Direct current arc fault detection method based on synchronous compressive wavelet transform and ARIMA algorithm

Abstract

The application discloses a DC fault arc detection method based on synchronous compression wavelet transform and ARIMA algorithm, utilizes electromagnetic radiation signals of a DC fault arc obtained by a Hilbert antenna, obtains time-frequency information of the electromagnetic signals through synchronous compression wavelet transform, adopts an autoregressive integrated moving average (ARIMA) model to perform fault detection, and if the fault detection result is fault in three continuous time windows, outputs a fault arc cutting signal, so that accurate prediction of the DC fault arc is realized by analyzing the characteristics of the electromagnetic radiation signals. Through synchronous compression wavelet transform, the application mines characteristic information of the electromagnetic radiation signals of the DC fault arc under different current levels, different electrode materials, different antenna measurement distances and angles, and more accurately and quickly cuts off the DC fault arc, and the universality of the fault arc detection method and the safe and stable operation ability of the system are improved.

Description

Technical Field

[0001] This invention belongs to the field of electrical fault detection technology for low-voltage systems, and specifically relates to a method for extracting the time-frequency characteristics of electromagnetic radiation signals generated by DC fault arcs under different conditions using synchronous compressed wavelet transform. Background Technology

[0002] Alternating current (AC) is widely used in various electrical equipment and loads, while direct current (DC) is mainly used for signal control and backup power. With the development of technologies such as solar energy, communication technology, and DC traction, the application of DC power and its related power distribution protection are receiving increasing attention. Due to its advantages such as high efficiency, easy connection, and small size, DC power supplies are widely used in aerospace, data centers, electric vehicles, photovoltaic power generation, and ships. Among various renewable energy sources, solar energy has a very broad application prospect, and the photovoltaic power generation industry has therefore developed rapidly. However, this has also brought some serious problems. In particular, after many photovoltaic power plants have been operating for a long time, the electrical system safety problems caused by the aging of photovoltaic modules have become increasingly apparent. Photovoltaic power generation failures and accidents occur frequently, among which fires caused by arc faults on the DC side of photovoltaic systems are a major cause. Photovoltaic systems have complex structures and numerous connected devices, and the DC side voltage may exceed 1000V. Once an arc fault occurs, surrounding flammable materials or photovoltaic modules may immediately ignite, leading to a fire. Currently, research on arc faults on the DC side of photovoltaic systems is not yet mature, and reliable protection strategies have not yet been formed.

[0003] With the rapid growth of communication traffic, the requirements for safe power supply to communication equipment are increasing. Research results show that using a DC power supply system is the most reliable, safe, and economical power supply solution for communication equipment. As the foundation of the communication system, the communication power supply, while possessing high safety, can have a fatal impact on communication equipment if it malfunctions, with an impact far exceeding that of other equipment. In DC power supply systems for communication, factors such as line damage, aging electronic components, loose connections, or animal bites can all lead to fault arcing. Therefore, studying DC fault arcing in communication equipment is of great significance.

[0004] Arc detection technology is essential for the prevention of DC arc. Although DC arc releases a large amount of energy during combustion, it does not generate overcurrent in the circuit, and DC series arc causes the line current to drop. Current protection measures based on circuit breakers and fuses cannot effectively and comprehensively detect small current arc faults. Compared with AC arc, DC arc has no zero point, and arc is more likely to extinguish at the zero point. Under the same combustion conditions, DC fault arc is more difficult to extinguish than AC fault arc, and is extremely prone to fire accidents. Therefore, in-depth study of the characteristics of DC arc and development of effective arc detection methods are of great significance for timely removal of DC fault arc and enhancing the reliability of DC power supply systems. SUMMARY

[0005] The purpose of the present application is to accurately, reliably and quickly detect DC fault arc by collecting electromagnetic radiation signals, and to provide a DC fault arc detection method based on synchronous compression wavelet transform and ARIMA algorithm.

[0006] To achieve the above purpose, the following technical solutions are adopted in the present application:

[0007] The DC fault arc detection method based on synchronous compression wavelet transform and ARIMA algorithm comprises the following steps:

[0008] 1) A specific type of Hilbert antenna is used to sample the electromagnetic radiation signals of DC fault arc in real time under different experimental measurement conditions according to the sampling frequency fs. After collecting a complete signal analysis length N, go to step 2);

[0009] 2) Analyze and reconstruct the collected electromagnetic radiation signals based on synchronous compression wavelet transform, decompose and reconstruct according to the set type of analysis wavelet and corresponding sampling frequency parameters, and perform feature processing on the data in each time window in the specific frequency band after reconstruction, to finally obtain the time-frequency feature quantity of the specific frequency band electromagnetic radiation signal, and go to step 3);

[0010] 3) Calibrate the DC fault arc feature quantity of step 2), set the time window label of the normal state to 0 and the time window label of the fault state to 1 in the selected feature frequency band. In order to improve the feature discrimination, integrate the fault arc feature data sets of different current levels, different antenna measurement distances, different antenna measurement angles and different electrode materials, and go to step 4);

[0011] 4) Train the autoregressive integrated moving average (ARIMA) model using the fault arc feature data set under the multiple conditions of step 3), and output the prediction after adjusting the parameters for optimization, to obtain the fault arc detection model and output the preliminary detection result of the DC fault arc, and go to step 5);

[0012] 5) According to the preliminary results of step 4), if the algorithm detection results of the continuous k time windows are all fault state 1, it is finally determined that an arc fault occurs, and an arc fault cutting signal is output.

[0013] The preferred value of the sampling frequency fs is 1 MHz, and the value of the sampling point N is 5000-10000.

[0014] The selected antenna type is a fourth-order Hilbert antenna, and the side length of the antenna is 8.0 cm. Different types and side lengths of Hilbert antennas will affect the collected electromagnetic radiation signals. In the experiment, the variables need to be controlled to obtain accurate electromagnetic radiation signal data.

[0015] The collected electromagnetic radiation signals are analyzed and reconstructed using synchronous compression wavelet transform. After reconstruction, the data in each time window of the specific frequency band are processed. The feature processing method is to square and process the data points in each time window. Finally, the time-frequency feature quantity of the specific frequency band electromagnetic radiation signal is obtained.

[0016] In order to improve the universality of the algorithm model, the fault arc characteristic data sets of different current levels, different antenna measurement distances, different antenna measurement angles and different electrode materials are integrated. For different antenna measurement angles, the electromagnetic radiation direction parallel to the antenna plate is defined as 0°, and the direction perpendicular to the antenna plate is defined as 90°.

[0017] Raw signal data is obtained through different experiments. Case one selects different current level ranges of 5A-20A. In the experiment, the applied DC voltage is kept at 220V, and the current level is changed by changing the resistor value. Case two selects different antenna measurement distances of 2m-5m. The distance between the antenna and the fault arc point is measured using a tape measure. Case three selects different antenna measurement angles of 0°-90°. Case four selects different electrode materials of aluminum and red copper.

[0018] The synchronous compression wavelet transform is a redistribution of the frequency spectrum obtained by wavelet transform processing of the electromagnetic radiation signal generated by the DC fault arc with a sampling rate fs. The phase of the signal after wavelet transform is not affected by the scale change. The same frequency scale is accumulated, and the coefficients around the same frequency are compressed and concentrated on the frequency. Through a special mapping relationship, the time scale plane is converted into a time-frequency plane, so that the coefficient concentrated frequency spectrum is obtained. The frequency band range of the frequency spectrum obtained by this experimental method is from 0 to half of the sampling frequency fs. Compared with the time-frequency resolution of the wavelet transform result, the energy concentration is higher, and the processed signal has a large distinction between fault state and normal state.

[0019] The synchronous compression wavelet transform algorithm uses the wsst function in MATLAB, which uses morlet wavelet for analytical calculation, and can also use the specified analytical wavelet 'amor' or 'bump' to perform synchronous compression wavelet transform, and the parameters in the algorithm need to input the to-be-tested signal data and the sampling frequency, and the synchronous compression wavelet transform output result is returned in the form of a matrix, reflecting the time-frequency characteristic information, and the synchronous compression frequency f is returned, corresponding to the row of the output matrix.

[0020] The ARIMA model prediction equation with p, d, and q as parameters can be shown as

[0021]

[0022] In the formula, y t is a sample value; and theta j (j=1, 2, …, q) are model parameters;F1 is a characteristic value of the sample;epsilon t is a white noise sequence subject to independent normal distribution. p, d, and q are the order of the model, and the prediction result of the model can be optimized by testing different combinations of p, d, and q, to find the most suitable model parameters. The fault detection model is based on the autoregressive moving average (ARIMA) model, and the performance of the model is optimized by setting the parameters p, d, and q in ARIMA(p, d, q), p represents the "autoregressive part": this parameter describes the lag value of the observation value used in the model. q represents the "moving average part": this parameter describes the lag value of the error term used in the model. d is the order of difference, and the goal of difference is to convert a non-stationary sequence into a stationary sequence. The prediction model can better learn the characteristic parameters of the electromagnetic radiation detection signal, and has strong data processing capability.

[0023] If the fault arc state is output for 3-7 consecutive time windows, it is judged that a direct current fault arc occurs, and the operation of cutting off the fault arc signal is performed.

[0024] The present application has the following beneficial technical effects:

[0025] 1) The method proposes an effective method for detecting direct current fault arcs, which uses Hilbert antennas to collect electromagnetic radiation signals generated by direct current fault arcs with different current sizes, different antenna measurement distances, different antenna measurement angles, and different electrode materials, uses a feature extraction method based on synchronous compression wavelet transform, and explores characteristic signals of direct current fault arcs under different conditions, improves the detection capability of direct current fault currents, ensures the safe and stable operation of the system, and makes the detection method more universal.

[0026] 2) The method uses a fault identification model of autoregressive integrated moving average (ARIMA) model, which has strong learning ability for different condition fault arc data sets, strong fault detection ability and data processing ability, can effectively identify the characteristics of DC fault arc, and the parameter of the ARIMA model is optimized to improve the fault identification accuracy, and the system can run safely and stably under various complex operating conditions. BRIEF DESCRIPTION OF DRAWINGS

[0027] Figure 1 The flow chart of the DC fault arc detection method of the application is shown in the figure.

[0028] Figure 2 The principle block diagram of the application hardware implementation containing the DC fault arc detection device is shown in the figure.

[0029] Figure 3 The column chart of the signal enhancement ratio of the characteristic value of the DC fault arc under different experimental conditions is shown in the figure.

[0030] Figure 4 The effect picture of the electromagnetic radiation signal of the DC fault arc and the characteristic quantity effect of the synchronous compression wavelet transform applied thereto is shown in the figure.

[0031] Figure 5 The effect picture of the electromagnetic radiation signal of the DC fault arc and the system state judgment output signal is shown in the figure. DETAILED DESCRIPTION

[0032] The method of the application will be described in detail below in combination with the drawings and examples.

[0033] In combination Figure 1 The steps of the DC fault arc detection method based on synchronous compression wavelet transform and ARIMA algorithm according to the application are described in detail.

[0034] Step 1: Set the parameter sampling rate to 1MHz, sample the time window of 0.01s of the electromagnetic signal to be measured, select a four-order Hilbert antenna with an edge length of 8.0cm and fix it at the back of the fault arc to verify the universality of the detection method; obtain the original signal data by using different experimental conditions, condition one selects different current levels 18.3A and 5.6A, the DC voltage size remains unchanged at 220V in the experiment, and the current level is changed by changing the resistor value. Condition two selects different antenna measurement distances 2m and 5m, and uses a tape measure to measure the distance between the antenna and the fault arc occurrence point. Condition three selects different antenna measurement angles 0° and 90°, for different antenna measurement angles, the electromagnetic radiation direction parallel to the antenna plate is defined as 0°, and the direction perpendicular to the antenna plate is defined as 90°. Condition four selects different electrode materials aluminum and red copper to conduct DC fault arc detection experiments.

[0035] Step two, the collected fault arc electromagnetic radiation signal is analyzed by using synchronous compression wavelet transform, the characteristic value of the signal frequency band in 79 kHz-86.3 kHz is extracted, the time window length selected is 0.01 s, the sampling frequency is 1 MHz, so there are 10000 data points in each time window, the characteristic value in each time window is processed by sum of squares, and the fault arc characteristics of the frequency band are constructed after processing. The characteristic value of the signal in the frequency band is large during the occurrence of fault arc and in the normal state, which is beneficial to the establishment of subsequent ARIMA algorithm model. In this experiment, morlet wavelet is used for analysis and calculation in the synchronous compression wavelet transform algorithm, and the parameters in the algorithm need to input the signal data to be tested and the signal sampling frequency. The output result of synchronous compression wavelet transform is returned in matrix form, reflecting the time-frequency characteristic information, and the synchronous compression frequency f is returned at the same time, corresponding to the row of the output matrix. Synchronous compression wavelet transform involves redistributing the frequency spectrum graph processed by wavelet transform. The phase of the signal after wavelet transform is not affected by the change of scale, the same frequency scale is accumulated, the coefficients around the same frequency are compressed and concentrated on the frequency, and finally the time scale plane is converted into time-frequency plane through special mapping relationship, so that the concentrated frequency spectrum graph is obtained.

[0036] Step three, the work of labeling the characteristic value is realized by using MATLAB software, the label value of each fault arc signal before the arc burning occurs and after the power is turned off is recorded as 0, and the label value of the time when the fault arc occurs is recorded as 1. The characteristic data and labels of different currents, different antenna measurement distances, different antenna measurement angles and different electrode materials are collected and constructed into fault arc characteristic data set, and the constructed arc characteristic data set is saved in csv file format, which is conducive to data import work in ARIMA model training process.

[0037] Step four, train autoregressive moving average (ARIMA) model, set p, d, q parameters in ARIMA (p, d, q) to optimize the performance of the model, the autoregressive part p describes the lag value of the observation value used in the model, the moving average part q describes the lag value of the error term used in the model, and d is the order of difference. The purpose of difference is to convert non-stationary sequence into stationary sequence. The prediction equation of ARIMA model with p, d and q as parameters can be shown as

[0038]

[0039] In the formula, y t is the sample value; and θ j(j = 1, 2, ..., q) are the model parameters; F1 is the feature value of the sample. p, d, and q are the order of the model. In this experiment, after adjustment and testing, the parameters p, d, and q were set to 1, 1, and 1 respectively. F1 is the feature value of the fault arc electromagnetic radiation signal obtained through synchronous compression transformation and sum of squares processing. t-1 y t-2 …y t-p+1 It is the label value of the p-1 time windows preceding this time window, y t The final output is the ARIMA model's prediction of the fault in the current time window.

[0040] Step 5: After training the model using the data collected from different experimental scenarios, predictions can be made on newly collected data, outputting prediction results of 0 (normal state) or 1 (fault state) for different time windows. In this experiment, when the DC fault arc identification model determines that three consecutive time windows are in a fault state, it ultimately determines that a fault arc has occurred in the system, and reports and clears the fault in a timely manner to avoid dangerous situations.

[0041] Combination Figures 2 to 5 This paper describes the effectiveness of the DC fault arc detection method based on electromagnetic radiation signals of the present invention.

[0042] like Figure 2 The diagram shows the principle block diagram of the application hardware implementation of the DC fault arc detection device. A DC voltage source provides 220V voltage, which is connected to the fault arc generator and a resistor. The current level can be adjusted by changing the resistance value of the resistor. A fourth-order Hilbert antenna is placed directly behind the fault arc generator. The antenna is connected to an oscilloscope to collect electromagnetic radiation signal data.

[0043] like Figure 3 As shown, to illustrate the universality of the synchronous compressed wavelet transform described in this invention for detecting DC fault arcs under different conditions, electromagnetic radiation signals obtained under four different conditions were selected for comparison. At the same time, the feature boost ratio was calculated to compare the effect of synchronous compressed wavelet features under different conditions. The boost ratio formula is shown in (2) below.

[0044]

[0045] In equation (2): F Ⅱ F represents the average value of characteristic quantities during the occurrence of a fault arc; ⅠThe average value of the characteristic quantity in the normal operation process of the system. The larger lifting ratio indicates that the characteristic difference before and after the occurrence of the fault arc is large, which is beneficial to improve the detection accuracy. There are four different experimental conditions, condition one: the current size is 18.3A and 5.6A respectively; condition two: the distance between the antenna and the arc occurrence is 2m and 5m; condition three: the antenna measures the angle of the fault arc 0° and 90°; condition four: the electrode material is aluminum and purple copper respectively.

[0046] As shown in Figure 4 , to illustrate the effect of the synchronous compression wavelet transform feature detection of the direct current fault arc, Figure 4 The electromagnetic radiation signal is combined with the characteristic value, compared with the original signal, the characteristic value more intuitively and effectively shows the occurrence of the direct current fault arc, and improves the detection efficiency of the subsequent algorithm for the fault arc.

[0047] As shown in Figure 5 , the system occurs arc phenomenon at 0.498s, before that the recognition result output by the ARIMA model is normal state, 0.12s after the occurrence of the fault arc, the detection level state changes from 0 to 1, and the output recognition result changes to fault state, when the output result of the continuous three time windows is fault state, it is judged that the fault arc occurs and the signal is cut off. The setting operation avoids the misjudgment phenomenon of the fault arc.

Claims

1. A method for detecting DC arc fault based on synchronous compressive wavelet transform and ARIMA algorithm, characterized in that: It comprises the following steps: 1) using a specific type of Hilbert antenna under different experimental measurement conditions, according to the sampling frequency fs, the electromagnetic radiation signal of direct current fault arc is sampled in real time, after collecting a complete signal analysis length N, turn to step 2); 2) based on the synchronous compression wavelet transform, the collected electromagnetic radiation signal is analyzed and reconstructed, the corresponding sampling frequency parameters are decomposed and reconstructed according to the set type of analysis wavelet, and the data in each time window in the reconstructed specific frequency band are processed, and finally the time-frequency characteristic quantity of the electromagnetic radiation signal in the specific frequency band is obtained, and the step 3) is turned to; 3) the characteristic quantity of step 2) is calibrated, the time window label of normal state is set to 0 and the time window label of fault state is set to 1 in the selected characteristic frequency band, in order to improve the characteristic degree, the fault arc characteristic data set of different current levels, different antenna measurement distances, different antenna measurement angles and different electrode materials is integrated, and the step 4) is turned to; 4) the fault arc characteristic data set under the condition of step 3) is used to train the autoregressive moving average ARIMA model, the prediction output is carried out after adjusting the parameters, the fault arc detection model is obtained, the preliminary detection result of fault arc is output, and the step 5) is turned to; 5) according to the preliminary result of step 4), the detection and judgment are continued, if the algorithm detection result of continuous k time windows is fault state 1, the fault arc is finally judged to occur, and the fault arc cutting signal is output.

2. The method for detecting DC arc fault based on synchronous compression wavelet transform and ARIMA algorithm according to claim 1, characterized in that: The preferred value of sampling frequency fs is 1MHz, and the value of sampling point number N is 5000-10000.

3. The method for detecting DC arc fault based on synchronized compression wavelet transform and ARIMA algorithm according to claim 1, characterized in that: The selected antenna type is four-order Hilbert antenna, and the side length of the antenna is 8.0cm. Different types and side lengths of Hilbert antenna will affect the collected electromagnetic radiation signal. In the experiment, the variable needs to be controlled to obtain accurate electromagnetic radiation signal data.

4. The method according to claim 1 or 2, wherein the method is characterized by: The collected electromagnetic radiation signal is analyzed and reconstructed by using synchronous compression wavelet transform, and the data in each time window in the reconstructed specific frequency band are processed. The feature processing method is to square and process the data points in each time window. Finally, the time-frequency characteristic quantity of the electromagnetic radiation signal in the specific frequency band is obtained.

5. The method for detecting DC arc fault based on synchronized compression wavelet transform and ARIMA algorithm according to claim 3, characterized in that: The universality of the algorithm model is improved, the fault arc characteristic data set of different current levels, different antenna measurement distances, different antenna measurement angles and different electrode materials is integrated, and for different antenna measurement angles, the electromagnetic radiation direction parallel to the antenna plate is defined as 0°, and the electromagnetic radiation direction perpendicular to the antenna plate is defined as 90°.

6. The method for detecting DC arc fault based on synchronized compression wavelet transform and ARIMA algorithm according to claim 5, characterized in that: The original signal data is obtained by different experiments, the current level range is selected as 5A-20A, the direct current voltage size remains unchanged in the experiment, the current level is changed by changing the resistor value, the antenna measurement distance is selected as 2m-5m, the distance between the antenna and the fault arc occurrence point is measured by using a tape measure, the antenna measurement angle is selected as 0°-90°, and the electrode material is selected as aluminum and red copper.

7. The method for detecting DC arc fault based on synchronized compression wavelet transform and ARIMA algorithm according to claim 1, characterized in that: The synchronous compression wavelet transform is a redistribution of the frequency spectrum diagram of the electromagnetic radiation signal generated by the direct current arc after wavelet transform processing, the phase of the signal after wavelet transform is not affected by the scale change, the scales under the same frequency are accumulated, and the coefficients around the same frequency are compressed and concentrated on the frequency; through a special mapping relationship, the time scale plane is converted into a time-frequency plane, thereby obtaining the concentrated coefficient spectrum diagram, and the frequency band range of the spectrum diagram obtained by the experimental method is 0 to half of the sampling frequency fs, and compared with the time-frequency resolution of the wavelet transform result, the energy concentration is higher, and the processed signal has a large distinction between the fault state and the normal state.

8. The method for detecting DC arc fault based on synchronized compression wavelet transform and ARIMA algorithm according to claim 4, characterized in that: The synchronous compression wavelet transform algorithm uses the wsst function in MATLAB, the function uses morlet wavelet for analytical calculation, and the specified analytical wavelets 'amor' or 'bump' can also be used for synchronous compression wavelet transform, the parameters in the algorithm need to input the signal data to be detected and the sampling frequency, the synchronous compression wavelet transform output result is returned in the form of a matrix, reflecting the time-frequency characteristic information, and the synchronous compression frequency f is returned, corresponding to the row of the output matrix.

9. The method for detecting DC arc fault based on synchronized compression wavelet transform and ARIMA algorithm according to claim 1, characterized in that: The ARIMA model prediction equation with p, d and q as parameters can be shown as In the formula, y t is a sample value; and θ j (j = 1, 2, …, q) are model parameters; F1 is a characteristic value of the sample; ε t is a white noise sequence subject to independent normal distribution, p, d, q are orders of the model, and the prediction result of the model can be optimized by testing different p, d, q combinations to find the most suitable model parameters; the fault detection model is based on an autoregressive integrated moving average ARIMA model, the performance of the model is optimized by setting p, d, q parameters in ARIMA (p, d, q), p represents an "autoregressive part": the parameter describes lag values of observations used in the model, q represents a "moving average part": the parameter describes lag values of error terms used in the model, d is an order of difference, and the target of difference is to transform a non-stationary sequence into a stationary sequence, the prediction model can better learn characteristic parameters of electromagnetic radiation detection signals and has strong data processing capability.

10. The method for detecting DC arc fault based on synchronized compression wavelet transform and ARIMA algorithm according to claim 1, characterized in that: The output of the fault arc state can be set to 3-7 time windows, and then the occurrence of the direct current fault arc is judged, and the operation of cutting off the fault arc signal is performed.

Images