Photovoltaic direct current arc fault detection method and device

By extracting the time and frequency domain characteristics of the current signal of the photovoltaic power grid-connected system and integrating it, combining Stacking models of multiple machine learning models for detection, the low accuracy of photovoltaic DC arc fault detection and large resource consumption in the existing technology is solved. The problem of this problem achieves more efficient fault detection and model robustness.

CN119001358APending Publication Date: 2024-11-22ZHANJIANG POWER SUPPLY BUREAU OF GUANGDONG POWER GRID CO LTD

Patent Information

Application Number
CN202411131034.4
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-16
Publication Date
2024-11-22

AI Technical Summary

Technical Problem

The existing photovoltaic DC arc fault detection methods have low accuracy, poor generalization capabilities of detection models, poor robust performance, excessive computing resources consumed, and low detection efficiency.

Method used

By obtaining the current signal of the photovoltaic power grid-connected system, extracting the time and frequency domain characteristics, and fusing them into the Stacking model for detection. The Stacking model includes a decision tree, a support vector machine, a K nearest neighbor model and a logistic regression model, using the main Component analysis method performs dimensionality reduction and combines cross-verified grid search method to optimize model parameters.

Benefits of technology

It improves the accuracy and efficiency of arc fault detection, enhances the generalization ability and robustness of the model, reduces the consumption of computing resources, and avoids the occurrence of overfitting.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119001358A_ABST
    Figure CN119001358A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of arc fault detection, and provides a photovoltaic direct current arc fault detection method and device, and the method comprises the steps: obtaining a to-be-detected current signal in a photovoltaic power generation grid-connected system; extracting a time domain feature and a frequency domain feature from the current signal, the time domain feature and the frequency domain feature respectively comprising a plurality of feature quantities; fusing the time domain feature and the frequency domain feature to obtain a fused feature; the fusion features are input into a Stacking model for arc fault detection, a primary classifier of the Stacking model comprises at least one of a decision tree, a support vector machine and a K neighbor model, and a secondary classifier of the Stacking model is a logistic regression model. The arc fault can be accurately detected by integrating multi-aspect current information of the time domain and the frequency domain, the detection efficiency is improved, meanwhile, the over-fitting phenomenon of the detection model is effectively avoided to a certain extent, and the generalization ability and robustness of the detection model are enhanced.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present disclosure relates to the technical field of arc fault detection, and in particular to a photovoltaic DC arc fault detection method and device. Background Art

[0002] There are many series connection points in the photovoltaic modules connected in series in the photovoltaic power generation grid-connected system. If these series connection points become loose or have poor contact due to unreliable installation, collision, or long-term aging, DC arcs may be generated at the poor connection points. DC arcs can generate high-temperature open flames, which can burn the connection terminals, cause photovoltaic power generation system failures, and affect power generation revenue. In severe cases, they may cause serious fires and casualties, which are very harmful. DC arcs are divided into series arcs and parallel arcs, of which series arcs are more harmful.

[0003] The accuracy of traditional photovoltaic DC arc fault detection methods is low. For example, although the threshold-based detection method can achieve good accuracy in specific cases, when the environment changes or the system is disturbed, the threshold may fail, resulting in a rapid decrease in accuracy. This method cannot adapt to environmental changes well and has poor flexibility. Some machine learning-based fault detection methods use a single machine learning algorithm, which has limited fault information mined and is sometimes prone to fall into local optimal points, resulting in overfitting. Therefore, when the data set changes, its accuracy will drop rapidly, the model generalization ability is not strong, and the robust performance is poor. In addition, in the existing photovoltaic DC arc fault detection model, the amount of data input for fault detection is too large, resulting in excessive computing resources consumed by the model. For example, some technologies use image recognition neural networks for fault detection, but images are two-dimensional information, which contain too much data and need to be continuously iterated to achieve a high accuracy. Therefore, the computing resources required are large and the detection time is long. Summary of the invention

[0004] The purpose of the embodiments of the present disclosure is to provide a photovoltaic DC arc fault detection method and device, so as to at least solve the technical problems existing in the prior art such as low accuracy of photovoltaic DC arc fault detection methods, weak generalization ability of detection models, poor robust performance, excessive consumption of computing resources, and low detection efficiency.

[0005] In order to solve the above technical problems, the embodiments of the present disclosure adopt the following technical solutions:

[0006] In a first aspect, an embodiment of the present disclosure provides a photovoltaic DC arc fault detection method, comprising:

[0007] Obtain the current signal to be detected in the photovoltaic power generation grid-connected system;

[0008] Extracting time domain features and frequency domain features from the current signal, wherein the time domain features and frequency domain features respectively include a plurality of feature quantities;

[0009] Fusing the time domain features and the frequency domain features to obtain fused features;

[0010] The fusion features are input into a Stacking model for arc fault detection, wherein the primary classifier of the Stacking model includes at least one of a decision tree, a support vector machine, and a K nearest neighbor model, and the secondary classifier of the Stacking model is a logistic regression model.

[0011] In some embodiments, extracting time domain features and frequency domain features from the current signal includes:

[0012] Windowing the current signal to extract time domain waveform features;

[0013] Performing fast Fourier transform on the current signal to extract frequency domain features.

[0014] In some embodiments, the time domain features and the frequency domain features are fused to obtain fused features, including:

[0015] Normalizing each of the feature quantities;

[0016] Merging the normalized feature quantities horizontally to obtain merged feature data;

[0017] The principal component analysis method is used to reduce the dimension of the combined feature data to obtain detection feature data.

[0018] In some embodiments, the training of the Stacking model includes:

[0019] Sampling current data in different operating states of the photovoltaic power generation grid-connected system to obtain a sampled data set, wherein the operating state includes a normal state, an interference state, and a fault state;

[0020] Extract the time domain features and frequency domain features from the sampled data set to construct a feature data set;

[0021] Extracting a training set from the feature data set;

[0022] The training set is input into the Stacking model to train the Stacking model.

[0023] In some embodiments, the method further comprises:

[0024] Optimizing the model parameters of the Stacking model using a grid search method with cross-validation;

[0025] Among them, the optimization parameters of the decision tree include the maximum depth and the maximum number of sub-leaf nodes; the optimization parameters of the support vector machine include the parameter penalty coefficient, the kernel function and the kernel function parameters; the optimization parameters of the K nearest neighbor model include the number of reference "neighbor" label values; the optimization parameters of the logistic regression model include the inverse of the regularization coefficient λ.

[0026] In some embodiments, the method further comprises:

[0027] Extracting a test set from the feature data set;

[0028] Inputting the test set into the Stacking model with optimal parameters to detect arc faults in the test set;

[0029] The Stacking model is evaluated according to the arc fault detection results of the test set.

[0030] In some embodiments, the evaluation indicators of the Stacking model include accuracy, precision, recall and F1 score.

[0031] The calculation formula of the accuracy is:

[0032]

[0033] The calculation formula of the accuracy is:

[0034]

[0035] The calculation formula of the recall rate is:

[0036]

[0037] The calculation formula of the F1 score is:

[0038]

[0039] In the formula, TP means that the actual positive sample is judged as a negative sample, TN means that the actual negative sample is judged as a negative sample, FP means that the actual negative sample is judged as a positive sample, and FN means that the actual negative sample is judged as a positive sample.

[0040] In some embodiments, the characteristic quantities of the time domain features include current peak-to-peak value, current variance, current peak factor, and current waveform factor; the characteristic quantities of the frequency domain features include the energy ratio of each frequency band, the ratio of the standard deviation of the amplitude of the high and low frequency bands, the spectrum average value, and the spectrum standard deviation.

[0041] In some embodiments, sampling current data in different operating states of the photovoltaic power generation grid-connected system to obtain a sampled data set includes:

[0042] Acquire normal current data and fault current data flowing through the detection circuit under different environmental conditions, wherein the environmental conditions include temperature and light;

[0043] The current data flowing through the detection circuit after the environmental condition suddenly changes is used as interference current data, and the interference current data under different sudden environmental conditions is obtained.

[0044] In a second aspect, an embodiment of the present disclosure provides a photovoltaic DC arc fault detection device, comprising:

[0045] An acquisition module, configured to acquire a current signal to be detected in a photovoltaic power generation grid-connected system;

[0046] An extraction module, configured to extract time domain features and frequency domain features from the current signal, wherein the time domain features and frequency domain features respectively include a plurality of feature quantities;

[0047] A fusion module, configured to fuse the time domain features and the frequency domain features to obtain fusion features;

[0048] The fault detection module is configured to input the fusion feature into a Stacking model for arc fault detection, wherein the primary classifier of the Stacking model includes at least one of a decision tree, a support vector machine, and a K nearest neighbor model, and the secondary classifier of the Stacking model is a logistic regression model.

[0049] In a third aspect, an embodiment of the present disclosure provides an electronic device, comprising at least a memory and a processor, wherein the memory stores a computer program, and the processor implements the above-mentioned photovoltaic DC arc fault detection method when executing the computer program on the memory.

[0050] In a fourth aspect, an embodiment of the present disclosure provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the above-mentioned photovoltaic DC arc fault detection method is implemented.

[0051] A photovoltaic DC arc fault detection method and device provided by the embodiments of the present disclosure obtains a current signal to be detected in a photovoltaic power generation grid-connected system; extracts time domain features and frequency domain features from the current signal, wherein the time domain features and frequency domain features respectively include multiple feature quantities; then, the time domain features and frequency domain features are fused to obtain fused features; thereafter, the fused features are input into a Stacking model obtained by fusing multiple machine learning models to perform arc fault detection, which can accurately detect arc faults by integrating various current information in the time domain and frequency domain, and improve detection efficiency. At the same time, it can effectively avoid the overfitting phenomenon of the detection model to a certain extent, and enhance the generalization ability and robustness of the detection model. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the embodiments of the present disclosure or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present disclosure. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0053] Figure 1 is a flow chart of a photovoltaic DC arc fault detection method according to an embodiment of the present disclosure;

[0054] Figure 2 Schematic diagram of the structure of a photovoltaic DC arc fault detection device according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0055] Various aspects and features of the present disclosure are described herein with reference to the drawings.

[0056] It should be understood that various modifications may be made to the embodiments of the present application. Therefore, the above description should not be considered as limiting, but only as an example of an embodiment. Other modifications within the scope and spirit of the present disclosure will occur to those skilled in the art.

[0057] The accompanying drawings, which are incorporated in and constitute a part of the specification, illustrate embodiments of the present disclosure and, together with the general description of the present disclosure given above and the detailed description of the embodiments given below, serve to explain the principles of the present disclosure.

[0058] These and other characteristics of the present disclosure will become apparent from the following description of a preferred form of embodiment given as a non-limiting example, with reference to the attached drawings.

[0059] It should also be understood that, although the present disclosure has been described with reference to some specific examples, those skilled in the art will be able to realize many other equivalent forms of the present disclosure that have the features described in the claims and are therefore within the scope of protection defined thereby.

[0060] The above and other aspects, features and advantages of the present disclosure will become more apparent from the following detailed description when taken in conjunction with the accompanying drawings.

[0061] Specific embodiments of the present disclosure are described hereinafter with reference to the accompanying drawings; however, it should be understood that the embodiments claimed are merely examples of the present disclosure, which may be implemented in a variety of ways. Well-known and / or repeated functions and structures are not described in detail to avoid obscuring the present disclosure with unnecessary or redundant details. Therefore, the specific structural and functional details claimed herein are not intended to be limiting, but merely serve as a basis and representative basis for the claims to teach those skilled in the art to use the present disclosure in a variety of ways with substantially any suitable detailed structure.

[0062] This specification may use the phrases "in one embodiment," "in another embodiment," "in a further embodiment," or "in other embodiments," all of which may refer to one or more of the same or different embodiments according to the present disclosure.

[0063] Embodiment 1

[0064] Figure 1 A flow chart of a photovoltaic DC arc fault detection method according to an embodiment of the present disclosure is shown. Figure 1 As shown, a photovoltaic DC arc fault detection method provided by an embodiment of the present disclosure includes:

[0065] S101: Acquire a current signal to be detected in a photovoltaic power generation grid-connected system.

[0066] In this step, the current flowing through the detection circuit in the photovoltaic power generation grid-connected system can be detected in real time and used as the current signal to be detected.

[0067] S102: extracting time domain features and frequency domain features from the current signal, wherein the time domain features and frequency domain features respectively include a plurality of feature quantities.

[0068] After acquiring the current signal, the time domain and frequency domain feature quantities of the current signal are extracted respectively to detect arc faults by integrating various information.

[0069] In step S102, extracting time domain features from the current signal includes:

[0070] The current signal is windowed to extract time domain waveform features.

[0071] By windowing the current signal, the time domain waveform features can be extracted, and the time domain waveform features can be processed to obtain different feature quantities. At the same time, windowing the current signal can make the time domain signal better meet the periodicity requirements of FFT processing and reduce leakage.

[0072] The characteristic quantities of the time domain features include current peak-to-peak value, current variance, current peak factor, and current waveform factor. The characteristic quantities of the time domain features are extracted in units of a certain time window size (eg, 0.02s).

[0073] The expression of the peak-to-peak current is:

[0074] I pp =max(I(k))-min(I(k))(k∈(1,N)) (1)

[0075] The expression of current variance is:

[0076]

[0077] The expression of current crest factor is:

[0078]

[0079] The expression of current waveform factor is:

[0080]

[0081] In the above formulas (1) to (4), I(k) is the amplitude of the kth sampled current data in a time window, max(I(k)) is the maximum current value in a time window, min(I(k)) is the minimum current value, N is the total number of sampled current data in the current time window, I mean Indicates the average value of current data in the current time window.

[0082] In step S102, extracting frequency domain features from the current signal includes:

[0083] Perform fast Fourier transform on the current signal to extract frequency domain features.

[0084] Performing fast Fourier transform on the current signal can obtain the spectrum of the current signal in the range of 0-500kHz, and then extract the characteristic quantities of the current in the frequency domain. The characteristic quantities of the frequency domain characteristics mainly include: the energy ratio of each frequency band, the ratio of the standard deviation of the amplitude of the high and low frequency bands, the spectrum average, and the spectrum standard deviation.

[0085] The expression of the energy ratio of each frequency band is:

[0086]

[0087] The expression for the ratio of the standard deviation of the high and low frequency amplitudes is:

[0088]

[0089] The expression for the spectrum average is:

[0090]

[0091] The expression of the spectrum standard deviation is:

[0092]

[0093] In the process of frequency domain feature extraction, the time window size is also 0.02s. When solving the energy ratio of each frequency band, the frequency band range of 0-450kHz in the spectrum obtained after FFT transformation of the current data within the time range of 0.02s is divided into 9 different sub-bands with an interval of 50kHz. Therefore, 9 different characteristic values ​​can be obtained, which are named F in order from low to high. ep1 、F ep2 、F ep3 、F ep4 、F ep5 、F ep6 、F ep7 、F ep8 、F ep9 In the process of solving the standard deviation of the amplitude of the high and low frequency bands, the frequency of 50kHz is used as the boundary between the high and low frequency bands.

[0094] In the above formulas (5) to (8), f i It represents the amplitude of the i-th discrete current data in the current frequency band, n represents the number of discrete current data sampled in the frequency band of 50kHz, N represents the number of current data collected in a time window, and the length of the time window is also 0.02s. μ1 represents the mean amplitude of the low frequency band after FFT decomposition, and μ2 represents the mean amplitude of the high frequency band.

[0095] In addition to the peak-to-peak value, variance, mean value, and standard deviation, the characteristic indicators used to characterize the characteristic quantity can also be mean square value, root mean square value (RMS), mean square error (MSE), root mean square error (RMSE), root square amplitude, etc. In a specific implementation, the characteristic quantity of the time domain feature can also be pulse factor, margin factor, kurtosis factor, etc.; the characteristic quantity of the frequency domain feature can also be root mean square frequency, etc.

[0096] S103: Fusing the time domain features and the frequency domain features to obtain fused features.

[0097] Since different feature quantities represent different information, the extracted time domain features and frequency domain features are fused to obtain more comprehensive feature information to judge the arc fault.

[0098] S104: Inputting the fusion feature into a Stacking model to perform arc fault detection.

[0099] In this step, at least one of the machine learning models such as decision tree, support vector machine (SVM), k-nearest neighbor (KNN) is used as the primary classifier, and the logistic regression model is used as the secondary classifier. Multiple machine learning models are integrated to obtain a Stacking model for DC series arc fault detection. The meta-features produced by the primary classifier are used as the input of the secondary classifier. The Stacking model can mine and analyze the fused feature data from multiple aspects based on comprehensive and reliable fusion features to accurately detect arc faults. At the same time, since the final detection and judgment result is obtained by combining multiple machine learning models, it can effectively avoid the occurrence of overfitting to a certain extent and enhance the generalization ability and robustness of the model.

[0100] In order to improve the classification accuracy of the primary classifier, the decision tree, support vector machine and K nearest neighbor model can be selected as the primary classifier at the same time, and the outputs of the above three models are used as the input of the secondary classifier, and are jointly input into the secondary classifier to obtain the final output result of the Stacking model (arc fault detection result).

[0101] The photovoltaic DC arc fault detection method provided by the embodiment of the present disclosure obtains the current signal to be detected in the photovoltaic power generation grid-connected system; extracts time domain features and frequency domain features from the current signal, wherein the time domain features and frequency domain features respectively include multiple feature quantities; then, the time domain features and frequency domain features are fused to obtain fused features; thereafter, the fused features are input into a Stacking model obtained by fusing multiple machine learning models to perform arc fault detection, which can accurately detect arc faults by integrating various current information in the time domain and frequency domain, and improve detection efficiency. At the same time, it can effectively avoid the overfitting phenomenon of the detection model to a certain extent, and enhance the generalization ability and robustness of the detection model.

[0102] In some embodiments, in step S103, the time domain feature and the frequency domain feature are fused to obtain a fused feature, including:

[0103] S1031: performing normalization processing on each of the feature quantities;

[0104] S1032: merging the normalized feature quantities horizontally to obtain merged feature data;

[0105] S1033: Use principal component analysis to reduce the dimension of the combined feature data to obtain detection feature data.

[0106] Since the numerical differences between different feature quantities are too large, each individual feature quantity is normalized, and the numerical range of each feature quantity is set to between [0, 1], and then the normalized feature quantities are horizontally merged to obtain merged feature data.

[0107] In order to reduce the computational burden of the Stacking model while keeping the information content of the feature data unchanged as much as possible to achieve a better detection effect, the principal component analysis (PCA) method is used to perform dimensionality reduction analysis on the merged feature data: the principal component cumulative variance share curve is obtained after the merged feature data is subjected to principal component analysis, and the first N (for example, the first 4) principal components with a cumulative variance share of 95% are extracted as the final detection feature data for arc fault detection. The principal component analysis method reveals the internal structure between multiple variables through a few principal components, which can retain as much information of the original variables as possible, and are independent of each other, thereby ensuring the accuracy of detection and reducing the computational burden of the model. In specific implementation, other methods can also be used for dimensionality reduction, such as the UMAP dimensionality reduction method, which has a faster calculation speed and is suitable for data with large amounts of data.

[0108] As can be seen from the above, in this embodiment, the time domain features and the frequency domain features are normalized, dimensionally reduced, etc., and then the time domain features and the frequency domain features are fused to obtain fused features. In specific implementations, the time domain features and the frequency domain features can also be fused by up-sampling, down-sampling, feature weighted fusion, etc.

[0109] In some embodiments, the training of the Stacking model includes:

[0110] S201: Sampling current data in different operating states of the photovoltaic power generation grid-connected system to obtain a sample data set, wherein the operating state includes a normal state, an interference state, and a fault state;

[0111] S202: extracting the time domain features and frequency domain features from the sample data set to construct a feature data set;

[0112] S203: extracting a training set from the feature data set;

[0113] S204: Input the training set into the Stacking model to train the Stacking model.

[0114] Before performing fault detection on the current signal to be detected, it is necessary to pre-train the Stacking model. Therefore, in this embodiment, the current data under different operating states in the photovoltaic power generation grid-connected system is pre-sampled to obtain a sampling data set, which is used as training data to train the Stacking model. The sampling data set includes normal current data under normal operating conditions of the power grid system, fault current data under fault conditions, and interference current data under interference conditions. Using interference current data to train the model helps to reduce the misjudgment of the model and improve the accuracy of model detection. After obtaining the above-mentioned sampling data set, time domain features and frequency domain features can be extracted from the sampling data set to obtain a feature data set, and the feature data set can be marked, the label of the fault sample is set to 1, and the label of the non-fault sample is set to 0, and then a part of the marked feature data set is used as a training set to input into the Stacking model for training. The other part can be used as a verification set or a test set. For example, the marked feature data set can be divided into a training set and a test set in a ratio of 7:3.

[0115] The extraction of time domain features and frequency domain features refers to the above step S102 and will not be repeated here. When extracting time domain features and frequency domain features from the sampled data set, it is also necessary to reduce the dimension of the extracted feature data, and mark the fault state data and non-fault state data after the dimension reduction to form the above feature data set. Through the dimension reduction process, the computational burden and training time of the model can be greatly reduced while basically keeping the fault information unchanged, thereby speeding up the training process of the model.

[0116] In some embodiments, in step S201, current data in different operating states of the photovoltaic power generation grid-connected system are sampled to obtain a sample data set, including:

[0117] S2011: Acquire normal current data and fault current data flowing through the detection circuit under different environmental conditions, wherein the environmental conditions include temperature and light;

[0118] S2012: Using the current data flowing through the detection circuit after the environmental condition suddenly changes as interference current data, and acquiring the interference current data under different sudden environmental conditions.

[0119] When acquiring the sampled data, at least one of the two environmental conditions of temperature and illumination can be changed to acquire the normal current data and fault current data flowing through the detection circuit under different environmental conditions. For example, under the condition that the temperature remains unchanged, the normal current data and fault current data under different illumination conditions can be collected, so as to obtain a Stacking model for photovoltaic DC arc fault detection according to the influence of illumination on the current; or, under the condition that the illumination remains unchanged, the normal current data and fault current data under different temperature conditions can be collected, so as to obtain a Stacking model for photovoltaic DC arc fault detection according to the influence of temperature on the current; or, the temperature and illumination can be changed at the same time, and the Stacking model for photovoltaic DC arc fault detection can be obtained according to the joint influence of temperature and illumination on the current. At the same time, the data flowing through the detection circuit after the environmental condition suddenly changes is used as the interference current data, the temperature and / or illumination before and after the sudden change is changed, and the interference current data under different sudden change environmental conditions are collected, so that the model finally obtained can adapt to more environments and maintain the stability of the model performance. In the specific implementation, the environmental conditions can also include air pressure, etc., and the specific type of environmental conditions is not specifically limited in this disclosure.

[0120] In some embodiments, the method further comprises:

[0121] S205: Optimizing the model parameters of the Stacking model using a grid search method with cross-validation.

[0122] After the Stacking model is trained, the model parameters of the Stacking model are optimized using the grid search method with cross-validation to find the Stacking model with the optimal parameters, thereby effectively improving the fault detection accuracy of the model and improving the detection performance of the model. The model parameters that need to be optimized in the Stacking model are: the maximum depth max_depth and the maximum number of leaf nodes max_leaf_nodes in the decision tree algorithm; the parameter penalty coefficient C1, kernel function kernel and kernel function parameter gama in the SVM algorithm; the number of reference "neighbor" label values ​​n_neighbors in the K nearest neighbor algorithm; the parameters in the logistic regression algorithm are the reciprocal of the regularization coefficient λ C2.

[0123] In some embodiments, the method further comprises:

[0124] S301: extracting a test set from the feature data set;

[0125] S302: inputting the test set into the Stacking model with optimal parameters to detect arc faults in the test set;

[0126] S303: Evaluate the Stacking model according to the arc fault detection result of the test set.

[0127] In this embodiment, after the Stacking model is obtained by training with the training set and the model parameters are optimized, the model with the optimal parameters is evaluated with the test set to evaluate the generalization ability of the final model.

[0128] The evaluation indicators of the Stacking model include accuracy, precision, recall and F1 score.

[0129] The calculation formula of the accuracy is:

[0130]

[0131] The calculation formula of the accuracy is:

[0132]

[0133] The calculation formula of the recall rate is:

[0134]

[0135] The calculation formula of the F1 score is:

[0136]

[0137] In formula (9) to formula (12), TP means that the actual positive sample is judged as a negative sample, TN means that the actual negative sample is judged as a negative sample, FP means that the actual negative sample is judged as a positive sample, and FN means that the actual negative sample is judged as a positive sample.

[0138] Table 1 Evaluation results of the Stacking model

[0139] Accuracy Accuracy Recall F1 score 0.996 1.000 0.991 0.995

[0140] Table 1 shows the evaluation results of the Stacking model with optimal parameters. As shown in Table 1, after evaluating the Stacking model with optimal parameters using the test set, its fault detection accuracy can reach 0.996, which is relatively high, and other evaluation indicators also perform well, and arc faults can be detected well.

[0141] Embodiment 2

[0142] Figure 2 FIG. 4 shows a schematic diagram of the structure of a photovoltaic DC arc fault detection device according to an embodiment of the present disclosure. Figure 2 As shown, the embodiment of the present disclosure provides a photovoltaic DC arc fault detection device, comprising:

[0143] An acquisition module 10 is configured to acquire a current signal to be detected in a photovoltaic power generation grid-connected system;

[0144] An extraction module 20, configured to extract time domain features and frequency domain features from the current signal, wherein the time domain features and frequency domain features respectively include a plurality of feature quantities;

[0145] A fusion module 30 is configured to fuse the time domain features and the frequency domain features to obtain fused features;

[0146] The fault detection module 40 is configured to input the fusion feature into the Stacking model for arc fault detection, wherein the primary classifier of the Stacking model includes at least one of a decision tree, a support vector machine, and a K nearest neighbor model, and the secondary classifier of the Stacking model is a logistic regression model.

[0147] In some embodiments, the extraction module 20 is further configured to:

[0148] Windowing the current signal to extract time domain waveform features;

[0149] Performing fast Fourier transform on the current signal to extract frequency domain features.

[0150] In some embodiments, the fusion module 30 is further configured to:

[0151] Normalizing each of the feature quantities;

[0152] Merging the normalized feature quantities horizontally to obtain merged feature data;

[0153] The principal component analysis method is used to reduce the dimension of the combined feature data to obtain detection feature data.

[0154] In some embodiments, a training module is further included, configured to:

[0155] Sampling current data in different operating states of the photovoltaic power generation grid-connected system to obtain a sampled data set, wherein the operating state includes a normal state, an interference state, and a fault state;

[0156] Extract the time domain features and frequency domain features from the sampled data set to construct a feature data set;

[0157] Extracting a training set from the feature data set;

[0158] The training set is input into the Stacking model to train the Stacking model.

[0159] In some embodiments, the training module is further configured to:

[0160] Optimizing the model parameters of the Stacking model using a grid search method with cross-validation;

[0161] Among them, the optimization parameters of the decision tree include the maximum depth and the maximum number of sub-leaf nodes; the optimization parameters of the support vector machine include the parameter penalty coefficient, the kernel function and the kernel function parameters; the optimization parameters of the K nearest neighbor model include the number of reference "neighbor" label values; the optimization parameters of the logistic regression model include the inverse of the regularization coefficient λ.

[0162] In some embodiments, the training module is further configured to:

[0163] Extracting a test set from the feature data set;

[0164] Inputting the test set into the Stacking model with optimal parameters to detect arc faults in the test set;

[0165] The Stacking model is evaluated according to the arc fault detection results of the test set.

[0166] In some embodiments, the evaluation indicators of the Stacking model include accuracy, precision, recall and F1 score.

[0167] The calculation formula of the accuracy is:

[0168]

[0169] The calculation formula of the accuracy is:

[0170]

[0171] The calculation formula of the recall rate is:

[0172]

[0173] The calculation formula of the F1 score is:

[0174]

[0175] In the formula, TP means that the actual positive sample is judged as a negative sample, TN means that the actual negative sample is judged as a negative sample, FP means that the actual negative sample is judged as a positive sample, and FN means that the actual negative sample is judged as a positive sample.

[0176] In some embodiments, the characteristic quantities of the time domain features include current peak-to-peak value, current variance, current peak factor, and current waveform factor; the characteristic quantities of the frequency domain features include the energy ratio of each frequency band, the ratio of the standard deviation of the amplitude of the high and low frequency bands, the spectrum average value, and the spectrum standard deviation.

[0177] In some embodiments, the training module includes a sampling unit configured to:

[0178] Acquire normal current data and fault current data flowing through the detection circuit under different environmental conditions, wherein the environmental conditions include temperature and light;

[0179] The current data flowing through the detection circuit after the environmental condition suddenly changes is used as interference current data, and the interference current data under different sudden environmental conditions is obtained.

[0180] The photovoltaic DC arc fault detection device provided in the embodiment of the present disclosure corresponds to the photovoltaic DC arc fault detection method in the above-mentioned embodiment. Any optional items in the embodiment of the photovoltaic DC arc fault detection method are also applicable to the embodiment of the photovoltaic DC arc fault detection device, which will not be repeated here.

[0181] Embodiment 3

[0182] An embodiment of the present disclosure also provides an electronic device, comprising at least a memory and a processor, wherein the memory stores a computer program, and the processor implements the above-mentioned photovoltaic DC arc fault detection method when executing the computer program on the memory.

[0183] In some embodiments, the processor executing the computer program may be a processing device including one or more general-purpose processing devices, such as a microprocessor, a central processing unit (CPU), a graphics processing unit (GPU), etc. More specifically, the processor may be a complex instruction set computing (CISC) microprocessor, a reduced instruction set computing (RISC) microprocessor, a very long instruction word (VLIW) microprocessor, a processor running other instruction sets, or a processor running a combination of instruction sets. The processor may also be one or more special-purpose processing devices, such as an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA), a digital signal processor (DSP), a system on a chip (SoC), etc.

[0184] The memory may be a read-only memory (ROM), a random access memory (RAM), a phase-change random access memory (PRAM), a static random access memory (SRAM), a dynamic random access memory (DRAM), an electrically erasable programmable read-only memory (EEPROM), other types of random access memory (RAM), a flash disk or other form of flash memory, a cache, a register, a static memory, a compact disk read-only memory (CD-ROM), a digital versatile disk (DVD) or other optical storage, a cassette or other magnetic storage device, or any other possible non-temporary medium used to store information or instructions that can be accessed by a computer device.

[0185] The electronic devices of the embodiments of the present disclosure may include but are not limited to fixed terminal devices such as servers, desktop computers, digital TVs, etc., and mobile terminal devices such as vehicle-mounted devices (such as head-up display devices), handheld devices (such as mobile phones, tablet computers, etc.), wearable devices (such as smart watches, smart bracelets, etc.), etc.

[0186] Embodiment 4

[0187] The embodiment of the present disclosure further provides a computer-readable storage medium, wherein the computer-readable medium stores a computer program, and when the computer program is executed by a processor, the above-mentioned photovoltaic DC arc fault detection method is implemented.

[0188] The computer-readable storage medium of the embodiments of the present disclosure may adopt any combination of one or more computer-readable media. The computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. In the embodiments of the present disclosure, the computer-readable storage medium may be any tangible medium containing or storing a program, which may be used by or in combination with an instruction execution system, device or device, for example, the above-mentioned memory.

[0189] The computer programs of the embodiments of the present disclosure may be organized into one or more computer executable components or modules. Any number and combination of such components or modules may be used to implement the various aspects of the present disclosure. For example, the various aspects of the present disclosure are not limited to the specific computer executable instructions or specific components or modules shown in the drawings and described herein. Other embodiments may include different computer executable instructions or components with more or less functions than those shown and described herein.

[0190] The above description is only a preferred embodiment of the present disclosure and an explanation of the technical principles used. Those skilled in the art should understand that the scope of disclosure involved in the present disclosure is not limited to the technical solutions formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above disclosed concept. For example, the above features are replaced with the technical features with similar functions disclosed in the present disclosure (but not limited to) by each other to form a technical solution.

Claims

1. A photovoltaic DC arc fault detection method, characterized in that: include: Obtain the current signal to be detected in the photovoltaic power generation grid-connected system; Extracting time domain features and frequency domain features from the current signal, wherein the time domain features and frequency domain features respectively include a plurality of feature quantities; Fusing the time domain features and the frequency domain features to obtain fused features; The fusion features are input into a Stacking model for arc fault detection, wherein the primary classifier of the Stacking model includes at least one of a decision tree, a support vector machine, and a K nearest neighbor model, and the secondary classifier of the Stacking model is a logistic regression model.

2. The method according to claim 1, characterized in that Extracting time domain features and frequency domain features from the current signal includes: Windowing the current signal to extract time domain waveform features; Perform fast Fourier transform on the current signal to extract frequency domain features.

3. The method according to claim 2, characterized in that The time domain features and the frequency domain features are fused to obtain fused features, including: Normalizing each of the feature quantities; Merging the normalized feature quantities horizontally to obtain merged feature data; The principal component analysis method is used to reduce the dimension of the combined feature data to obtain detection feature data.

4. The method according to claim 1, characterized in that: The training of the Stacking model includes: Sampling current data in different operating states of the photovoltaic power generation grid-connected system to obtain a sampled data set, wherein the operating state includes a normal state, an interference state, and a fault state; Extract the time domain features and frequency domain features from the sampled data set to construct a feature data set; Extracting a training set from the feature data set; The training set is input into the Stacking model to train the Stacking model.

5. The method according to claim 4, characterized in that The method further comprises: Optimizing the model parameters of the Stacking model using a grid search method with cross-validation; Among them, the optimization parameters of the decision tree include the maximum depth and the maximum number of sub-leaf nodes; the optimization parameters of the support vector machine include the parameter penalty coefficient, the kernel function and the kernel function parameters; the optimization parameters of the K nearest neighbor model include the number of reference "neighbor" label values; the optimization parameters of the logistic regression model include the inverse of the regularization coefficient λ.

6. The method according to claim 5, characterized in that The method further comprises: Extracting a test set from the feature data set; Inputting the test set into the Stacking model with optimal parameters to detect arc faults in the test set; The Stacking model is evaluated according to the arc fault detection results of the test set.

7. The method according to claim 6, characterized in that The evaluation indicators of the Stacking model include accuracy, precision, recall and F1 score. The calculation formula of the accuracy is: The calculation formula of the accuracy is: The calculation formula of the recall rate is: The calculation formula of the F1 score is: In the formula, TP means that the actual positive sample is judged as a negative sample, TN means that the actual negative sample is judged as a negative sample, FP means that the actual negative sample is judged as a positive sample, and FN means that the actual negative sample is judged as a positive sample.

8. The method according to claim 1, characterized in that The characteristic quantities of the time domain features include current peak-to-peak value, current variance, current peak factor, and current waveform factor; the characteristic quantities of the frequency domain features include the energy ratio of each frequency band, the ratio of the standard deviation of the amplitude of the high and low frequency bands, the spectrum average value, and the spectrum standard deviation.

9. The method according to claim 4, characterized in that Sampling current data in different operating states of the photovoltaic power generation grid-connected system to obtain a sample data set includes: Acquire normal current data and fault current data flowing through the detection circuit under different environmental conditions, wherein the environmental conditions include temperature and light; The current data flowing through the detection circuit after the environmental condition suddenly changes is used as interference current data, and the interference current data under different sudden environmental conditions is obtained.

10. A photovoltaic DC arc fault detection device, characterized in that: include: An acquisition module, configured to acquire a current signal to be detected in a photovoltaic power generation grid-connected system; An extraction module, configured to extract time domain features and frequency domain features from the current signal, wherein the time domain features and frequency domain features respectively include a plurality of feature quantities; A fusion module, configured to fuse the time domain features and the frequency domain features to obtain fusion features; The fault detection module is configured to input the fusion feature into a Stacking model for arc fault detection, wherein the primary classifier of the Stacking model includes at least one of a decision tree, a support vector machine, and a K nearest neighbor model, and the secondary classifier of the Stacking model is a logistic regression model.

Citation Information

Patent Citations

  • A multi-classifier fusion method based on PCA dimension reduction

    CN109447099A

  • Data classification method and device based on base classifier, medium and electronic equipment

    CN110097098A

  • Method and device for determining classification result, equipment and storage medium

    CN111104980A

  • DC fault arc detection method and device

    CN115905921A

  • Partial discharge detection method and system applied to field linkage control

    CN117214627A

Cited By

  • Online monitoring and fault diagnosis method and system for laser cutting machine

    CN120244308A

  • A laser cutting machine online monitoring and fault diagnosis method and system

    CN120244308B

  • Power cable fault discharge sound recognition method and system based on multi-characteristic fusion

    CN120544610A

  • Multi-type fault arc edge identification and positioning method

    CN121917907A

  • Multi-type fault arc edge recognition and positioning method

    CN121917907B