Photovoltaic system DC series arc fault feature extraction method and system
Through the CUSUM-VMD-CNN method, the characteristics of DC series arc faults of photovoltaic systems are detected, and the problem of poor adaptability of feature extraction algorithms in the existing technology is solved, and the accuracy of identification of DC series arc faults of photovoltaic systems is realized, and the safety of the power grid is improved. stability.
Patent Information
- Application Number
- CN202411570712.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-11-06
- Publication Date
- 2024-12-03
AI Technical Summary
The DC series arc fault feature content of photovoltaic systems is small, and the existing feature extraction algorithm is poorly adaptable to different photovoltaic systems, which leads to difficulty in identifying faults and affects the safety and stability of the power grid.
The CUSUM-VMD-CNN method is used to detect mutation points through accumulation and algorithms, and the trend components are removed through singular spectrum analysis, and the variational modal decomposition algorithm is improved to extract signal characteristics, and the arc fault type is identified by using convolutional neural network.
Effectively extract the DC series arc fault characteristics of photovoltaic system, improve the accurate identification of fault types, and enhance the safety and stability of the power grid.
Smart Images

Figure CN119066400A_ABST
Abstract
Description
Technical Field
[0001] The present invention belongs to the field of direct current signal feature extraction, and in particular to a method and system for extracting features of direct current series arc faults in a photovoltaic system. Background Art
[0002] Solar photovoltaic power generation systems have been widely promoted and applied due to their small geographical restrictions and large power generation capacity. However, photovoltaic systems may also experience various faults in actual operation, among which DC series arc faults caused by environmental and aging factors are the most serious. Arc faults will not only lead to a significant decrease in power generation efficiency, but may even cause fires, seriously affecting property and life safety. Therefore, in the context of the continuous increase in photovoltaic installed capacity, timely detection and control of arc faults generated by photovoltaic systems is of great significance to ensure the safe and stable operation of the power grid.
[0003] At present, the feature content of DC series arc fault in photovoltaic system is relatively small, and the feature extraction algorithm has poor adaptability to different photovoltaic systems. Summary of the invention
[0004] In view of the problems that the current photovoltaic system DC series arc fault feature content is relatively small and the feature extraction algorithm has poor adaptability to different photovoltaic systems, the present invention provides a photovoltaic system DC series arc fault feature extraction method and system based on CUSUM-VMD-CNN, so as to effectively extract the characteristics of the photovoltaic system DC series arc fault.
[0005] To this end, the present invention adopts the following technical solution.
[0006] In a first aspect, the present invention provides a method for extracting features of a DC series arc fault in a photovoltaic system, comprising: Step 1: Collect DC bus current data, use Cumulative Sum (CUSUM) algorithm to detect mutation points, and extract current sequence; Step 2, using singular spectrum analysis (SSA) to remove the trend component in the current series; Step 3: Improve the variational mode decomposition (VMD) algorithm to decompose the current sequence after removing the trend component and extract the signal characteristics under different mode components; Step 4: Based on CNN (convolutional neural network), the signal characteristics under the modal component are used to identify the type of DC series arc fault in the photovoltaic system.
[0007] Furthermore, the step 1 specifically includes: Step 1.1, using the cumulative sum algorithm to detect mutation points; The specific steps of the cumulative sum algorithm are as follows: 1) Define the mean calculation windowW 0 Event Detection Window W a Two consecutive sliding windows, both with a window length of n The average currents of the two windows are and ; 2) Introduction k Statistics at time , calculate its cumulative sum: (1) In the formula, z k for k The current value at the moment; Random noise introduced by the outside world; represents the initial value of the statistic; When there is no fault, the DC bus current does not fluctuate significantly. f k Fluctuates around zero; if an arc fault occurs, the DC bus current signal will suddenly change. hour, Indicates the normal fluctuation value of current; f k Start accumulating from zero, and introduce the cumulative sum threshold G , when 0< f k < G , define the time delay factor d ,when f k+d > G When the current signal changes suddenly, the moment of occurrence is determined by t = k - d get; Step 1.2: Extract the current sequence before and after the mutation .
[0008] Furthermore, step 2 specifically includes: The singular spectrum analysis is used to remove the trend component of the current sequence and reduce the difference of current signals of different photovoltaic systems, specifically: 21) Embedding: The current sequence I Map to L × K The trajectory matrix of X , as shown below: (2) In the formula, L = m / 2, K= m - L +1; 22) Matrix X Perform singular value decomposition as shown below: (3) In the formula, express L OK L An orthogonal matrix of columns; represents an orthogonal matrix with K rows and K columns; Rewrite the above formula as a column vector multiplied by a row vector: (4) In the formula, r =rank( X ), which is a matrix X rank; 23) Reconstruction: For the submatrix in equation (4) Perform anti-diagonal mean calculation and restore each sub-matrix to a length of m The sequence of; let the submatrix The elements in are x ij ,but The reconstruction sequence is expressed as: (5) in, (6) In the formula, when L < K hour, ;otherwise ; ; ; 24) Remove current sequence I The trend term in is calculated based on formula (6): X 1 The reconstruction sequence is: (7) The current sequence is detrended using equation (7) to obtain the current sequence after the trend component is removed: , as shown below: (8) In the formula, .
[0009] Furthermore, step 3 specifically includes: Step 3.1: Set the penalty factor of the parameter to be optimized in the variational mode decomposition algorithm and the range of decomposition levels R, the parameters in the variational mode decomposition algorithm are optimized using the grasshopper optimization algorithm, and the envelope entropy is selected as the evaluation function; Step 3.2: Input the current series after removing the trend , the variational mode decomposition algorithm is used to remove the trend of the current series Decompose, set the number of iterations, and find the penalty factor And the optimal value of the number of decomposition levels R; Step 3.3, obtain and save the optimal parameters of the variational mode decomposition algorithm, and calculate the optimal parameters of the current series after removing the trend. Decompose and extract signal features under different modal components.
[0010] Furthermore, step 4 specifically includes: The convolutional neural network is used to train and identify the signal characteristic waveforms of each modal component obtained in step 3. The training label is the type of DC series arc fault in the photovoltaic system: 41) For the signal characteristics of each modal component obtained in step 3, draw characteristic waveform diagrams under different modal components; 42) Set the number of CNN convolutional layers and pooling layers, convert the feature waveform into a matrix form, perform convolution operation with the convolution kernel of the convolution layer, input the calculation result into the activation function, obtain the feature map, and input it into the pooling layer; 43) Finally, the Softmax function is used to identify the type of DC series arc fault in the current photovoltaic system.
[0011] In a second aspect, the present invention provides a photovoltaic system DC series arc fault feature extraction system, comprising: Current sequence extraction unit: collects DC bus current data, uses cumulative sum algorithm to detect mutation points, and extracts current sequences before and after the mutation occurs; Trend component removal unit: Use singular spectrum analysis to remove trend components in the current sequence; Signal feature extraction unit: Improve the variational mode decomposition algorithm, decompose the current sequence after removing the trend component, and extract the signal features under different mode components; Fault type identification unit: Based on CNN, the signal characteristics under the modal component are used to identify the DC series arc fault type of the photovoltaic system.
[0012] The beneficial effects of the present invention are as follows: the present invention can effectively extract the characteristics of a DC series arc fault in a photovoltaic system, thereby accurately identifying the type of a DC series arc fault in the photovoltaic system. BRIEF DESCRIPTION OF THE DRAWINGS
[0013] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.
[0014] Figure 1 It is a flow chart of the method for extracting characteristics of DC series arc fault in photovoltaic system of the present invention; Figure 2 A diagram showing the structure of a photovoltaic system DC series arc fault feature extraction system according to the present invention; Figure 3 This is a simulation diagram of a photovoltaic system built using MATLAB / Simulink in the present invention; Figure 4 The present invention performs modal decomposition on the collected bus current data to draw grayscale characteristic diagrams under different modes; Figure 5 This is a diagram of the series arc fault type identification result (confusion matrix) of the present invention. DETAILED DESCRIPTION
[0015] In order to make the purpose, technical scheme and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention, that is, the embodiments described are only part of the embodiments of the present invention, rather than all of the embodiments. The components of the embodiments of the present invention described and shown in the drawings herein can be arranged and designed in various different configurations.
[0016] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative work are within the scope of protection of the present invention.
[0017] Example 1 This embodiment is a method for extracting characteristics of DC series arc faults in photovoltaic systems. Figure 1 As shown, the steps are as follows: Step 1: Collect DC bus current data, use Cumulative Sum (CUSUM) algorithm to detect mutation points, and extract current sequences before and after the mutation occurs.
[0018] Step 1.1 Use the CUSUM algorithm to remove mutation point detection, specifically: When the photovoltaic system is operating normally, the DC bus current will not fluctuate significantly. However, when a fault occurs, the current will change suddenly, with obvious transient processes. The CUSUM algorithm based on the sliding window can effectively judge the moment when the mutation occurs. In actual situations, due to the volatility of the current and the randomness of the occurrence of series arc faults, the non-parametric CUSUM algorithm can be used as an effective algorithm for detecting the mutation point of the DC bus current of the photovoltaic system. The specific process of the algorithm is as follows: (1) Define the mean calculation window W 0 Event Detection Window W a Two consecutive sliding windows, both with a window length of n The average currents of the two windows are and .
[0019] (2) Introduction k Statistics at time , calculate its cumulative sum: (1) In the formula, z k for k The current value at the moment; Random noise introduced by the outside world; represents the initial value of the statistic; When there is no fault, the DC bus current does not fluctuate significantly. f k Fluctuates around zero; if an arc fault occurs, the DC bus current signal will suddenly change. hour, Indicates the normal fluctuation value of current; f k Start accumulating from zero, and introduce the cumulative sum threshold G , when 0< f k < G , define the time delay factor d ,when f k+d > G When the current signal changes suddenly, the moment of occurrence is determined by t = k - d get; Step 1.2: Extract the current sequence before and after the mutation .
[0020] Step 2: Use singular spectrum analysis (SSA) to remove the trend component in the current series.
[0021] Step 2 specifically includes: The singular spectrum analysis is used to remove the trend component of the current sequence and reduce the difference of current signals of different photovoltaic systems, specifically: 21) Embedding: The current sequence I Map to L × K The trajectory matrix of X , as shown below: (2) In the formula, L = m / 2, K = m - L +1; 22) Matrix X Perform singular value decomposition as shown below: (3) In the formula, express L OK L An orthogonal matrix of columns; represents an orthogonal matrix with K rows and K columns; Rewrite the above formula as a column vector multiplied by a row vector: (4) In the formula, r =rank( X ), which is a matrix X rank; 23) Reconstruction: For the submatrix in equation (4) Perform anti-diagonal mean calculation and restore each sub-matrix to a length of m The sequence of; let the submatrix The elements in are x ij ,but The reconstruction sequence is expressed as: (5) in, (6) In the formula, when L < K hour, ;otherwise ; ; ; 24) Remove current sequence I The trend term in is calculated based on formula (6): X 1 The reconstruction sequence is: (7) The current sequence is detrended using equation (7) to obtain the current sequence after the trend component is removed: , as shown below: (8) In the formula, .
[0022] Step 3. Based on the improved VMD algorithm, the trend-removed current series Perform modal decomposition, the steps include: Step 3.1: Set the penalty factor of the parameter to be optimized in the variational mode decomposition algorithm and the range of decomposition levels R, the parameters in the variational mode decomposition algorithm are optimized using the grasshopper optimization algorithm, and the envelope entropy is selected as the evaluation function; Step 3.2: Input the current series after removing the trend , the variational mode decomposition algorithm is used to remove the trend of the current series Decompose, set the number of iterations, and find the penalty factor And the optimal value of the number of decomposition levels R; Step 3.3, obtain and save the optimal parameters of the variational mode decomposition algorithm, and calculate the optimal parameters of the current series after removing the trend. Decompose and extract signal features under different modal components.
[0023] Step 4: Based on CNN, the signal characteristics under the modal component are used to identify the type of DC series arc fault in the photovoltaic system.
[0024] The convolutional neural network is used to train and identify the signal characteristic waveforms of each modal component obtained in step 3. The training label is the type of DC series arc fault in the photovoltaic system: 41) For the signal characteristics of each modal component obtained in step 3, draw characteristic waveform diagrams under different modal components; 42) Set the number of CNN convolutional layers and pooling layers, convert the feature waveform into a matrix form, perform convolution operation with the convolution kernel of the convolution layer, input the calculation result into the activation function, obtain the feature map, and input it into the pooling layer; 43) Finally, the Softmax function is used to identify the type of DC series arc fault in the current photovoltaic system.
[0025] Example 2 This embodiment provides a photovoltaic system DC series arc fault feature extraction system. Figure 2 As shown, it consists of a current sequence extraction unit, a trend component removal unit, a signal feature extraction unit and a fault type identification unit.
[0026] Current sequence extraction unit: collects DC bus current data, uses cumulative sum algorithm to detect mutation points, and extracts current sequences before and after the mutation occurs; Trend component removal unit: Use singular spectrum analysis to remove trend components in the current sequence; Signal feature extraction unit: Improve the variational mode decomposition algorithm, decompose the current sequence after removing the trend component, and extract the signal features under different mode components; Fault type identification unit: Based on CNN, the signal characteristics under the modal component are used to identify the DC series arc fault type of the photovoltaic system.
[0027] The current sequence extraction unit is specifically used for: Use cumulative sum algorithm to detect mutation points; The specific steps of the cumulative sum algorithm are as follows: 1) Define the mean calculation window W 0 Event Detection Window W a Two consecutive sliding windows, both with a window length of n The average currents of the two windows are and ; 2) Introduction k Statistics at time , calculate its cumulative sum: (1) In the formula, z k for k The current value at the moment; Random noise introduced by the outside world; represents the initial value of the statistic; When there is no fault, the DC bus current does not fluctuate significantly. f k Fluctuates around zero; if an arc fault occurs, the DC bus current signal will suddenly change. hour, Indicates the normal fluctuation value of current; f k Start accumulating from zero, and introduce the cumulative sum threshold G , when 0< f k < G , define the time delay factor d ,when f k+d > G When the current signal changes suddenly, the moment of occurrence is determined byt = k - d get; Extract the current sequence before and after the mutation .
[0028] The trend component removal unit is specifically used for: The singular spectrum analysis is used to remove the trend component of the current sequence and reduce the difference of current signals of different photovoltaic systems, specifically: 21) Embedding: The current sequence I Map to L × K The trajectory matrix of X , as shown below: (2) In the formula, L = m / 2, K = m - L +1; 22) Matrix X Perform singular value decomposition as shown below: (3) In the formula, express L OK L An orthogonal matrix of columns; represents an orthogonal matrix with K rows and K columns; Rewrite the above formula as a column vector multiplied by a row vector: (4) In the formula, r =rank( X ), which is a matrix X rank; 23) Reconstruction: For the submatrix in equation (4) Perform anti-diagonal mean calculation and restore each sub-matrix to a length of m The sequence of; let the submatrix The elements in are x ij ,but The reconstruction sequence is expressed as: (5) in, (6) In the formula, when L < K hour, ;otherwise ; ; ; 24) Remove current sequence I The trend term in is calculated based on formula (6): X 1 The reconstruction sequence is: (7) The current sequence is detrended using equation (7) to obtain the current sequence after the trend component is removed: , as shown below: (8) In the formula, .
[0029] The signal feature extraction unit is specifically used for: Set the penalty factor for the parameter to be optimized in the variational mode decomposition algorithm and the range of decomposition levels R, the parameters in the variational mode decomposition algorithm are optimized using the grasshopper optimization algorithm, and the envelope entropy is selected as the evaluation function; Enter the current series after removing the trend , the variational mode decomposition algorithm is used to remove the trend of the current series Decompose, set the number of iterations, and find the penalty factor And the optimal value of the number of decomposition levels R; Obtain and save the optimal parameters of the variational mode decomposition algorithm for the current series after detrending Decompose and extract signal features under different modal components.
[0030] The fault type identification unit is specifically used for: The convolutional neural network is used to train and identify the signal characteristic waveforms of each modal component obtained by the signal feature extraction unit. The training label is the type of DC series arc fault in the photovoltaic system: 1) For the signal features of each modal component obtained by the signal feature extraction unit, a characteristic waveform diagram under different modal components is drawn; 2) Construct a 7-layer convolutional neural network (CNN), which is convolution layer 1 with a step size of 10; pooling layer 2 with a step size of 2; convolution layer 3 with a step size of 1; pooling layer 2 with a step size of 2; convolution layer 3 with a step size of 1; pooling layer 3 with a step size of 1; fully connected layer. In the present invention, the reLu function is selected as the network activation function, the batch size is selected as 128, and the initial learning rate is set to 0.99.
[0031] 3) After setting the number of CNN convolutional layers and pooling layers, the feature waveform is converted into a matrix form, convolved with the convolution kernel of the convolutional layer, and the calculation result is input into the activation function to obtain the feature map, which is then input into the pooling layer; 4) Finally, the Softmax function is used to identify the type of DC series arc fault in the current photovoltaic system.
[0032] Simulation Example According to the principle of photovoltaic power generation, a photovoltaic system simulation model was built using MATLAB / Simulink. The system consists of a photovoltaic array, a photovoltaic bus, a maximum power point tracking control module and a DC load. Figure 3 In the figure, T represents temperature, Ir represents light intensity, and the photovoltaic array is composed of 9 photovoltaic modules. The MPPT control uses the conductance increment method to achieve the matching of the photovoltaic array and the load impedance to achieve maximum power tracking. Figure 3 As shown in the figure, the fault area is divided. According to the different locations of the series arc fault, the series arc fault is divided into: series arc fault between strings, series arc fault of photovoltaic busbar, and series arc fault of DC load side. The fault locations correspond to Figure 3 Arc model 1, arc model 2, and arc model 3 in the embodiment 1 of the present invention. After dividing the fault area, the Cassie arc model is connected in series to the corresponding fault location, and a large amount of various types of series arc fault data of the photovoltaic system are collected (the sampling frequency is 100KHz). Then, the feature extraction and identification of the series arc fault are realized according to steps 1 to 4 described in embodiment 1 of the present invention. The collected linear bus current data is subjected to modal decomposition, and the characteristic waveforms under different modes are plotted as follows: Figure 4 As shown in the figure, the series arc fault type identification result is as follows: Figure 5 shown. Figure 5 In the figure, 8 indicates the number of times the real fault category is fault 1 (i.e., arc fault in series between strings) and the identified fault category is also fault 1; 1 indicates the number of times the real fault category is fault 1 but the identified fault category is fault 3 (i.e., arc fault in series on the DC load side); 19 indicates the number of times the real fault category is fault 2 (i.e., arc fault in series on the PV busbar) and the identified fault category is also fault 2; 22 indicates the number of times the real fault category is fault 3 and the identified fault category is also fault 3.
[0033] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the present invention.
Claims
1. A method for extracting features of DC series arc faults in photovoltaic systems. It is characterized in that include: Step 1: Collect DC bus current data, use cumulative sum algorithm to detect mutation points, and extract current sequences before and after the mutation occurs; Step 2: using singular spectrum analysis to remove trend components in the current sequence; Step 3: Improve the variational mode decomposition algorithm, decompose the current sequence after removing the trend component, and extract the signal characteristics under different mode components; Step 4: Based on CNN, the signal characteristics under the modal component are used to identify the type of DC series arc fault in the photovoltaic system.
2. A method for extracting characteristics of DC series arc faults in photovoltaic systems according to claim 1, It is characterized in that The step 1 specifically includes: Step 1.1, using the cumulative sum algorithm to detect mutation points; The specific steps of the cumulative sum algorithm are as follows: 1) Define the mean calculation window W 0 With event detection window W a Two consecutive sliding windows, both with a window length of n The average currents of the two windows are and ; 2) Introduction k Statistics at time , calculate its cumulative sum: (1) In the formula, z k for k The current value at the moment; Random noise introduced by the outside world; represents the initial value of the statistic; When there is no fault, the DC bus current does not fluctuate significantly. f k Fluctuates around zero; if an arc fault occurs, the DC bus current signal will suddenly change. hour, Indicates the normal fluctuation value of current; f k Start accumulating from zero, and introduce the cumulative sum threshold at this time G , when 0< f k < G , define the time delay factor d ,when f k+d > G When the current signal changes suddenly, the moment of occurrence is determined by t = k - d get; Step 1.2: Extract the current sequence before and after the mutation .
3. A method for extracting characteristics of DC series arc faults in photovoltaic systems according to claim 2, It is characterized in that Step 2 specifically includes: The singular spectrum analysis is used to remove the trend component of the current sequence and reduce the difference of current signals of different photovoltaic systems, specifically: 21) Embedding: The current sequence I Map to L × K The trajectory matrix of X , as shown below: (2) In the formula, L = m / 2, K = m - L +1; 22) Pair Matrix X Perform singular value decomposition as shown below: (3) In the formula, express L OK L An orthogonal matrix of columns; represents an orthogonal matrix with K rows and K columns; Rewrite the above formula as a column vector multiplied by a row vector: (4) In the formula, r =rank( X ), which is a matrix X rank; 23) Reconstruction: For the submatrix in equation (4) Perform anti-diagonal mean calculation and restore each sub-matrix to a length of m The sequence of; let the submatrix The elements in are x ij ,but The reconstruction sequence is expressed as: (5) in, (6) In the formula, when L < K hour, ;otherwise ; ; ; 24) Remove current sequence I The trend term in is calculated based on formula (6): X 1 The reconstruction sequence is: (7) The current sequence is detrended using equation (7) to obtain the current sequence after the trend component is removed: , as shown below: (8) In the formula, .
4. A method for extracting characteristics of DC series arc faults in photovoltaic systems according to claim 3, It is characterized in that Step 3 specifically includes: Step 3.1: Set the penalty factor of the parameter to be optimized in the variational mode decomposition algorithm and the range of decomposition levels R, the parameters in the variational mode decomposition algorithm are optimized using the grasshopper optimization algorithm, and the envelope entropy is selected as the evaluation function; Step 3.2: Input the current series after removing the trend , the variational mode decomposition algorithm is used to remove the trend of the current series Decompose, set the number of iterations, and find the penalty factor And the optimal value of the number of decomposition levels R; Step 3.3, obtain and save the optimal parameters of the variational mode decomposition algorithm, and calculate the optimal parameters of the current series after removing the trend. Decompose and extract signal features under different modal components.
5. A method for extracting characteristics of DC series arc faults in photovoltaic systems according to claim 1, It is characterized in that Step 4 specifically includes: The convolutional neural network is used to train and identify the signal characteristic waveforms of each modal component obtained in step 3. The training label is the type of DC series arc fault in the photovoltaic system: 41) For the signal characteristics of each modal component obtained in step 3, draw characteristic waveform diagrams under different modal components; 42) Set the number of CNN convolutional layers and pooling layers, convert the feature waveform into a matrix form, perform convolution operation with the convolution kernel of the convolution layer, input the calculation result into the activation function, obtain the feature map, and input it into the pooling layer; 43) Finally, the Softmax function is used to identify the type of DC series arc fault in the current photovoltaic system.
6. A photovoltaic system DC series arc fault feature extraction system, It is characterized in that include: Current sequence extraction unit: collects DC bus current data, uses cumulative sum algorithm to detect mutation points, and extracts current sequences before and after the mutation occurs; Trend component removal unit: Use singular spectrum analysis to remove trend components in the current sequence; Signal feature extraction unit: Improve the variational mode decomposition algorithm, decompose the current sequence after removing the trend component, and extract the signal features under different mode components; Fault type identification unit: Based on CNN, the signal characteristics under the modal component are used to identify the DC series arc fault type of the photovoltaic system.
7. A photovoltaic system DC series arc fault feature extraction system according to claim 6, It is characterized in that The current sequence extraction unit is specifically used for: Use cumulative sum algorithm to detect mutation points; The specific steps of the cumulative sum algorithm are as follows: 1) Define the mean calculation window W 0 With event detection window W a Two consecutive sliding windows, both with a window length of n The average currents of the two windows are and ; 2) Introduction k Statistics at time , calculate its cumulative sum: (1) In the formula, z k for k The current value at the moment; Random noise introduced by the outside world; represents the initial value of the statistic; When there is no fault, the DC bus current does not fluctuate significantly. f k Fluctuates around zero; if an arc fault occurs, the DC bus current signal will suddenly change. hour, Indicates the normal fluctuation value of current; f k Start accumulating from zero, and introduce the cumulative sum threshold at this time G , when 0< f k < G , define the time delay factor d ,when f k+d > G When the current signal changes suddenly, the moment of occurrence is determined by t = k - d get; Extract the current sequence before and after the mutation .
8. A photovoltaic system DC series arc fault feature extraction system according to claim 7, It is characterized in that The trend component removal unit is specifically used for: The singular spectrum analysis is used to remove the trend component of the current sequence and reduce the difference of current signals of different photovoltaic systems, specifically: 21) Embedding: The current sequence I Map to L × K The trajectory matrix of X , as shown below: (2) In the formula, L = m / 2, K = m - L +1; 22) Pair Matrix X Perform singular value decomposition as shown below: (3) In the formula, express L OK L An orthogonal matrix of columns; represents an orthogonal matrix with K rows and K columns; Rewrite the above formula as a column vector multiplied by a row vector: (4) In the formula, r =rank( X ), which is a matrix X rank; 23) Reconstruction: For the submatrix in equation (4) Perform anti-diagonal mean calculation and restore each sub-matrix to a length of m The sequence of; let the submatrix The elements in are x ij ,but The reconstruction sequence is expressed as: (5) in, (6) In the formula, when L < K hour, ;otherwise ; ; ; 24) Remove current sequence I The trend term in is calculated based on formula (6): X 1 The reconstruction sequence is: (7) The current sequence is detrended using equation (7) to obtain the current sequence after the trend component is removed: , as shown below: (8) In the formula, .
9. A photovoltaic system DC series arc fault feature extraction system according to claim 8, It is characterized in that The signal feature extraction unit is specifically used for: Set the penalty factor for the parameter to be optimized in the variational mode decomposition algorithm and the range of decomposition levels R, the parameters in the variational mode decomposition algorithm are optimized using the grasshopper optimization algorithm, and the envelope entropy is selected as the evaluation function; Enter the current series after removing the trend , the variational mode decomposition algorithm is used to remove the trend of the current series Decompose, set the number of iterations, and find the penalty factor And the optimal value of the number of decomposition levels R; Obtain and save the optimal parameters of the variational mode decomposition algorithm for the current series after detrending Decompose and extract signal features under different modal components.
10. A photovoltaic system DC series arc fault feature extraction system according to claim 6, It is characterized in that The fault type identification unit is specifically used for: The convolutional neural network is used to train and identify the signal characteristic waveforms of each modal component obtained by the signal feature extraction unit. The training label is the type of DC series arc fault in the photovoltaic system: 1) Based on the signal characteristics of each modal component obtained by the signal feature extraction unit, draw characteristic waveform diagrams under different modal components; 2) Set the number of CNN convolutional layers and pooling layers, convert the feature waveform into a matrix form, perform convolution operation with the convolution kernel of the convolution layer, input the calculation result into the activation function, obtain the feature map, and input it into the pooling layer; 3) Finally, the Softmax function is used to identify the type of DC series arc fault in the current photovoltaic system.
Citation Information
Patent Citations
Photovoltaic system direct current side arc fault detection apparatus and detection method
CN107294494A
Method and system for testing series arc fault of alternating-current and direct-current power supply circuit
CN118549771A
Cited By
Hydroelectric generating set operation state online monitoring method and device based on big data processing
CN120974205A
Photovoltaic direct current side arc fault identification method and system
CN121831428A
