flow chart of another embodiment of the present invention. DETAILED DESCRIPTION

[0038] The technical solution of the present invention is further described below in conjunction with the accompanying drawings and embodiments.

[0039] like Figures 1 to 3 As shown, the embodiment of the present invention proposes a photovoltaic array fault diagnosis method based on composite model soft voting integration, comprising the following steps:

[0040] S1: Obtain some PV curve data when the PV array is operating normally and when a fault occurs;

[0041] S2: Preprocess several PV curve data, formulate fault type labels and divide them, and obtain training data sets and test data sets;

[0042] S3: Construct multiple composite sub-models, integrate the composite sub-models based on the soft voting integration algorithm, and determine the hyperparameters of the integrated diagnosis model based on the global optimization algorithm to complete the construction of the integrated diagnosis model;

[0043] S4: inputting the training data set into the integrated diagnosis model for training to obtain a trained integrated diagnosis model;

[0044] S7: Input the test data set into the trained integrated diagnosis model to test and evaluate its performance;

[0045] S6: diagnose the photovoltaic array through the integrated diagnosis model to determine the fault type of the photovoltaic array;

[0046] The working principle of the above technical solution is as follows: in actual use, multiple composite sub-models are integrated by combining the soft voting integration algorithm to obtain an integrated diagnostic model; the parameters of the multiple composite sub-models after combination are determined by combining all optimization algorithms to improve the combination effect;

[0047] The beneficial effects of the above technical solution are as follows: by combining the post-integrated diagnosis model and adopting the soft voting integration algorithm, multiple composite sub-models with different characteristics are cleverly combined, giving full play to the advantages of each model and using multiple composite sub-models to jointly determine the fault diagnosis results, which can improve the accuracy and stability of photovoltaic single fault and concurrent fault diagnosis, and provide strong technical support for the efficient operation and maintenance of photovoltaic arrays; at the same time, the model performance is improved by optimizing the parameter settings of multiple composite sub-models and the weight distribution settings of soft voting.

[0048] In one embodiment, several PV curve data when the photovoltaic array is operating normally and when a fault occurs are obtained, including:

[0049] Through Matlab or Simulink, three photovoltaic strings are connected in parallel, each photovoltaic string is a photovoltaic array including three photovoltaic modules, and then simulation is performed to obtain several PV curve data when the photovoltaic array operates normally and fails.

[0050] The working principle and beneficial effects of the above technical solution are as follows: to obtain the PV curve data of the photovoltaic array during normal operation or failure, Matlab / Simulink is used to build a photovoltaic array with 3 photovoltaic strings in series and parallel, with 3 photovoltaic modules in each string, and then simulate and collect the data; the controlled voltage source uses a ramp signal source to continuously scan the output voltage of the photovoltaic array, and the comprehensiveness of the data is increased by setting the irradiance range to 100-1000W / m2; the temperature range is 0-45℃.

[0051] In one embodiment, a plurality of PV curve data are preprocessed, and fault type labels and division processes are formulated to obtain a training data set and a test data set, including:

[0052] Based on the linear normalization method, several PV curve data are normalized, and the calculation formula is as follows:

[0053]

[0054] Among them, X is the PV curve data, Xmin is the minimum value of the PV curve data, X max is the maximum value of the PV curve data, X norm is the normalized data;

[0055] The normalized data is given corresponding fault labels, and the data with fault labels are divided according to the preset division rules to obtain the training data set and the test data set; the fault labels include: normal operation, open circuit fault, short circuit fault, aging fault, occlusion fault, occlusion and open circuit concurrent fault, occlusion and aging concurrent fault, and occlusion and short circuit concurrent fault;

[0056] The working principle and beneficial effects of the above technical solution are as follows: fault labels are formulated by labeling different types of data from 1 to 8, including normal operation of the photovoltaic array, open circuit fault, short circuit fault, aging fault, shading fault, shading and open circuit fault, shading and aging fault, and shading and short circuit fault; the preset division rule is that 70% of the data is used to train the integrated diagnostic model (integrated data set), and 30% of the data is used to test the performance of the diagnostic model (test data set); at the same time, by using the photovoltaic array PV curve data as the input parameter, the tedious steps of feature extraction are reduced.

[0057] In one embodiment, multiple composite sub-models are constructed, the composite sub-models are integrated based on a soft voting integration algorithm, and various hyperparameters of the integrated diagnosis model are determined based on a global optimization algorithm to complete the construction of the integrated diagnosis model, including:

[0058] Construct multiple composite sub-models; wherein the composite sub-models include: CNN-SVM model, CNN-BiGRU model and DBN-ELM model;

[0059] Based on the soft voting integration algorithm, multiple composite sub-models are integrated to obtain an integrated diagnosis model;

[0060] Based on the whale optimization algorithm, various parameters of multiple composite sub-models are optimized and adjusted to determine various parameters of multiple composite sub-models;

[0061] Determine the weight values ​​of multiple composite sub-models in the soft voting integrated algorithm and complete the construction of the integrated diagnosis model;

[0062] The working principle of the above technical solution is:

[0063] like Figures 2 to 3As shown in the figure, a CNN-SVM model of a composite sub-model is constructed: Convolutional neural network (CNN) can learn feature representation from data, and through the combination of convolutional layer and pooling layer, it can effectively capture the local structure and pattern of input data. The structure of support vector machine (SVM) is relatively simple, mainly including a decision function to replace the softmax classifier of CNN output layer. The combination of the two can better handle the problem of diagnosis classification. The formula is as follows:

[0064]

[0065] in, is the jth feature map of the lth convolutional layer, M j is the set of input feature maps, is the weight, is the bias term and f is the activation function.

[0066]

[0067] Among them, P ij is the maximum pooling output, A ij is the pooling window, d k is the subset within the window.

[0068]

[0069] Among them, α i and is the Lagrange multiplier, G(x i ,x j ) is the kernel function and a is the bias term.

[0070]

[0071] in, is x i The transpose of

[0072] Building a CNN-BiGRU model of a composite sub-model: The bidirectional gated recurrent unit (BiGRU) is composed of a forward gated recurrent unit (GRU) and a backward GRU. It can process both the forward and reverse information of the data and capture the long-term dependencies in the data. Combined with CNN, CNN performs feature extraction and uses these feature vector sequences as the input of the BiGRU. After forward and backward propagation and calculation of hidden states, the final output is the result. The CNN formula is the same as (1), and the BiGRU formula is as follows:

[0073] z t =σ(W z ·[h t-1 ,x t ])

[0074] Among them, z t is the value of the update gate, σ is the sigmoid function, W z is the weight matrix, h t-1 is the hidden state at the previous moment, x t is the current moment input.

[0075] r t =σ(W r ·[h t-1 ,x t ])

[0076] Among them, r t is the value of the reset gate, W r is the weight matrix.

[0077]

[0078] in, is a candidate hidden state, W is a weight matrix, and * denotes element-wise multiplication.

[0079]

[0080] Among them, h t is the current hidden state.

[0081]

[0082] in, When propagating forward, the hidden state of each moment is calculated in chronological order to obtain the forward hidden state sequence; When propagating backward, the hidden state of each moment is calculated in reverse time order to obtain the backward hidden state sequence; H t is the final hidden state;

[0083] Building a DBN-ELM model of a composite sub-model: The deep belief network (DBN) uses stacked restricted Boltzmann machines (RBM) to achieve excellent feature learning and hierarchical representation, can extract complex features from large-scale data, and is suitable for dealing with nonlinear problems; while the extreme learning machine (ELM) is known for its efficient training and inference speed, and has the characteristics of fast convergence and ease of use. The BP neural network classifier in the traditional DBN model is replaced with the ELM network as the classifier. The formula is as follows:

[0084]

[0085] Where: i is the hidden layer unit, j is the visible layer unit, w i is the weight of the n-1th to nth hidden layers, b iis the bias from the n-1th layer to the nth hidden layer, g(x) is the activation function, H n-1 is the output of the n-1th layer, β i is the output weight from the nth hidden layer to the output layer;

[0086] The global optimization algorithm is used to optimize the parameters of the composite sub-model, including: Global optimization is to use the whale optimization algorithm (WOA) to optimize, adjust and determine the parameters of the diagnostic model. The WOA algorithm is a heuristic optimization algorithm based on the hunting behavior of whale groups. It seeks the optimal parameters by imitating the whale group to surround, attack and search for prey. The formula is as follows:

[0087] X(t+1)=X * (t)-A·D

[0088] D=|C·X * (t)-X(t)|

[0089] A=2a·r1-a

[0090] C=2·r2

[0091] Where t is the current iteration number, X(t+1) is the updated position, and X * (t) is the current best position, A and C are coefficient vectors, D is the distance between the whale and the prey, r1 and r2 are random variables between [0,1];

[0092] X(t+1)=D'·e bl ·cos(2πl)+X * (t)

[0093] D'=|X * (t)-X(t)|

[0094] Where X(t+1) is the updated position of the spiral, D' is the distance between the whale and the prey, b is a constant defining the shape of the spiral, and l is a random number in the range [-1, 1].

[0095] X(t+1)=X rand (t)-A·D

[0096] Among them, X(t+1) is the whale searching, selecting a whale at a random position to update the position, X rand is the position of a whale randomly selected from the current population; the optimal solution (model parameters) of the diagnostic model is found through the WOA optimization algorithm;

[0097] The beneficial effect of the above technical solution is: by setting the parameters of multiple composite sub-models and setting the weight distribution of soft voting, the model performance is improved.

[0098] In one embodiment, the test data set is input into the trained integrated diagnostic model to test and evaluate its performance, including:

[0099] Input the test data set into the trained integrated diagnosis model for testing and obtain the test results;

[0100] Based on the test results, the accuracy, precision, recall and F1-Score are calculated respectively. The calculation formula is as follows:

[0101]

[0102] Among them, T P is the number of samples predicted as positive examples in the test results, T N is the number of samples predicted as negative examples in the test results, P is the number of samples in the test results where the true negative examples are predicted as positive examples, F N is the number of samples predicted as negative examples from the true positive examples in the test results; A is the accuracy, R is the recall, P is the precision, and F1 is the F1-Score;

[0103] The working principle and beneficial effects of the above technical solution: by collecting data from the integrated diagnostic model and then comparing them, the integrated diagnostic model is the best in all evaluation indicators; the model comparison table is as follows:

[0104] Table 1 Model comparison table

[0105]

[0106]

[0107] Among them, the samples of true positive examples are the elements whose fault type labels in the test set are the fault types reflected by the PV curve data, and the samples of true negative examples are the elements whose fault type labels in the test set are not the fault types reflected by the PV curve data. The samples predicted as positive examples indicate that the prediction results of the integrated diagnosis model are correct, and the samples predicted as positive examples indicate that the prediction results of the integrated diagnosis model are wrong.

[0108] In one embodiment, the output layers of the plurality of composite sub-models all adopt a softmax function so as to process the probability of the fault type of the photovoltaic array being diagnosed;

[0109] In the process of diagnosis through the integrated diagnosis model, the soft voting algorithm is used to perform weighted average of the probabilities of the fault types output by the composite sub-models according to the determined weights to obtain the final diagnosis probability, and the one with the highest probability is determined as the sample fault type. The calculation formula is as follows:

[0110]

[0111] E(x)=maxP n (x)

[0112] Among them, K is the number of composite sub-models, k is the serial number of the composite sub-model, x is the data serial number of the entire photovoltaic array to be tested, n is the serial number of the fault type, ω k is the weight value assigned to the kth composite sub-model, P kn (x) is the probability that the kth composite sub-model x is judged as n, P n (x) is the probability that the integrated diagnosis model judges it to be n, and E(x) is the final judgment result of x.

[0113] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solution of the present invention can be modified or replaced by equivalents without departing from the purpose and scope of the technical solution of the present invention, which should be included in the scope of the claims of the present invention.

Claims

1. A photovoltaic array fault diagnosis method based on composite model soft voting integration, characterized in that: The following steps are involved: Obtain some PV curve data when the PV array is operating normally and when a fault occurs; Preprocess several PV curve data, formulate fault type labels and divide them into different categories, and obtain training data sets and test data sets; Construct multiple composite sub-models, integrate the composite sub-models based on the soft voting integration algorithm, and determine the hyperparameters of the integrated diagnosis model based on the global optimization algorithm to complete the construction of the integrated diagnosis model; Inputting the training data set into the integrated diagnostic model for training to obtain a trained integrated diagnostic model; Input the test data set into the trained integrated diagnostic model to test and evaluate its performance; The photovoltaic array is diagnosed through an integrated diagnostic model to determine the fault type of the photovoltaic array.

2. A photovoltaic array fault diagnosis method based on composite model soft voting integration as claimed in claim 1, characterized in that: Obtain several PV curve data when the PV array is operating normally or has a fault, including: Through Matlab or Simulink, three photovoltaic strings are connected in parallel. Each photovoltaic string is a photovoltaic array including three photovoltaic modules. Then simulation is performed to obtain several PV curve data when the photovoltaic array operates normally and fails.

3. A photovoltaic array fault diagnosis method based on composite model soft voting integration as claimed in claim 1, characterized in that: Preprocess several PV curve data, formulate fault type labels and divide them into different categories, and obtain training and test data sets, including: Based on the linear normalization method, several PV curve data are normalized, and the calculation formula is as follows: Among them, X is the PV curve data, X min is the minimum value of the PV curve data, X max is the maximum value of the PV curve data, X norm is the normalized data; Corresponding fault labels are set for the normalized data, and the data with fault labels are divided according to the preset division rules to obtain training data sets and test data sets; the fault labels include: normal operation, open circuit fault, short circuit fault, aging fault, occlusion fault, occlusion and open circuit concurrent fault, occlusion and aging concurrent fault, and occlusion and short circuit concurrent fault.

4. A photovoltaic array fault diagnosis method based on composite model soft voting integration as claimed in claim 1, characterized in that: Construct multiple composite sub-models, integrate the composite sub-models based on the soft voting integration algorithm, and determine the hyperparameters of the integrated diagnosis model based on the global optimization algorithm to complete the construction of the integrated diagnosis model, including: Construct multiple composite sub-models; wherein the composite sub-models include: CNN-SVM model, CNN-BiGRU model and DBN-ELM model; Based on the soft voting integration algorithm, multiple composite sub-models are integrated to obtain an integrated diagnosis model; Based on the whale optimization algorithm, various parameters of multiple composite sub-models are optimized and adjusted to determine various parameters of multiple composite sub-models; Determine the weight values ​​of multiple composite sub-models in the soft voting integration algorithm and complete the construction of the integrated diagnosis model.

5. A photovoltaic array fault diagnosis method based on composite model soft voting integration as claimed in claim 1, characterized in that: Input the test data set into the trained integrated diagnostic model to test and evaluate its performance, including: Input the test data set into the trained integrated diagnosis model for testing and obtain the test results; Based on the test results, the accuracy, precision, recall and F1-Score are calculated respectively. The calculation formula is as follows: Among them, T P is the number of samples predicted as positive examples in the test results, T N is the number of samples predicted as negative examples in the test results, P is the number of samples in the test results where the true negative examples are predicted as positive examples, F N is the number of samples predicted as negative examples from the true positive examples in the test results; A is the accuracy, R is the recall, P is the precision, and F1 is the F1-Score.

6. A photovoltaic array fault diagnosis method based on composite model soft voting integration as claimed in claim 1, characterized in that: The output layers of the plurality of composite sub-models all adopt a softmax function so as to process the probability of the fault type of the diagnosed photovoltaic array.

7. A photovoltaic array fault diagnosis method based on composite model soft voting integration as claimed in claim 1, characterized in that: In the process of diagnosis through the integrated diagnosis model, the soft voting algorithm is used to perform weighted average of the probabilities of the fault types output by the composite sub-models according to the determined weights to obtain the final diagnosis probability, and the one with the highest probability is determined as the sample fault type. The calculation formula is as follows: E(x)=maxP n (x) Among them, K is the number of composite sub-models, k is the serial number of the composite sub-model, x is the data serial number of the entire photovoltaic array to be tested, n is the serial number of the fault type, ω k is the weight value assigned to the kth composite sub-model, P kn (x) is the probability that the kth composite sub-model x is judged as n, P n (x) is the probability that the integrated diagnosis model judges it to be n, and E(x) is the final judgment result of x.

Images