A method and apparatus for detecting direct current arc faults

Abstract

The application provides a direct current fault arc detection method and device, which combines wavelet packet decomposition and reconstruction, intrinsic time scale decomposition and reconstruction and classical modal decomposition and reconstruction three kinds of feature extraction technologies to process an electrical signal sequence to be analyzed, and respectively obtains first, second and third electrical characteristic signal sequences. The method can reveal the fault characteristics of the electrical signal from multiple dimensions, realizes feature fusion, effectively improves the accuracy and robustness of fault detection, reduces false positives and false negatives, has strong adaptability, is suitable for various electrical signal analysis, and provides accurate and reliable technical support for arc fault detection of a direct current power supply system.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electric power, in particular to a direct current fault arc detection method and device. BACKGROUND

[0002] In recent years, due to the rapid increase of direct current electrical equipment and the increase of distributed energy, direct current power supply is increasingly widely used in distribution networks. Compared with alternating current power supply, direct current power supply has the advantages of small energy loss, high power supply efficiency, low line cost, etc., and is widely used in large data centers, electric vehicles, spacecraft, electrified ships and other fields. However, in the direct current distribution network, due to the small cross section of the conductor, low mechanical strength, low line insulation level and other reasons, it is easy to cause insulation damage, line aging and conductor fracture during system operation, and then lead to the generation of direct current arc. The power supply system of spacecraft, the electrical system of automobile and other power electronic systems are also the areas where direct current arc faults occur most frequently. Direct current arc has no current zero point and cannot extinguish itself. If the fault cannot be detected and eliminated in time, the direct current arc will endanger the power supply system and the control system, and in severe cases may cause fire and explosion accidents, which will threaten the stable operation of the equipment and the personal safety of the operators. SUMMARY

[0003] The technical problem to be solved by the embodiments of the present application is to provide a direct current fault arc detection method and device to improve the accuracy and efficiency of direct current fault arc detection.

[0004] In order to solve the above technical problems, the embodiments of the present application provide a direct current fault arc detection method, comprising the following steps:

[0005] The electrical signals of the monitored load are sampled according to a preset sampling frequency to obtain an electrical signal sequence, the electrical signal sequence comprising a loop current signal sequence and an arc voltage signal sequence;

[0006] The electrical signal sequence is taken out from the electrical signal sequence to obtain an electrical signal sequence to be analyzed matched with a preset time window, the electrical signal sequence to be analyzed comprising an electrical current signal sequence to be analyzed and a voltage signal sequence to be analyzed;

[0007] The electrical signal sequence to be analyzed is wavelet packet decomposed and reconstructed to obtain a first electrical characteristic signal sequence, the electrical signal sequence to be analyzed is intrinsic time scale decomposed and reconstructed to obtain a second electrical characteristic signal sequence, and the electrical signal sequence to be analyzed is classic modal decomposed and reconstructed to obtain a third electrical characteristic signal sequence;

[0008] When the load type of the monitored load is unknown, similarity metrics are respectively performed on the first, second and third electrical characteristic signal sequences and electrical characteristic sample sequences corresponding to a plurality of preset types of loads to obtain first, second and third similarity metric values of electrical characteristics between the monitored load and the plurality of preset types of loads, and the first, second and third similarity metric values of electrical characteristics between the monitored load and the plurality of preset types of loads are input into a first pre-trained probability neural network to obtain the load type of the monitored load.

[0009] When the load type of the monitored load is known, similarity metrics are respectively performed on the first, second and third electrical characteristic signal sequences and electrical characteristic sample sequences corresponding to the monitored load to obtain first, second and third similarity metric values for fault analysis, and the first, second and third similarity metric values for fault analysis are input into a second pre-trained probability neural network to obtain a direct current fault arc detection result of the monitored load; the direct current fault arc detection result is normal operation or existence of a direct current fault arc.

[0010] Optionally, the first electrical characteristic signal sequence includes a first current characteristic signal sequence and a first voltage characteristic signal sequence, the second electrical characteristic signal sequence includes a second current characteristic signal sequence and a second voltage characteristic signal sequence, and the third electrical characteristic signal sequence includes a third current characteristic signal sequence and a third voltage characteristic signal sequence.

[0011] The electrical characteristic sample sequence corresponding to each preset type of load includes a first current characteristic sample sequence, a second current characteristic sample sequence, a third current characteristic sample sequence, a first voltage characteristic sample sequence, a second voltage characteristic sample sequence and a third voltage characteristic sample sequence when the preset type of load is in normal operation.

[0012] Optionally, the similarity metrics performed on the first, second and third electrical characteristic signal sequences and electrical characteristic sample sequences corresponding to a plurality of preset types of loads to obtain first, second and third similarity metric values of electrical characteristics between the monitored load and the plurality of preset types of loads further include:

[0013] calculating Euclidean distances between the first current characteristic signal sequence and the first current characteristic sample sequence corresponding to each of the preset types of loads, and outputting the calculated Euclidean distances as first current similarity measurement values; and calculating Euclidean distances between the first voltage characteristic signal sequence and the first voltage characteristic sample sequence corresponding to each of the preset types of loads, and outputting the calculated Euclidean distances as first voltage similarity measurement values;

[0014] calculating Euclidean distances between the second current characteristic signal sequence and the second current characteristic sample sequence corresponding to each of the preset types of loads, and outputting the calculated Euclidean distances as second current similarity measurement values; and calculating Euclidean distances between the second voltage characteristic signal sequence and the second voltage characteristic sample sequence corresponding to each of the preset types of loads, and outputting the calculated Euclidean distances as second voltage similarity measurement values;

[0015] calculating Euclidean distances between the third current characteristic signal sequence and the third current characteristic sample sequence corresponding to each of the preset types of loads, and outputting the calculated Euclidean distances as third current similarity measurement values; and calculating Euclidean distances between the third voltage characteristic signal sequence and the third voltage characteristic sample sequence corresponding to each of the preset types of loads, and outputting the calculated Euclidean distances as third voltage similarity measurement values;

[0016] The similarity measurement values of the electrical characteristics between the monitored load and the plurality of preset types of loads include first current similarity measurement values, second current similarity measurement values, third current similarity measurement values, first voltage similarity measurement values, second voltage similarity measurement values, and third voltage similarity measurement values.

[0017] Optionally, the similarity measurement of the first electrical characteristic signal sequence, the second electrical characteristic signal sequence, and the third electrical characteristic signal sequence with respect to the electrical characteristic sample sequence corresponding to the monitored load obtains first similarity measurement values, second similarity measurement values, and third similarity measurement values for fault analysis, and further includes:

[0018] calculating Euclidean distances between the first current characteristic signal sequence and the first current characteristic sample sequence corresponding to the monitored load, and outputting the calculated Euclidean distances as first current similarity measurement values; and calculating Euclidean distances between the first voltage characteristic signal sequence and the first voltage characteristic sample sequence corresponding to the monitored load, and outputting the calculated Euclidean distances as first voltage similarity measurement values;

[0019] calculating the Euclidean distance between the second current feature signal sequence and the second current feature sample sequence corresponding to the monitored load, and outputting the calculated Euclidean distance as a second current similarity measurement value; and calculating the Euclidean distance between the second voltage feature signal sequence and the second voltage feature sample sequence corresponding to the monitored load, and outputting the calculated Euclidean distance as a second voltage similarity measurement value;

[0020] calculating the Euclidean distance between the third current feature signal sequence and the third current feature sample sequence corresponding to the monitored load, and outputting the calculated Euclidean distance as a third current similarity measurement value; and calculating the Euclidean distance between the third voltage feature signal sequence and the third voltage feature sample sequence corresponding to the monitored load, and outputting the calculated Euclidean distance as a third voltage similarity measurement value;

[0021] The similarity measurement values of the first, second and third electrical feature signal sequences of the monitored load and the electrical feature sample sequence include first, second and third current similarity measurement values, and first, second and third voltage similarity measurement values.

[0022] Optionally, the method further includes:

[0023] After the wavelet packet decomposition and reconstruction and the intrinsic time-scale decomposition and reconstruction of the to-be-analyzed electrical signal sequence and the obtaining of the first, second and third electrical feature signal sequences, the first, second and third current feature signal sequences, the first, second and third voltage feature signal sequences are further divided into equal-length sub-segments, and then the average values of each sub-segment are calculated to correspond to represent each sub-segment, so as to convert the first, second and third current feature signal sequences, the first, second and third voltage feature signal sequences into low-dimensional signal sequences.

[0024] The embodiment of the present application also provides a direct-current fault arc detection device, which comprises:

[0025] A sampling module is configured to sample an electrical signal of a monitored load according to a preset sampling frequency to obtain an electrical signal sequence, wherein the electrical signal sequence comprises a loop current signal sequence and an arc voltage signal sequence.

[0026] a preprocessing module configured to extract a to-be-analyzed electrical signal sequence matching a preset time window from the electrical signal sequence, the to-be-analyzed electrical signal sequence including a to-be-analyzed current signal sequence and a to-be-analyzed voltage signal sequence;

[0027] a feature extraction module configured to perform wavelet packet decomposition and reconstruction on the to-be-analyzed electrical signal sequence to obtain a first electrical feature signal sequence, perform intrinsic time-scale decomposition and reconstruction on the to-be-analyzed electrical signal sequence to obtain a second electrical feature signal sequence, and perform classical mode decomposition and reconstruction on the to-be-analyzed electrical signal sequence to obtain a third electrical feature signal sequence;

[0028] a load type identification module configured to, when the load type of the monitored load is unknown, perform similarity measurement on the first electrical feature signal sequence, the second electrical feature signal sequence, and the third electrical feature signal sequence, respectively, and electrical feature sample sequences corresponding to a plurality of preset type loads to obtain first, second, and third similarity measurement values of electrical features between the monitored load and the plurality of preset type loads, input the first, second, and third similarity measurement values of electrical features between the monitored load and the plurality of preset type loads into a first pre-trained probability neural network to obtain the load type of the monitored load;

[0029] a load fault identification module configured to, when the load type of the monitored load is known, perform similarity measurement on the first electrical feature signal sequence, the second electrical feature signal sequence, and the third electrical feature signal sequence, respectively, and electrical feature sample sequences corresponding to the monitored load to obtain first, second, and third similarity measurement values for fault analysis, and input the first, second, and third similarity measurement values for fault analysis into a second pre-trained probability neural network to obtain a direct current fault arc detection result of the monitored load; the direct current fault arc detection result is normal operation or existence of a direct current fault arc.

[0030] Optionally, the first electrical feature signal sequence includes a first current feature signal sequence and a first voltage feature signal sequence, the second electrical feature signal sequence includes a second current feature signal sequence and a second voltage feature signal sequence, and the third electrical feature signal sequence includes a third current feature signal sequence and a third voltage feature signal sequence.

[0031] The electrical feature sample sequence corresponding to each preset type load includes a first current feature sample sequence, a second current feature sample sequence, a third current feature sample sequence, a first voltage feature sample sequence, a second voltage feature sample sequence, and a third voltage feature sample sequence when the preset type load is in normal operation.

[0032] Optionally, the load type identification module is further configured to:

[0033] calculate the Euclidean distance between the first current characteristic signal sequence and the first current characteristic sample sequence corresponding to each of the preset type loads, and output the calculated Euclidean distance as a first current similarity measurement value; and calculate the Euclidean distance between the first voltage characteristic signal sequence and the first voltage characteristic sample sequence corresponding to each of the preset type loads, and output the calculated Euclidean distance as a first voltage similarity measurement value.

[0034] calculate the Euclidean distance between the second current characteristic signal sequence and the second current characteristic sample sequence corresponding to each of the preset type loads, and output the calculated Euclidean distance as a second current similarity measurement value; and calculate the Euclidean distance between the second voltage characteristic signal sequence and the second voltage characteristic sample sequence corresponding to each of the preset type loads, and output the calculated Euclidean distance as a second voltage similarity measurement value.

[0035] calculate the Euclidean distance between the third current characteristic signal sequence and the third current characteristic sample sequence corresponding to each of the preset type loads, and output the calculated Euclidean distance as a third current similarity measurement value; and calculate the Euclidean distance between the third voltage characteristic signal sequence and the third voltage characteristic sample sequence corresponding to each of the preset type loads, and output the calculated Euclidean distance as a third voltage similarity measurement value.

[0036] The similarity measurement values of the electrical characteristics between the monitored load and the plurality of preset type loads each include a first current similarity measurement value, a second current similarity measurement value, a third current similarity measurement value, a first voltage similarity measurement value, a second voltage similarity measurement value, and a third voltage similarity measurement value.

[0037] Optionally, the load fault identification module is further configured to:

[0038] calculate the Euclidean distance between the first current characteristic signal sequence and the first current characteristic sample sequence corresponding to the monitored load, and output the calculated Euclidean distance as a first current similarity measurement value; and calculate the Euclidean distance between the first voltage characteristic signal sequence and the first voltage characteristic sample sequence corresponding to the monitored load, and output the calculated Euclidean distance as a first voltage similarity measurement value.

[0039] calculating the Euclidean distance between the second current feature signal sequence and the second current feature sample sequence corresponding to the monitored load, and outputting the calculated Euclidean distance as a second current similarity measurement value; and calculating the Euclidean distance between the second voltage feature signal sequence and the second voltage feature sample sequence corresponding to the monitored load, and outputting the calculated Euclidean distance as a second voltage similarity measurement value;

[0040] calculating the Euclidean distance between the third current feature signal sequence and the third current feature sample sequence corresponding to the monitored load, and outputting the calculated Euclidean distance as a third current similarity measurement value; and calculating the Euclidean distance between the third voltage feature signal sequence and the third voltage feature sample sequence corresponding to the monitored load, and outputting the calculated Euclidean distance as a third voltage similarity measurement value;

[0041] The similarity measurement values of the first, second and third electrical feature signal sequences of the monitored load and the electrical feature sample sequence include first, second and third current similarity measurement values and first, second and third voltage similarity measurement values.

[0042] Optionally, the feature extraction module is further configured to:

[0043] After the wavelet packet decomposition and reconstruction and the intrinsic time-scale decomposition and reconstruction of the to-be-analyzed electrical signal sequence and the obtaining of the first, second and third electrical feature signal sequences, the first, second and third current feature signal sequences, the first, second and third voltage feature signal sequences are further divided into equal-length sub-segments, and then the average values of each sub-segment are calculated to correspond to each sub-segment, so as to convert the first, second and third current feature signal sequences and the first, second and third voltage feature signal sequences into low-dimensional signal sequences.

[0044] The embodiments of the present application have the following beneficial effects:

[0045] The embodiments of the present application adopt various feature extraction methods, which are wavelet packet decomposition and reconstruction, intrinsic time-scale decomposition and reconstruction, and classical mode decomposition and reconstruction respectively, to process the electrical signal sequence to be analyzed, so as to obtain electrical characteristic signal sequences in multiple dimensions, and fuse the characteristics to improve the accuracy of fault detection. The embodiments of the present application can effectively reduce the risk of fire, explosion and other safety accidents caused by arc fault through real-time monitoring and rapid processing of the direct current arc fault, and protect the stable operation of the equipment and system. The embodiments of the present application are suitable for various direct current power supply systems, such as large data centers, electric vehicles, spacecraft, electrified ships and the like, and have wide application prospects. BRIEF DESCRIPTION OF DRAWINGS

[0046] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or the prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings obtained according to these drawings without creative labor are still within the scope of the present application.

[0047] Figure 1 A flow chart of a direct current fault arc detection method in an embodiment of the present application.

[0048] Figure 2 A structural diagram of a direct current fault arc detection device in an embodiment of the present application. DETAILED DESCRIPTION

[0049] The detailed description of the drawings is intended as a description of the current embodiments of the present application and is not intended to represent the only forms in which the present application can be implemented. It should be understood that the same or equivalent functions can be accomplished by different embodiments intended to be included within the spirit and scope of the present application.

[0050] Reference Figure 1 , an embodiment of the present application provides a direct current fault arc detection method, comprising the following steps:

[0051] Step S10: sampling the electrical signal of the monitored load according to a preset sampling frequency to obtain an electrical signal sequence, wherein the electrical signal sequence comprises a loop current signal sequence and an arc voltage signal sequence.

[0052] Specifically, the description in step S10 relates to the basic process of electrical signal sampling. The sampling frequency refers to the number of samples per unit time, usually in units of hertz (Hz). The preset sampling frequency is a parameter set before signal sampling, which determines the time interval of sampling. The frequency needs to be high enough to accurately restore the original signal.

[0053] The electrical signal of the monitored load refers to the electrical signal generated by the monitoring target (i.e. the load) during operation, including current and voltage signals. The load can be any device or system using direct current.

[0054] Sampling refers to recording the value of the electrical signal at a specific time point according to a preset sampling frequency. This process is the first step of digitization, which converts continuous analog signals into discrete digital signals.

[0055] The series of values obtained by sampling are arranged in chronological order to form the electrical signal sequence. In this step, the loop current signal sequence and the arc voltage signal sequence are obtained. The loop current signal sequence refers to the variation of current during the operation of the load, which reflects the electrical behavior and power demand of the load. The arc voltage signal sequence refers to the variation of voltage measured across the load, which can reveal the existence and characteristics of the arc, as the arc causes specific changes in the voltage signal.

[0056] Step S20: Extracting the to-be-analyzed electrical signal sequence matching the preset time window from the electrical signal sequence, the to-be-analyzed electrical signal sequence including the to-be-analyzed current signal sequence and the to-be-analyzed voltage signal sequence.

[0057] Specifically, a specific time window has been set before analysis, which is determined based on the analysis and monitoring requirements of the fault arc characteristics. It represents the duration of the signal segment extracted from the electrical signal sequence. The to-be-analyzed electrical signal sequence refers to the part of the entire electrical signal sequence that matches the preset time window. This part of the signal sequence will be the object of analysis for detecting the characteristics of the fault arc. The to-be-analyzed current signal sequence is a part of the to-be-analyzed electrical signal sequence, specifically representing the current variation within the preset time window. The to-be-analyzed voltage signal sequence is a part of the to-be-analyzed electrical signal sequence, representing the voltage variation within the preset time window.

[0058] Step S30: Wavelet packet decomposition and reconstruction of the to-be-analyzed electrical signal sequence to obtain the first electrical characteristic signal sequence, intrinsic time scale decomposition and reconstruction of the to-be-analyzed electrical signal sequence to obtain the second electrical characteristic signal sequence, and classical modal decomposition and reconstruction of the to-be-analyzed electrical signal sequence to obtain the third electrical characteristic signal sequence.

[0059] Specifically, wavelet packet decomposition is a more detailed signal processing method, which not only decomposes the low-frequency part of the signal, but also further decomposes the high-frequency part. This method can more effectively extract the detailed information in the signal. Specifically, the electrical signal sequence to be analyzed is decomposed into multiple subbands by wavelet packet transform, each subband contains signal components of different frequency ranges. The decomposed subbands are selected and reconstructed to extract specific frequency features, which help identify the signal characteristics of arc faults. The first electrical feature signal sequence is obtained through this process, which contains important time-frequency information of the signal and helps subsequent fault detection and analysis.

[0060] Intrinsic time-scale decomposition is a nonlinear signal processing technique that can decompose signals based on their intrinsic characteristics, suitable for time-frequency analysis of non-stationary signals. Specifically, the electrical signal sequence to be analyzed is decomposed into several proper rotating components (PROD) and a trend term, each PROD component represents different time scale characteristics of the signal. The PROD components and trend terms obtained by decomposition are reconstructed to obtain the second electrical feature signal sequence, which can reflect the time scale changes of the signal and is very useful for detecting the transient characteristics of arc faults.

[0061] Classical mode decomposition is an adaptive time-frequency analysis technique that can decompose complex signals into a series of intrinsic mode functions (IMF), each IMF represents an intrinsic vibration mode of the signal. Specifically, the electrical signal sequence to be analyzed is decomposed by EMD to obtain several IMFs and a residual component. The IMFs obtained by decomposition are selected and reconstructed to form the third electrical feature signal sequence. This sequence can reveal the internal vibration characteristics of the signal and help identify weak signals of arc faults.

[0062] Through these three different decomposition and reconstruction techniques, the method of the present embodiment can extract features of electrical signals from different angles, providing rich feature information for subsequent fault detection. These feature signal sequences will be used as input data for subsequent similarity measurement and fault detection analysis, thereby improving the accuracy and reliability of detection.

[0063] Step S40, when the load type of the monitored load is unknown, similarity measures are performed between the first, second and third electrical characteristic signal sequences and electrical characteristic sample sequences corresponding to a plurality of preset type loads, respectively, to obtain first, second and third similarity measure values of electrical characteristics between the monitored load and the plurality of preset type loads, and the first, second and third similarity measure values of electrical characteristics between the monitored load and the plurality of preset type loads are input into a pre-trained first probability neural network to obtain the load type of the monitored load.

[0064] Specifically, step S40 is to determine the type of the monitored load by comparing the similarity between the electrical characteristic signal sequences and the electrical characteristic sample sequences of a plurality of preset type loads when the load type of the monitored load is unknown.

[0065] Similarity measure is a method for evaluating the degree of similarity between two data sets. In step S40, the three extracted electrical characteristic signal sequences need to be compared with the electrical characteristic sample sequences of a plurality of preset type loads for similarity measure. The similarity measure can use various methods, such as Euclidean distance, cosine similarity, correlation coefficient, etc.

[0066] The first electrical characteristic signal sequence (feature obtained by wavelet packet decomposition and reconstruction) is compared with the first electrical characteristic sample sequences of a plurality of preset type loads for similarity measure, respectively, to obtain first similarity measure values corresponding to the plurality of preset type loads.

[0067] The second electrical characteristic signal sequence (feature obtained by intrinsic time-scale decomposition and reconstruction) is compared with the second electrical characteristic sample sequences of a plurality of preset type loads for similarity measure, respectively, to obtain second similarity measure values corresponding to the plurality of preset type loads.

[0068] The third electrical characteristic signal sequence (feature obtained by classical mode decomposition and reconstruction) is compared with the third electrical characteristic sample sequences of a plurality of preset type loads for similarity measure, respectively, to obtain third similarity measure values corresponding to the plurality of preset type loads.

[0069] The obtained similarity metric values are input as input features into a pre-trained first probabilistic neural network model. The probabilistic neural network is a neural network based on probability statistics, which is suitable for classification problems and performs well in handling uncertain and probabilistic data. Before implementing step S40, the first probabilistic neural network needs to be trained using electrical feature samples of known load types. During the training process, the network learns how to distinguish different load types based on the input similarity metric values. The first probabilistic neural network outputs the type of the monitored load based on the input first, second, and third similarity metric values. The network output is based on the pattern recognition ability learned during the training process, which can identify the load type that best matches the input features.

[0070] For example, assuming there are n preset types of loads, there are n groups of similarity metric values, each group including first, second, and third similarity metric values. The n groups of similarity metric values are input into the first probabilistic neural network, which outputs the probability that the monitored load belongs to the n preset types of loads. The load type with the highest probability is the identified load type.

[0071] Step S50, when the load type of the monitored load is known, similarity metrics are performed on the first, second, and third electrical feature signal sequences and the electrical feature sample sequence corresponding to the monitored load to obtain first, second, and third similarity metric values for fault analysis, and the first, second, and third similarity metric values for fault analysis are input into a pre-trained second probabilistic neural network to obtain a direct current fault arc detection result of the monitored load; the direct current fault arc detection result is normal operation or existence of a direct current fault arc.

[0072] Specifically, step S50 is to detect the existence of DC fault arc by comparing the extracted electrical feature signal sequences with the electrical feature sample sequences corresponding to the type of the monitored load in the case where the type of the monitored load is known. In this step, the type of the monitored load has been determined, which means that the electrical feature sample sequences related to this specific load type can be used as a reference for fault detection. According to the previous steps (e.g. S30), three different electrical feature signal sequences have been extracted from the original electrical signal, next, similarity measurement needs to be performed on these three electrical feature signal sequences respectively with the electrical feature sample sequences of the known load type to obtain the first, second and third similarity measurement values for fault analysis, the purpose of which is to quantify the degree of similarity between the two sequences. The three similarity measurement values obtained will be input features, which will be sent to the second pre-trained probability neural network (PNN), which has learned how to distinguish between normal operation and fault state according to these similarity measurement values during the training process. The second probability neural network will process the input similarity measurement values and output a detection result. This detection result is usually a classification label, which indicates whether the monitored load is in a "normal operation" state or "DC fault arc exists". If the output indicates "normal operation", the monitored load has no detected fault arc. If the output indicates "DC fault arc exists", it means that the monitored load may have a fault and needs to be further checked or maintenance measures need to be taken.

[0073] Further, the first electrical feature signal sequence includes a first current feature signal sequence and a first voltage feature signal sequence, the second electrical feature signal sequence includes a second current feature signal sequence and a second voltage feature signal sequence, and the third electrical feature signal sequence includes a third current feature signal sequence and a third voltage feature signal sequence.

[0074] The electrical feature sample sequence corresponding to each preset type of load includes a first current feature sample sequence, a second current feature sample sequence, a third current feature sample sequence, a first voltage feature sample sequence, a second voltage feature sample sequence and a third voltage feature sample sequence of the preset type of load in normal operation.

[0075] Specifically, the first, second and third electrical feature signal sequences further include current feature signal sequences and voltage feature signal sequences, and similarly, the electrical feature sample sequence corresponding to each preset type of load also includes current and voltage feature sample sequences.

[0076] The first current feature signal sequence is a feature sequence extracted from the original current signal through the wavelet packet decomposition and reconstruction method. It contains the key information in the current signal, which is used to characterize the current characteristics of the load. The first voltage feature signal sequence is a feature sequence extracted from the original voltage signal through the wavelet packet decomposition and reconstruction method. It reflects the important features of the voltage signal.

[0077] The second current feature signal sequence is a feature sequence extracted from the original current signal through the intrinsic time-scale decomposition and reconstruction method. It provides another perspective of the current signal, which may capture information that wavelet packet decomposition fails to extract. The second voltage feature signal sequence is a feature sequence extracted from the original voltage signal through the intrinsic time-scale decomposition and reconstruction method, which is also used to reveal the characteristics of the voltage signal.

[0078] The third current feature signal sequence is a feature sequence extracted from the original current signal through the classical mode decomposition and reconstruction method. It provides an additional dimension for the analysis of the current signal. The third voltage feature signal sequence is a feature sequence extracted from the original voltage signal through the classical mode decomposition and reconstruction method, which further enriches the feature information of the voltage signal.

[0079] For each type of preset load, there is a corresponding electrical feature sample sequence, which is recorded when the load is running normally and used as a reference standard.

[0080] The first current feature sample sequence is a sample of the first current feature signal sequence recorded when the preset type load is running normally, which is obtained through the wavelet packet decomposition and reconstruction method. The second current feature sample sequence is a sample of the second current feature signal sequence recorded when the preset type load is running normally, which is obtained through the intrinsic time-scale decomposition and reconstruction method. The third current feature sample sequence is a sample of the third current feature signal sequence recorded when the preset type load is running normally, which is obtained through the classical mode decomposition and reconstruction method.

[0081] The first voltage feature sample sequence is a sample of the first voltage feature signal sequence recorded when the preset type load is running normally, which is obtained through the wavelet packet decomposition and reconstruction method. The second voltage feature sample sequence is a sample of the second voltage feature signal sequence recorded when the preset type load is running normally, which is obtained through the intrinsic time-scale decomposition and reconstruction method. The third voltage feature sample sequence is a sample of the third voltage feature signal sequence recorded when the preset type load is running normally, which is obtained through the classical mode decomposition and reconstruction method.

[0082] In step S50, for each type of preset load, the following similarity measure values are calculated respectively:

[0083] The similarity measure value of the first current characteristic signal sequence and the corresponding first current characteristic sample sequence.

[0084] The similarity measure value of the second current characteristic signal sequence and the corresponding second current characteristic sample sequence.

[0085] The similarity measure value of the third current characteristic signal sequence and the corresponding third current characteristic sample sequence.

[0086] The similarity measure value of the first voltage characteristic signal sequence and the corresponding first voltage characteristic sample sequence.

[0087] The similarity measure value of the second voltage characteristic signal sequence and the corresponding second voltage characteristic sample sequence.

[0088] The similarity measure value of the third voltage characteristic signal sequence and the corresponding third voltage characteristic sample sequence.

[0089] These similarity measure values will be input features, which are sent to the second pre-trained probability neural network to detect the existence of DC fault arc. Through this refined feature extraction and comparison, the accuracy and reliability of fault detection can be improved.

[0090] Further, the similarity measure of the first, second and third electrical characteristic signal sequences and the electrical characteristic sample sequences corresponding to the plurality of preset types of loads respectively obtains the first, second and third similarity measure values of the electrical characteristics between the monitored load and the plurality of preset types of loads, further comprising:

[0091] Calculate the Euclidean distance between the first current characteristic signal sequence and the first current characteristic sample sequence corresponding to each preset type of load, and output the calculated Euclidean distance as the first current similarity measure value; and calculate the Euclidean distance between the first voltage characteristic signal sequence and the first voltage characteristic sample sequence corresponding to each preset type of load, and output the calculated Euclidean distance as the first voltage similarity measure value;

[0092] Calculate the Euclidean distance between the second current characteristic signal sequence and the second current characteristic sample sequence corresponding to each preset type of load, and output the calculated Euclidean distance as the second current similarity measure value; and calculate the Euclidean distance between the second voltage characteristic signal sequence and the second voltage characteristic sample sequence corresponding to each preset type of load, and output the calculated Euclidean distance as the second voltage similarity measure value;

[0093] calculating the Euclidean distance between the third current characteristic signal sequence and the third current characteristic sample sequence corresponding to each preset type of load, and outputting the calculated Euclidean distance as a third current similarity measurement value; and calculating the Euclidean distance between the third voltage characteristic signal sequence and the third voltage characteristic sample sequence corresponding to each preset type of load, and outputting the calculated Euclidean distance as a third voltage similarity measurement value;

[0094] The similarity measurement values of the electrical characteristics between the monitored load and the plurality of preset types of load each include a first current similarity measurement value, a second current similarity measurement value, a third current similarity measurement value, a first voltage similarity measurement value, a second voltage similarity measurement value, and a third voltage similarity measurement value.

[0095] Specifically, for the similarity measurement of each preset type of load, the following six similarity measurement values are obtained: a first current similarity measurement value, a second current similarity measurement value, a third current similarity measurement value, a first voltage similarity measurement value, a second voltage similarity measurement value, and a third voltage similarity measurement value. These similarity measurement values collectively constitute a comprehensive evaluation of the similarity of electrical characteristics between the monitored load and the corresponding preset type of load. These measurement values are then input into the pre-trained probabilistic neural network to identify the type of the monitored load. Through such detailed comparison and measurement, the load type can be more accurately identified, thereby improving the operation efficiency and safety of the electrical system.

[0096] Further, the similarity measurement of the first electrical characteristic signal sequence, the second electrical characteristic signal sequence, and the third electrical characteristic signal sequence with respect to the electrical characteristic sample sequence corresponding to the monitored load to obtain first, second, and third similarity measurement values for fault analysis further includes:

[0097] calculating the Euclidean distance between the first current characteristic signal sequence and the first current characteristic sample sequence corresponding to the monitored load, and outputting the calculated Euclidean distance as a first current similarity measurement value; and calculating the Euclidean distance between the first voltage characteristic signal sequence and the first voltage characteristic sample sequence corresponding to the monitored load, and outputting the calculated Euclidean distance as a first voltage similarity measurement value;

[0098] calculating the Euclidean distance between the second current characteristic signal sequence and the second current characteristic sample sequence corresponding to the monitored load, and outputting the calculated Euclidean distance as a second current similarity measurement value; and calculating the Euclidean distance between the second voltage characteristic signal sequence and the second voltage characteristic sample sequence corresponding to the monitored load, and outputting the calculated Euclidean distance as a second voltage similarity measurement value;

[0099] calculating the Euclidean distance between the third current feature signal sequence and the third current feature sample sequence corresponding to the monitored load, and outputting the calculated Euclidean distance as a third current similarity measurement value; and calculating the Euclidean distance between the third voltage feature signal sequence and the third voltage feature sample sequence corresponding to the monitored load, and outputting the calculated Euclidean distance as a third voltage similarity measurement value;

[0100] The similarity measurement values of the first, second and third electrical feature signal sequences of the monitored load and the electrical feature sample sequence include first, second and third current similarity measurement values, and first, second and third voltage similarity measurement values.

[0101] Specifically, similarity measurement is performed on the first, second and third electrical feature signal sequences and the electrical feature sample sequence corresponding to the monitored load, respectively, to obtain the following six similarity measurement values for fault analysis: first, second and third current similarity measurement values, and first, second and third voltage similarity measurement values. These similarity measurement values collectively constitute a comprehensive evaluation of the electrical feature similarity between the monitored load and the corresponding known type load. These measurement values are then input into the pre-trained probabilistic neural network to detect whether a direct current fault arc exists.

[0102] Further, the step S30 further includes:

[0103] After wavelet packet decomposition and reconstruction and intrinsic time-scale decomposition and reconstruction are performed on the electrical signal sequence to be analyzed to obtain the first, second and third electrical feature signal sequences, the first, second and third current feature signal sequences, and the first, second and third voltage feature signal sequences are further divided into equal-length sub-segments, and then the average value of each sub-segment is calculated to correspond to each sub-segment, so as to convert the first, second and third current feature signal sequences and the first, second and third voltage feature signal sequences into low-dimensional signal sequences.

[0104] Specifically, in step S30, in addition to wavelet packet decomposition and reconstruction and intrinsic time-scale decomposition and reconstruction of the electrical signal sequence to obtain the electrical feature signal sequence, further processing of the feature signal sequence is included, i.e. segmentation and dimension reduction. The first current feature signal sequence, the second current feature signal sequence, the third current feature signal sequence, the first voltage feature signal sequence, the second voltage feature signal sequence, and the third voltage feature signal sequence are segmented into equal-length sub-segments. The length of the sub-segments depends on the total length of the signal and the required resolution. For example, if the original signal sequence has 1000 data points, and we decide that each sub-segment contains 100 data points, the entire sequence will be segmented into 10 equal-length sub-segments. For each sub-segment, the average value of its data points is calculated. The average value, as the representative value of the sub-segment, can effectively reduce the dimension of the data. By calculating the average value of each sub-segment, the original electrical feature signal sequence is converted into a low-dimensional signal sequence. This low-dimensional sequence contains the key information of the original signal, but the number of data points is greatly reduced. For example, if each original feature signal sequence has 1000 data points, after segmentation and average calculation, the new low-dimensional signal sequence will only have 10 data points (assuming each sub-segment has 100 data points). Dimension reduction can reduce the amount of data for subsequent similarity measurement calculation and neural network processing, improving computational efficiency. By calculating the average value of the sub-segment, the main features of the signal can be highlighted, and the impact of noise and details on analysis can be reduced. Low-dimensional data is easier to analyze and understand, which helps to improve the accuracy of fault detection and load type identification. Through this series of processing, the original electrical signal is converted into a low-dimensional feature signal sequence that is easier to process and analyze, providing effective data input for subsequent fault detection and load type identification.

[0105] Referring to Figure 2 Another embodiment of the present application also provides a direct current fault arc detection device, comprising:

[0106] A sampling module 1 is configured to sample the electrical signal of the monitored load according to a preset sampling frequency to obtain an electrical signal sequence, wherein the electrical signal sequence comprises a loop current signal sequence and an arc voltage signal sequence;

[0107] A preprocessing module 2 is configured to extract a to-be-analyzed electrical signal sequence matching a preset time window from the electrical signal sequence, wherein the to-be-analyzed electrical signal sequence comprises a to-be-analyzed current signal sequence and a voltage signal sequence;

[0108] a feature extraction module 3 configured to perform wavelet packet decomposition and reconstruction on the to-be-analyzed electrical signal sequence to obtain a first electrical feature signal sequence, perform intrinsic time-scale decomposition and reconstruction on the to-be-analyzed electrical signal sequence to obtain a second electrical feature signal sequence, and perform classical mode decomposition and reconstruction on the to-be-analyzed electrical signal sequence to obtain a third electrical feature signal sequence;

[0109] a load type identification module 4 configured to, when the load type of the monitored load is unknown, perform similarity measurement on the first electrical feature signal sequence, the second electrical feature signal sequence, and the third electrical feature signal sequence, respectively, and electrical feature sample sequences corresponding to a plurality of preset type loads to obtain first, second, and third similarity measurement values of electrical features between the monitored load and the plurality of preset type loads, input the first, second, and third similarity measurement values of electrical features between the monitored load and the plurality of preset type loads into a first pre-trained probability neural network to obtain the load type of the monitored load;

[0110] a load fault identification module 5 configured to, when the load type of the monitored load is known, perform similarity measurement on the first electrical feature signal sequence, the second electrical feature signal sequence, and the third electrical feature signal sequence, respectively, and electrical feature sample sequences corresponding to the monitored load to obtain first, second, and third similarity measurement values for fault analysis, and input the first, second, and third similarity measurement values for fault analysis into a second pre-trained probability neural network to obtain a direct current fault arc detection result of the monitored load; the direct current fault arc detection result is normal operation or existence of a direct current fault arc.

[0111] Further, the first electrical feature signal sequence includes a first current feature signal sequence and a first voltage feature signal sequence, the second electrical feature signal sequence includes a second current feature signal sequence and a second voltage feature signal sequence, and the third electrical feature signal sequence includes a third current feature signal sequence and a third voltage feature signal sequence.

[0112] The electrical feature sample sequence corresponding to each preset type load includes a first current feature sample sequence, a second current feature sample sequence, a third current feature sample sequence, a first voltage feature sample sequence, a second voltage feature sample sequence, and a third voltage feature sample sequence of the preset type load in normal operation.

[0113] Further, the load type identification module 4 is further configured to:

[0114] calculating the Euclidean distance between the first current characteristic signal sequence and the first current characteristic sample sequence corresponding to each of the preset type loads, and outputting the calculated Euclidean distance as a first current similarity measurement value; and calculating the Euclidean distance between the first voltage characteristic signal sequence and the first voltage characteristic sample sequence corresponding to each of the preset type loads, and outputting the calculated Euclidean distance as a first voltage similarity measurement value;

[0115] calculating the Euclidean distance between the second current characteristic signal sequence and the second current characteristic sample sequence corresponding to each of the preset type loads, and outputting the calculated Euclidean distance as a second current similarity measurement value; and calculating the Euclidean distance between the second voltage characteristic signal sequence and the second voltage characteristic sample sequence corresponding to each of the preset type loads, and outputting the calculated Euclidean distance as a second voltage similarity measurement value;

[0116] calculating the Euclidean distance between the third current characteristic signal sequence and the third current characteristic sample sequence corresponding to each of the preset type loads, and outputting the calculated Euclidean distance as a third current similarity measurement value; and calculating the Euclidean distance between the third voltage characteristic signal sequence and the third voltage characteristic sample sequence corresponding to each of the preset type loads, and outputting the calculated Euclidean distance as a third voltage similarity measurement value;

[0117] The similarity measurement values of the electrical characteristics between the monitored load and the plurality of preset type loads include a first current similarity measurement value, a second current similarity measurement value, a third current similarity measurement value, a first voltage similarity measurement value, a second voltage similarity measurement value, and a third voltage similarity measurement value.

[0118] Further, the load fault identification module 5 is further configured to:

[0119] calculating the Euclidean distance between the first current characteristic signal sequence and the first current characteristic sample sequence corresponding to the monitored load, and outputting the calculated Euclidean distance as a first current similarity measurement value; and calculating the Euclidean distance between the first voltage characteristic signal sequence and the first voltage characteristic sample sequence corresponding to the monitored load, and outputting the calculated Euclidean distance as a first voltage similarity measurement value;

[0120] calculating the Euclidean distance between the second current characteristic signal sequence and the second current characteristic sample sequence corresponding to the monitored load, and outputting the calculated Euclidean distance as a second current similarity measurement value; and calculating the Euclidean distance between the second voltage characteristic signal sequence and the second voltage characteristic sample sequence corresponding to the monitored load, and outputting the calculated Euclidean distance as a second voltage similarity measurement value;

[0121] calculating the Euclidean distance between the third current feature signal sequence and the third current feature sample sequence corresponding to the monitored load, and outputting the calculated Euclidean distance as a third current similarity measurement value; and calculating the Euclidean distance between the third voltage feature signal sequence and the third voltage feature sample sequence corresponding to the monitored load, and outputting the calculated Euclidean distance as a third voltage similarity measurement value;

[0122] The similarity measurement values of the first, second and third electrical feature signal sequences of the monitored load and the electrical feature sample sequence include first, second and third current similarity measurement values and first, second and third voltage similarity measurement values.

[0123] Further, the feature extraction module 3 is further configured to:

[0124] After the wavelet packet decomposition and reconstruction and the intrinsic time-scale decomposition and reconstruction of the to-be-analyzed electrical signal sequence and the obtaining of the first, second and third electrical feature signal sequences, the first, second and third current feature signal sequences, the first, second and third voltage feature signal sequences are further divided into equal-length sub-segments, and then the average values of each sub-segment are calculated to correspond to each sub-segment, so as to convert the first, second and third current feature signal sequences and the first, second and third voltage feature signal sequences into low-dimensional signal sequences.

[0125] It should be noted that the device of the present embodiment corresponds to the above-mentioned method of the embodiment, and therefore, the contents not described in detail in the present embodiment can be obtained by referring to the contents of the above-mentioned method of the embodiment, which will not be described herein again.

[0126] The above only discloses the preferred embodiments of the present application, and of course cannot limit the scope of the rights of the present application, therefore, the equivalent changes made according to the claims of the present application still fall within the scope of the present application.

Claims

1. A method of detecting a direct current fault arc, the method comprising: The method comprises the following steps: According to the preset sampling frequency, the electrical signals of the monitored load are sampled to obtain an electrical signal sequence, which comprises a loop current signal sequence and an arc voltage signal sequence; The electrical signal sequence is taken out to obtain an electrical signal sequence to be analyzed which matches a preset time window, and the electrical signal sequence to be analyzed comprises a current signal sequence to be analyzed and a voltage signal sequence to be analyzed; The electrical signal sequence to be analyzed is subjected to wavelet packet decomposition and reconstruction to obtain a first electrical characteristic signal sequence, subjected to intrinsic time-scale decomposition and reconstruction to obtain a second electrical characteristic signal sequence, and subjected to classical modal decomposition and reconstruction to obtain a third electrical characteristic signal sequence; When the load type of the monitored load is unknown, the first electrical characteristic signal sequence, the second electrical characteristic signal sequence and the third electrical characteristic signal sequence are subjected to similarity measurement with electrical characteristic sample sequences corresponding to a plurality of preset types of loads to obtain first, second and third similarity measurement values of electrical characteristics between the monitored load and the plurality of preset types of loads, and the first, second and third similarity measurement values of electrical characteristics between the monitored load and the plurality of preset types of loads are input into a first pre-trained probability neural network to obtain the load type of the monitored load; When the load type of the monitored load is known, the first electrical characteristic signal sequence, the second electrical characteristic signal sequence and the third electrical characteristic signal sequence are subjected to similarity measurement with electrical characteristic sample sequences corresponding to the monitored load to obtain first, second and third similarity measurement values for fault analysis, and the first, second and third similarity measurement values for fault analysis are input into a second pre-trained probability neural network to obtain a direct current fault arc detection result of the monitored load; the direct current fault arc detection result is normal operation or existence of a direct current fault arc.

2. The DC fault arc detection method of claim 1, wherein, The first electrical characteristic signal sequence comprises a first current characteristic signal sequence and a first voltage characteristic signal sequence, the second electrical characteristic signal sequence comprises a second current characteristic signal sequence and a second voltage characteristic signal sequence, and the third electrical characteristic signal sequence comprises a third current characteristic signal sequence and a third voltage characteristic signal sequence; The electrical characteristic sample sequence corresponding to each preset type of load comprises a first current characteristic sample sequence, a second current characteristic sample sequence, a third current characteristic sample sequence, a first voltage characteristic sample sequence, a second voltage characteristic sample sequence and a third voltage characteristic sample sequence when the preset type of load is in normal operation.

3. The DC fault arc detection method of claim 2, wherein, The similarity measurement of the first electrical characteristic signal sequence, the second electrical characteristic signal sequence and the third electrical characteristic signal sequence with respect to the electrical characteristic sample sequences corresponding to the plurality of preset types of loads respectively obtains the first similarity measurement value, the second similarity measurement value and the third similarity measurement value of the electrical characteristic between the monitored load and the plurality of preset types of loads, and further comprises: The Euclidean distance between the first current characteristic signal sequence and the first current characteristic sample sequence corresponding to each preset type of load is calculated, and the calculated Euclidean distance is output as the first current similarity measurement value; and the Euclidean distance between the first voltage characteristic signal sequence and the first voltage characteristic sample sequence corresponding to each preset type of load is calculated, and the calculated Euclidean distance is output as the first voltage similarity measurement value; The Euclidean distance between the second current characteristic signal sequence and the second current characteristic sample sequence corresponding to each preset type of load is calculated, and the calculated Euclidean distance is output as the second current similarity measurement value; and the Euclidean distance between the second voltage characteristic signal sequence and the second voltage characteristic sample sequence corresponding to each preset type of load is calculated, and the calculated Euclidean distance is output as the second voltage similarity measurement value; The Euclidean distance between the third current characteristic signal sequence and the third current characteristic sample sequence corresponding to each preset type of load is calculated, and the calculated Euclidean distance is output as the third current similarity measurement value; and the Euclidean distance between the third voltage characteristic signal sequence and the third voltage characteristic sample sequence corresponding to each preset type of load is calculated, and the calculated Euclidean distance is output as the third voltage similarity measurement value; The similarity measurement of the first electrical characteristic signal sequence, the second electrical characteristic signal sequence and the third electrical characteristic signal sequence with respect to the electrical characteristic sample sequences corresponding to the plurality of preset types of loads respectively obtains the first similarity measurement value, the second similarity measurement value and the third similarity measurement value of the electrical characteristic between the monitored load and the plurality of preset types of loads, and further comprises:

4. The DC fault arc detection method of claim 2, wherein, The similarity measurement of the first electrical characteristic signal sequence, the second electrical characteristic signal sequence and the third electrical characteristic signal sequence with respect to the electrical characteristic sample sequences corresponding to the plurality of preset types of loads respectively obtains the first similarity measurement value, the second similarity measurement value and the third similarity measurement value of the electrical characteristic between the monitored load and the plurality of preset types of loads, and further comprises: The Euclidean distance between the first current characteristic signal sequence and the first current characteristic sample sequence corresponding to each preset type of load is calculated, and the calculated Euclidean distance is output as the first current similarity measurement value; and the Euclidean distance between the first voltage characteristic signal sequence and the first voltage characteristic sample sequence corresponding to each preset type of load is calculated, and the calculated Euclidean distance is output as the first voltage similarity measurement value; calculating the Euclidean distance between the second current feature signal sequence and the second current feature sample sequence corresponding to the monitored load, and outputting the calculated Euclidean distance as a second current similarity measure value; and calculating the Euclidean distance between the second voltage feature signal sequence and the second voltage feature sample sequence corresponding to the monitored load, and outputting the calculated Euclidean distance as a second voltage similarity measure value; calculating the Euclidean distance between the third current feature signal sequence and the third current feature sample sequence corresponding to the monitored load, and outputting the calculated Euclidean distance as a third current similarity measure value; and calculating the Euclidean distance between the third voltage feature signal sequence and the third voltage feature sample sequence corresponding to the monitored load, and outputting the calculated Euclidean distance as a third voltage similarity measure value; wherein the similarity measure values of the first, second and third electrical feature signal sequences of the monitored load and the electrical feature sample sequence include first, second and third current similarity measure values, and first, second and third voltage similarity measure values.

5. The DC fault arc detection method of claim 2, wherein, The method further comprises: After wavelet packet decomposition and reconstruction, intrinsic time-scale decomposition and reconstruction of the to-be-analyzed electrical signal sequence, and obtaining the first, second and third electrical feature signal sequences, the first, second, third current feature signal sequences, the first, second and third voltage feature signal sequences are respectively divided into equal-length sub-segments, and then the average value of each sub-segment is calculated to correspond to represent each sub-segment, so as to convert the first, second, third current feature signal sequences, the first, second and third voltage feature signal sequences into low-dimensional signal sequences.

6. A DC fault arc detection apparatus characterized by, It comprises: a sampling module for sampling the electrical signal of the monitored load according to a preset sampling frequency to obtain an electrical signal sequence, the electrical signal sequence including a loop current signal sequence and an arc voltage signal sequence; a preprocessing module for taking out a to-be-analyzed electrical signal sequence matching a preset time window from the electrical signal sequence, the to-be-analyzed electrical signal sequence including a to-be-analyzed current signal sequence and a voltage signal sequence; a feature extraction module for decomposing and reconstructing the to-be-analyzed electrical signal sequence by wavelet packet to obtain a first electrical feature signal sequence, decomposing and reconstructing the to-be-analyzed electrical signal sequence by intrinsic time-scale to obtain a second electrical feature signal sequence, and decomposing and reconstructing the to-be-analyzed electrical signal sequence by classical modal to obtain a third electrical feature signal sequence; The load type identification module is configured to, when the load type of the monitored load is unknown, respectively perform similarity measurement on the first electrical characteristic signal sequence, the second electrical characteristic signal sequence, and the third electrical characteristic signal sequence and electrical characteristic sample sequences corresponding to a plurality of preset type loads to obtain first, second, and third similarity measurement values of electrical characteristics between the monitored load and the plurality of preset type loads, input the first, second, and third similarity measurement values of electrical characteristics between the monitored load and the plurality of preset type loads into a first pre-trained probability neural network to obtain the load type of the monitored load. The load fault identification module is configured to, when the load type of the monitored load is known, respectively perform similarity measurement on the first electrical characteristic signal sequence, the second electrical characteristic signal sequence, and the third electrical characteristic signal sequence and electrical characteristic sample sequences corresponding to the monitored load to obtain first, second, and third similarity measurement values for fault analysis, and input the first, second, and third similarity measurement values for fault analysis into a second pre-trained probability neural network to obtain a direct current fault arc detection result of the monitored load; the direct current fault arc detection result is normal operation or existence of a direct current fault arc.

7. The DC fault arc detection apparatus of claim 6, wherein, The first electrical characteristic signal sequence includes a first current characteristic signal sequence and a first voltage characteristic signal sequence, the second electrical characteristic signal sequence includes a second current characteristic signal sequence and a second voltage characteristic signal sequence, and the third electrical characteristic signal sequence includes a third current characteristic signal sequence and a third voltage characteristic signal sequence. The electrical characteristic sample sequence corresponding to each preset type load includes a first current characteristic sample sequence, a second current characteristic sample sequence, a third current characteristic sample sequence, a first voltage characteristic sample sequence, a second voltage characteristic sample sequence, and a third voltage characteristic sample sequence of the preset type load in normal operation.

8. The DC fault arc detection apparatus of claim 7, wherein, The load type identification module is further configured to: calculate Euclidean distances between the first current characteristic signal sequence and the first current characteristic sample sequence corresponding to each preset type load, and output the calculated Euclidean distances as first current similarity measurement values; and calculate Euclidean distances between the first voltage characteristic signal sequence and the first voltage characteristic sample sequence corresponding to each preset type load, and output the calculated Euclidean distances as first voltage similarity measurement values; calculate Euclidean distances between the second current characteristic signal sequence and the second current characteristic sample sequence corresponding to each preset type load, and output the calculated Euclidean distances as second current similarity measurement values; and calculate Euclidean distances between the second voltage characteristic signal sequence and the second voltage characteristic sample sequence corresponding to each preset type load, and output the calculated Euclidean distances as second voltage similarity measurement values; and calculate Euclidean distances between the third current characteristic signal sequence and the third current characteristic sample sequence corresponding to each preset type load, and output the calculated Euclidean distances as third current similarity measurement values; and calculate Euclidean distances between the third voltage characteristic signal sequence and the third voltage characteristic sample sequence corresponding to each preset type load, and output the calculated Euclidean distances as third voltage similarity measurement values. calculating the Euclidean distance between the third current characteristic signal sequence and the third current characteristic sample sequence corresponding to each preset type of load, and outputting the calculated Euclidean distance as a third current similarity measurement value; and calculating the Euclidean distance between the third voltage characteristic signal sequence and the third voltage characteristic sample sequence corresponding to each preset type of load, and outputting the calculated Euclidean distance as a third voltage similarity measurement value; The similarity measurement values of the electrical characteristics between the monitored load and the plurality of preset types of load include first current similarity measurement values, second current similarity measurement values, third current similarity measurement values, first voltage similarity measurement values, second voltage similarity measurement values, and third voltage similarity measurement values.

9. The DC fault arc detection apparatus of claim 7, wherein, The load fault identification module is further configured to: calculate the Euclidean distance between the first current characteristic signal sequence and the first current characteristic sample sequence corresponding to the monitored load, and output the calculated Euclidean distance as a first current similarity measurement value; and calculate the Euclidean distance between the first voltage characteristic signal sequence and the first voltage characteristic sample sequence corresponding to the monitored load, and output the calculated Euclidean distance as a first voltage similarity measurement value; calculate the Euclidean distance between the second current characteristic signal sequence and the second current characteristic sample sequence corresponding to the monitored load, and output the calculated Euclidean distance as a second current similarity measurement value; and calculate the Euclidean distance between the second voltage characteristic signal sequence and the second voltage characteristic sample sequence corresponding to the monitored load, and output the calculated Euclidean distance as a second voltage similarity measurement value; calculate the Euclidean distance between the third current characteristic signal sequence and the third current characteristic sample sequence corresponding to the monitored load, and output the calculated Euclidean distance as a third current similarity measurement value; and calculate the Euclidean distance between the third voltage characteristic signal sequence and the third voltage characteristic sample sequence corresponding to the monitored load, and output the calculated Euclidean distance as a third voltage similarity measurement value; The similarity measurement values of the first electrical characteristic signal sequence, the second electrical characteristic signal sequence, and the third electrical characteristic signal sequence of the monitored load and the electrical characteristic sample sequence include first current similarity measurement values, second current similarity measurement values, third current similarity measurement values, first voltage similarity measurement values, second voltage similarity measurement values, and third voltage similarity measurement values.

10. The DC fault arc detection apparatus of claim 7, wherein, The feature extraction module is further configured to: After wavelet packet decomposition and reconstruction, intrinsic time-scale decomposition and reconstruction are performed on the to-be-analyzed electrical signal sequence, and first, second and third electrical characteristic signal sequences are obtained, the first, second and third current characteristic signal sequences, the first, second and third voltage characteristic signal sequences are then divided into equal-length subsegments, and the average value of each subsegment is calculated to correspond to each subsegment, so as to convert the first, second and third current characteristic signal sequences and the first, second and third voltage characteristic signal sequences into low-dimensional signal sequences.

Images