High-voltage circuit breaker fault identification system based on knowledge graph deep learning

Abstract

The application belongs to the technical field of high-voltage circuit breaker diagnosis, and particularly relates to a high-voltage circuit breaker fault identification system based on knowledge graph deep learning, which comprises the following steps: S1, data acquisition, obtaining original data of circuit breakers in a power grid system through monitoring the power grid system; S2, data analysis, original data classification processing; data grabbing is performed in the classified data, and data with high similarity to a preset threshold value is collected; classified data is sampled at intervals; S3, data reasoning, model reasoning is performed according to the action information of the circuit breaker and the suspected fault section; S4, DTW algorithm reasoning, the similarity of the original data and the grabbed data is calculated by using the DTW algorithm, and re-calculation is performed by using the DTW algorithm optimization; more accurate data information can be better obtained; the probability of false positives and false negatives is reduced through iterative learning of the knowledge graph, and through a large amount of data training and learning in the later period, complex power grid data and fault conditions can be processed, and the fault point can be quickly found.

Description

Technical Field

[0001] This invention relates to the field of high-voltage circuit breaker diagnostic technology, specifically a high-voltage circuit breaker fault identification system based on knowledge graph deep learning. Background Technology

[0002] The power system is an indispensable infrastructure of modern society, used to generate, transmit and distribute electrical energy. With the continuous growth of energy demand and the expansion of the power system, the reliability and stability of power grid equipment have become particularly important. A circuit breaker is a switching device that can close, carry and interrupt current under normal circuit conditions and close, carry and interrupt current under abnormal circuit conditions within a specified time. Circuit breakers are classified into high-voltage circuit breakers and low-voltage circuit breakers according to their application range. The boundary between high and low voltage is relatively blurred. Generally, those above 3kV are referred to as high-voltage electrical appliances. Circuit breakers are used to distribute electrical energy, infrequently start asynchronous motors, and protect power lines and motors. They can automatically disconnect the circuit when serious overloads, short circuits, or undervoltage faults occur. Their function is equivalent to a combination of a fuse switch and an over / under-temperature relay. Moreover, they generally do not require replacement of parts after interrupting the fault current. They have been widely used. In the generation, transmission, and use of electricity, power distribution is an extremely important link. The power distribution system includes transformers and various high and low voltage electrical equipment, and low-voltage circuit breakers are a widely used type of electrical appliance. However, high-voltage circuit breakers are diverse in type and structure, and are greatly affected by their own quality and external environmental factors. When they fail, they can affect the safety and stability of the power system. Therefore, it is particularly important to be able to detect faults in a timely and accurate manner. To this end, we propose a high-voltage circuit breaker fault identification system based on deep learning and knowledge graph. Summary of the Invention

[0003] The purpose of this invention is to provide a high-voltage circuit breaker fault identification system based on knowledge graph deep learning, so as to solve the problems mentioned in the background art.

[0004] To achieve the above objectives, the present invention provides the following technical solution: a high-voltage circuit breaker fault identification system based on knowledge graph deep learning, comprising the following steps: S1, Data Acquisition, Monitoring the power grid system to obtain raw data of circuit breakers within the power grid system; S2, Data Analysis, Raw Data Classification Processing; Data Extraction in Classified Data, Focusing on Collecting Data with High Similarity to Preset Thresholds; Interval Sampling of Classified Data; S3, Data Reasoning, performs model reasoning based on the circuit breaker's action information and suspected faulty road sections; S4, DTW algorithm inference, uses the DTW algorithm to calculate the similarity between the original data and the crawled data, and performs a recalculation optimized by the DTW algorithm; S5, Fault Identification: Based on the algorithm-optimized comparison data, the faulty line is identified according to the best data, and the task is terminated, returning to the initial program.

[0005] Preferably, the high-voltage circuit breakers are tested in groups, and the test parameters of the high-voltage circuit breakers under various conditions are recorded. The recorded data are then classified according to the conditions, including normal conditions and multiple fault conditions. Preferably, a clustering algorithm is used to process the sample data for each fault condition to obtain the cluster center signal and the maximum cluster distance for each fault condition sample data. The distance value is compared with the distance value of each cluster center signal. If the distance value is less than the maximum cluster distance, the distance value is compared with the average cluster distance. If the distance value is less than the average cluster distance, the corresponding vector element is adjusted by replacing the data of the dissimilar segments in turn. Preferably, in step S1, the collected data parameters include one of the coil current, contact vibration signal, contact current and contact travel of the high-voltage circuit breaker. For a suspected faulty line, the comprehensive current sample sequence of the five cycles before the fault time and the comprehensive current sample sequence of the three to five cycles after the fault time are obtained.

[0006] Preferably, in step S4, the similarity is compared with a preset threshold, and the parameters are adjusted according to the comparison results; the comparison data is optimized again by the DTE algorithm to calculate the horizontal and vertical differences of the captured data; and then the similarity, differences and preset thresholds are compared to obtain the best data.

[0007] Preferably, in the classification data interval sampling in step S2, the original reference signal and the sampled signal are segmented according to the operating time of the circuit breaker, feature matching is performed on each segment, the most similar sampled signal is selected for each segment, and finally the signal source with the most similar sampled signals is used as the new reference signal, and the new reference signal is used as the adjustment parameter.

[0008] Preferably, the identification system includes a cloud server and three storage areas. Steps S1 and S2 involve constructing a basic knowledge graph, collecting and initially analyzing basic data parameters within the power grid system, and then storing them in the first storage area. The basic knowledge graph is used to generate a time-sensitive reasoning knowledge graph. Steps S3, S4, and S5 involve obtaining a reasoning rule set, importing the collected basic knowledge graph into the reasoning rule set to generate a reasoning knowledge graph, and then storing it in the second storage area. The reasoning knowledge graph is the basis for steps S3, S4, and S5.

[0009] Preferably, within the reasoning knowledge graph, based on the data parameters derived from the reasoning rule set, the first data is labeled as (N1), and the first data (N1) is used as the initial data. According to the time interval, the subsequently derived data is sequentially labeled as (N2)...(Nn). The first data (N1) is sequentially overlapped and compared with the subsequent labeled data (N2)...(Nn). The data after each overlap is updated as the first knowledge graph, up to the Nth knowledge graph, and the updated knowledge graph is stored in the third storage area.

[0010] Preferably, the obtained reasoning knowledge graph is sampled to obtain an initial keyframe; using the initial keyframe as the initial value, an approximate keyframe with high similarity to the original keyframe is calculated from the initial keyframe by an algorithm. The similarity between the initial keyframe and the original keyframe is measured by the DTW distance. The smaller the DTW distance, the higher the similarity between the initial keyframe and the original keyframe.

[0011] Compared with the prior art, the beneficial effects of the present invention are: This solution provides a knowledge graph-based identification system. Based on keyframe sequences obtained through keyframe sampling, the system uses the DTW optimization algorithm to measure the similarity between keyframes and the original keyframes. This allows for iterative comparison with the original keyframes without losing the keyframe sequence, forming a knowledge learning graph. It better captures the rhythm of the entire data acquisition process, obtaining more accurate data information. Through iterative learning of the knowledge graph, the probability of false alarms and missed alarms is reduced. With extensive training and learning on large amounts of data in the later stages, it can handle complex power grid data and fault conditions, helping us quickly locate fault points. Attached Figure Description

[0012] Figure 1 This is a logic block diagram of the present invention; Figure 2 This is a comparison diagram of the data and knowledge graph of this invention. Detailed Implementation

[0013] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0014] In the description of this invention, it should be understood that the terms "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0015] Example: Please see Figure 1 This invention provides a technical solution for a high-voltage circuit breaker fault identification system based on knowledge graph deep learning, comprising the following steps: S1, Data Acquisition, Monitoring the power grid system to obtain raw data of circuit breakers within the power grid system; The collected data parameters include one of the following: coil current, contact vibration signal, contact current and contact travel of the high-voltage circuit breaker. For suspected faulty lines, the comprehensive current sample sequence of the five cycles before the fault time and the comprehensive current sample sequence of at least three to five cycles after the fault are obtained. S2, Data Analysis, Raw Data Classification Processing; Data Extraction in Classified Data, Focusing on Collecting Data with High Similarity to Preset Thresholds; Interval Sampling of Classified Data; High-voltage circuit breakers were tested in groups, and the test parameters of the high-voltage circuit breakers under various conditions were recorded. The recorded data were then classified according to the conditions, including normal conditions and multiple fault conditions. In the classification data interval sampling, the original reference signal and the sampled signal are segmented according to the circuit breaker's operating time. Feature matching is performed on each segment, and the most similar sampled signal is selected for each segment. Finally, the signal source with the most similar sampled signals is used as the new reference signal, and the new reference signal is used as the adjustment parameter. S3, Data Reasoning, performs model reasoning based on the circuit breaker's action information and suspected faulty road sections; The identification system includes a cloud server and three storage areas. Steps S1 and S2 involve constructing a basic knowledge graph, collecting and initially analyzing basic data parameters within the power grid system, and then storing them in the first storage area. The basic knowledge graph is used to generate a time-sensitive reasoning knowledge graph. Next, a reasoning rule set is obtained, and the collected basic knowledge graph is imported into the reasoning rule set to generate the reasoning knowledge graph, which is then stored in the second storage area. Steps S3, S4, and S5 define the reasoning knowledge graph. Within the reasoning knowledge graph, based on the data parameters derived from the reasoning rule set, the first data is labeled as (N1), and the first data (N1) is used as the initial data. According to the time interval, the subsequently derived data is labeled as (N2)...(Nn). The first data (N1) is overlapped and compared with the subsequent labeled data (N2)...(Nn) in turn. The data after each overlap is updated as the first knowledge graph, up to the Nth knowledge graph, and the updated knowledge graph is stored in the third storage area. S4, DTW algorithm inference, uses the DTW algorithm to calculate the similarity between the original data and the crawled data, and performs a recalculation optimized by the DTW algorithm; The diagnostic identification is implemented based on an artificial neural network on a cloud server. The artificial algorithm includes multiple neurons, each corresponding to the identification of different working conditions. The identification algorithm for different working conditions is the DTW algorithm. The input of the artificial neural algorithm is the data to be classified, and the output of the artificial neural algorithm is a unit vector containing only one element. The unit vector corresponds to the working condition identification result. The data is compared with a preset threshold based on similarity, and the parameters are adjusted according to the comparison results. The comparison data is then optimized again using the DTE algorithm to calculate the horizontal and vertical differences of the captured data. Finally, the data is compared with the preset threshold based on similarity and differences to obtain the best data. In the DTW algorithm optimization in step S4, the obtained reasoning knowledge graph is sampled to obtain an initial key frame; using the initial key frame as the initial value, the algorithm calculates an approximate key frame that is highly similar to the original key frame from the initial key frame. The similarity between the initial key frame and the original key frame is measured by the DTW distance. The smaller the DTW distance, the higher the similarity between the initial key frame and the original key frame. S5, Fault Identification: Based on the algorithm-optimized comparison data, the faulty line is identified according to the best data, and the task is terminated, returning to the initial program; Clustering algorithms are used to process the sample data for each fault condition to obtain the cluster center signal and the maximum cluster distance for each fault condition sample data. The distance value is compared with the distance value of each cluster center signal. If the distance value is less than the maximum cluster distance, the distance value is compared with the average cluster distance. If the distance value is less than the average cluster distance, the corresponding vector element is adjusted by replacing the data of the dissimilar segment in turn. All the data and knowledge graphs that need to be compared are listed as sequences. Data is compared with data, and knowledge graphs are compared with knowledge graphs. Suppose we have two time series Q and C, with lengths n and m respectively; where: Q = q1 - q2 - q3 - qn; C = c1 - c2 - c3 - cn; Calculate the similarity between sequences Q and C, with lengths n and m respectively. The normalized path form is as follows: wk has the form (i,j), where i represents the coordinate of the x-axis and j represents the coordinate of the y-axis, as shown in the appendix. Figure 2 As shown; If two time series have unequal lengths and can have many possible correspondences, we need to find the shortest matching path. We obtain the following through dynamic programming: According to the dynamic programming algorithm, the final regular path can be found by working from the bottom up. After a sample is identified, the sample data obtained will be fed back to the sample storage area of ​​the corresponding category. When the rate of change of the sample variable in the sample storage area is greater than 2%, the baseline signal will be recalculated. Compared to traditional methods, this approach uses keyframe sequences obtained through keyframe sampling as a foundation. The DTW optimization algorithm measures the similarity between keyframes and the original keyframes, enabling iterative comparison with the original keyframes without losing the keyframe sequence, thus forming a knowledge learning graph. This allows for better capture of the entire data acquisition process, resulting in more accurate data information. Iterative learning of the knowledge graph reduces the probability of false alarms and missed alarms. Through extensive training and learning with large amounts of subsequent data, it can handle complex power grid data and fault conditions, helping us quickly locate fault points.

[0016] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or basic characteristics. Therefore, the embodiments should be considered exemplary and non-limiting in all respects. The scope of the invention is defined by the appended claims rather than the foregoing description. Therefore, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention, and no reference numerals in the claims should be construed as limiting the scope of the claims.

[0017] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A high-voltage circuit breaker fault identification system based on knowledge graph deep learning, characterized in that, Includes the following steps: S1, Data Acquisition, Monitoring the power grid system to obtain raw data of circuit breakers within the power grid system; S2, Data Analysis, Raw Data Classification Processing; Data Extraction in Classified Data, Focusing on Collecting Data with High Similarity to Preset Thresholds; Interval Sampling of Classified Data; S3, Data Reasoning, performs model reasoning based on the circuit breaker's action information and suspected faulty road sections; S4, DTW algorithm inference, uses the DTW algorithm to calculate the similarity between the original data and the crawled data, and performs a recalculation optimized by the DTW algorithm; S5, Fault Identification: Based on the algorithm-optimized comparison data, the faulty line is identified according to the best data, and the task is terminated, returning to the initial program.

2. The high-voltage circuit breaker fault identification system based on knowledge graph deep learning according to claim 1, characterized in that: In step S1, the collected data parameters include one of the following: coil current of the high-voltage circuit breaker, contact vibration signal, contact current and contact travel. For a suspected faulty line, the comprehensive current sample sequence of the five cycles before the fault time and the comprehensive current sample sequence of the three to five cycles after the fault time are obtained.

3. The high-voltage circuit breaker fault identification system based on knowledge graph deep learning according to claim 1, characterized in that: In step S4, the similarity is compared with a preset threshold, and the parameters are adjusted according to the comparison results; the comparison data is optimized again by the DTE algorithm to calculate the horizontal and vertical differences of the captured data; then the similarity, differences and preset thresholds are compared to obtain the best data.

4. The high-voltage circuit breaker fault identification system based on knowledge graph deep learning according to claim 1, characterized in that: In the classification data interval sampling in step S2, the original reference signal and the sampled signal are segmented according to the circuit breaker's operating time. Feature matching is performed on each segment, and the most similar sampled signal is selected for each segment. Finally, the signal source with the most similar sampled signals is used as the new reference signal, and the new reference signal is used as the adjustment parameter.

5. The high-voltage circuit breaker fault identification system based on knowledge graph deep learning according to claim 1, characterized in that: The identification system includes a cloud server and three storage areas. Steps S1 and S2 involve constructing a basic knowledge graph, collecting basic data parameters within the power grid system and performing preliminary analysis, and then storing them in the first storage area. The basic knowledge graph is used to generate a time-sensitive reasoning knowledge graph. Obtain the inference rule set, generate an inference knowledge graph based on the collected basic knowledge graph and import it into the inference rule set, and then store it in the second storage area; the inference knowledge graph is the result of steps S3, S4 and S5.

6. The high-voltage circuit breaker fault identification system based on knowledge graph deep learning according to claim 1, characterized in that: Within the reasoning knowledge graph, based on the data parameters derived from the reasoning rule set, the first data is labeled as (N1), and the first data (N1) is used as the initial data. According to the time interval, the subsequently derived data is sequentially labeled as (N2)...(Nn). The first data (N1) is sequentially overlapped and compared with the subsequent labeled data (N2)...(Nn). The data after each overlap is updated as the first knowledge graph, up to the Nth knowledge graph, and the updated knowledge graph is stored in the third storage area.

7. The high-voltage circuit breaker fault identification system based on knowledge graph deep learning according to claim 6, characterized in that: The obtained reasoning knowledge graph is sampled to obtain initial keyframes. Using the initial keyframes as initial values, an algorithm is used to calculate approximate keyframes that are highly similar to the original keyframes. The similarity between the initial keyframes and the original keyframes is measured by the DTW distance. The smaller the DTW distance, the higher the similarity between the initial keyframes and the original keyframes.

Images