A method for diagnosing a failure of a circuit breaker operating mechanism

Abstract

The application discloses a kind of fault diagnosis methods of circuit breaker operating mechanism, comprising the following steps: step S1, the current signal generated in the process of circuit breaker operating mechanism is collected;Step S2, the current signal obtained by collection is preprocessed, including removing basic offset noise and normalizing current signal;Step S3, the original signal is decomposed into a plurality of product functions and a residual signal using local mean decomposition algorithm;Step S4, each product function is extracted by fuzzy entropy feature, and the feature vector of component composed of multiple fuzzy entropies is obtained;Step S5, clustering analysis is carried out on multiple feature vectors to classify, when new current signal is collected, the current state of circuit breaker operating mechanism is diagnosed by the category of current signal corresponding feature vector. The beneficial effects of the application are that the fault diagnosis process of circuit breaker operating mechanism is quantified, the accuracy and efficiency are relatively high, and it is convenient for subsequent improvement.

Description

Technical Field

[0001] This invention belongs to the field of power transmission and distribution technology, and more specifically, relates to a fault diagnosis method for a circuit breaker operating mechanism. Background Technology

[0002] Circuit breakers are critical control and protection devices in power systems. Their main function is to close and open lines under normal load, and to quickly interrupt fault current when a short-circuit fault occurs in the system, ensuring the safe and stable operation of the power grid.

[0003] The circuit breaker operating mechanism is the core component responsible for the opening and closing of the circuit breaker. Its normal operation determines whether the circuit breaker can function properly. Conversely, when the circuit breaker operating mechanism malfunctions, it is also necessary to diagnose the type of fault in order to determine the cause of the malfunction and facilitate subsequent maintenance and testing of the circuit breaker.

[0004] However, most current fault diagnosis methods for circuit breaker operating mechanisms still rely on experience-based judgment. That is, technicians diagnose the fault type of the circuit breaker based on the characteristics of previous faults in the operating mechanism. This usually leads to the following problems: (1) Because the subjective influence of technicians is significant during fault diagnosis, different technicians at the same time or the same technician at different times may make deviations in their judgment of the same fault, ultimately affecting the accuracy of fault diagnosis; (2) When multiple circuit breaker operating mechanisms are involved in fault diagnosis, there is a shortage of technicians, resulting in low fault diagnosis efficiency; (3) There is a lack of quantitative analysis of faults, which makes it difficult to improve the fault diagnosis process in the future. Summary of the Invention

[0005] Existing circuit breaker operating mechanisms mainly rely on the experience and judgment of technicians to diagnose faults, resulting in low accuracy and efficiency in fault diagnosis, and making it difficult to improve the fault diagnosis process in the future.

[0006] A fault diagnosis method for a circuit breaker operating mechanism includes the following steps: Step S1: Collect the current signal generated by the tripping coil of the circuit breaker operating mechanism during the tripping process; Step S2: Preprocess the acquired current signal, including removing basic offset noise and normalizing the current signal; Step S3: Use the local mean decomposition algorithm to decompose the original signal into multiple product functions and a residual signal; Step S4: Extract fuzzy entropy features for each product function and obtain a feature vector composed of multiple fuzzy entropy components; Step S5: Perform cluster analysis on multiple feature vectors obtained in step S4 to generate multiple categories. Each category corresponds to a state of the circuit breaker operating mechanism. When a new current signal is acquired, the current state of the circuit breaker operating mechanism is diagnosed by calculating the category of the feature vector corresponding to the current signal.

[0007] Furthermore, a closed-loop Hall current sensor is used to collect the current signal, and the sampling frequency of the sensor is at least 20kHz. The sampling triggering method is set so that when the current of the trip coil is detected to exceed 0.1A, 20ms before that time point is taken as the initial recording time point, and when the current of the trip coil is detected to reach 0A and last for 10ms, this time point is taken as the final recording time point.

[0008] Furthermore, the basic offset noise is removed using the following calculation formula: The original current signal is The current signal after removing the basic offset noise is t is a time point, This refers to the c-th time point within a certain time interval before the initial recording time point. represent The initial current signal at a given time point, where A represents the number of time points within the aforementioned time period prior to the initial recording time point.

[0009] Furthermore, the normalization of the current signal is achieved using the following formula: , This represents the normalized current signal.

[0010] Furthermore, the steps of the local mean decomposition method are as follows: Let n i Current signal The i-th extreme point, given the formula for calculating the local mean. Formula for calculating local envelope estimates Connect all m using the moving average method i and a i The local mean function is obtained. and local envelope function Given frequency modulation signal ,satisfy ;like envelope function satisfy Then the product function for: ; obtain residual signal ,satisfy After that As a new current signal, steps (1) to (3) are repeated k times to obtain k signals. and 1 residual signal .

[0011] Furthermore, in step S3, if the following conditions are not met... Then As a new input signal, the frequency modulation signal is repeatedly obtained until... envelope function satisfy Then the product function satisfy .

[0012] Furthermore, the steps of fuzzy entropy feature extraction include, using the product function... Represented as a time series u, construct a vector. , Let the mean of the vector be denoted; define the similarity. ,have , For vectors and Calculate the Chebyshev distance between them and the average similarity. and its average value The calculation formula is: and Regarding the fuzzy entropy of the time series u satisfy The three product functions After replacing the time series u, the feature vector S is obtained, which is... .

[0013] Furthermore, the clustering algorithm is the K-means algorithm, whose steps include: randomly generating K cluster centers with the same dimension as the feature vectors; calculating the distance between each feature vector and the cluster center; assigning the feature vector to the cluster center with the smallest distance; selecting the centroid of each cluster as the new cluster center; recalculating the distance between the feature vector and all newly generated cluster centers; assigning the feature vector to the cluster center with the smallest distance; repeating this process until the cluster centers no longer change, thus dividing the feature vectors into K categories.

[0014] Furthermore, the method for determining the category of the feature vector corresponding to the current signal is to identify the category of the cluster whose center is closest to the feature vector.

[0015] This invention proposes a fault diagnosis method for circuit breaker operating mechanisms. After preprocessing the current signal acquired during the circuit breaker's opening process, a local mean decomposition algorithm is used to decompose the original signal into multiple product functions and a residual signal. Fuzzy entropy feature extraction is performed on each product function to obtain a feature vector composed of multiple fuzzy entropy components. Cluster analysis is then performed on these feature vectors to classify the circuit breaker operating mechanism's state into multiple categories. When a new current signal is acquired, the current state of the circuit breaker operating mechanism is diagnosed by calculating the category of the feature vector corresponding to that current signal. The advantages of this invention are that it constructs a quantitative method for fault diagnosis of circuit breaker operating mechanisms, achieving high accuracy and reliability, and facilitating further optimization of the fault diagnosis process after acquiring new current signals. Attached Figure Description

[0016] Figure 1 This is a flowchart of the fault diagnosis method in this invention. Detailed Implementation

[0017] The technical solution of the present invention will be clearly and completely described below with reference to the embodiments.

[0018] like Figure 1 As shown, the present invention proposes a fault diagnosis method for a circuit breaker operating mechanism, which specifically includes the following steps.

[0019] Step S1: Collect the current signal generated by the circuit breaker operating mechanism's tripping coil during the tripping process.

[0020] In step S1, the core task is to collect the electrical signals generated by the circuit breaker operating mechanism during operation. The specific selection of the electrical signal collection process can be considered from three parts: the operating process, the operating coil, and the electrical signals. The operating process can be divided into operating mechanism opening or operating mechanism closing, the operating coil can be divided into opening coil and closing coil, and the electrical signals can be divided into current signals and voltage signals.

[0021] In this embodiment, the current signal generated by the circuit breaker operating mechanism's tripping coil during the tripping process is selected for acquisition. The reason is that after the circuit breaker operating mechanism has completed tripping, it has entered the closed state. Based on the acquired electrical signal results, it is possible to directly select whether to perform maintenance and testing. The current signal is selected for acquisition because when a fault occurs, the fluctuation of the current signal is much larger than that of the voltage fluctuation.

[0022] The current signal generated by a complete circuit breaker operating mechanism during the tripping process includes at least the following four stages; Rapid rise of current Let T0 and T1 be the starting points of this stage. At this time, the trip coil is energized and the current increases monotonically. However, the electromagnetic force generated by the trip coil is not enough to overcome the reaction force on the iron core, and the iron core remains stationary.

[0023] Slow rise of current Let the starting time points of this stage be T1 and T2 respectively. When the current reaches a certain threshold, its electromagnetic force is greater than the reaction force on the iron core, the iron core begins to move and generates a significant back electromotive force. This back electromotive force is opposite to the current direction, causing the current to be in a slow rising state, or even to decrease. When it decreases...

[0024] Current saturation section The starting time points of this stage are recorded as T2 and T3 respectively. The iron core hits the operating mechanism and stops moving. The back electromotive force disappears. During this process, the current will gradually rise until it reaches the saturation value. When the current drops in (2), a trough will appear on the current signal.

[0025] Current cutting section The starting time points of this stage are denoted as T3 and T4 respectively. The circuit breaker operating mechanism quickly completes the tripping process and disconnects the tripping coil, and its current drops rapidly to zero.

[0026] In summary, the current signal generated during the circuit breaker tripping process is a non-stationary and non-linear current signal, and any deviation at any stage will cause a large change in the current signal, so it cannot be directly classified.

[0027] This embodiment uses a closed-loop Hall current sensor connected in series with the trip coil to acquire the current signal. This method features high and low voltage electrical isolation and high safety. The sensor's sampling frequency is set to at least 20kHz. The sampling triggering method is configured such that when the trip coil current exceeds 0.1A, it serves as the start recording point, with the initial recording time T0 set 20ms before the start recording point. When the trip coil current reaches 0A and remains there for 10ms, this moment is set as the final recording time T4.

[0028] Step S2: Preprocessing of current signal The acquired raw current signal often contains baseline offset noise. Furthermore, to eliminate the absolute value differences in current signal amplitude caused by different circuit breaker models or different power supply voltages and to improve applicability, preprocessing of the acquired raw current data is necessary. This mainly includes the following preprocessing steps.

[0029] (1) Remove basic offset noise Sensors may experience zero-point drift due to temperature variations, resulting in baseline offset noise. This zero-point drift can be removed using the current signal recorded within the first 40ms of the initial recording time. Since the time required for the trip coil current to exceed 0.1A typically does not exceed 20ms, the trip coil can be considered essentially unenergized before the initial recording time. If the original current signal is... The current signal after removing the basic offset noise is Where t is a time point, the following formula can be used for calculation:

[0030] in This refers to the c-th time point within 40ms before the initial recording time point. represent The initial current signal at the given time point, where A represents the number of time points within 40ms prior to the initial recording time point. This 40ms period can be set to other values, but it must not be less than 20ms.

[0031] (2) Normalization of current signal To eliminate the difference in absolute current amplitude caused by different circuit breaker models or different power supply voltages, and to improve the applicability of the subsequently established fault diagnosis method, the Min-Max normalization method was used to obtain (1). Preprocessing is performed using the following formula:

[0032] This represents the normalized current signal.

[0033] Step S3, Local Mean Decomposition Algorithm Local mean decomposition (LMD) is an iterative decomposition algorithm used to analyze non-stationary and nonlinear signals. Its core idea is to decompose a complex signal into the sum of several product functions PF, where each PF is obtained by multiplying an envelope signal by a purely frequency-modulated signal. It performs well in handling endpoint effects and preserving the physical meaning of instantaneous frequencies.

[0034] The Local Mean Decomposition algorithm specifically includes the following steps: (1) Determine the local extreme points and calculate the local mean and local envelope. Find the preprocessed signal All local maxima and local minima, assuming the current signal The i-th extreme point is n i For each pair of adjacent extreme points n i and n i+1 Calculate their local mean m iand local envelope estimate a i At the same time: and .

[0035] (2) Smoothing process, and elimination of mean and demodulation Connect all m using the moving average method i and a i , respectively, yielding continuous local mean functions and local envelope function Subtracting the local mean function from the original signal yields the mean-free signal. That is: ;Will Divide by the local envelope function Amplitude modulation and demodulation are performed to obtain the frequency modulated signal. That is: .

[0036] (3) Iteratively check and construct the product function. Ideally, if It is a pure frequency modulated signal, and its envelope function is... It should always be equal to 1. envelope function The method for obtaining and the envelope function Same; if ,but This is the pure FM signal we're looking for; if it doesn't meet the requirements... Then As the new input signal, repeat steps (1) and (2) until a pure frequency-modulated signal is obtained. Its envelope function satisfy The final envelope signal The product of the envelope functions generated during all iterations is: .

[0037] First product function From the final envelope signal and the final pure frequency modulation signal Multiplying them together gives: .this It contains the highest frequency components of the original signal and has physical meaning in terms of instantaneous amplitude and instantaneous frequency.

[0038] (4) Separation, circulation, termination Then the first product function From current signal The residual signal was separated from the source. ,have: After that, As a new current signal, steps (1) to (3) are repeated in a loop. The loop stops when the following condition is met, and the residual signal is... It becomes a monotonic function or a constant, or its amplitude is so small as to be negligible compared to the original signal.

[0039] Ultimately, the original signal was decomposed into k elements. Product function and a residual signal sum: .

[0040] In this embodiment, the decomposition of the original signal is stopped when k is 3, and the decomposition results generally exhibit the following pattern: This includes high-frequency electromagnetic noise and minute vibrations at the moment the iron core starts. and This includes information on the main dynamic changes during the current drop process within the trip coil. Residual component. This mainly reflects the macroscopic trend of the current signal.

[0041] Step S4: Fuzzy Entropy Feature Extraction The main function of fuzzy entropy is to measure the degree of disorder. The larger the value of fuzzy entropy, the more complex and irregular the signal is; the smaller the value, the more regular the signal is. For some circuit breakers that have malfunctioned, the fuzzy entropy of their current signals often increases significantly.

[0042] For a given product function Both can be represented as time series u. , Represents the product function The value at the nth time point, and the following steps are performed sequentially: (1) Phase space reconstruction and distance calculation Given an m-dimensional vector ,have , This is the mean of all components within the vector. Chebyshev distance is used to calculate the mean of the two vectors. and Distance between ,have: .

[0043] (2) Definition of fuzzy similarity and calculation of global probability A fuzzy membership function is introduced to define similarity. Since this invention uses an exponential function, we have:

[0044] n is the boundary gradient parameter, which determines how quickly the similarity decays with distance. In this embodiment, it is set to 3. r is the similarity tolerance, which is generally related to the standard deviation of the time series and satisfies the following conditions: , This is the proportionality coefficient. The standard deviation is denoted as .

[0045] For each i, calculate its average similarity with all the remaining j. , Then calculate average ,have .

[0046] (3) Increase the dimension and calculate the entropy value Increase the dimension to m+1 and repeat (1) and (2) to obtain It can also calculate the fuzzy entropy of the time series u. ,have: .

[0047] By using three product functions After generating a time series u, its corresponding fuzzy entropy can be calculated according to the above steps (1)-(3), and then the feature vector S can be constructed. .

[0048] Step S5: Fault Diagnosis and Classification Each time the circuit breaker operating mechanism trips, a corresponding current signal is generated. Each current signal can generate a corresponding feature vector S through steps S1-S4. When there are enough feature vectors S, these feature vectors S can be classified first. Each category corresponds to a fault type (including no fault). After a new current signal is collected, the fault type can be determined by judging which category its feature vector S falls into, thereby achieving the purpose of fault diagnosis of the circuit breaker operating mechanism.

[0049] In this embodiment, the K-means algorithm is used to classify the collected feature vector S. The basic steps are as follows: (1) Let the feature vector corresponding to the i-th current signal be S. i And there are: , The representative eigenvector is S i The kth fuzzy entropy component; (2) Selecting the value of K is equivalent to dividing the circuit breaker operating mechanism into K states and randomly generating K cluster centers, with the jth cluster center denoted as Q. j And calculate the distance between each feature vector and the cluster center, using the following formula: , Cluster center Q j The k-th component is denoted as S.i Assign the eigenvectors of the eigenvectors of the j-th cluster to the cluster center with the smallest distance. The eigenvectors of the same cluster center are the cluster relations. Let S be the eigenvector of the i-th cluster. ij (3) Select the centroid of each cluster as the new cluster center. The centroid formula is: , Let q be the centroid of the j-th cluster, and q be the number of feature vectors in the current cluster. Recalculate the distance between the feature vector and all newly generated cluster centers, and assign the feature vector to the cluster center with the smallest distance from it; (4) Repeat step (3) until the cluster center no longer changes. At this time, the feature vector is divided into K clusters, that is, K categories, and each category corresponds to the state of a circuit breaker operating mechanism.

[0050] For any given cluster, determining the state of any one of its circuit breaker operating mechanisms is sufficient to determine whether the circuit breaker operating mechanism has malfunctioned and the type of malfunction. However, to improve the accuracy of fault diagnosis, it is usually preferable to determine the feature vectors that are closer to the cluster center.

[0051] After detecting a new current signal generated during the opening process of the circuit breaker operating mechanism, the current signal only needs to be processed through steps S1-S5 to determine its category. This allows for the diagnosis of whether a fault has occurred in the circuit breaker operating mechanism and the type of fault. After diagnosis, the classification cluster is updated through step S5 after expanding the new feature vector. Specifically, the distance between the feature vector and the cluster center can also be considered as the probability of a fault of that category occurring; the shorter the distance, the higher the probability.

[0052] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A fault diagnosis method for a circuit breaker operating mechanism, characterized in that, Includes the following steps: Step S1: Collect the current signal generated by the tripping coil of the circuit breaker operating mechanism during the tripping process; Step S2: Preprocess the acquired current signal, including removing basic offset noise and normalizing the current signal; Step S3: Use the local mean decomposition algorithm to decompose the original signal into multiple product functions and a residual signal; Step S4: Extract fuzzy entropy features for each product function and obtain a feature vector composed of multiple fuzzy entropy components; Step S5: Perform cluster analysis on multiple feature vectors obtained in step S4 to generate multiple categories. Each category corresponds to a state of the circuit breaker operating mechanism. When a new current signal is acquired, the current state of the circuit breaker operating mechanism is diagnosed by calculating the category of the feature vector corresponding to the current signal.

2. The fault diagnosis method according to claim 1, characterized in that: The current signal is acquired using a closed-loop Hall current sensor with a sampling frequency of at least 20kHz. The sampling triggering method is set such that when the current of the trip coil is detected to exceed 0.1A, the 20ms before that time point is taken as the initial recording time point, and when the current of the trip coil is detected to reach 0A and last for 10ms, this time point is taken as the final recording time point.

3. The fault diagnosis method according to claim 2, characterized in that: The basic offset noise is removed using the following calculation formula. The original current signal is The current signal after removing the basic offset noise is t is a time point, This refers to the c-th time point within a certain time interval before the initial recording time point. represent The initial current signal at a given time point, where A represents the number of time points within the aforementioned time period prior to the initial recording time point.

4. The fault diagnosis method according to claim 3, characterized in that: The normalization of the current signal is achieved using the following formula: , This represents the normalized current signal.

5. The fault diagnosis method according to claim 4, characterized in that: The steps of the local mean decomposition method are as follows: Let n i Current signal The i-th extreme point, given the formula for calculating the local mean. Formula for calculating local envelope estimates Connect all m using the moving average method i and a i The local mean function is obtained. and local envelope function Given frequency modulation signal ,satisfy ;like envelope function satisfy Then the product function for: ; obtain residual signal ,satisfy After that As a new current signal, steps (1) to (3) are repeated k times to obtain k signals. and 1 residual signal .

6. The fault diagnosis method according to claim 5, characterized in that: In step S3, if the condition is not met... Then As a new input signal, the frequency modulation signal is repeatedly obtained until... envelope function satisfy Then the product function satisfy .

7. The fault diagnosis method according to claim 6, characterized in that: The steps of fuzzy entropy feature extraction include: multiplying the product function... Represented as a time series u, construct a vector. , Let the mean of the vector be denoted; define the similarity. ,have , For vectors and Calculate the Chebyshev distance between them and the average similarity. and its average value The calculation formula is: and Regarding the fuzzy entropy of the time series u satisfy The three product functions After replacing the time series u, the feature vector S is obtained, which is... .

8. The fault diagnosis method according to claim 7, characterized in that: The clustering algorithm is the K-means algorithm. Its steps include: randomly generating K cluster centers with the same dimension as the feature vectors; calculating the distance between each feature vector and a cluster center; assigning the feature vector to the cluster center with the smallest distance; selecting the centroid of each cluster as the new cluster center; recalculating the distance between the feature vector and all newly generated cluster centers; assigning the feature vector to the cluster center with the smallest distance; repeating this process until the cluster centers no longer change, thus dividing the feature vectors into K categories.

9. The fault diagnosis method according to claim 8, characterized in that: The method for determining the category of the feature vector corresponding to the current signal is to identify the category of the cluster whose center is closest to the feature vector.

Images