An ac series arc fault detection method using voltage information

Abstract

An alternating current series arc fault detection method using voltage information belongs to the field of electrical engineering measurement. Step a, analog band pass filtering processing; Step b, whether the maximum value of the instantaneous characteristic energy in the half wave is greater than or equal to the preset threshold value; Step c, continuously calculating the instantaneous characteristic energy of each half wave; Step d, calculating the corresponding phase of the maximum value of the instantaneous characteristic energy of each half wave; Step e, calculating the mapping proportion of each phase; Step f, constructing the feature vector of the maximum instantaneous characteristic energy phase mapping proportion; Step g, calculating the Euclidean feature distance of the feature vector phase mapping proportion and the reference vector; Step h, judging whether the fault classification basis is met. In the alternating current series arc fault detection method using voltage information, the whole phase domain is divided into multiple phase mapping areas, and a specific phase domain is selected as the phase area of the fault signal mapping, which is not only suitable for resistive load, but also suitable for low power factor load and various nonlinear loads, and has higher practical value.

Description

TECHNICAL FIELD

[0001] The application discloses an alternating current series arc fault detection method using voltage information and belongs to the field of electrical engineering measurement. BACKGROUND

[0002] The series arc fault is caused by loose connection terminals, poor wire contact or damaged conductors caused by the construction process. Such a fault usually has the characteristics of continuous occurrence, and only 20V voltage is needed to make it periodically re-ignite. Since the arc is essentially a voltage breakdown of air ionization discharge phenomenon, the temperature of the arc fault point is extremely high, and the central temperature of the 0.5A arc can reach 2000-3000℃. Therefore, if such a fault persists, it will cause the wire insulation layer to decompose and eventually cause a fire. If there are flammable materials around the arc fault point, the danger is more urgent.

[0003] In the prior art, the detection of series arc faults in the past five years mainly has the following methods:

[0004] 1. Current detection method. When the series arc fault occurs, the fault current is not much different from the normal load current, so it is impossible to detect such a fault by overcurrent detection method. In addition, since the series arc fault is not a line-to-ground fault, it is also impossible to detect it by using residual current detection technology. When the series arc fault occurs, the fault information is contained in the current waveform, so the existing methods are mostly used to extract the fault information in the current for fault detection. When the arc fault occurs, the arc current usually shows the characteristics of current waveform, such as increased high-frequency noise, current zero-hibernation area, high current rise rate, increased current harmonic content, and unsymmetrical positive and negative half-wave. Therefore, a large number of existing research methods at home and abroad extract one or more of the above characteristics in time domain, frequency domain or wavelet domain, and then set a fault detection threshold to achieve detection. In recent years, with the popularization of machine learning technology in various fields, a large number of black box model detection methods using neural networks to fuse the above characteristics have appeared. At present, the arc fault detection products on the market, such as AFCI or AFDD, are also designed based on the above principles.

[0005] In existing technologies, various current detection methods essentially detect fault currents by detecting their zero-wave distortion, high-frequency characteristics, and harmonics. However, in terms of detection effectiveness, since the load current of a nonlinear load can have the same or similar current waveform characteristics as the arc fault current, current detection methods suffer from the inability to select fault features. Furthermore, even when current features can be selected, line parameters influence these features, making it difficult to set detection thresholds. For example, zero-wave characteristics are weakened by the inductive component content of the load and the inductance of the line parameters, further complicating the setting of fault detection thresholds. While neural network-based multi-feature fusion methods can construct fault detection criteria through prior sample training, their detection accuracy is only applicable to the selected training load; their performance under unknown loads or combined loads remains unreliable.

[0006] Meanwhile, the current detection method detects faults in downstream lines by installing a detection device in the distribution box at the entrance of the house (existing AFCI or AFDD are based on this method). Setting up a monitoring point upstream of the fault point has the advantage of being easy to implement in engineering. The load-side voltage detection method detects faults by setting up a monitoring point at the load end. However, due to its high actual application cost, this type of method is difficult to implement in engineering. The voltage detection method upstream of the fault point has the disadvantages of difficulty in setting the total energy amplitude of the half-wave characteristic and being limited by the load power factor.

[0007] 2. Voltage Detection Methods. Compared to current detection methods, voltage fault detection methods offer an alternative approach to fault detection. When an arc fault occurs, the load impedance accounts for the majority of the circuit impedance, thus downstream of the fault point, detection has high sensitivity. The main idea behind voltage detection methods is to detect arc faults by setting up monitoring points on the load side.

[0008] For example: (1) In the technical solution disclosed in the literature [Miao Xiren, Guo Yinting, Tang Jincheng, et al. Load-end arc fault voltage detection and morphological wavelet identification], the fault arc features are extracted by wavelet decomposition of the load-end voltage. (2) In the technical solution disclosed in the literature [Zhang Liping, Miao Xiren, Shi Dunyi. Research on low-voltage arc fault identification method based on EMD and ELM], the load-end fault features are extracted by empirical mode decomposition.

[0009] Both methods essentially utilize the arc initiation and phase break angle of the fault point to detect the waveform distortion of the load-side voltage. In practice, the arc initiation and phase break angle characteristics are affected by the load type and line parameters, making it difficult to set a characteristic threshold. Furthermore, these methods cannot distinguish the load-side voltage distortion caused by the nonlinear voltage drop of the line.

[0010] (3) In the technical solution disclosed in the literature [Gao Hongxin, Guo Fengyi, Tang Aixia, et al. Predicting series fault arcs using load-end voltage], wavelet packets are used to decompose the load-end voltage and the prediction variance of extreme learning machine is used to extract fault arc features. The actual detection effect of this method depends on the type and number of training loads.

[0011] (4) The technical solution disclosed in the literature [Zhao Yuan, Zhang Guanying, Wang Yao, et al. Series Fault Arc Detection Method Based on Load-End Voltage Analysis] proposes an upstream line fault detection method based on the differential mean of adjacent voltage waveforms at the load end. This method assumes that the voltage waveforms of adjacent cycles at the load end have large differences when a fault occurs, and that the fault voltage cycles at the load end can also have high similarity when a fault occurs. Therefore, the accuracy of this method is poor.

[0012] Therefore, most existing voltage detection methods utilize load-side voltage for detection, essentially relying on the arc extinction-initiation process of the arc voltage at the fault point to detect the fault distortion point generated by the load-side voltage. In actual systems, higher harmonics can also cause distortion points in the load-side voltage, and their fault characteristics may be similar to those generated by the arc voltage, thus easily leading to misjudgment. Furthermore, the voltage change caused by the arc voltage's arc extinction-initiation process is related to the load type and line parameters, meaning that the fault characteristics of the fault distortion point can lie in multiple frequency bands. Therefore, these algorithms face the problem of difficulty in selecting the characteristic frequency band in application.

[0013] 3. Besides the approach of setting monitoring points on the load side, the proposed method in the paper "Clearing Series AC Arc Faults and Avoiding False Alarms using Only Voltage Waveforms" (Jonathan C. Kim et al., IEEE Transactions on Power Delivery) suggests a method for detecting series arc faults in downstream lines using voltage information upstream of the fault point. This method uses the voltage peak value as the dividing point between the left and right half-wave voltage windows. It assumes that when a fault occurs, the total energy of the left and right half-waves is roughly equal, the total energy of the half-wave is relatively large, and the total energy of the half-waves of adjacent cycles varies significantly. A comprehensive criterion is constructed for detection: the difference in the total energy of the right half-wave is less than a set threshold, the amplitude of the total energy of the half-wave is greater than a set threshold, and the difference in the total energy of adjacent half-waves is greater than a threshold. However, this method still has the following shortcomings:

[0014] (1) This method uses the peak position of the power supply voltage as the basis for dividing the left and right windows of the half-wave voltage, and further calculates the energy of the left and right windows of the half-wave based on this condition. The condition for this application is that the phase of the arc voltage at the fault point is the same as the phase of the power supply voltage. In reality, the phase of the arc voltage depends on the phase of the current rather than the power supply voltage. Since the power factor of the load is between 0.8 and 1, the peak position of the power supply is not the corresponding half-wave center position of the arc voltage when the load power factor is not 1. When the load power factor is small, the deviation between the peak position of the power supply and the half-wave center position of the arc voltage can reach tens of degrees. At this time, the difference in symmetrical energy between the left and right windows is very large, and it cannot meet the criterion of small difference in symmetrical energy between the left and right windows required by this method. Therefore, this method is only applicable to the case where the load power factor is high or close to 1.

[0015] (2) The criterion of this method uses the amplitude of the half-wave characteristic total energy. However, the amplitude of the half-wave characteristic total energy is closely related to the inductance parameter of the line and the equivalent impedance of the load. Under different application scenarios, the amplitude of the half-wave characteristic total energy during a fault is different, so its detection threshold is also different. Therefore, the method has the disadvantage of making it difficult to set the amplitude of the half-wave characteristic total energy in practical applications. Summary of the Invention

[0016] The technical problem to be solved by the present invention is to overcome the shortcomings of the prior art and provide an AC series arc fault detection method that utilizes voltage information. The entire phase domain is divided into multiple phase mapping regions, and a specific phase domain is selected as the phase region for fault signal mapping. This method fully considers the characteristics of the diversity of load power factors in practical applications. It is not only applicable to resistive loads, but also to low power factor loads and various nonlinear loads. Compared with the existing methods, it has a wider applicability to load types and higher practical value.

[0017] The technical solution adopted by this invention to solve its technical problem is: an AC series arc fault detection method utilizing voltage information, characterized by comprising the following steps:

[0018] Step a: Perform analog bandpass filtering on the voltage signal;

[0019] Step b, using the maximum instantaneous characteristic energy within the half-wave. Whether the limit is exceeded is used as the trigger condition for detection; if the maximum instantaneous characteristic energy within half-wave... If the value is greater than or equal to the preset threshold, proceed to step c; otherwise, return to step a.

[0020] Step c: Calculate the instantaneous characteristic energy of each half-wave within 1 second.

[0021] Step d: Calculate the maximum instantaneous characteristic energy in half-wave units. The corresponding maximum instantaneous characteristic energy corresponds to the phase.

[0022] Step e: Calculate the phase corresponding to the maximum instantaneous characteristic energy of each half-wave within 1 second. The mapping ratio;

[0023] Step f: Construct the phase corresponding to the maximum instantaneous feature energy. The feature vector of the mapping proportion;

[0024] Step g: Calculate the phase corresponding to the maximum instantaneous characteristic energy. The Euclidean feature distance between the feature vector of the mapping proportion and the reference vector;

[0025] Step h: Determine whether the fault classification criteria are met. If they are met, it is determined that an AC series arc fault has occurred. If they are not met, it means that no AC series arc fault has occurred.

[0026] Preferably, when performing step d,

[0027] First, calculate the instantaneous energy within 1 second, using half-wave as the unit. Maximum value: Maximum instantaneous characteristic energy

[0028]

[0029] Then, calculate the maximum instantaneous characteristic energy of each half-wave within 1 second. Corresponding phase: Phase corresponding to the maximum instantaneous characteristic energy

[0030]

[0031] Where θ is the sampled signal u M (n) is the initial phase angle of the current power frequency signal obtained by performing FFT operation, u M (n) represents the voltage u at the upstream monitoring point of the line. M The time-domain discrete representation of (t); P is the maximum instantaneous characteristic energy. The corresponding sampling point number; mod[] 2π This indicates that the result of the operation is moduloed with a period of 2π, and N is the number of sampling points per cycle.

[0032] Preferably, during step c, the instantaneous characteristic energy of each bandpass filter within 1 second is calculated in half-wave units.

[0033]

[0034] Where i is the half-wave index (i = 1 to 100), and N is the number of sampling points per wave cycle. This is a half-wave sequence after bandpass filtering.

[0035] Preferably, when performing step e, the interval [0, 2π] is uniformly divided into K phase domains, and the phases corresponding to all maximum instantaneous characteristic energies within 1 second are calculated. Mapped into A1, A2, ..., A K The proportion of R j :

[0036]

[0037] in, This indicates the phase corresponding to the maximum instantaneous characteristic energy.

[0038] Preferably, during step f, the phase corresponding to the maximum instantaneous characteristic energy is constructed. The feature vector V representing the mapping proportion:

[0039] V = [R1,R2,…,R] K ].

[0040] Preferably, during step g, a reference vector of 1×K mapping ratio of the phase corresponding to the characteristic energy before the fault is constructed based on the full-phase domain uniformity mapping characteristics of the characteristic energy phase before the fault and the sensitive domain concentration mapping characteristics after the fault. Reference vector of the mapping ratio of the characteristic energy corresponding to the phase after the fault

[0041]

[0042] Calculate the eigenvector V and the reference vector respectively. Reference vector Euclidean feature distance D UnFault and D Fault :

[0043]

[0044] Where K represents vector V, and The dimension of is , where m is an element in the K-dimensional vector.

[0045] Preferably, in step h, the fault classification is based on:

[0046]

[0047] In the formula, β is the threshold coefficient for distinguishing between faults and non-faults.

[0048] Preferably, the 3k-45kHz frequency band is selected as the characteristic frequency band for AC series arc fault detection.

[0049] Compared with the prior art, the beneficial effects of this invention are:

[0050] 1. In this AC series arc fault detection method utilizing voltage information, the entire phase domain is divided into multiple phase mapping regions, and a specific phase domain is selected as the phase region for fault signal mapping. This approach fully considers the diverse characteristics of load power factors in practical applications. It is not only applicable to resistive loads, but also to low power factor loads and various nonlinear loads. Compared with existing methods, it has a wider applicability to load types and higher practical value.

[0051] 2. In this AC series arc fault detection method utilizing voltage information, the fault detection criterion is not set by the amplitude of the voltage characteristic energy. Instead, the fault criterion is converted into the phase distribution information of the maximum instantaneous value of each half-wave characteristic energy. Compared with existing methods, it has the outstanding advantage of not being affected by the magnitude of line parameters or load impedance.

[0052] 3. In this AC series arc fault detection method utilizing voltage information, the feature distances to be detected are calculated between the feature vector to be detected and the ideal fault feature vector and the ideal non-fault feature vector, respectively, transforming the 1D fault criterion into a 2D fault criterion. Because two-dimensional distance features are used, the difference in feature distances before and after the fault is more pronounced, significantly increasing the reliability of the detection method and reducing the difficulty of setting the fault determination threshold.

[0053] 4. In this AC series arc fault detection method that utilizes voltage information, the fault detection criteria are extracted using statistical information within 1 second rather than in units of cycles, which has the advantage of strong anti-interference capability. Attached Figure Description

[0054] Figure 1 A flowchart for an AC series arc fault detection method utilizing voltage information.

[0055] Figure 2 This is a waveform diagram of voltage and current during an electric heater malfunction.

[0056] Figure 3 This is a waveform diagram of the fault voltage and current of an induction cooker.

[0057] Figure 4 This is a waveform diagram of voltage and current during a microwave oven malfunction.

[0058] Figure 5 This is a waveform diagram of voltage and current during a vacuum cleaner malfunction.

[0059] Figure 6 This is the equivalent circuit diagram for a series arc fault.

[0060] Figure 7This is a schematic diagram of the phase mapping of the maximum instantaneous energy of the characteristic signal.

[0061] Figure 8 This is a comparison chart of characteristic energy levels before and after an electric heater malfunction.

[0062] Figure 9 This is a comparison chart of characteristic energy levels before and after an induction cooker malfunction.

[0063] Figure 10 This is a comparison chart of characteristic energy levels before and after a microwave oven malfunction.

[0064] Figure 11 This is a comparison chart of characteristic energy levels before and after a vacuum cleaner malfunction. Detailed Implementation

[0065] Figures 1-11 This is the preferred embodiment of the present invention, which is described below in conjunction with the accompanying drawings. Figures 1-11 The present invention will be further described below.

[0066] like Figure 1 As shown, an AC series arc fault detection method utilizing voltage information includes the following steps:

[0067] Step 1001, Begin;

[0068] Unlike the diverse characteristics of load current, the voltage waveform of a series arc fault under different load currents exhibits relatively uniform characteristics. For example... Figures 2-5 Different types of loads shown Figures 2-5 The fault voltage and current waveforms of electric heaters, induction cookers, microwave ovens, and vacuum cleaners are shown in sequence. For stable combustion arcs, their waveforms have a common pattern similar to square waves. Specifically, square waves are composed of high-frequency rising / falling edges and low-frequency DC arc components.

[0069] Combination Figure 6 The equivalent circuit diagram of a series arc fault is shown below.

[0070] For voltage fault characteristics, at high frequencies, due to the presence of transformer windings and line inductance, the power supply and line impedance can reach a level comparable to the load impedance. In this case, the high-frequency components in the rising / falling edges of the square wave will have a clear characterization before the fault point. Therefore, this characteristic can be used to construct fault detection criteria. For example... Figure 6 in, u M (t) represents the voltage at the upstream monitoring point, R1 represents the resistance of the upstream line, R2 represents the resistance of the downstream line, L1 represents the inductance of the upstream line, L2 represents the inductance of the downstream line, and u arc (t) represents the arc voltage, R L L L Indicates load.

[0071] In terms of frequency composition, the arc voltage source contains frequency components spanning a wide frequency range from 50Hz to 90kHz, with the highest amplitude in the 50Hz to 45kHz band. Since the harmonics contained in actual power supplies are generally below the 40th order, in order to minimize the interference of power supply harmonics, the 3kHz to 45kHz frequency band is selected as the characteristic frequency band for fault detection.

[0072] As mentioned above, its characteristic frequency band signal theoretically occurs at the rising / falling edges of the arc voltage. The phase of the arc voltage waveform at the fault point is constrained by the load current; therefore, the rising / falling edges of the arc voltage correspond to the zero-crossing points of the arc current. Considering the load power factor... The difference, for the monitoring point, will be reflected in u by the characteristic frequency band signal. M (t) and Phase interval; considering the phase shift introduced by the fault propagation network to the characteristic frequency band signal of low power factor loads. The impact, in reality, can and This is a sensitive area for fault detection of non-switching power supply loads.

[0073] like Figure 7 As shown, if the detection interval of the monitoring point is divided into equal intervals according to the phase, the fault characteristic energy signal generated by the rising / falling edge of the arc voltage will have a relatively high amplitude and be mapped to A. sen1 and A sen2 The phase domain is a typical characteristic; the fault characteristic energy interference signal that may exist in the load component is characterized by a lower amplitude and can theoretically appear in A in the figure. sen1 A sen2 A1, A2, A K-2 The arbitrary phase domain is a typical feature.

[0074] like Figures 8-11 The following are some common loads ( Figures 8-11 The comparison shows the characteristic energy of an electric heater, induction cooker, microwave oven, and vacuum cleaner before and after a fault (faults begin at the 10th cycle). Observations reveal that after a fault, the characteristic energy amplitude near the rising / falling edge of the arc voltage increases most significantly compared to before the fault, exhibiting a periodic pattern. Furthermore, the randomness of the arc electrode carbonization degree causes the arc voltage waveform to display random characteristics, resulting in random differences in the characteristic energy amplitude of the monitoring point voltage. In some cases, the characteristic energy after a fault is even similar to that before the fault. Therefore, relying solely on the characteristic energy amplitude is insufficient to guarantee detection reliability; a comprehensive detection strategy integrating characteristic energy amplitude and phase information is necessary.

[0075] Step 1002: Perform analog bandpass filtering on the voltage signal.

[0076] Step 1003: Determine whether the instantaneous characteristic energy within the half-wave is greater than or equal to a preset threshold.

[0077] Maximum instantaneous characteristic energy within half-wave Whether the limit is exceeded is used as the trigger condition for detection, based on the maximum instantaneous characteristic energy within 1 second. The phase mapping distance distribution enables fault detection of the load. If the maximum instantaneous characteristic energy within half-wave... Greater than or equal to the preset threshold E SET2 Proceed to step 1004; otherwise, return to step 1001.

[0078] Step 1004: Calculate the instantaneous characteristic energy of each half-wave within 1 second.

[0079] Calculate the instantaneous characteristic energy after bandpass filtering within 1 second, using half-wave as the unit.

[0080]

[0081] Where i is the half-wave index (i = 1 to 100), and N is the number of sampling points per wave cycle. This is a half-wave sequence after bandpass filtering.

[0082] Step 1005: Calculate the phase of the maximum instantaneous characteristic energy in half-wave units;

[0083] Calculate the instantaneous energy within 1 second, using half-wave as the unit. Maximum value: Maximum instantaneous characteristic energy

[0084]

[0085] Then calculate the maximum instantaneous characteristic energy of each half-wave within 1 second. Corresponding phase: Phase corresponding to the maximum instantaneous characteristic energy

[0086]

[0087] Where θ is the sampled signal u M (n) is the initial phase angle of the current power frequency signal obtained by performing FFT operation, u M (n) represents the voltage u at the upstream monitoring point of the line. M The time-domain discrete representation of (t); P is the maximum instantaneous characteristic energy. The corresponding sampling point number; mod[] 2π This indicates that the result of the operation is moduloed with a period of 2π, and N is the number of sampling points per cycle.

[0088] Step 1006: Calculate the mapping percentage of each phase within 1 second;

[0089] Divide the interval [0, 2π] uniformly into K phase domains, and calculate the phase corresponding to all maximum instantaneous characteristic energies within 1 second. Mapped into A1, A2, ..., A K The proportion of R j :

[0090]

[0091] Step 1007: Calculate the phase corresponding to the maximum instantaneous characteristic energy of each half-wave within 1 second. The mapping ratio;

[0092] Construct the proportion of phase mapping corresponding to the maximum instantaneous feature energy Feature vector V:

[0093] V = [R1,R2,…,R] K (5)

[0094] Step 1008: Calculate the Euclidean feature distance between the feature energy phase mapping proportion feature vector and the reference vector;

[0095] Based on the uniformity of the phase mapping of the characteristic energy phase before the fault and the concentrated mapping of the sensitive domain after the fault, a reference vector of 1×K is constructed for the mapping ratio of the phase corresponding to the characteristic energy before the fault. Reference vector of the mapping ratio of the characteristic energy corresponding to the phase after the fault

[0096]

[0097] Calculate the eigenvector V and the reference vector respectively. Reference vector Euclidean feature distance D UnFault and D Fault :

[0098]

[0099] Where K represents vector V, and The dimension of , where m is the m-th element in the K-dimensional vector.

[0100] Step 1009: Does it meet the criteria for fault classification?

[0101] Determine whether the fault classification criteria are met. If they are met, proceed to step 1010; otherwise, proceed to step 1011.

[0102] Euclidean distance reflects the relationship between the eigenvector V and the Euclidean distance. The smaller the Euclidean distance, the higher the similarity to the ideal non-fault / fault characteristic energy phase mapping ratio. Using D... UnFault and D Fault Based on the two-dimensional feature distance distribution, the classification strategy is constructed as follows:

[0103]

[0104] In the formula, β is the threshold coefficient for distinguishing between faults and non-faults.

[0105] The process for determining the threshold coefficient β is as follows:

[0106] By load power factor Considering the load, the fault detection sensitive area corresponds to the intervals [0, π / 6] and [π, 7π / 6]. Taking into account the phase offset, the actual sensitive phase areas for fault detection are [0, π / 5) and [π, 6π / 5). The uniform phase mapping domain is constructed as shown in Table 1.

[0107] Table 1 Uniform Phase Mapping Domain

[0108]

[0109] Constructing the feature energy phase mapping ratio to the reference vector Calculate the Euclidean distance between the feature energy phase mapping ratio vector before and after the fault and the reference vector, and construct its feature distance distribution plane. Since two-dimensional distance features are used, the difference in feature distance before and after the fault is more obvious, which is beneficial for setting the fault determination threshold. To avoid misjudgment, in practical applications, the threshold coefficient β > 1 can be set.

[0110] Step 1010: An AC series arc fault occurs.

[0111] Step 1011: No AC series arc fault occurred.

[0112] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for detecting AC series arc faults using voltage information, characterized in that: Includes the following steps: Step a: Perform analog bandpass filtering on the voltage signal; Step b, using the maximum instantaneous characteristic energy within the half-wave. Whether the limit is exceeded is used as the trigger condition for detection; if the maximum instantaneous characteristic energy within half-wave... If the value is greater than or equal to the preset threshold, proceed to step c; otherwise, return to step a. Step c: Calculate the instantaneous characteristic energy of each half-wave within 1 second. Step d: Calculate the maximum instantaneous characteristic energy in half-wave units. The corresponding maximum instantaneous characteristic energy corresponds to the phase. Step e: Calculate the phase corresponding to the maximum instantaneous characteristic energy of each half-wave within 1 second. The mapping ratio; Step f: Construct the phase corresponding to the maximum instantaneous feature energy. The feature vector of the mapping proportion; Step g: Calculate the phase corresponding to the maximum instantaneous characteristic energy. The Euclidean feature distance between the feature vector of the mapping proportion and the reference vector; Step h: Determine whether the fault classification criteria are met. If they are met, it is determined that an AC series arc fault has occurred. If they are not met, it means that no AC series arc fault has occurred. During step c, the instantaneous characteristic energies after bandpass filtering within 1 second are calculated in half-wave units. Where i is the half-wave index (i = 1 to 100), and N is the number of sampling points per wave cycle. This is a half-wave sequence after bandpass filtering; When performing step e, the interval [0, 2π] is uniformly divided into K phase domains, and the phase corresponding to all maximum instantaneous characteristic energies within 1 second is calculated. Mapped into A1, A2, ..., A K The proportion of R j : in, This indicates the phase corresponding to the maximum instantaneous characteristic energy.

2. The AC series arc fault detection method utilizing voltage information according to claim 1, characterized in that: When performing step d, First, calculate the instantaneous energy within 1 second, using half-wave as the unit. Maximum value: Maximum instantaneous characteristic energy Then, calculate the maximum instantaneous characteristic energy of each half-wave within 1 second. Corresponding phase: Phase corresponding to the maximum instantaneous characteristic energy Where θ is the sampled signal u M (n) is the initial phase angle of the current power frequency signal obtained by performing FFT operation, u M (n) represents the voltage u at the upstream monitoring point of the line. M The time-domain discrete representation of (t); P is the maximum instantaneous characteristic energy. The corresponding sampling point number; mod[] 2π This indicates that the result of the operation is moduloed with a period of 2π, and N is the number of sampling points per cycle.

3. The AC series arc fault detection method utilizing voltage information according to claim 1, characterized in that: During step f, the phase corresponding to the maximum instantaneous characteristic energy is constructed. The feature vector V representing the mapping proportion: V=[R1,R2,…,R K ]。 4. The AC series arc fault detection method utilizing voltage information according to claim 1, characterized in that: During step g, based on the full-phase domain uniformity mapping characteristics of the characteristic energy phase before the fault and the sensitive domain concentration mapping characteristics after the fault, a 1×K reference vector is constructed for the mapping ratio of the characteristic energy corresponding to the phase before the fault. Reference vector of the mapping ratio of the characteristic energy corresponding to the phase after the fault Calculate the eigenvector V and the reference vector respectively. Reference vector Euclidean feature distance D UnFault and D Fault : Where K represents vector V, and The dimension of is , where m is an element in the K-dimensional vector.

5. The AC series arc fault detection method utilizing voltage information according to claim 4, characterized in that: In step h, the fault classification is based on: In the formula, β is the threshold coefficient for distinguishing between faults and non-faults.

6. The AC series arc fault detection method utilizing voltage information according to claim 1, characterized in that: The 3k-45kHz frequency band was selected as the characteristic frequency band for AC series arc fault detection.

Images