Modular vacuum circuit breaker adaptive fault identification method based on multi-optimization algorithm

Abstract

The invention discloses a modular vacuum circuit breaker adaptive fault identification method based on a multi-optimization algorithm, relates to the technical field of power equipment, and can improve the fault identification accuracy of a modular vacuum circuit breaker. According to the method, adaptive optimization is carried out on the SVM model by using three optimization algorithms, a random forest model is introduced to extract a nonlinear relation of output results of the three optimized SVM models, and comprehensive fault identification and judgment are carried out on the modular vacuum circuit breaker in combination with the output results of the three optimized SVM models. The fault identification accuracy of the modular vacuum circuit breaker can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of power equipment, and particularly relates to a modular vacuum circuit breaker adaptive fault identification method based on multiple optimization algorithms. Background Art

[0002] As an important switching element in the distribution network, the modular vacuum circuit breaker is responsible for connecting and disconnecting the normal current, which plays a very important role in people's production and life. However, when the circuit breaker fails, it will cause damage to the protected circuits and equipment, and then lead to general power grid accidents or the expansion of power grid accidents. In the lightest case, it affects national production and life, and in the worst case, it affects social stability and causes serious economic losses to the country. Therefore, monitoring and diagnosing the mechanical faults of modular vacuum circuit breakers has great value for ensuring the safety of the power system and national production and life.

[0003] The traditional fault identification scheme of modular vacuum circuit breakers uses artificial neural networks to diagnose the faults of vacuum circuit breakers. However, due to the easy occurrence of overfitting and poor generalization ability of artificial neural networks, the accuracy of fault identification is relatively low.

[0004] In view of this, a modular vacuum circuit breaker adaptive fault identification method based on multiple optimization algorithms is needed. Summary of the Invention

[0005] Aiming at the problem of low accuracy in the existing fault identification scheme of modular vacuum circuit breakers, the present invention provides a modular vacuum circuit breaker adaptive fault identification method based on multiple optimization algorithms, which can improve the accuracy of modular vacuum circuit breaker fault identification. The specific technical solutions are as follows:

[0006] In a first aspect, an embodiment of the present application provides a modular vacuum circuit breaker adaptive fault identification method based on multiple optimization algorithms, including:

[0007] S1: Obtain the operation data of the modular vacuum circuit breaker, where the operation data includes multiple data samples;

[0008] S2: Construct a first feature vector for each of the data samples based on the operation data;

[0009] S3: Input the first feature vector into a preset first support vector machine (SVM) model, a second SVM model, and a third SVM model respectively for training to obtain a first output result output by the first SVM model, a second output result output by the second SVM model, and a third output result output by the third SVM model; wherein the training of the first SVM model is optimized based on the genetic algorithm, the training of the second SVM model is optimized based on the Bayesian algorithm, and the training of the third SVM model is optimized based on the optimal particle swarm algorithm;

[0010] S4: Input the second feature vector into a preset random forest model for training to obtain a fault identification result output by the random forest model; wherein, the training of the random forest model is optimized based on the optimal particle swarm optimization algorithm, and the second feature vector is obtained based on the first output result, the second output result, and the third output result corresponding to the same data sample.

[0011] S5: Based on the fault identification model, perform fault identification on the modular vacuum circuit breaker. The first layer of the fault identification model includes the first SVM model, the second SVM model, and the third SVM model, and the second layer of the fault identification model includes the random forest model; the input of the first layer is the first feature vector, the output of the first layer is the second feature vector, the input of the second layer is the output of the first layer, and the output of the second layer is the fault identification result.

[0012] Preferably, the data sample includes the closing and opening coil current signals within a preset time window, and the first feature vector includes a first feature; constructing the first feature vector of each data sample based on the operation data includes:

[0013] Calculate the slope of the data points on the change curve of the closing and opening coil current signals. The calculation formula of the slope is:

[0014]

[0015] wherein, x is time, y is the intensity of the closing and opening coil current signal, x0 is the initial moment, Δx is the change amount of time, Δy is the change amount of the closing and opening coil current signal, f(x) is an approximate function of the closing and opening coil current signal changing with time, and f'(x) is the slope of the data point corresponding to time x on the change curve of the closing and opening coil current signal;

[0016] When the slope of the i-th data point is 0, and the slopes of the (i - 1)-th data point and the (i + 1)-th data point are not all 0, determine the midpoint coordinates of the i-th data point and the (i + 1)-th data point as the extreme point coordinates; when the slope of the i-th data point is not 0, and the slopes of the (i - 1)-th data point and the (i + 1)-th data point belong to positive and negative values respectively, determine the (i + 1)-th data point as the extreme point coordinates; based on the extreme point coordinates, determine the peaks and valleys of the change curve; based on the time and current corresponding to the first peak of the change curve, the time and current corresponding to the first valley, the time and current corresponding to the second peak, and the initial moment, construct the first feature.

[0017] Preferably, the first feature vector includes a second feature; constructing the first feature vector of each data sample based on the operation data includes: calculating the mean, standard deviation, and kurtosis of the switching coil current signal; where the formula for kurtosis is:

[0018]

[0019] where K is the kurtosis of the current signal, N is the number of data points of the switching coil current signal in the data sample, y i is the current signal strength of the i-th data point, μ is the mean of the switching coil current signal, and σ is the standard deviation of the switching coil current signal; constructing the second feature based on the mean, the standard deviation, and the kurtosis.

[0020] Preferably, the data sample includes the switching coil voltage signal within a preset time window, and the first feature vector includes a third feature; constructing the first feature vector of each data sample based on the operation data includes: calculating the work done by the closing current based on the switching coil voltage signal and the switching coil current signal; the formula for calculating the work done by the closing current is:

[0021]

[0022] where W is the work done by the closing current, U is the switching coil voltage signal, I is the switching coil current signal, t is the length of the preset time window, is the integral symbol, and dt represents an infinitesimal increment in time; constructing the third feature based on the work done.

[0023] Preferably, the data sample includes the three-phase switching times, the transient values and amplitudes of the three-phase currents, the harmonic amplitudes of the three-phase currents, and the rated current of the modular vacuum circuit breaker; the first feature vector includes a fourth feature; constructing the first feature vector of each data sample based on the operation data includes: determining the switching time of the modular vacuum circuit breaker based on the maximum value of the three-phase switching times; calculating the three-phase difference degree based on the transient values and peak values of the three-phase currents; the formula for calculating the three-phase difference degree is:

[0024]

[0025] where λ represents the three-phase difference degree, i U 、i V and i W are respectively the amplitudes of the U-phase current, V-phase current, and I-phase current in the three-phase current; I U 、I V and I Ware the transient values of the U-phase current, V-phase current, and I-phase current in the three-phase current. 2πj and 2π(j + 1) represent points on the complex plane. Based on the harmonic amplitudes of the three-phase current, calculate the harmonic maximum value of the three-phase current. Based on the harmonic amplitudes of the three-phase current and the rated current, calculate the harmonic distortion rate of the three phases. Based on the switching time, the three-phase difference, the harmonic maximum value, and the harmonic distortion rate, construct the fourth feature.

[0026] Preferably, after step S5, the method further includes: S6: Transmit the recognition result of the fault recognition model to the host computer, so that the host computer visually displays the recognition result.

[0027] Preferably, after step S5, the method further includes: S7: Based on the recognition result of the fault recognition model, perform incremental optimization on the fault recognition model.

[0028] In a second aspect, an embodiment of the present application provides a modular vacuum circuit breaker adaptive fault recognition system based on multiple optimization algorithms. The system includes a host computer, a lower computer, and a modular vacuum circuit breaker. The host computer is connected to the lower computer, and the lower computer is connected to the modular vacuum circuit breaker. The lower computer is used to apply the method described in the first aspect to perform fault recognition on the modular vacuum circuit breaker and transmit the result of the fault recognition to the host computer.

[0029] In a third aspect, an embodiment of the present application provides a computing device, including: a memory for storing a program; a processor for loading the program to execute the method described in the first aspect.

[0030] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium. The computer-readable storage medium includes a stored program. When the program runs, it controls the device where the computer-readable storage medium is located to execute the method described in the first aspect.

[0031] Compared with the prior art, the beneficial effects of the present invention are as follows: The method of the present invention adaptively optimizes the SVM model by using multiple optimization algorithms, introduces a random forest model to extract the non-linear relationship of the output results of the three optimized SVM models, and combines the output results of the three optimized SVM models for comprehensive judgment, which can improve the accuracy of modular vacuum circuit breaker fault recognition. Description of the Drawings

[0032] In order to more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for the specific embodiments or the description of the prior art. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts do not necessarily draw according to the actual scale.

[0033] Figure 1 Schematic diagram of the structure of a modular vacuum circuit breaker adaptive fault recognition system based on multiple optimization algorithms provided by an embodiment of the present application;

[0034] Figure 2 Schematic diagram of the process of a modular vacuum circuit breaker adaptive fault recognition method based on multiple optimization algorithms provided by an embodiment of the present application;

[0035] Figure 3 Schematic diagram of the process of constructing a first feature vector provided by an embodiment of the present application;

[0036] Figure 4 Schematic diagram of the process of an SVM model training process provided by an embodiment of the present application;

[0037] Figure 5 Schematic diagram of the process of a fault recognition model provided by an embodiment of the present application;

[0038] Figure 6 Comparison diagram of recognition results provided by an embodiment of the present application;

[0039] Figure 7 Flowchart of the processing of recognition results provided by an embodiment of the present application;

[0040] Figure 8 Visualization display diagram of recognition results provided by an embodiment of the present application;

[0041] Figure 9 Schematic diagram of model incremental optimization provided by an embodiment of the present application;

[0042] Figure 10 Schematic diagram of the structure of a computing device provided by an embodiment of the present application. Detailed implementation manners

[0043] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are some, but not all, of the embodiments of the present invention. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0044] It should be understood that when used in this specification and the appended claims, the terms "include" and "comprise" indicate the presence of the described features, wholes, steps, operations, elements, and / or components, but do not exclude the presence or addition of one or more other features, wholes, steps, operations, elements, components, and / or their combinations.

[0045] It should also be understood that the terms used in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. As used in the specification of the present invention and the appended claims, unless the context clearly indicates otherwise, the singular forms "a", "an" and "the" are intended to include the plural forms.

[0046] It should be further understood that the term "and / or" used in the specification of the present invention and the appended claims refers to any combination and all possible combinations of one or more of the associated listed items, and includes these combinations.

[0047] To solve the problem of low accuracy in the existing fault identification scheme of modular vacuum circuit breakers, the present invention provides an adaptive fault identification method and related equipment for modular vacuum circuit breakers based on multiple optimization algorithms, which can improve the accuracy of fault identification of modular vacuum circuit breakers.

[0048] Please refer to Figure 1 , Figure 1 which is a schematic structural diagram of an adaptive fault identification system for modular vacuum circuit breakers based on multiple optimization algorithms provided by an embodiment of the present application. As Figure 1 shown, the system 10 includes a host computer 11, a slave computer 12, and a modular vacuum circuit breaker 13; wherein the host computer 11 is connected to the slave computer 12, and the slave computer 12 is connected to the modular vacuum circuit breaker 13.

[0049] Among them, the host computer 11 is at a higher level in the control system and is mainly used for a device or software platform that centrally monitors, manages, and operates the entire system where the modular vacuum circuit breaker 13 is located. It usually has strong computing power, display functions, and a human-computer interaction interface, can interact with operators for information, and send instructions to and receive data from the slave computer 12.

[0050] Specifically, the host computer 11 can be the total control device of the computer room where the control system is located, or an intelligent terminal beside the management personnel. Specifically, the host computer 11 can be a server, or an intelligent terminal such as a personal computer (PC), a laptop computer, or a tablet computer. The specific form of the host computer 11 in the embodiments of the present application is not limited.

[0051] The slave computer 12 is at a lower level in the control system and is a device that directly interacts with on-site devices or controlled objects. It is mainly responsible for receiving instructions sent by the host computer 11, controlling on-site devices according to the instructions, and uploading the status and data of on-site devices to the host computer after collection.

[0052] Specifically, the slave computer 12 may include a power supply module, a data acquisition module, a data filtering module, and a signal transmission module.

[0053] The modular vacuum circuit breaker 13 is used to control the on / off of the circuit and adjust the operating state of the power system; it is also used to protect the power system. Among them, the modular vacuum circuit breaker adopts a detachable and replaceable independent module design. Each module is quickly connected through an interface. When maintaining, the faulty component can be replaced separately without disassembling the whole device, making the maintenance more convenient.

[0054] Preferably, the modular vacuum circuit breaker 13 further includes access devices such as a main shaft angle sensor, a voltage transformer, a current transformer, and a vibration sensor. These access devices are connected to the appropriate positions of the modular vacuum circuit breaker as component modules of the modular vacuum circuit breaker 13.

[0055] Based on the above access devices, the slave computer 12 can obtain the operating data of the modular vacuum circuit breaker 13, which may specifically include current data, voltage data, and vibration data. Since multiple source signals are usually affected simultaneously when a fault occurs in the modular vacuum circuit breaker 13, there is usually a problem of low accuracy in the analysis and diagnosis of single-source signals. In the embodiment of the present application, the slave computer 12 can identify faults in the modular vacuum circuit breaker 13 by combining multiple source signals.

[0056] It can be understood that the specific fault identification method will be described in the method embodiment below and will not be elaborated here.

[0057] It can be understood that, it should be noted that, in specific implementation, the system architecture of the above system provided by the embodiment of the present application can be any architecture including Figure 1 a similar structure. The embodiment of the present application does not limit the specific composition of the system architecture. In addition, Figure 1 the architecture composition shown in Figure 1 does not constitute a limitation to the system architecture. Except for

[0058] Please refer to Figures 2 to 9 , Figure 2 which is a schematic flow chart of a modular vacuum circuit breaker adaptive fault identification method based on multiple optimization algorithms provided by the embodiment of the present application. This method can be applied to the slave computer in the above system or other computing devices with computing capabilities. In this embodiment, the case where the slave computer executes this method is taken as an example for description. As Figure 2 shown, this method includes the following steps:

[0059] S1: The slave computer obtains the operating data of the modular vacuum circuit breaker.

[0060] Before step S1, the following steps are further included:

[0061] S01A: The management or maintenance personnel connect devices such as the spindle angle sensor, voltage transformer, current transformer, and vibration sensor to the appropriate positions of the modular vacuum circuit breaker.

[0062] S01B: Connect the modular vacuum circuit breaker to the lower computer and conduct debugging.

[0063] S01C: Turn on the upper computer and check whether there is an error in the communication between the lower computer and the upper computer through the preset button of the lower computer. Check whether there is packet loss in the data transmitted from the lower computer to the upper computer through the preset data packet transmission. If there is packet loss, check the reason. Otherwise, start normal operation.

[0064] Among them, the operation data includes multiple data samples. Specifically, the lower computer can obtain the operation data of the modular vacuum circuit breaker through a preset sliding window, and the operation data within the window duration obtained based on the sliding window each time is used as a data sample.

[0065] S2: The lower computer constructs a first feature vector for each of the data samples based on the operation data.

[0066] As Figure 3 shown, constructing the first feature vector includes the following steps:

[0067] S201: Due to the sensitivity problem of the system, the lower computer can perform step S201 to filter and denoise the closing and opening coil current signals.

[0068] S201A: There are tip noises in the closing and opening coil current signals. The lower computer can use the sliding median method to remove these tip signals; assume the sliding window length is 2a + 1, and the current signal data within the window time is expressed as:

[0069] X0,...,X a ,...,X 2a , then the sliding median filtering operation is:

[0070] x medium =Medium(X0,...,X a ,...,X 2a );

[0071] Among them: Medium represents the operation of taking the median value of the current signals within the window, x medium is the median value of the data within the window, and X i is the current signal within the window.

[0072] S201B: However, since there are still some small irregular fluctuations in the current signal after passing through the sliding median method, the lower computer can use the moving average filtering method to further process the current signal waveform. Let the window length be 2b + 1, and the current signal data within the window is expressed as: X m0 ,..., X mb ,..., X m2b , and the operation of the moving average filtering method is as follows:

[0073]

[0074] Where: is the average value of the data within the window, and X mi is the current signal after median filtering within the window.

[0075] It can be understood that only the preprocessing of the closing and opening coil current signal is described here. For other types of operation data, the lower computer can use the same or similar methods for preprocessing, such as data cleaning, outlier detection, and compensation of the operation data.

[0076] S202: After obtaining the operation data after filtering and noise reduction, the lower computer can extract the features of the operation data to construct the first feature vector.

[0077] The current signal after sliding median filtering and moving average filtering is considered a smooth curve, and the lower computer can obtain the local features of the current signal curve based on this curve. Since the current signal can be regarded as composed of a series of discrete points, it is difficult to represent the curve in the form of an expression, so the derivative method cannot be used to obtain the extreme points of the current curve. However, as long as the interval between two points is small enough, the slope of adjacent points can be used to approximately replace the derivative of that point.

[0078] Preferably, the data sample includes the closing and opening coil current signals within a preset time window, and the first feature vector includes the first feature; calculate the slope of the data points on the change curve of the closing and opening coil current signal, and the calculation formula of the slope is:

[0079]

[0080] Where x is the time, y is the intensity of the closing and opening coil current signal, x0 is the initial time, Δx is the change in time, Δy is the change in the closing and opening coil current signal, f(x) is the approximate function of the closing and opening coil current signal changing with time, and f'(x) is the slope of the data point corresponding to the time x on the change curve of the closing and opening coil current signal;

[0081] When the slope at the i-th data point is 0, and the slopes of the (i - 1)-th and (i + 1)-th data points are not all 0, determine the midpoint coordinates of the i-th and (i + 1)-th data points as the extreme point coordinates; when the slope of the i-th data point is not 0, and the slopes of the (i - 1)-th and (i + 1)-th data points belong to positive and negative values respectively, determine the (i + 1)-th data point as the extreme point coordinates; based on the extreme point coordinates, determine the peaks and valleys of the variation curve; based on the time and current corresponding to the first peak of the variation curve, the time and current corresponding to the first valley, the time and current corresponding to the second peak, and the initial moment, construct the first feature.

[0082] Different types of power faults often cause specific peak-valley variation characteristics in the current curve. For example, a short-circuit fault usually causes the current to increase sharply, showing a peak significantly higher than normal operation; while an open-circuit fault may cause the current to suddenly become zero, manifested as an abnormal valley. In the embodiments of the present application, by constructing the first feature with the peaks and valleys of the current signal, the fault type can be intuitively reflected through the first feature, facilitating subsequent fault identification. In addition, the peaks and valleys of the current curve can accurately reflect the instantaneous change of the current. When a fault occurs, abnormal peaks or valleys will appear in the current curve at the corresponding moment. By capturing these abnormal features, the time point of the fault occurrence can be accurately determined, which helps to further analyze the cause and process of the fault and is of great significance for quickly eliminating the fault and restoring the normal operation of the power system.

[0083] Preferably, the first feature vector includes a second feature; the lower computer can calculate the mean, standard deviation, and kurtosis of the switching coil current signal;

[0084] The calculation formula for the mean is:

[0085]

[0086] where μ is the mean of the switching coil current signal, N is the number of data points of the switching coil current signal in the data sample, and is the current signal intensity of the i-th data point.

[0087] The calculation formula for the standard deviation is:

[0088]

[0089] where σ is the standard deviation of the switching coil current signal.

[0090] The calculation formula for the kurtosis is:

[0091]

[0092] Wherein, K is the kurtosis of the current signal; then, the lower computer can construct the second feature based on the mean value, the standard deviation, and the kurtosis.

[0093] The current kurtosis is very sensitive to the impact components in the current signal. When a power equipment fails, such as bearing failure or gear failure of a motor, the current signal often generates instantaneous impact pulses, and these impacts will significantly increase the current kurtosis value. In the embodiments of the present application, by constructing the second feature based on the kurtosis, the abnormal changes caused by these faults can be more significantly highlighted, so that even tiny and early fault signs can be clearly reflected in the kurtosis value, which helps to improve the accuracy of subsequent fault identification.

[0094] Preferably, the data sample includes the opening and closing coil voltage signals within a sliding window, and the first feature vector includes a third feature; the lower computer can calculate the work done by the closing current based on the opening and closing coil voltage signals and the opening and closing coil current signals; the calculation formula for calculating the work done by the closing current is:

[0095]

[0096] Wherein, W is the work done by the closing current, U is the opening and closing coil voltage signal, I is the opening and closing coil current signal, t is the length of a preset time window, is the integral symbol, and dt represents a tiny increment in time; based on the work done, the third feature is constructed.

[0097] Different types of faults have different characteristics in their influence on the current work done. For example, resistive faults may mainly affect the effective value of the current work done, while inductive or capacitive faults may cause changes in the phase and frequency of the current work done. By analyzing the details of the current work done, including changes in its amplitude, phase, frequency, etc., important clues can be provided for fault location and diagnosis, helping technicians to more accurately determine the location and cause of the fault.

[0098] Preferably, the data sample includes the three-phase opening and closing times, the transient values and amplitudes of the three-phase currents, the harmonic amplitudes of the three-phase currents, and the rated current of the modular vacuum circuit breaker; the first feature vector includes a fourth feature; the lower computer can determine the opening and closing time of the modular vacuum circuit breaker based on the maximum value of the three-phase opening and closing times;

[0099] Specifically, the opening and closing time is the maximum value of the three-phase opening and closing times:

[0100] T = max(T U T v T w );

[0101] Wherein, T is the opening and closing time, T U 、Tv and T w are the opening and closing times of the U-phase, V-phase, and W-phase of the modular vacuum circuit breaker, respectively.

[0102] The opening and closing time is an important parameter of the modular vacuum circuit breaker, which can directly reflect the performance of its mechanical operating mechanism. Under normal circumstances, the opening and closing time should be within the specified range. If the opening and closing time is too long or too short, it may mean that there are mechanical faults such as jamming, poor lubrication, and spring fatigue in the operating mechanism. By monitoring the opening and closing time, these potential problems can be detected in time to avoid power system failures caused by untimely opening and closing.

[0103] The lower computer can calculate the three-phase difference degree based on the transient values and peak values of the three-phase currents; the calculation formula for the three-phase difference degree is:

[0104]

[0105] where λ represents the three-phase difference degree, i U , i V and i W are the amplitudes of the U-phase current, V-phase current, and I-phase current in the three-phase currents, respectively; I U , I V and I W are the transient values of the U-phase current, V-phase current, and I-phase current in the three-phase currents, 2πj and 2π(j + 1) represent points on the complex plane;

[0106] The three-phase difference degree is mainly used to measure the difference degree of parameters such as opening and closing time, current, and voltage between the three phases of the modular vacuum circuit breaker. Under normal operating conditions, the three-phase parameters should be basically balanced. If the three-phase difference degree exceeds the normal range, it may indicate the existence of three-phase unbalance faults, such as poor contact of a certain phase contact, open circuit or short circuit in a certain phase line, etc. By monitoring the three-phase difference degree, the faulty phase can be quickly located, providing an important basis for fault diagnosis and repair.

[0107] The lower computer can calculate the harmonic maximum value of the three-phase currents based on the harmonic amplitudes of the three-phase currents; the calculation formula is:

[0108] I Hn = max{I Un , I Vn , I Wn};

[0109] I H,max = max{I H1 / I N , I H2 / I N ,..., I Hn / I N};

[0110] Among them, I H.max is the maximum harmonic value; n is the harmonic order, and I Un , I Vn , I Wn are the harmonic effective values of the U-phase current, V-phase current, and I-phase current in the corresponding order respectively; I Hn is the maximum harmonic effective value in the corresponding order, and I N is the rated current.

[0111] The maximum harmonic value can help determine whether there are abnormal harmonic sources and possible fault types in the power system. When non-linear loads such as frequency converters and rectifiers are connected to the power system, harmonic currents will be generated. If the maximum harmonic value exceeds the normal range, it may mean that there are faults in these non-linear loads, such as component damage, abnormal control circuits, etc., resulting in a significant increase in the harmonic content. In addition, faults inside power equipment, such as transformer core saturation, capacitor faults, etc., will also cause changes in the maximum harmonic value. By monitoring the maximum harmonic value and combining the analysis of other electrical parameters, the harmonic source and fault type can be accurately identified, providing a direction for troubleshooting.

[0112] The slave computer can calculate the harmonic distortion rates of the three phases respectively based on the harmonic amplitudes of the three-phase current and the rated current; the calculation formula for the harmonic distortion rate of the U-phase is:

[0113]

[0114] Among them, θ THD，U represents the harmonic distortion rate of the U-phase, and I U,k represents the harmonic effective value of the K-th harmonic of the U-phase current; it can be understood that the calculation formulas for the harmonic distortion rates of the V-phase and W-phase are similar to the calculation formula for the harmonic distortion rate of the U-phase, and will not be elaborated here.

[0115] The harmonic distortion rate is a comprehensive index to measure the harmonic content in the power system. It reflects the proportional relationship of each harmonic component relative to the fundamental wave component, and can more comprehensively reflect the harmonic pollution degree of the power system. Compared with the maximum harmonic value, the harmonic distortion rate not only considers the maximum amplitude of the harmonics, but also considers the distribution of each harmonic. Even if the maximum harmonic value is within a certain range, but if the harmonic distortion rate is high, it also indicates that there are relatively complex harmonic problems in the power system, which may pose potential hazards to the operation of power equipment and the system. By monitoring the harmonic distortion rate, the operation status of the power system can be more accurately evaluated, and potential harmonic fault hidden dangers can be discovered in time.

[0116] The slave computer can construct the fourth feature based on the switching time, the three-phase difference degree, the maximum harmonic value, and the harmonic distortion rate.

[0117] S203: The lower computer can construct a first feature vector based on one or more of the above-mentioned first feature, second feature, third feature, and fourth feature.

[0118] S3: The lower computer inputs the first feature vector into a preset first support vector machine (SVM) model, a second SVM model, and a third SVM model respectively for training, and obtains a first output result output by the first SVM model, a second output result output by the second SVM model, and a third output result output by the third SVM model.

[0119] Among them, the training of the first SVM model is optimized based on the genetic algorithm, the training of the second SVM model is optimized based on the Bayesian algorithm, and the training of the third SVM model is optimized based on the optimal particle swarm algorithm.

[0120] As Figure 4 shown, the process of training the SVM model includes the following steps:

[0121] S301: Mark the fault labels for the first feature vector of the modular vacuum circuit breaker obtained in S2, which are circuit breaker normal - 0, iron core jamming - 1, main shaft rotation jamming - 2, closing spring missing - 3, voltage too high

[0122] - 4, and voltage too low - 5.

[0123] S302A: Build an initial SVM mathematical model, and the expression of the SVM mathematical model is:

[0124]

[0125] Among them, w is the hyperplane normal vector, b is the bias term, ξ i is the slack variable, C is the penalty parameter, x i is the feature vector of the i-th training sample, y i is the class label of the i-th training sample, and n is the number of training samples.

[0126] S302B: Solve the above problem using the Lagrange multiplier method to obtain the SVM classification mathematical model:

[0127]

[0128] Among them, f(x) is the decision function, sgn(·) is the sign function, which is a piecewise function; K(x i , x j ) is the kernel function. α i and are the Lagrange multipliers.

[0129] S302C: K(x i , x j ) uses the RBF kernel function, and its expression is:

[0130]

[0131] where K represents the value of the kernel function, which is used to indicate the similarity between sample x i and sample x j ; the exp function is the exponential function, and σ represents the scale parameter of the kernel function.

[0132] Step S302 can establish three SVM models, namely the first SVM model, the second SVM model, and the third SVM model.

[0133] S303: The lower computer can use the genetic algorithm to optimize the first SVM model and find the optimal training hyperparameters of the first SVM model.

[0134] S303A: Define the crossover probability adjustment method of the genetic optimization algorithm as:

[0135]

[0136] where: P cmin and P cmax are the minimum and maximum values of the crossover probability; f i is the fitness value of individual i; f avg is the average fitness value of all individuals in the population; N is the population size; P ca is the crossover probability adjustment parameter.

[0137] S303B: Define the mutation probability adjustment method of the genetic optimization algorithm as:

[0138]

[0139] where: P mmax and P mmin are the maximum and minimum values of the mutation probability; P ma is the mutation probability adjustment parameter.

[0140] S303C: Load the operation data of the modular vacuum circuit breaker.

[0141] S303D: Divide the operation data into a training set and a test set.

[0142] S303E: Set the relevant parameters of the genetic algorithm, including the population size, the maximum number of iterations, the minimum value and adjustment coefficient of the crossover probability, the minimum value and adjustment coefficient of the mutation probability, etc., and initialize the population.

[0143] S303F: Load the first SVM model built in S302, set the initial values of C and σ, and set the search range.

[0144] S303G: Assign the initial values of C and σ to the first SVM model, and calculate the initial fitness of the first SVM model.

[0145] S303H: Start iteration. In each iteration, calculate the fitness value and decide whether to continue iteration according to the result of the fitness value. If the maximum iteration number is reached or the function converges, output the optimal solutions of C and σ; otherwise, continue iteration. Perform selection operation, crossover operation, and mutation operation in sequence, and then update the population.

[0146] S303J: Save the first SVM model obtained after training and optimization in the lower computer.

[0147] S304: The lower computer can use the Bayesian optimization algorithm to optimize the second SVM model and find the optimal training hyperparameters of the second SVM model.

[0148] S304A: Load the operation data of the modular vacuum circuit breaker.

[0149] S304B: Divide the operation data into a training set and a test set.

[0150] S304C: Define the number of iterations as 100 and the number of cross-validation folds as 5.

[0151] S304D: Load the second SVM model built in S302, set the initial values of C and σ, and set the search range.

[0152] S304E: Assign the initial values of C and σ to the second SVM model, and calculate the initial judgment accuracy of the second SVM model.

[0153] S304G: Start the optimization iteration to update the parameters C and σ and obtain the optimal model parameters.

[0154] S304H: Save the second SVM model obtained after training and optimization in the lower computer.

[0155] S305: The lower computer can use the optimal particle swarm optimization (PSO) algorithm to optimize the third SVM model and find the optimal training hyperparameters of the third SVM model.

[0156] S305A: Load the operation data of the modular vacuum circuit breaker.

[0157] S305B: Divide the operation data into a training set and a test set.

[0158] S305C: Set the relevant parameters of the PSO algorithm, including the particle swarm size, the maximum number of iterations, the inertia weight, the individual learning factor, the social learning factor, etc., and initialize the particle swarm.

[0159] S305D: Load the third SVM model built in S202, set the initial values of C and σ, and set the search range.

[0160] S305E: Assign the initial values of C and to the SVM, and calculate the initial fitness of the third SVM model.

[0161] S305F: Start the iteration. In each iteration, calculate the fitness value, and decide whether to continue the iteration according to the result of the fitness value. If the maximum number of iterations is reached or the fitness accuracy requirement is met, output the optimal solutions of C and σ; otherwise, continue the iteration. Update the particle position and speed in turn, and then update the particle swarm.

[0162] S305G: Save the optimized third SVM model in the lower computer.

[0163] S4: The lower computer inputs the second feature vector into the preset random forest model for training, and obtains the fault identification result output by the random forest model.

[0164] Among them, the training of the random forest model is optimized based on the optimal particle swarm algorithm, and the second feature vector is obtained based on the first output result, the second output result, and the third output result corresponding to the same data sample.

[0165] As Figure 5 shown, the process of training the random forest model includes the following steps:

[0166] S401: The lower computer builds an initial random forest model.

[0167] S402: Call the first SVM model, the second SVM model, and the third SVM model saved in S303, S304, and S305. Input the running data into these three models to obtain the first output result, the second output result, and the third output result respectively. Combine the results corresponding to the same data sample in the three groups of output results to form the second feature vector.

[0168] S403A: Set the relevant parameters of the PSO algorithm, including the particle swarm size, the maximum number of iterations, the inertia weight, the individual learning factor, the social learning factor, etc., and initialize the particle swarm.

[0169] S403B: Load the random forest model, set the initial parameters such as the number of trees and the maximum depth, and define the search range.

[0170] S403C: Apply the initial parameters to the random forest model and calculate the initial fitness of the random forest model.

[0171] S403D: Start the iteration. In each iteration, calculate the fitness value and decide whether to continue the iteration based on the result of the fitness value. If the maximum number of iterations is reached or the fitness accuracy requirement is met, output the optimal random forest model parameters; otherwise, update the particle position and velocity in sequence, and then update the particle swarm to continue the iteration.

[0172] S403E: Save the trained random forest model in the lower computer.

[0173] After obtaining the trained random forest model, the lower computer can construct and save a fault identification model based on the first SVM model, the second SVM model, and the third SVM model. Among them, the first layer of the fault identification model includes the first SVM model, the second SVM model, and the third SVM model, and the second layer of the fault identification model includes the random forest model; the input of the first layer is the first feature vector, the output of the first layer is the second feature vector, the input of the second layer is the output of the first layer, and the output of the second layer is the fault identification result.

[0174] In one experiment, the lower computer inputs the same test set into the first SVM model, the second SVM model, the third SVM model, and the fault identification model respectively, and obtains the Figure 6 recognition rate comparison results as shown. As Figure 6 shown, the fault recognition rate of the first SVM model (GA+SVM) is 87.8%, the fault recognition rate of the second SVM model (Bayes+SVM) is 91%, the fault recognition rate of the third SVM model (PSO+SVM) is 89.5%, and the fault recognition rate of the fault identification model (3SVM+RF) is 87.8%. It can be seen that using the fault identification model obtained by the method of this embodiment to identify faults in the modular vacuum circuit breaker can improve the accuracy of fault identification.

[0175] S5: The lower computer performs fault identification on the modular vacuum circuit breaker based on the fault identification model.

[0176] Among them, the lower computer can obtain the real-time operation data of the modular vacuum circuit breaker based on a preset sliding window and input the real-time operation data into the fault identification model to obtain the fault identification result.

[0177] Preferably, after step S5, the method further includes: S6: Transmit the identification result of the fault identification model to the upper computer so that the upper computer can visually display the identification result.

[0178] The processing flow of the identification result is asFigure 7 As shown in the figure, it includes the following steps:

[0179] S601: The lower computer receives the operation data of the modular vacuum circuit breaker in the current environment, and calls the fault identification model constructed in S4 to perform fault identification based on this operation data.

[0180] S602A: The lower computer saves the fault identification result and transmits the fault identification result to the upper computer.

[0181] S602B: Combine and save the fault identification result with the input data as the corresponding fault label and data sample.

[0182] S603: The upper computer visually displays the fault identification result.

[0183] The style of the upper computer visually displaying the fault identification result can be as Figure 8 shown. In a single fault identification, the fault identification result is: based on the obtained operation data, it is identified that there is a 96.8% probability that the current modular vacuum circuit breaker has a stuck iron core fault.

[0184] Preferably, after step S5, the method further includes: S7: Based on the identification result of the fault identification model, perform incremental optimization on the fault identification model.

[0185] The process of incremental optimization of the fault identification model is as Figure 8 shown, including the following steps:

[0186] S701: The lower computer randomly extracts 80% of the historical operation data and merges it with the incrementally obtained real-time data as a new data set, and randomly selects 80% of the data in this data set as the training set, and the remaining 20% of the data as the test set.

[0187] S702: Call the fault identification model and its parameters used last time in S5 as the model and parameters to be optimized.

[0188] S703: Perform incremental update on the model parameters to be optimized based on the training set and the test set; the update formula includes:

[0189] θ new =θ old +η▽ θ L(θ; D new )

[0190] where θ old and θ new respectively represent the model parameters before and after the update, η is the learning rate, ▽ θ represents the gradient of the loss function with respect to the model parameters, and L is based on the new data set D newCalculated loss function.

[0191] S704: Store the updated fault identification model.

[0192] In the embodiments of the present application, by adaptively optimizing the SVM model using a variety of optimization algorithms, introducing a random forest model to extract the non-linear relationship of the output results of three optimized SVM models, and making a comprehensive judgment by combining the output results of the three optimized SVM models, the model accuracy can be improved, and the accuracy of modular vacuum circuit breaker fault identification can be improved.

[0193] As Figure 10 shown, Figure 10 FIG. 13 is a possible schematic logical structure diagram of a computing device provided by an embodiment of the present application. The computing device 1000 includes: a processor 1001, a communication interface 1002, a memory 1003, and a bus 1004. The processor 1001, the communication interface 1002, and the memory 1003 are interconnected through the bus 1004. In the embodiments of the present application, the processor 1001 is used to control and manage the actions of the computing device 1000. For example, the processor 1001 is used to execute Figure 2 the steps in the embodiments and / or other processes for the technologies described herein. The communication interface 1002 is used to support the computing device 1000 to communicate. The memory 1003 is used to store the program code and data of the computing device 1000.

[0194] Among them, the processor 1001 may be a central processing unit, a general-purpose processor, a digital signal processor, an application-specific integrated circuit, a field-programmable gate array, or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute various exemplary logical blocks, modules, and circuits described in connection with the disclosure of the present application. The processor may also be a combination of computing functions, such as a combination of one or more microprocessors, a combination of a digital signal processor and a microprocessor, and so on. The bus 1004 may be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. The bus may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 10 only a thick line is shown in FIG. 13, but it does not mean that there is only one bus or one type of bus.

[0195] In another embodiment of the present application, a computer-readable storage medium is further provided. The computer-readable storage medium includes instructions that, when run on a computer, cause the computer to execute the above Figure 2The method described in the embodiments.

[0196] Those of ordinary skill in the art can realize that the units of each example described in combination with the embodiments disclosed herein can be implemented by electronic hardware, computer software, or a combination of the two. To clearly illustrate the interchangeability of hardware and software, the composition of each example has been generally described according to functions in the above description. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present invention.

[0197] Those skilled in the art can clearly understand that for the convenience and brevity of description, the specific working processes of the systems, devices, and units described above can refer to the corresponding processes in the foregoing method embodiments and will not be elaborated here.

[0198] In the several embodiments provided in the embodiments of the present application, it should be understood that the disclosed systems, devices, and methods can be implemented in other ways. For example, the device embodiments described above are merely illustrative. For example, the division of the units is only a logical function division, and there can be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be through some interfaces. The indirect couplings or communication connections of the devices or units can be in electrical, mechanical, or other forms.

[0199] The units described as separate components may or may not be physically separated. The components displayed as units may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0200] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above integrated units can be implemented in the form of hardware or in the form of software functional units.

[0201] When the above-mentioned functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as USB flash drives, read-only memories (ROMs), random access memories (RAMs), mobile hard disks, magnetic disks, or optical discs.

[0202] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some or all of the technical features. These modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the various embodiments of the present invention, and they should all be covered by the scope of the claims and the description of the present invention.

Claims

1. A modular vacuum circuit breaker adaptive fault identification method based on multiple optimization algorithms, characterized in that, Including: S1: Obtain the operation data of the modular vacuum circuit breaker, where the operation data includes multiple data samples; S2: Construct the first feature vector of each data sample based on the operation data; S3: Input the first feature vector into a preset first support vector machine (SVM) model, a second SVM model, and a third SVM model respectively for training, to obtain a first output result output by the first SVM model, a second output result output by the second SVM model, and a third output result output by the third SVM model; Wherein the training of the first SVM model is optimized based on a genetic algorithm, the training of the second SVM model is optimized based on a Bayesian algorithm, and the training of the third SVM model is optimized based on an optimal particle swarm algorithm; S4: Input the second feature vector into a preset random forest model for training, to obtain a fault identification result output by the random forest model; wherein, the training of the random forest model is optimized based on the optimal particle swarm algorithm, and the second feature vector is obtained based on the first output result, the second output result, and the third output result corresponding to the same data sample; S5: Conduct fault identification on the modular vacuum circuit breaker based on a fault identification model, where the first layer of the fault identification model includes the first SVM model, the second SVM model, and the third SVM model, and the second layer of the fault identification model includes the random forest model; the input of the first layer is the first feature vector, the output of the first layer is the second feature vector, the input of the second layer is the output of the first layer, and the output of the second layer is the fault identification result.

2. The method according to claim 1, wherein The data sample includes the closing and opening coil current signals within a preset time window, and the first feature vector includes a first feature; constructing the first feature vector of each data sample based on the operation data includes: Calculate the slope of the data points on the change curve of the closing and opening coil current signals, and the calculation formula of the slope is: Where x is time, y is the intensity of the closing and opening coil current signal, x0 is the initial moment, Δx is the change amount of time, Δy is the change amount of the closing and opening coil current signal, f(x) is the approximate function of the closing and opening coil current signal changing with time, and f'(x) is the slope of the data point corresponding to time x on the change curve of the closing and opening coil current signal; When the slope of the i-th data point is 0, and the slopes of the (i - 1)-th data point and the (i + 1)-th data point are not all 0, determine the midpoint coordinates of the i-th data point and the (i + 1)-th data point as the extreme point coordinates; When the slope of the i-th data point is not 0, and the slopes of the (i - 1)-th data point and the (i + 1)-th data point belong to positive and negative values respectively, determine the (i + 1)-th data point as the extreme point coordinates; Based on the extreme point coordinates, determine the peaks and valleys of the change curve; Construct the first feature based on the time and current corresponding to the first peak of the change curve, the time and current corresponding to the first trough, the time and current corresponding to the second peak, and the initial moment.

3. The method according to claim 2, characterized in that, The first feature vector includes a second feature; constructing the first feature vector of each data sample based on the operation data includes: Calculating the mean, standard deviation, and kurtosis of the switching coil current signal; wherein, the calculation formula of the kurtosis is: where K is the kurtosis of the current signal, N is the number of data points of the opening and closing coil current signals in the data sample, y i is the current signal strength of the i-th data point, μ is the mean of the opening and closing coil current signals, and σ is the standard deviation of the opening and closing coil current signals; Construct the second feature based on the mean, the standard deviation, and the kurtosis.

4. The method according to claim 2, characterized in that The data sample includes the switching coil voltage signal within a preset time window, and the first feature vector includes a third feature; constructing the first feature vector of each data sample based on the operation data includes: Based on the switching coil voltage signal and the switching coil current signal, calculate the work done by the closing current; the calculation formula for calculating the work done by the closing current is: Wherein, W is the work done by the closing current, U is the voltage signal of the closing and opening coil, I is the current signal of the closing and opening coil, and t is the length of the preset time window. is the integral symbol, and dt represents an infinitesimal increment in time; Construct the third feature based on the work done.

5. The method according to claim 2, wherein The data sample includes the three-phase switching time, the transient values and amplitudes of the three-phase currents, the harmonic amplitudes of the three-phase currents, and the rated current of the modular vacuum circuit breaker; the first feature vector includes a fourth feature; constructing the first feature vector of each data sample based on the operation data includes: Determine the switching time of the modular vacuum circuit breaker based on the maximum value of the three-phase switching time; Based on the transient values and peak values of the three-phase currents, calculate the three-phase difference degree; the calculation formula for calculating the three-phase difference degree is: Among them, λ represents the three-phase difference degree, and i U , i V , and i W are respectively the amplitudes of the U-phase current, V-phase current, and I-phase current in the three-phase current; I U , I V , and I W are the transient values of the U-phase current, V-phase current, and I-phase current in the three-phase current, and 2πj and 2π(j + 1) represent points on the complex plane; Based on the harmonic amplitudes of the three-phase currents, calculate the maximum harmonic value of the three-phase currents; Based on the harmonic amplitudes of the three-phase currents and the rated current, calculate the harmonic distortion rate of the three phases; Construct the fourth feature based on the switching time, the three-phase difference degree, the maximum harmonic value, and the harmonic distortion rate.

6. The method according to any one of claims 1-5, characterized in that, After step S5, the method further includes: S6: Transmit the recognition result of the fault recognition model to the upper computer so that the upper computer visually displays the recognition result.

7. The method according to any one of claims 1-5, characterized in that, After step S5, the method further includes: S7: Based on the recognition result of the fault recognition model, perform incremental optimization on the fault recognition model.

8. A modular vacuum circuit breaker adaptive fault identification system based on multiple optimization algorithms, characterized in that, The system includes an upper computer, a lower computer, and a modular vacuum circuit breaker. The upper computer is connected to the lower computer, and the lower computer is connected to the modular vacuum circuit breaker; The lower computer is configured to apply the method according to any one of claims 1 to 7 to perform fault recognition on the modular vacuum circuit breaker and transmit the fault recognition result to the upper computer.

9. A computing device, characterized in that, Including: A memory for storing programs; A processor for loading the program to execute the method according to any one of claims 1 - 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored program, wherein when the program runs, it controls the device where the computer-readable storage medium is located to execute the method according to any one of claims 1 - 7.

Images