A method and apparatus for monitoring the state of a direct current fault arc, computer program product

Abstract

The application provides a direct-current fault arc state monitoring method and device and a computer program product, obtains current and voltage signal sequences by sampling electrical signals of a monitored load. The electrical characteristic signal sequences are extracted by using wavelet packet decomposition and reconstruction technology. When the load type is unknown, similarity measurement is performed on the electrical characteristic sample sequences of multiple preset types of loads, and a deep belief network model is used to determine the load type. When the load type is known, fault analysis is performed by using the similarity measurement value, and a second deep belief network model is used to output a direct-current fault arc state monitoring result, so as to determine whether the load is normally operated or a direct-current fault arc exists, thereby improving the accuracy and efficiency of monitoring.

Description

Technical Field

[0001] This invention relates to the field of power technology, and in particular to a method and device for monitoring the state of DC fault arcing, and a computer program product. Background Technology

[0002] With the continuous development of society, the improvement of living standards, and the advancement of technology, the development of electricity is closely related to the power industry. However, the existence or occurrence of fault arcs can seriously affect the safe and stable operation of the power system. Fault arcs are classified into DC fault arcs and AC fault arcs according to current type, both of which are major factors causing electrical fires. In traditional power systems, AC power accounts for a larger proportion, and research on the monitoring and diagnosis of AC arcs is relatively mature. However, DC fault arcs are currently not given enough attention, and research on their diagnostic methods is still in the development stage.

[0003] A direct current (DC) arc is a high-energy instantaneous current, typically manifested as a gas discharge. The main difference between a DC arc and an alternating current (AC) arc is that a DC arc does not experience zero-crossing. Therefore, traditional circuit protection devices are largely ineffective. In other words, once a DC arc fault occurs, the faulty section will remain in a stable arcing state for an extended period without extinguishing, potentially leading to fires and other electrical accidents. Summary of the Invention

[0004] The technical problem to be solved by the embodiments of the present invention is to provide a method and device for monitoring the state of DC fault arc, and a computer program product, so as to improve the accuracy and efficiency of monitoring the state of DC fault arc.

[0005] To address the aforementioned technical problems, embodiments of the present invention provide a method for monitoring the state of a DC fault arc, comprising the following steps:

[0006] The electrical signal sequence is obtained by sampling the electrical signal of the monitored load according to a preset sampling frequency. The electrical signal sequence includes a loop current signal sequence and an arc voltage signal sequence.

[0007] Extract the electrical signal sequence to be analyzed from the electrical signal sequence that matches the preset time window. The electrical signal sequence to be analyzed includes the current signal sequence and the voltage signal sequence to be analyzed.

[0008] The electrical signal sequence to be analyzed is decomposed and reconstructed by wavelet packet to obtain an electrical feature signal sequence, which includes a current feature signal sequence and a voltage feature signal sequence.

[0009] When the load type of the monitored load is unknown, a similarity metric is performed between the electrical feature signal sequence and the electrical feature sample sequences corresponding to multiple preset load types to obtain a similarity metric value between the monitored load and the multiple preset load types. The similarity metric value between the monitored load and the multiple preset load types is then input into a pre-trained first deep belief network model to obtain the load type of the monitored load. Each preset load type's corresponding electrical feature sample sequence includes current feature sample sequences and voltage feature sample sequences of the preset load under various operating conditions.

[0010] When the load type of the monitored load is known, a similarity metric is performed between the electrical feature signal sequence and the electrical feature sample sequence corresponding to the monitored load to obtain a similarity metric value for fault analysis. The similarity metric value for fault analysis is then input into a pre-trained second deep belief network model to obtain the DC fault arc state monitoring result of the monitored load. The DC fault arc state monitoring result indicates normal operation or the presence of a DC fault arc.

[0011] Optionally, the method further includes:

[0012] After performing wavelet packet decomposition and reconstruction on the electrical signal sequence to be analyzed to obtain the electrical feature signal sequence, the current feature signal sequence and the voltage feature signal sequence are further divided into equal-length sub-segments. Then, the average value of each sub-segment is calculated to represent each sub-segment, thereby converting the current feature signal sequence and the voltage feature signal sequence into a low-dimensional signal sequence.

[0013] Optionally, the step of performing a similarity measurement on the electrical feature signal sequence and the electrical feature sample sequences corresponding to multiple preset types of loads to obtain a similarity measurement value of the electrical features between the monitored load and the multiple preset types of loads further includes:

[0014] A similarity metric is performed between the current feature signal sequence and the current feature sample sequence corresponding to each preset type of load. A distance matrix is ​​created to store the distance between each pair of points between the current feature signal sequence and the current feature sample sequence currently being measured for similarity. The distance between each point in the current feature signal sequence and each point in the current feature sample sequence is calculated. A path from the upper left corner to the lower right corner is found in the distance matrix such that the cumulative distance on the path is minimized. This path represents the best alignment between the current feature signal sequence and each current feature sample sequence. Finally, the cumulative distance on the path is output as the current similarity metric value between the current feature signal sequence and the current feature sample sequence currently being measured for similarity.

[0015] A similarity metric is performed between the voltage feature signal sequence and the voltage feature sample sequence corresponding to each preset type of load. A distance matrix is ​​created to store the distance between each pair of points between the voltage feature signal sequence and the voltage feature sample sequence currently being measured for similarity. The distance between each point in the voltage feature signal sequence and each point in the voltage feature sample sequence is calculated. A path from the upper left corner to the lower right corner is found in the distance matrix such that the cumulative distance on the path is minimized. This path represents the best alignment between the voltage feature signal sequence and each voltage feature sample sequence. Finally, the cumulative distance on the path is output as the voltage similarity metric value between the voltage feature signal sequence and the voltage feature sample sequence currently being measured for similarity.

[0016] The similarity measurement values ​​of electrical characteristics between the monitored load and the various preset types of loads all include current similarity measurement values ​​and voltage similarity measurement values.

[0017] Optionally, the step of performing a similarity measurement on the electrical characteristic signal sequence and the electrical characteristic sample sequence corresponding to the monitored load to obtain a similarity measurement value for fault analysis further includes:

[0018] A similarity metric is performed between the current feature signal sequence and the current feature sample sequence corresponding to the monitored load. A distance matrix is ​​created to store the distance between each pair of points between the current feature signal sequence and the current feature sample sequence corresponding to the monitored load. The distance between each point in the current feature signal sequence and each point in the current feature sample sequence is calculated. A path from the upper left corner to the lower right corner is found in the distance matrix such that the cumulative distance on the path is minimized. This path represents the best alignment between the current feature signal sequence and each current feature sample sequence. Finally, the cumulative distance on the path is output as the current similarity metric between the current feature signal sequence and the current feature sample sequence corresponding to the monitored load.

[0019] When measuring the similarity between the voltage feature signal sequence and the voltage feature sample sequence corresponding to the monitored load, a distance matrix is ​​created to store the distance between each pair of points between the voltage feature signal sequence and the voltage feature sample sequence corresponding to the monitored load. The distance between each point in the voltage feature signal sequence and each point in the voltage feature sample sequence is calculated. A path from the upper left corner to the lower right corner is found in the distance matrix such that the cumulative distance on the path is minimized. This path represents the best alignment between the voltage feature signal sequence and each voltage feature sample sequence. Finally, the cumulative distance on the path is output as the voltage similarity metric between the voltage feature signal sequence and the voltage feature sample sequence corresponding to the monitored load.

[0020] The similarity measure of electrical characteristics between the current characteristic signal sequence and the current characteristic sample sequence corresponding to the monitored load includes current similarity measure and voltage similarity measure.

[0021] This invention also provides a DC fault arc condition monitoring device, comprising:

[0022] The sampling module is used to sample the electrical signals 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;

[0023] A preprocessing module is used to extract an electrical signal sequence to be analyzed that matches a preset time window from the electrical signal sequence. The electrical signal sequence to be analyzed includes a current signal sequence and a voltage signal sequence to be analyzed.

[0024] The feature extraction module is used to perform wavelet packet decomposition and reconstruction on the electrical signal sequence to be analyzed to obtain an electrical feature signal sequence, which includes a current feature signal sequence and a voltage feature signal sequence.

[0025] The load type identification module is used to, when the load type of the monitored load is unknown, perform a similarity measurement on the electrical feature signal sequence and the electrical feature sample sequences corresponding to multiple preset load types to obtain a similarity measurement value of the electrical features between the monitored load and the multiple preset load types. The similarity measurement value of the electrical features between the monitored load and the multiple preset load types is then input into a pre-trained first deep belief network model to obtain the load type of the monitored load. Each preset load type's corresponding electrical feature sample sequence includes a current feature sample sequence and a voltage feature sample sequence of the preset load under various operating conditions.

[0026] The DC fault identification module is used to, when the load type of the monitored load is known, perform a similarity measurement on the electrical feature signal sequence and the electrical feature sample sequence corresponding to the monitored load to obtain a similarity measurement value for fault analysis, and input the similarity measurement value for fault analysis into a pre-trained second deep belief network model to obtain the DC fault arc state monitoring result of the monitored load; the DC fault arc state monitoring result is normal operation or the presence of a DC fault arc.

[0027] Optionally, the feature extraction module is further configured to, after performing wavelet packet decomposition and reconstruction on the electrical signal sequence to be analyzed to obtain the electrical feature signal sequence, further divide the current feature signal sequence and the voltage feature signal sequence into equal-length sub-segments, and then calculate the average value of each sub-segment to represent each sub-segment, thereby converting the current feature signal sequence and the voltage feature signal sequence into a low-dimensional signal sequence.

[0028] Optionally, the load type identification module is further configured to:

[0029] A similarity metric is performed between the current feature signal sequence and the current feature sample sequence corresponding to each preset type of load. A distance matrix is ​​created to store the distance between each pair of points between the current feature signal sequence and the current feature sample sequence currently being measured for similarity. The distance between each point in the current feature signal sequence and each point in the current feature sample sequence is calculated. A path from the upper left corner to the lower right corner is found in the distance matrix such that the cumulative distance on the path is minimized. This path represents the best alignment between the current feature signal sequence and each current feature sample sequence. Finally, the cumulative distance on the path is output as the current similarity metric value between the current feature signal sequence and the current feature sample sequence currently being measured for similarity.

[0030] A similarity metric is performed between the voltage feature signal sequence and the voltage feature sample sequence corresponding to each preset type of load. A distance matrix is ​​created to store the distance between each pair of points between the voltage feature signal sequence and the voltage feature sample sequence currently being measured for similarity. The distance between each point in the voltage feature signal sequence and each point in the voltage feature sample sequence is calculated. A path from the upper left corner to the lower right corner is found in the distance matrix such that the cumulative distance on the path is minimized. This path represents the best alignment between the voltage feature signal sequence and each voltage feature sample sequence. Finally, the cumulative distance on the path is output as the voltage similarity metric value between the voltage feature signal sequence and the voltage feature sample sequence currently being measured for similarity.

[0031] The similarity measurement values ​​of electrical characteristics between the monitored load and the various preset types of loads all include current similarity measurement values ​​and voltage similarity measurement values.

[0032] Optionally, the DC fault identification module is further configured to:

[0033] A similarity metric is performed between the current feature signal sequence and the current feature sample sequence corresponding to the monitored load. A distance matrix is ​​created to store the distance between each pair of points between the current feature signal sequence and the current feature sample sequence corresponding to the monitored load. The distance between each point in the current feature signal sequence and each point in the current feature sample sequence is calculated. A path from the upper left corner to the lower right corner is found in the distance matrix such that the cumulative distance on the path is minimized. This path represents the best alignment between the current feature signal sequence and each current feature sample sequence. Finally, the cumulative distance on the path is output as the current similarity metric between the current feature signal sequence and the current feature sample sequence corresponding to the monitored load.

[0034] When measuring the similarity between the voltage feature signal sequence and the voltage feature sample sequence corresponding to the monitored load, a distance matrix is ​​created to store the distance between each pair of points between the voltage feature signal sequence and the voltage feature sample sequence corresponding to the monitored load. The distance between each point in the voltage feature signal sequence and each point in the voltage feature sample sequence is calculated. A path from the upper left corner to the lower right corner is found in the distance matrix such that the cumulative distance on the path is minimized. This path represents the best alignment between the voltage feature signal sequence and each voltage feature sample sequence. Finally, the cumulative distance on the path is output as the voltage similarity metric between the voltage feature signal sequence and the voltage feature sample sequence corresponding to the monitored load.

[0035] The similarity measure of electrical characteristics between the current characteristic signal sequence and the current characteristic sample sequence corresponding to the monitored load includes current similarity measure and voltage similarity measure.

[0036] This invention also provides a computer program product, including computer program instructions, which instruct a computer device to perform operations corresponding to the methods described above.

[0037] The above-mentioned DC fault arc condition monitoring method, device, and computer program product have the following beneficial effects:

[0038] (1) By performing wavelet packet decomposition and reconstruction on the electrical signal sequence to be analyzed, the current feature signal sequence and voltage feature signal sequence are extracted, effectively capturing the subtle features of DC fault arc, making the monitoring results more accurate.

[0039] (2) It can identify the load type of the load to be monitored. When the load type is unknown, it can make intelligent judgment on the load type by measuring the similarity with the electrical characteristic sample sequence of multiple preset load types, thereby improving the adaptability of the monitoring method to different loads.

[0040] (3) The deep belief network model is used to monitor the state of the fault arc. Compared with traditional methods, the deep learning model has stronger feature extraction and classification capabilities, and can provide accurate monitoring results in a short time, thus improving monitoring efficiency.

[0041] (4) By monitoring the DC fault arc status in real time, potential fault arcs can be detected and dealt with in a timely manner, which effectively reduces the risk of equipment damage, power outages, and electrical fires caused by DC fault arcs, ensuring the safe and stable operation of the power system, reducing the cost of equipment maintenance and replacement caused by fault arcs, and lowering the operation and maintenance cost of the power system. Attached Figure Description

[0042] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, obtaining other drawings based on these drawings without creative effort still falls within the scope of the present invention.

[0043] Figure 1 This is a flowchart of a DC fault arc state monitoring method according to one embodiment of the present invention.

[0044] Figure 2 This is a schematic diagram of a deep belief network model in one embodiment of the present invention.

[0045] Figure 3 This is a structural diagram of a DC fault arc condition monitoring device according to one embodiment of the present invention. Detailed Implementation

[0046] The detailed description of the accompanying drawings is intended to illustrate the present embodiments of this application and is not intended to represent only the forms in which this 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 this application.

[0047] See Figure 1An embodiment of the present invention provides a method for monitoring the state of a DC fault arc, comprising the following steps:

[0048] Step S10: Sample 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.

[0049] Specifically, the description in step S10 involves the basic process of electrical signal sampling. The sampling frequency refers to the number of samples taken per unit time, usually in Hertz (Hz). The preset sampling frequency is a parameter set before signal acquisition, which determines the sampling time interval. This frequency needs to be high enough to ensure that the original signal can be accurately recovered.

[0050] The electrical signals of the monitored load refer to the electrical signals generated by the monitored target (i.e., the load) during operation, including current and voltage signals. The load can be any device or system that uses direct current.

[0051] Sampling refers to recording the values ​​of electrical signals at specific points in time according to a preset sampling frequency. This process is the first step in digitization, converting continuous analog signals into discrete digital signals.

[0052] A series of values ​​obtained through sampling are arranged in chronological order to form an 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 change 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 change of voltage measured at both ends of the load, which can reveal the existence and characteristics of the arc, because the generation of the arc will cause specific changes in the voltage signal.

[0053] Step S20: Extract the electrical signal sequence to be analyzed that matches the preset time window from the electrical signal sequence. The electrical signal sequence to be analyzed includes the current signal sequence and the voltage signal sequence to be analyzed.

[0054] Specifically, a specific time window is set before analysis. This time window is determined based on the needs of analyzing and monitoring the characteristics of the fault arc, and it represents the duration of the signal segment extracted from the electrical signal sequence. The electrical signal sequence to be analyzed refers to the portion extracted from the entire electrical signal sequence that matches the preset time window. This portion of the signal sequence will be used as the object of analysis to detect the characteristics of the fault arc. The current signal sequence to be analyzed is a part of the electrical signal sequence to be analyzed, specifically representing the current changes within the preset time window. The voltage signal sequence to be analyzed is a part of the electrical signal sequence to be analyzed, representing the voltage changes within the preset time window.

[0055] Step S30: Perform wavelet packet decomposition and reconstruction on the electrical signal sequence to be analyzed to obtain an electrical feature signal sequence, which includes a current feature signal sequence and a voltage feature signal sequence.

[0056] Specifically, wavelet packet decomposition is a more refined signal processing method that decomposes not only the low-frequency components of a signal but also further decomposes the high-frequency components. This method can more effectively analyze the detailed parts of a signal, especially useful for detecting transient changes and non-stationary characteristics. Wavelet packet decomposition breaks down the original signal into multiple sub-bands of different frequencies, which can represent the signal characteristics in greater detail. After wavelet packet decomposition, certain sub-bands can be selectively reconstructed as needed to extract features useful for fault diagnosis. The reconstruction process involves recombining the decomposed sub-band signals to form new signal sequences that highlight specific features in the original signal.

[0057] The new signal sequences obtained through wavelet packet decomposition and reconstruction contain fault diagnosis-related feature information from the original signal. These feature signal sequences are usually simpler and easier to analyze and identify fault modes than the original signal. The current feature signal sequence is the current signal after wavelet packet decomposition and reconstruction; it contains current-related fault features such as current abrupt changes and spikes. The voltage feature signal sequence is the voltage signal after wavelet packet decomposition and reconstruction; it contains voltage-related fault features such as voltage fluctuations and drops.

[0058] Specifically, the operation procedure for step S30 is as follows:

[0059] Step 1: Perform wavelet packet decomposition on the current signal sequence and voltage signal sequence to be analyzed.

[0060] Step 2: Based on the characteristics of the fault arc, select an appropriate sub-band for reconstruction in order to extract the characteristics of current and voltage.

[0061] Step 3: Reconstruct the current characteristic signal sequence and voltage characteristic signal sequence, which will be used for subsequent fault analysis and monitoring.

[0062] This step effectively extracts features useful for fault arc detection from complex electrical signals, providing important data support for subsequent fault diagnosis.

[0063] Step S40: When the load type of the monitored load is unknown, a similarity measurement is performed on the electrical feature signal sequence and the electrical feature sample sequences corresponding to multiple preset load types to obtain the similarity measurement value of the electrical features between the monitored load and the multiple preset load types. The similarity measurement value of the electrical features between the monitored load and the multiple preset load types is input into a pre-trained first deep belief network model to obtain the load type of the monitored load. Among them, the electrical feature sample sequence corresponding to each preset load type includes the current feature sample sequence and voltage feature sample sequence of the preset load under various operating conditions.

[0064] Specifically, the unknown load type of the monitored load means that during the monitoring process, we do not know what type of equipment or system the monitored load belongs to. Different load types may produce different electrical characteristics, so it is necessary to determine the load type first in order to more accurately monitor fault arcs.

[0065] The electrical characteristic signal sequence is obtained in step S30 through wavelet packet decomposition and reconstruction, and includes the current characteristic signal sequence and the voltage characteristic signal sequence.

[0066] The electrical characteristic sample sequence corresponding to various preset load types refers to the electrical characteristic samples of different types of loads that have been collected and preprocessed during the system design and training phases. These sample sequences include current characteristic sample sequences and voltage characteristic sample sequences under various operating conditions (such as normal operation, different load conditions, etc.).

[0067] Similarity measurement refers to calculating the similarity between the electrical characteristic signal sequence to be analyzed and the electrical characteristic sample sequence of each preset type of load. For example, similarity measurement can be a statistical method such as Euclidean distance, cosine similarity, or correlation coefficient, used to evaluate the degree of matching between the two sequences.

[0068] The first deep belief network model is a pre-trained deep learning model that can predict or classify load types based on input similarity metrics. Deep belief networks are neural networks with multiple hidden layers and are suitable for feature learning and classification tasks.

[0069] Step S50: When the load type of the monitored load is known, a similarity metric is performed between the electrical feature signal sequence and the electrical feature sample sequence corresponding to the monitored load to obtain a similarity metric value for fault analysis. The similarity metric value for fault analysis is then input into a pre-trained second deep belief network model to obtain the DC fault arc state monitoring result of the monitored load. The DC fault arc state monitoring result indicates normal operation or the presence of a DC fault arc.

[0070] Specifically, knowing the load type of the monitored load means that we have determined the type of the monitored load during the monitoring process. This information is crucial for monitoring fault arcs because different types of loads may have different electrical characteristics and fault modes.

[0071] The electrical feature signal sequence is the electrical feature signal sequence obtained by wavelet packet decomposition and reconstruction in step S30. It includes the current feature signal sequence and the voltage feature signal sequence.

[0072] The electrical characteristic sample sequence corresponding to the monitored load refers to the electrical characteristic sample sequence pre-collected and prepared for known load types. These sample sequences represent the electrical characteristics of that type of load under normal operation and various fault conditions.

[0073] Similarity measurement refers to calculating the similarity between the electrical characteristic signal sequence to be analyzed and the electrical characteristic sample sequence of known load types. The similarity measurement value can reflect the degree of matching between the current signal and the sample sequence, thus providing a basis for fault analysis.

[0074] The second deep belief network model is a pre-trained deep learning model specifically designed to determine whether a load is in a DC fault arc state based on the similarity metric of the input. This model has been trained with a large amount of data and is able to identify the characteristics of fault arcs.

[0075] Specifically, Deep Belief Network (DBN) is a probabilistic generative model. Compared to traditional neural network models, DBN establishes a joint distribution of labeled observations among data, thus enabling it not only to identify features and classify data but also to generate training data with the highest probability. A DBN network consists of many layers of Restricted Boltzmann Machines (RBMs) and one Backpropagation (BP) layer. During training, each RBM layer is first trained unsupervised to obtain as much feature information as possible. Its final layer is a BP network that receives the feature vectors from the preceding layers, making it a semi-supervised probabilistic prediction model. The structure of DBN is as follows: Figure 2As shown, a Deep Belief Network (DBN) consists of multiple layers of neurons, which are divided into dominant and recessive types. Dominant neurons receive input signals, while recessive neurons extract features. The objectives of this invention can be achieved by training a first deep belief network and a second deep belief network according to different needs.

[0076] Furthermore, step S30 further includes:

[0077] After performing wavelet packet decomposition and reconstruction on the electrical signal sequence to be analyzed to obtain the electrical feature signal sequence, the current feature signal sequence and the voltage feature signal sequence are further divided into equal-length sub-segments. Then, the average value of each sub-segment is calculated to represent each sub-segment, thereby converting the current feature signal sequence and the voltage feature signal sequence into a low-dimensional signal sequence.

[0078] Specifically, the above description further refines step S30, illustrating how to process the obtained electrical feature signal sequence after wavelet packet decomposition and reconstruction to reduce the signal dimensionality and extract key information. Dividing into equal-length segments means dividing the current and voltage feature signal sequences into several equal-length parts. The length of each segment is preset to ensure the uniformity and comparability of signal processing. For each segment, the average of all its values ​​is calculated. This average will serve as the representative value of that segment, reflecting its general signal characteristics. By calculating the average of each segment, the original current and voltage feature signal sequences are transformed into a series of averages, which constitute a new, lower-dimensional signal sequence. This process is essentially a dimensionality reduction, simplifying the signal data and making subsequent analysis and calculation more efficient. Further, step S40, which involves measuring the similarity between the electrical feature signal sequence and the electrical feature sample sequences corresponding to various preset load types to obtain the similarity measurement value of the electrical features between the monitored load and the various preset load types, further includes:

[0079] Step S401: Perform similarity measurement on the current feature signal sequence and the current feature sample sequence corresponding to each preset type of load, create a distance matrix to store the distance between each pair of points between the current feature signal sequence and the current feature sample sequence currently being measured for similarity, calculate the distance between each point in the current feature signal sequence and each point in the current feature sample sequence, and find a path from the upper left corner to the lower right corner in the distance matrix such that the cumulative distance on the path is minimized. This path represents the best alignment between the current feature signal sequence and each current feature sample sequence. Finally, output the cumulative distance on the path as the current similarity measurement value between the current feature signal sequence and the current feature sample sequence currently being measured for similarity.

[0080] Specifically, in step S401, the number of rows and columns of the matrix correspond to the number of points in the two sequences, respectively. For each point in the current feature signal sequence, the distance between it and each point in the current feature sample sequence is calculated. This distance can be Euclidean distance, Manhattan distance, etc., depending on the application requirements. In the distance matrix, a path is found from the top left corner (representing the start of the two sequences) to the bottom right corner (representing the end of the two sequences). The cumulative distance on this path is minimized. This path is not a straight line; it can be curved to find the best alignment between the two time series. In this embodiment, the time series are allowed to stretch and twist on the time axis to find the best match between the two sequences. This means that even if the two sequences are different in time (e.g., one sequence may be faster or slower than the other), a correspondence can still be found between them. Along the found path, the sum of the distances in all cells on the path is calculated. This cumulative distance represents the similarity measure between the two sequences. The smaller the cumulative distance, the more similar the two sequences are.

[0081] Step S402: Perform similarity measurement on the voltage feature signal sequence and the voltage feature sample sequence corresponding to each preset type of load, create a distance matrix to store the distance between each pair of points between the voltage feature signal sequence and the voltage feature sample sequence currently being measured for similarity, calculate the distance between each point in the voltage feature signal sequence and each point in the voltage feature sample sequence, and find a path from the upper left corner to the lower right corner in the distance matrix such that the cumulative distance on the path is minimized. This path represents the best alignment between the voltage feature signal sequence and each voltage feature sample sequence. Finally, output the cumulative distance on the path as the voltage similarity measurement value between the voltage feature signal sequence and the voltage feature sample sequence currently being measured for similarity.

[0082] Specifically, in step S402, the number of rows and columns of the matrix correspond to the number of points in the two sequences, respectively. For each point in the voltage feature signal sequence, the distance between it and each point in the voltage feature sample sequence is calculated. This distance can be Euclidean distance, Manhattan distance, etc., depending on the application requirements. In the distance matrix, a path is found from the top left corner (representing the start of the two sequences) to the bottom right corner (representing the end of the two sequences). The cumulative distance on this path is minimized. This path is not a straight line; it can be curved to find the best alignment between the two time series. In this embodiment, the time series are allowed to stretch and twist on the time axis to find the best match between the two sequences. This means that even if the two sequences are different in time (e.g., one sequence may be faster or slower than the other), a correspondence can still be found between them. Along the found path, the sum of the distances in all cells on the path is calculated. This cumulative distance represents the similarity measure between the two sequences. The smaller the cumulative distance, the more similar the two sequences are.

[0083] The electrical similarity measurement values ​​between the monitored load and the various preset types of loads include both current similarity measurement values ​​and voltage similarity measurement values.

[0084] Further, step S50, which involves performing a similarity measurement on the electrical characteristic signal sequence and the electrical characteristic sample sequence corresponding to the monitored load to obtain a similarity measurement value for fault analysis, further includes:

[0085] Step S501: Perform a similarity measurement on the current feature signal sequence and the current feature sample sequence corresponding to the monitored load, create a distance matrix to store the distance between each pair of points between the current feature signal sequence and the current feature sample sequence corresponding to the monitored load, calculate the distance between each point in the current feature signal sequence and each point in the current feature sample sequence, and find a path from the upper left corner to the lower right corner in the distance matrix such that the cumulative distance on the path is minimized. This path represents the best alignment between the current feature signal sequence and each current feature sample sequence. Finally, output the cumulative distance on the path as the current similarity measurement value between the current feature signal sequence and the current feature sample sequence corresponding to the monitored load.

[0086] Specifically, in step S501, the number of rows and columns of the matrix correspond to the number of points in the two sequences, respectively. For each point in the current feature signal sequence, the distance between it and each point in the current feature sample sequence is calculated. This distance can be Euclidean distance, Manhattan distance, etc., depending on the application requirements. In the distance matrix, a path is found from the top left corner (representing the start of the two sequences) to the bottom right corner (representing the end of the two sequences). The cumulative distance on this path is minimized. This path is not a straight line; it can be curved to find the best alignment between the two time series. In this embodiment, the time series are allowed to stretch and twist on the time axis to find the best match between the two sequences. This means that even if the two sequences are different in time (e.g., one sequence may be faster or slower than the other), a correspondence between them can still be found. Along the found path, the sum of the distances in all cells on the path is calculated. This cumulative distance represents the similarity measure between the two sequences. The smaller the cumulative distance, the more similar the two sequences are.

[0087] Step S502: When performing a similarity measurement between the voltage feature signal sequence and the voltage feature sample sequence corresponding to the monitored load, a distance matrix is ​​created to store the distance between each pair of points between the voltage feature signal sequence and the voltage feature sample sequence corresponding to the monitored load. The distance between each point in the voltage feature signal sequence and each point in the voltage feature sample sequence is calculated. A path from the upper left corner to the lower right corner is found in the distance matrix such that the cumulative distance on the path is minimized. This path represents the best alignment between the voltage feature signal sequence and each voltage feature sample sequence. Finally, the cumulative distance on the path is output as the voltage similarity measurement value between the voltage feature signal sequence and the voltage feature sample sequence corresponding to the monitored load.

[0088] Specifically, in step S502, the number of rows and columns of the matrix correspond to the number of points in the two sequences, respectively. For each point in the voltage feature signal sequence, the distance between it and each point in the voltage feature sample sequence is calculated. This distance can be Euclidean distance, Manhattan distance, etc., depending on the application requirements. In the distance matrix, a path is found from the top left corner (representing the start of the two sequences) to the bottom right corner (representing the end of the two sequences). The cumulative distance on this path is minimized. This path is not a straight line; it can be curved to find the best alignment between the two time series. In this embodiment, the time series are allowed to stretch and twist on the time axis to find the best match between the two sequences. This means that even if the two sequences are different in time (e.g., one sequence may be faster or slower than the other), a correspondence can still be found between them. Along the found path, the sum of the distances in all cells on the path is calculated. This cumulative distance represents the similarity measure between the two sequences. The smaller the cumulative distance, the more similar the two sequences are.

[0089] The similarity metric between the electrical characteristics of the current characteristic signal sequence and the current characteristic sample sequence corresponding to the monitored load includes a current similarity metric and a voltage similarity metric.

[0090] See Figure 3 Another embodiment of the present invention also provides a DC fault arc condition monitoring device, comprising:

[0091] Sampling module 1 is used to sample 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;

[0092] Preprocessing module 2 is used to extract the electrical signal sequence to be analyzed that matches a preset time window from the electrical signal sequence, wherein the electrical signal sequence to be analyzed includes the current signal sequence and the voltage signal sequence to be analyzed;

[0093] Feature extraction module 3 is used to perform wavelet packet decomposition and reconstruction on the electrical signal sequence to be analyzed to obtain an electrical feature signal sequence, wherein the electrical feature signal sequence includes a current feature signal sequence and a voltage feature signal sequence;

[0094] The load type identification module 4 is used to, when the load type of the monitored load is unknown, perform a similarity measurement on the electrical feature signal sequence and the electrical feature sample sequences corresponding to multiple preset load types to obtain a similarity measurement value of the electrical features between the monitored load and the multiple preset load types, and input the similarity measurement value of the electrical features between the monitored load and the multiple preset load types into a pre-trained first deep belief network model to obtain the load type of the monitored load; wherein, the electrical feature sample sequence corresponding to each preset load type includes the current feature sample sequence and voltage feature sample sequence of the preset load under various operating conditions;

[0095] The DC fault identification module 5 is used to, when the load type of the monitored load is known, perform a similarity measurement on the electrical feature signal sequence and the electrical feature sample sequence corresponding to the monitored load to obtain a similarity measurement value for fault analysis, and input the similarity measurement value for fault analysis into a pre-trained second deep belief network model to obtain the DC fault arc state monitoring result of the monitored load; the DC fault arc state monitoring result is normal operation or the presence of a DC fault arc.

[0096] Furthermore, the feature extraction module 3 is further configured to, after performing wavelet packet decomposition and reconstruction on the electrical signal sequence to be analyzed to obtain the electrical feature signal sequence, further divide the current feature signal sequence and the voltage feature signal sequence into equal-length sub-segments, and then calculate the average value of each sub-segment to represent each sub-segment, thereby converting the current feature signal sequence and the voltage feature signal sequence into a low-dimensional signal sequence.

[0097] Furthermore, the load type identification module 4 is further configured to:

[0098] A similarity metric is performed between the current feature signal sequence and the current feature sample sequence corresponding to each preset type of load. A distance matrix is ​​created to store the distance between each pair of points between the current feature signal sequence and the current feature sample sequence currently being measured for similarity. The distance between each point in the current feature signal sequence and each point in the current feature sample sequence is calculated. A path from the upper left corner to the lower right corner is found in the distance matrix such that the cumulative distance on the path is minimized. This path represents the best alignment between the current feature signal sequence and each current feature sample sequence. Finally, the cumulative distance on the path is output as the current similarity metric value between the current feature signal sequence and the current feature sample sequence currently being measured for similarity.

[0099] A similarity metric is performed between the voltage feature signal sequence and the voltage feature sample sequence corresponding to each preset type of load. A distance matrix is ​​created to store the distance between each pair of points between the voltage feature signal sequence and the voltage feature sample sequence currently being measured for similarity. The distance between each point in the voltage feature signal sequence and each point in the voltage feature sample sequence is calculated. A path from the upper left corner to the lower right corner is found in the distance matrix such that the cumulative distance on the path is minimized. This path represents the best alignment between the voltage feature signal sequence and each voltage feature sample sequence. Finally, the cumulative distance on the path is output as the voltage similarity metric value between the voltage feature signal sequence and the voltage feature sample sequence currently being measured for similarity.

[0100] The similarity measurement values ​​of electrical characteristics between the monitored load and the various preset types of loads all include current similarity measurement values ​​and voltage similarity measurement values.

[0101] Furthermore, the DC fault identification module 5 is further used for:

[0102] A similarity metric is performed between the current feature signal sequence and the current feature sample sequence corresponding to the monitored load. A distance matrix is ​​created to store the distance between each pair of points between the current feature signal sequence and the current feature sample sequence corresponding to the monitored load. The distance between each point in the current feature signal sequence and each point in the current feature sample sequence is calculated. A path from the upper left corner to the lower right corner is found in the distance matrix such that the cumulative distance on the path is minimized. This path represents the best alignment between the current feature signal sequence and each current feature sample sequence. Finally, the cumulative distance on the path is output as the current similarity metric between the current feature signal sequence and the current feature sample sequence corresponding to the monitored load.

[0103] When measuring the similarity between the voltage feature signal sequence and the voltage feature sample sequence corresponding to the monitored load, a distance matrix is ​​created to store the distance between each pair of points between the voltage feature signal sequence and the voltage feature sample sequence corresponding to the monitored load. The distance between each point in the voltage feature signal sequence and each point in the voltage feature sample sequence is calculated. A path from the upper left corner to the lower right corner is found in the distance matrix such that the cumulative distance on the path is minimized. This path represents the best alignment between the voltage feature signal sequence and each voltage feature sample sequence. Finally, the cumulative distance on the path is output as the voltage similarity metric between the voltage feature signal sequence and the voltage feature sample sequence corresponding to the monitored load.

[0104] The similarity measure of electrical characteristics between the current characteristic signal sequence and the current characteristic sample sequence corresponding to the monitored load includes current similarity measure and voltage similarity measure.

[0105] It should be noted that the device in this embodiment corresponds to the method in the above embodiments. Therefore, any content not described in detail in this embodiment can be obtained by referring to the content of the method in the above embodiments, and will not be repeated in this embodiment.

[0106] This invention also provides a computer program product, including computer program instructions, which instruct a computer device to perform operations corresponding to the methods described above.

[0107] Specifically, the computer program product includes a series of computer program instructions that can instruct a computer device to execute the DC fault arc state monitoring method described in this application. These instructions are codes written in a computer program that define how to perform specific operations. In this embodiment, these instructions are used to execute the DC fault arc state monitoring method of the above embodiments.

[0108] These program instructions are designed to be loaded onto a computer device and to instruct the device to perform specific operations, which refer to the various steps in the DC fault arc state monitoring method described in the above embodiments.

[0109] In this way, the computer program product provides a complete software solution that can run on various computer devices to implement the DC fault arc state monitoring method described in the above embodiments.

[0110] The above description discloses only preferred embodiments of the present invention and should not be construed as limiting the scope of the present invention. Therefore, equivalent variations made in accordance with the claims of the present invention are still within the scope of the present invention.

Claims

1. A method for monitoring the state of a DC fault arc, characterized in that, Includes the following steps: The electrical signal sequence is obtained by sampling the electrical signal of the monitored load according to a preset sampling frequency. The electrical signal sequence includes a loop current signal sequence and an arc voltage signal sequence. Extract the electrical signal sequence to be analyzed from the electrical signal sequence that matches the preset time window. The electrical signal sequence to be analyzed includes the current signal sequence and the voltage signal sequence to be analyzed. The electrical signal sequence to be analyzed is decomposed and reconstructed by wavelet packet to obtain an electrical feature signal sequence, which includes a current feature signal sequence and a voltage feature signal sequence. When the load type of the monitored load is unknown, a similarity metric is performed between the electrical feature signal sequence and the electrical feature sample sequences corresponding to multiple preset load types to obtain a similarity metric value between the monitored load and the multiple preset load types. The similarity metric value between the monitored load and the multiple preset load types is then input into a pre-trained first deep belief network model to obtain the load type of the monitored load. Each preset load type's corresponding electrical feature sample sequence includes current feature sample sequences and voltage feature sample sequences of the preset load under various operating conditions. When the load type of the monitored load is known, the similarity measurement between the electrical feature signal sequence and the electrical feature sample sequence corresponding to the monitored load is performed to obtain a similarity measurement value for fault analysis. The similarity measurement value for fault analysis is then input into a pre-trained second deep belief network model to obtain the monitoring result of the DC fault arc state of the monitored load. The DC fault arc status monitoring result indicates normal operation or the presence of a DC fault arc.

2. The DC fault arc condition monitoring method according to claim 1, characterized in that, The method further includes: After performing wavelet packet decomposition and reconstruction on the electrical signal sequence to be analyzed to obtain the electrical feature signal sequence, the current feature signal sequence and the voltage feature signal sequence are further divided into equal-length sub-segments. Then, the average value of each sub-segment is calculated to represent each sub-segment, thereby converting the current feature signal sequence and the voltage feature signal sequence into a low-dimensional signal sequence.

3. The DC fault arc condition monitoring method according to claim 1, characterized in that, The step of performing a similarity measurement on the electrical feature signal sequence and the electrical feature sample sequences corresponding to multiple preset types of loads to obtain a similarity measurement value of the electrical features between the monitored load and the multiple preset types of loads further includes: A similarity metric is performed between the current feature signal sequence and the current feature sample sequence corresponding to each preset type of load. A distance matrix is ​​created to store the distance between each pair of points between the current feature signal sequence and the current feature sample sequence currently being measured for similarity. The distance between each point in the current feature signal sequence and each point in the current feature sample sequence is calculated. A path from the upper left corner to the lower right corner is found in the distance matrix such that the cumulative distance on the path is minimized. This path represents the best alignment between the current feature signal sequence and each current feature sample sequence. Finally, the cumulative distance on the path is output as the current similarity metric value between the current feature signal sequence and the current feature sample sequence currently being measured for similarity. A similarity metric is performed between the voltage feature signal sequence and the voltage feature sample sequence corresponding to each preset type of load. A distance matrix is ​​created to store the distance between each pair of points between the voltage feature signal sequence and the voltage feature sample sequence currently being measured for similarity. The distance between each point in the voltage feature signal sequence and each point in the voltage feature sample sequence is calculated. A path from the upper left corner to the lower right corner is found in the distance matrix such that the cumulative distance on the path is minimized. This path represents the best alignment between the voltage feature signal sequence and each voltage feature sample sequence. Finally, the cumulative distance on the path is output as the voltage similarity metric value between the voltage feature signal sequence and the voltage feature sample sequence currently being measured for similarity. The similarity measurement values ​​of electrical characteristics between the monitored load and the various preset types of loads all include current similarity measurement values ​​and voltage similarity measurement values.

4. The DC fault arc condition monitoring method according to claim 1, characterized in that, The step of performing a similarity measurement on the electrical characteristic signal sequence and the electrical characteristic sample sequence corresponding to the monitored load to obtain a similarity measurement value for fault analysis further includes: A similarity metric is performed between the current feature signal sequence and the current feature sample sequence corresponding to the monitored load. A distance matrix is ​​created to store the distance between each pair of points between the current feature signal sequence and the current feature sample sequence corresponding to the monitored load. The distance between each point in the current feature signal sequence and each point in the current feature sample sequence is calculated. A path from the upper left corner to the lower right corner is found in the distance matrix such that the cumulative distance on the path is minimized. This path represents the best alignment between the current feature signal sequence and each current feature sample sequence. Finally, the cumulative distance on the path is output as the current similarity metric between the current feature signal sequence and the current feature sample sequence corresponding to the monitored load. When measuring the similarity between the voltage feature signal sequence and the voltage feature sample sequence corresponding to the monitored load, a distance matrix is ​​created to store the distance between each pair of points between the voltage feature signal sequence and the voltage feature sample sequence corresponding to the monitored load. The distance between each point in the voltage feature signal sequence and each point in the voltage feature sample sequence is calculated. A path from the upper left corner to the lower right corner is found in the distance matrix such that the cumulative distance on the path is minimized. This path represents the best alignment between the voltage feature signal sequence and each voltage feature sample sequence. Finally, the cumulative distance on the path is output as the voltage similarity metric between the voltage feature signal sequence and the voltage feature sample sequence corresponding to the monitored load. The similarity measure of electrical characteristics between the current characteristic signal sequence and the current characteristic sample sequence corresponding to the monitored load includes current similarity measure and voltage similarity measure.

5. A DC fault arc condition monitoring device, characterized in that, include: The sampling module is used to sample the electrical signals 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 is used to extract an electrical signal sequence to be analyzed that matches a preset time window from the electrical signal sequence. The electrical signal sequence to be analyzed includes a current signal sequence and a voltage signal sequence to be analyzed. The feature extraction module is used to perform wavelet packet decomposition and reconstruction on the electrical signal sequence to be analyzed to obtain an electrical feature signal sequence, which includes a current feature signal sequence and a voltage feature signal sequence. The load type identification module is used to, when the load type of the monitored load is unknown, perform a similarity measurement on the electrical feature signal sequence and the electrical feature sample sequences corresponding to multiple preset load types to obtain a similarity measurement value of the electrical features between the monitored load and the multiple preset load types. The similarity measurement value of the electrical features between the monitored load and the multiple preset load types is then input into a pre-trained first deep belief network model to obtain the load type of the monitored load. Each preset load type's corresponding electrical feature sample sequence includes a current feature sample sequence and a voltage feature sample sequence of the preset load under various operating conditions. The DC fault identification module is used to, when the load type of the monitored load is known, perform a similarity measurement on the electrical feature signal sequence and the electrical feature sample sequence corresponding to the monitored load to obtain a similarity measurement value for fault analysis, and input the similarity measurement value for fault analysis into a pre-trained second deep belief network model to obtain the DC fault arc state monitoring result of the monitored load; the DC fault arc state monitoring result is normal operation or the presence of a DC fault arc.

6. The DC fault arc condition monitoring device according to claim 5, characterized in that, The feature extraction module is further configured to, after performing wavelet packet decomposition and reconstruction on the electrical signal sequence to be analyzed to obtain the electrical feature signal sequence, divide the current feature signal sequence and the voltage feature signal sequence into equal-length sub-segments respectively, and then calculate the average value of each sub-segment to represent each sub-segment, thereby converting the current feature signal sequence and the voltage feature signal sequence into a low-dimensional signal sequence.

7. The DC fault arc condition monitoring device according to claim 5, characterized in that, The load type identification module is further used for: A similarity metric is performed between the current feature signal sequence and the current feature sample sequence corresponding to each preset type of load. A distance matrix is ​​created to store the distance between each pair of points between the current feature signal sequence and the current feature sample sequence currently being measured for similarity. The distance between each point in the current feature signal sequence and each point in the current feature sample sequence is calculated. A path from the upper left corner to the lower right corner is found in the distance matrix such that the cumulative distance on the path is minimized. This path represents the best alignment between the current feature signal sequence and each current feature sample sequence. Finally, the cumulative distance on the path is output as the current similarity metric value between the current feature signal sequence and the current feature sample sequence currently being measured for similarity. A similarity metric is performed between the voltage feature signal sequence and the voltage feature sample sequence corresponding to each preset type of load. A distance matrix is ​​created to store the distance between each pair of points between the voltage feature signal sequence and the voltage feature sample sequence currently being measured for similarity. The distance between each point in the voltage feature signal sequence and each point in the voltage feature sample sequence is calculated. A path from the upper left corner to the lower right corner is found in the distance matrix such that the cumulative distance on the path is minimized. This path represents the best alignment between the voltage feature signal sequence and each voltage feature sample sequence. Finally, the cumulative distance on the path is output as the voltage similarity metric value between the voltage feature signal sequence and the voltage feature sample sequence currently being measured for similarity. The similarity measurement values ​​of electrical characteristics between the monitored load and the various preset types of loads all include current similarity measurement values ​​and voltage similarity measurement values.

8. The DC fault arc condition monitoring device according to claim 5, characterized in that, The DC fault identification module is further used for: A similarity metric is performed between the current feature signal sequence and the current feature sample sequence corresponding to the monitored load. A distance matrix is ​​created to store the distance between each pair of points between the current feature signal sequence and the current feature sample sequence corresponding to the monitored load. The distance between each point in the current feature signal sequence and each point in the current feature sample sequence is calculated. A path from the upper left corner to the lower right corner is found in the distance matrix such that the cumulative distance on the path is minimized. This path represents the best alignment between the current feature signal sequence and each current feature sample sequence. Finally, the cumulative distance on the path is output as the current similarity metric between the current feature signal sequence and the current feature sample sequence corresponding to the monitored load. When measuring the similarity between the voltage feature signal sequence and the voltage feature sample sequence corresponding to the monitored load, a distance matrix is ​​created to store the distance between each pair of points between the voltage feature signal sequence and the voltage feature sample sequence corresponding to the monitored load. The distance between each point in the voltage feature signal sequence and each point in the voltage feature sample sequence is calculated. A path from the upper left corner to the lower right corner is found in the distance matrix such that the cumulative distance on the path is minimized. This path represents the best alignment between the voltage feature signal sequence and each voltage feature sample sequence. Finally, the cumulative distance on the path is output as the voltage similarity metric between the voltage feature signal sequence and the voltage feature sample sequence corresponding to the monitored load. The similarity measure of electrical characteristics between the current characteristic signal sequence and the current characteristic sample sequence corresponding to the monitored load includes current similarity measure and voltage similarity measure.

9. A computer program product, characterized in that, It includes computer program instructions that instruct a computer device to perform an operation corresponding to the method as described in any one of claims 1 to 4.

Images