High-voltage circuit breaker spring state identification method and system, storage medium and equipment

Abstract

The embodiment of the invention discloses a method for identifying the spring state of a high-voltage circuit breaker. The method comprises the following steps: acquiring a contact closing stroke curve in the closing process of the high-voltage circuit breaker; feature extraction is conducted on the contact closing stroke curve, curve feature values are obtained, and the curve feature values comprise the time when the contact stroke reaches the maximum value and the contact movement speed maximum point; constructing a deep forest model, and training the deep forest model according to the curve characteristic value; and identifying the spring state of the high-voltage circuit breaker by using the trained deep forest model. According to the method, the state of the spring of the high-voltage circuit breaker can be effectively identified by acquiring the contact closing stroke curve, extracting the characteristic value and utilizing the deep forest model for training and identification, and technical support is provided for maintenance and fault diagnosis of the high-voltage circuit breaker.

Description

Technical Field

[0001] The present invention relates to the technical field of circuit breaker detection, and in particular to a method, system, storage medium and equipment for identifying a spring state of a high-voltage circuit breaker. Background Art

[0002] As a key equipment in the power system, high-voltage circuit breakers bear the important responsibility of protecting and controlling circuits. The stability and reliability of their performance are directly related to the safe operation of the power system. However, in actual applications, mechanical failures frequently occur in the spring operating mechanism of high-voltage circuit breakers. These failures not only affect the normal operation of the circuit breaker, but also seriously reduce the reliability of power supply. Therefore, accurate identification and monitoring of the spring status of high-voltage circuit breakers is of great significance for preventing faults and ensuring the stable operation of the power system.

[0003] At present, in the field of high-voltage circuit breaker fault detection, although some online sensing technologies have been applied, the data collected by these technologies often have great limitations, which makes the machine learning algorithms trained based on these data have low confidence and it is difficult to achieve ideal results in actual online monitoring. Summary of the invention

[0004] Based on this, it is necessary to propose a high-voltage circuit breaker spring state identification method to address the above problems.

[0005] A method for identifying a high-voltage circuit breaker spring state, the method comprising the following steps:

[0006] Obtaining a contact closing stroke curve during the closing process of a high-voltage circuit breaker;

[0007] Extracting features of the contact closing stroke curve to obtain curve feature values, wherein the curve feature values ​​include the time when the contact stroke reaches a maximum value and the maximum point of the contact movement speed;

[0008] Constructing a deep forest model, and training the deep forest model according to the curve characteristic values;

[0009] The trained deep forest model is used to identify the spring state of high-voltage circuit breaker.

[0010] In the above scheme, the step of obtaining the contact closing stroke curve during the closing process of the high-voltage circuit breaker specifically includes:

[0011] Collect the acceleration signal in the closing process in real time and filter it to obtain the processed acceleration signal;

[0012] Decomposing the processed acceleration signal to obtain a plurality of intrinsic mode functions;

[0013] Integrating the plurality of intrinsic mode functions to obtain a velocity signal;

[0014] The speed signal is integrated to obtain a contact closing stroke curve.

[0015] In the above scheme, the feature extraction of the contact closing stroke curve is performed to obtain the curve feature value, and the curve feature value includes the time when the contact stroke reaches the maximum value and the maximum point of the contact movement speed, specifically including:

[0016] Traversing the contact closing stroke curve, and determining the time when the contact stroke reaches the maximum value according to the local maximum value thereof;

[0017] Numerically integrating the contact closing stroke curve to obtain a speed curve;

[0018] The speed curve is traversed, and the maximum point of the contact movement speed is determined according to its local maximum value.

[0019] In the above scheme, the step of constructing a deep forest model and training the deep forest model according to the curve characteristic value specifically includes:

[0020] Construct a feature set X based on the curve eigenvalues i , where Xi = [t S_max ,t v_max ], t S_max is the time when the contact stroke reaches the maximum value, t v_max is the time when the speed reaches the maximum value;

[0021] The feature set X i With the spring status label y i Make associations;

[0022] Set the sampling window and sampling step size, and i Perform sliding sampling;

[0023] Inputting a plurality of subsamples after the sliding sampling into the deep forest model, and obtaining corresponding spring state labels;

[0024] Get the output of the deep forest model and the feature set X i The accuracy of the corresponding spring state label, when the accuracy meets the threshold condition, it is determined that the deep forest model training is completed.

[0025] In the above scheme, the step of inputting the plurality of subsamples after the sliding sampling into the deep forest model and obtaining the corresponding spring state labels specifically includes:

[0026] The deep forest model includes a multi-granularity scanning module and a layer cascade forest module, and the multi-granularity scanning module includes a random forest layer and a common random forest layer;

[0027] Input the plurality of subsamples into the random forest layer and the common random forest layer respectively, obtain corresponding probability vectors and concatenate them to form a first feature vector;

[0028] Using the first feature vector as input to a layer-by-layer cascade forest module;

[0029] The layer cascade forest module includes a plurality of layer cascade forests, wherein each layer of the cascade forest receives the output features of the previous layer as input and outputs the features corresponding to the layer;

[0030] The layer cascade forest module associates the features output by each layer of the cascade forest with the corresponding spring state label, and outputs a probability vector, wherein the probability vector represents the probability that the sample belongs to the spring state label category;

[0031] When the probability vector meets the threshold condition, the corresponding spring state label is determined according to the probability vector currently output by the layer cascade forest module.

[0032] In the above scheme, the use of the trained deep forest model to identify the state of the high-voltage circuit breaker spring specifically includes:

[0033] Determine the characteristic vector Ai according to the characteristic value of the curve, Ai=[t S_max ,t v_max ];

[0034] Input the feature vector Ai into the trained deep forest model;

[0035] The deep forest model outputs a corresponding probability vector, wherein the probability vector represents the probability that the feature vector belongs to each spring state category;

[0036] The category with the highest probability that the feature vector belongs to the spring state category is taken as the final prediction result to determine the spring state of the high-voltage circuit breaker.

[0037] The above scheme also includes:

[0038] Obtaining a contact closing stroke curve during the closing process of a high-voltage circuit breaker;

[0039] Extracting features of the contact closing stroke curve to obtain a curve feature value;

[0040] Constructing a validation set according to the curve characteristic values;

[0041] Use the validation set to validate the trained deep forest model and evaluate the accuracy of the deep forest model;

[0042] If the accuracy of the deep forest model does not reach the set threshold range, adjust the parameters in the deep forest model or increase the number of subsamples, and retrain the deep forest model until the accuracy meets the set threshold range.

[0043] The present application also proposes a high-voltage circuit breaker spring state identification system, the system comprising: a stroke acquisition unit, a feature value acquisition unit, a model training unit and a spring state identification unit;

[0044] The stroke acquisition unit is used to acquire a contact closing stroke curve during the closing process of the high-voltage circuit breaker;

[0045] The characteristic value acquisition unit is used to extract the characteristics of the contact closing stroke curve to obtain the curve characteristic value, wherein the curve characteristic value includes the time when the contact stroke reaches the maximum value and the maximum point of the contact movement speed;

[0046] The model training unit is used to construct a deep forest model and train the deep forest model according to the curve characteristic value;

[0047] The spring state recognition unit is used to identify the state of the high-voltage circuit breaker spring using the trained deep forest model.

[0048] The present application also proposes a readable storage medium storing a computer program. When the computer program is executed by a processor, the processor executes the following steps:

[0049] Obtaining a contact closing stroke curve during the closing process of a high-voltage circuit breaker;

[0050] Extracting features of the contact closing stroke curve to obtain curve feature values, wherein the curve feature values ​​include the time when the contact stroke reaches a maximum value and the maximum point of the contact movement speed;

[0051] Constructing a deep forest model, and training the deep forest model according to the curve characteristic values;

[0052] The trained deep forest model is used to identify the spring state of high-voltage circuit breaker.

[0053] The present application also proposes a computer device, comprising a memory and a processor, wherein the memory stores a computer program, and the processor executes the following steps of the computer program:

[0054] Obtaining a contact closing stroke curve during the closing process of a high-voltage circuit breaker;

[0055] Extracting features of the contact closing stroke curve to obtain curve feature values, wherein the curve feature values ​​include the time when the contact stroke reaches a maximum value and the maximum point of the contact movement speed;

[0056] Constructing a deep forest model, and training the deep forest model according to the curve characteristic values;

[0057] The trained deep forest model is used to identify the spring state of high-voltage circuit breaker.

[0058] The embodiment of the present invention has the following beneficial effects: first, obtain the contact closing stroke curve during the closing process of the high-voltage circuit breaker; extract features from the contact closing stroke curve to obtain curve feature values, the curve feature values ​​include the time when the contact stroke reaches the maximum value and the maximum point of the contact movement speed; construct a deep forest model, and train the deep forest model according to the curve feature values; use the trained deep forest model to identify the state of the high-voltage circuit breaker spring. This method combines the deep forest model to obtain the contact closing stroke curve, extract feature values, and use the deep forest model for training and identification, which can effectively identify the state of the high-voltage circuit breaker spring and provide technical support for the maintenance and fault diagnosis of high-voltage circuit breakers. BRIEF DESCRIPTION OF THE DRAWINGS

[0059] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0060] in:

[0061] Figure 1 A schematic diagram of a flow chart of a method for identifying a spring state of a high-voltage circuit breaker in an embodiment;

[0062] Figure 2 Schematic diagram of contact closing stroke curves in different states in one embodiment. DETAILED DESCRIPTION

[0063] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, rather than all the embodiments; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0064] In the following description, a large number of specific details are given in order to provide a more thorough understanding of the present invention; however, it is obvious to those skilled in the art that the present invention can be implemented without one or more of these details; in other examples, in order to avoid confusion with the present invention, some technical features known in the art are not described, and it should be understood that the present invention can be implemented in different forms and should not be interpreted as limited to the embodiments presented here; on the contrary, providing these embodiments will make the disclosure thorough and complete and fully convey the scope of the present invention to those skilled in the art.

[0065] The purpose of the terms used herein is only to describe specific embodiments and is not intended to be limiting of the present invention. When used herein, the singular forms "a", "an" and "said / the" are also intended to include the plural forms, unless the context clearly indicates otherwise, and it should also be understood that the terms "consisting of" and / or "comprising", when used in this specification, determine the presence of the features, integers, steps, operations, elements and / or parts, but do not exclude the presence or addition of one or more other features, integers, steps, operations, elements, parts and / or groups. When used herein, the term "and / or" includes any and all combinations of the relevant listed items.

[0066] In order to thoroughly understand the present invention, a detailed structure will be presented in the following description to illustrate the technical solution proposed by the present invention; optional embodiments of the present invention are described in detail as follows, but in addition to these detailed descriptions, the present invention may also have other implementation methods.

[0067] like Figure 1 As shown, in one embodiment, a method for identifying a high-voltage circuit breaker spring state is provided, and the method for identifying a high-voltage circuit breaker spring state includes steps S101 to S104, which are described in detail as follows:

[0068] S101, obtaining a contact closing stroke curve during a closing process of a high-voltage circuit breaker;

[0069] In some embodiments, obtaining a contact closing stroke curve during a high-voltage circuit breaker closing process specifically includes:

[0070] Collect the acceleration signal in the closing process in real time and filter it to obtain the processed acceleration signal;

[0071] Decomposing the processed acceleration signal to obtain several intrinsic mode functions;

[0072] Integrate several intrinsic mode functions to obtain velocity signals;

[0073] Integrate the speed signal to obtain the contact closing stroke curve.

[0074] Preferably, the acceleration sensor is installed at a position that can reflect the mechanical behavior of the spring, and the collected acceleration signal needs to be processed in real time to extract the acceleration change during the contact closing process. The collected acceleration signal is filtered using the ensemble empirical mode decomposition method to remove high-frequency noise. By adding Gaussian white noise to the original signal as the scale of the supplementary signal, signals of various scales can be automatically mapped to the appropriate scale related to the Gaussian white noise, changing the interval between extreme points so that the signal is evenly distributed over the entire frequency band, supplementing the lost signal scale, and ensuring that the local mean of the upper and lower envelopes of the signal can be accurately obtained each time the signal is calculated, thereby avoiding modal aliasing.

[0075] Specifically, the acceleration signal is the second-order derivative of displacement with respect to time, so to obtain the velocity signal from the acceleration signal, it is actually to integrate the acceleration signal. The formula for obtaining the velocity signal is:

[0076] v(t)=∫a(t)dt

[0077] Then, the displacement signal (contact closing stroke curve) is obtained by integrating the speed signal:

[0078] S(t)=∫v(t)dt

[0079] Among them, v(t) is the velocity signal, a(t) is the acceleration signal, and S(t) is the displacement signal.

[0080] By collecting the acceleration signal during the closing process and filtering, decomposing and integrating it, the contact closing stroke curve can be obtained. This curve reflects the movement trajectory of the contact during the closing process and can be used to analyze the spring state.

[0081] S102, extracting features from the contact closing stroke curve to obtain curve feature values, where the curve feature values ​​include the time when the contact stroke reaches a maximum value and the maximum point of the contact movement speed;

[0082] By analyzing the contact closing stroke curve, some key characteristic values ​​can be extracted, such as the time when the contact stroke reaches the maximum value and the maximum point of the contact movement speed. These characteristic values ​​can reflect the influence of factors such as spring elasticity and friction on the closing process, and are an important basis for identifying the spring state.

[0083] In some embodiments, feature extraction is performed on the contact closing stroke curve to obtain a curve feature value, which includes the time when the contact stroke reaches a maximum value and the maximum point of the contact movement speed, specifically including:

[0084] Traversing the contact closing stroke curve, and determining the time when the contact stroke reaches the maximum value according to its local maximum value;

[0085] Perform numerical integration on the contact closing stroke curve to obtain the speed curve;

[0086] Traverse the velocity curve and determine the maximum point of the contact movement velocity based on its local maximum value.

[0087] Preferably, the time point when the contact travel reaches the maximum value is the local maximum value in the S(t) signal. By traversing the travel signal S(t), the local maximum value is found, that is: S(t[i])>S(t[i-1]) and S(t[i])>S(t[i+1]). At this time, t[i] is considered to be the time when the contact travel reaches the maximum.

[0088] If the spring is in good condition, the speed of the contact is usually relatively smooth and in line with the expected movement law; if the spring condition deteriorates (for example, spring fatigue, relaxation or damage), the speed of the contact may be abnormal, such as too slow or too fast, impact changes, etc. Under normal circumstances, the speed of closing and opening should be faster at the beginning of the contact stroke, and then gradually slow down. When the spring condition is poor, it may appear that the initial speed is too low, or the speed changes irregularly near the end. By comparing the time when the maximum speed occurs during the closing and opening of the contact, the condition of the spring can be revealed. For example, the maximum speed of a normal spring usually occurs in the middle of the contact stroke, while spring problems may cause the maximum speed to be advanced or delayed.

[0089] We have already obtained the velocity curve v(t) by integrating the acceleration. We can also traverse the velocity curve to find the local maximum point: v(t[i])>v(t[i-1]) and v(t[i])>v(t[i+1]). At this time, t[i] is considered to be the time when the velocity reaches the maximum.

[0090] It can be seen that the state of the spring can be determined by combining the acceleration signal with the time when the contact stroke reaches the maximum value and the time when the speed reaches the maximum value.

[0091] S103, constructing a deep forest model, and training the deep forest model according to the curve eigenvalues;

[0092] The deep forest model is a powerful machine learning model that can effectively process nonlinear and high-dimensional data. By taking the extracted curve eigenvalues ​​as input and training the deep forest model, the model can learn the mapping relationship between the eigenvalues ​​and the spring state, thereby realizing the recognition of the spring state.

[0093] In some embodiments, a deep forest model is constructed, and the deep forest model is trained according to the curve feature value, specifically including:

[0094] Construct feature set X based on curve eigenvalues i , where Xi = [t S_max ,tv_max ], t S_max is the time when the contact stroke reaches the maximum value, t v_max is the time when the speed reaches the maximum value;

[0095] The feature set X i With the spring status label y i Make associations;

[0096] Set the sampling window and sampling step size, and i Perform sliding sampling;

[0097] Input several subsamples after sliding sampling into the deep forest model and obtain the corresponding spring state labels;

[0098] Get the output of the deep forest model and the feature set X i The accuracy of the corresponding spring state label. When the accuracy meets the threshold condition, it is determined that the deep forest model training is completed.

[0099] The label yi represents the state of the spring, for example, 1 represents the normal state and 0 represents the fault state.

[0100] In some embodiments, a plurality of subsamples after sliding sampling are input into a deep forest model, and corresponding spring state labels are obtained, specifically including:

[0101] The deep forest model includes a multi-granularity scanning module and a layer cascade forest module. The multi-granularity scanning module includes a random forest layer and a common random forest layer.

[0102] Input several subsamples into the random forest layer and the common random forest layer respectively, obtain the corresponding probability vectors and concatenate them to form the first feature vector;

[0103] The first eigenvector is used as the input of the layer-wise cascade forest module;

[0104] The layer cascade forest module contains several layers of cascade forests, where each layer of cascade forest receives the output features of the previous layer as input and outputs the features corresponding to the layer;

[0105] The layer cascade forest module associates the features output by each layer of the cascade forest with the corresponding spring state label and outputs a probability vector, which represents the probability that the sample belongs to the spring state label category;

[0106] When the probability vector meets the threshold condition, the corresponding spring state label is determined according to the probability vector currently output by the layer cascade forest module.

[0107] The threshold condition is the target accuracy range (such as the target accuracy is greater than 95%). When the accuracy of the validation set during training reaches or exceeds this threshold, the training is stopped.

[0108] Specifically, the multi-granularity scanning module obtains different enhanced features through convolution kernel sampling operations of different scales, and references a sliding window for local sampling to enhance the original features. After inputting a complete N-dimensional sample, a sliding sampling is performed through a sampling window of length L. If the number of sub-sample vectors is X and the sampling step is F, then:

[0109] X=(NL) / F+1

[0110] Each subsample is used for training of completely random forest and ordinary random forest, and a probability vector of length A is obtained in each forest. In this way, each forest will produce a representation vector of length X*A (that is, the probability vector converted and concatenated by the random forest). Finally, the results of the K forests in each layer are concatenated together to obtain the output of this layer.

[0111] In the cascade forest stage of the deep forest algorithm, there are multiple completely random forests and random forest models in each layer of the cascade forest. The deep forest model uses two different forest models, the random forest model and the completely random tree forest model, to ensure the generalization of the prediction model and reduce the test variance of the model. By enhancing the data layer by layer in a forest cascade manner, it is beneficial to improve the prediction accuracy and enhance the representation learning ability of the algorithm layer by layer.

[0112] In the process of generating cascade forests, the input of the first layer is the feature vector generated after multi-granularity scanning sampling, and the output is the concatenation of the original input feature vector and the feature category vector generated by this layer of forest. After that, the input of each layer is the output of the previous layer of cascade forest, and new feature information is generated and passed to the next layer using a method similar to the first layer. In addition, when a new forest layer is expanded, the performance of the entire cascade will be evaluated using the validation set. If there is no significant performance gain, the training process will be terminated.

[0113] S104. Use the trained deep forest model to identify the state of the high-voltage circuit breaker spring.

[0114] Using the trained deep forest model, the new contact closing stroke curve eigenvalues ​​can be input into the model and the corresponding spring state prediction results can be obtained. This can quickly and accurately judge the state of the spring and perform corresponding maintenance or replacement operations to ensure the normal operation of the high-voltage circuit breaker.

[0115] In some embodiments, the trained deep forest model is used to identify the state of the high-voltage circuit breaker spring, specifically including:

[0116] Determine the eigenvector Ai according to the eigenvalue of the curve, Ai=[t S_max ,t v_max ];

[0117] Input the feature vector Ai into the trained deep forest model;

[0118] The deep forest model outputs the corresponding probability vector, which represents the probability that the feature vector belongs to each spring state category;

[0119] The category with the highest probability that the feature vector belongs to the spring state category is taken as the final prediction result to determine the spring state of the high-voltage circuit breaker.

[0120] In some embodiments, the label yi has 5 states, (1) normal state, (2) closing spring pre-compression is reduced by 10mm, (3) closing spring pre-compression is reduced by 15mm, (4) opening spring pre-compression is reduced by 10mm, and (5) opening spring pre-compression is reduced by 15mm; if the model inputs 150-dimensional sequence data, the sampling window is set to 90 dimensions, and the sampling is performed step by step. The default sampling step is 1, and 61 sub-samples can be obtained in the end. The length of a single probability is 5, so the length of each forest is 305. If the number of forests in this layer is two, the output vector dimension is 610. This is used as the input of the first layer of the cascade forest. After training in the cascade forest stage, a model for identifying the state of the high-voltage circuit breaker spring is obtained. 20 groups of data are randomly selected from the above 150 groups of experimental data and identified using the model.

[0121] like Figure 2As shown in Figure 3, line S1 represents the contact closing stroke curve under normal conditions, line S2 represents a 10mm reduction in pre-compression, line S3 represents a 15mm reduction in pre-compression, line S4 represents a 10mm reduction in opening spring pre-compression, and line S5 represents a 15mm reduction in opening spring pre-compression. It can be seen from the lines S2 and S3 in Figure 3 that the reduction in the pre-compression of the closing spring reduces the stored closing energy. The reduced work done during the closing process causes a longer time to complete the closing process compared with the normal state, and its stroke curve will shift to the right as a whole. The more the pre-compression is reduced, the less the closing energy is stored, and the longer the time required to complete the closing process is, so the S3 line is more obviously shifted to the right than the S2 line; it can be seen from the lines S4 and S5 in Figure 3 that the pre-compression of the opening spring is reduced, that is, the output of the opening spring is reduced. At this time, the resistance of the opening spring is reduced during the closing process, which causes the contact speed to increase during the closing process. The time required to complete the closing process is shorter than the normal state, and its stroke curve will shift to the left as a whole. The S4 line represents a 10mm reduction in pre-compression, and the S5 line represents a 15mm reduction in pre-compression. The more the pre-compression is reduced, the more the output of the opening spring is reduced, resulting in a greater reduction in the resistance of the opening spring. Therefore, the contact speed increases more during the closing process, so the S5 line is more obviously shifted to the left than the S4 line.

[0122] In some embodiments, it also includes:

[0123] Obtaining a contact closing stroke curve during the closing process of a high-voltage circuit breaker;

[0124] Extract features of the contact closing stroke curve to obtain curve characteristic values;

[0125] Construct a validation set based on the curve eigenvalues;

[0126] Use the validation set to validate the trained deep forest model and evaluate the accuracy of the deep forest model;

[0127] If the accuracy of the deep forest model does not reach the set threshold range, adjust the parameters in the deep forest model or increase the number of subsamples, and retrain the deep forest model until the accuracy meets the set threshold range.

[0128] In summary, the present invention proposes a state recognition method for a spring in a high-voltage circuit breaker, and obtains a contact closing stroke curve during the closing process of the high-voltage circuit breaker through an acceleration sensor. The information contained in the stroke curve is analyzed to obtain the contact closing stroke curve characteristics of the spring in the high-voltage circuit breaker under different states. By processing the characteristics of the contact closing stroke curve of the spring in different states, such as time and shape, a state recognition model for evaluating the high-voltage circuit breaker spring is obtained through data training, which improves the data limitations, low algorithm confidence, insufficient state monitoring technology, and insufficient targeted detection of the existing technology in the high-voltage circuit breaker.

[0129] The present application also proposes a high-voltage circuit breaker spring state identification system, the system comprising: a stroke acquisition unit, a feature value acquisition unit, a model training unit and a spring state identification unit;

[0130] A stroke acquisition unit, used to acquire a contact closing stroke curve during the closing process of the high-voltage circuit breaker;

[0131] A characteristic value acquisition unit is used to extract characteristics of the contact closing stroke curve to obtain a curve characteristic value, wherein the curve characteristic value includes the time when the contact stroke reaches a maximum value and the maximum point of the contact movement speed;

[0132] A model training unit, used to construct a deep forest model and train the deep forest model according to the curve characteristic values;

[0133] The spring state recognition unit is used to identify the state of the high-voltage circuit breaker spring using the trained deep forest model.

[0134] The present application also proposes a readable storage medium storing a computer program. When the computer program is executed by a processor, the processor executes the following steps:

[0135] Obtaining a contact closing stroke curve during the closing process of a high-voltage circuit breaker;

[0136] Extract features of the contact closing stroke curve to obtain curve characteristic values, which include the time when the contact stroke reaches the maximum value and the maximum point of the contact movement speed;

[0137] Construct a deep forest model and train it according to the curve eigenvalues;

[0138] The trained deep forest model is used to identify the spring state of high-voltage circuit breaker.

[0139] The present application also proposes a computer device, including a memory and a processor, wherein the memory stores a computer program, and the computer program is executed by the processor to perform the following steps:

[0140] Obtaining a contact closing stroke curve during the closing process of a high-voltage circuit breaker;

[0141] Extract features of the contact closing stroke curve to obtain curve characteristic values, which include the time when the contact stroke reaches the maximum value and the maximum point of the contact movement speed;

[0142] Construct a deep forest model and train it according to the curve eigenvalues;

[0143] The trained deep forest model is used to identify the spring state of high-voltage circuit breaker.

[0144] Those skilled in the art will understand that all or part of the processes in the above-mentioned embodiments can be completed by instructing the relevant hardware through a computer program, and the program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in the present application may include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0145] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0146] The above-mentioned embodiments only express several implementation methods of the present application, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the patent of the present application. It should be pointed out that, for ordinary technicians in this field, several modifications and improvements can be made without departing from the concept of the present application, and these all belong to the protection scope of the present application. The above disclosure is only a preferred embodiment of the present invention, and of course it cannot be used to limit the scope of the rights of the present invention. Therefore, equivalent changes made according to the claims of the present invention are still within the scope covered by the present invention.

Claims

1. A method for identifying the state of a high-voltage circuit breaker spring, characterized in that: The method comprises: Obtaining a contact closing stroke curve during the closing process of a high-voltage circuit breaker; Extracting features of the contact closing stroke curve to obtain curve feature values, wherein the curve feature values ​​include the time when the contact stroke reaches a maximum value and the maximum point of the contact movement speed; Constructing a deep forest model, and training the deep forest model according to the curve characteristic values; The trained deep forest model is used to identify the spring state of high-voltage circuit breaker.

2. The high-voltage circuit breaker spring state identification method according to claim 1, characterized in that: The step of obtaining a contact closing stroke curve during the closing process of a high-voltage circuit breaker specifically includes: Collect the acceleration signal in the closing process in real time and filter it to obtain the processed acceleration signal; Decomposing the processed acceleration signal to obtain a plurality of intrinsic mode functions; Integrating the plurality of intrinsic mode functions to obtain a velocity signal; The speed signal is integrated to obtain a contact closing stroke curve.

3. The high-voltage circuit breaker spring state identification method according to claim 1, characterized in that: The feature extraction of the contact closing stroke curve is performed to obtain a curve feature value, wherein the curve feature value includes the time when the contact stroke reaches the maximum value and the maximum point of the contact movement speed, specifically including: Traversing the contact closing stroke curve, and determining the time when the contact stroke reaches the maximum value according to the local maximum value thereof; Numerically integrating the contact closing stroke curve to obtain a speed curve; The speed curve is traversed, and the maximum point of the contact movement speed is determined according to its local maximum value.

4. The method for identifying the state of a high-voltage circuit breaker spring according to claim 1, characterized in that: The constructing of the deep forest model and training the deep forest model according to the curve characteristic value specifically includes: Construct a feature set X based on the curve eigenvalues i , where Xi = [t S_max ,t v_max ], t S_max is the time when the contact stroke reaches the maximum value, t v_max is the time when the speed reaches the maximum value; The feature set X i With the spring status label y i Make associations; Set the sampling window and sampling step size, and i Perform sliding sampling; Inputting a plurality of subsamples after the sliding sampling into the deep forest model, and obtaining corresponding spring state labels; Get the output of the deep forest model and the feature set X i The accuracy of the corresponding spring state label, when the accuracy meets the threshold condition, it is determined that the deep forest model training is completed.

5. The method for identifying the state of a high-voltage circuit breaker spring according to claim 4, characterized in that: The step of inputting a plurality of subsamples after the sliding sampling into the deep forest model and obtaining corresponding spring state labels specifically includes: The deep forest model includes a multi-granularity scanning module and a layer cascade forest module, and the multi-granularity scanning module includes a random forest layer and a common random forest layer; Input the plurality of subsamples into the random forest layer and the common random forest layer respectively, obtain corresponding probability vectors and concatenate them to form a first feature vector; Using the first feature vector as input to a layer-by-layer cascade forest module; The layer cascade forest module includes a plurality of layer cascade forests, wherein each layer of the cascade forest receives the output features of the previous layer as input and outputs the features corresponding to the layer; The layer cascade forest module associates the features output by each layer of the cascade forest with the corresponding spring state label, and outputs a probability vector, wherein the probability vector represents the probability that the sample belongs to the spring state label category; When the probability vector meets the threshold condition, the corresponding spring state label is determined according to the probability vector currently output by the layer cascade forest module.

6. The method for identifying the state of a high-voltage circuit breaker spring according to claim 1, characterized in that: The method of using the trained deep forest model to identify the state of the high-voltage circuit breaker spring specifically includes: Determine the characteristic vector Ai according to the characteristic value of the curve, Ai=[t S_max ,t v_max ]; Input the feature vector Ai into the trained deep forest model; The deep forest model outputs a corresponding probability vector, wherein the probability vector represents the probability that the feature vector belongs to each spring state category; The category with the highest probability that the feature vector belongs to the spring state category is taken as the final prediction result to determine the spring state of the high-voltage circuit breaker.

7. The method for identifying the state of a high-voltage circuit breaker spring according to claim 1, characterized in that: Also includes: Obtaining a contact closing stroke curve during the closing process of a high-voltage circuit breaker; Extracting features of the contact closing stroke curve to obtain a curve feature value; Constructing a validation set according to the curve characteristic values; Use the validation set to validate the trained deep forest model and evaluate the accuracy of the deep forest model; If the accuracy of the deep forest model does not reach the set threshold range, adjust the parameters in the deep forest model or increase the number of subsamples, and retrain the deep forest model until the accuracy meets the set threshold range.

8. A high voltage circuit breaker spring state identification system, characterized in that: The system comprises: a stroke acquisition unit, a characteristic value acquisition unit, a model training unit and a spring state recognition unit; The stroke acquisition unit is used to acquire a contact closing stroke curve during the closing process of the high-voltage circuit breaker; The characteristic value acquisition unit is used to extract the characteristics of the contact closing stroke curve to obtain the curve characteristic value, wherein the curve characteristic value includes the time when the contact stroke reaches the maximum value and the maximum point of the contact movement speed; The model training unit is used to construct a deep forest model and train the deep forest model according to the curve characteristic value; The spring state recognition unit is used to identify the state of the high-voltage circuit breaker spring using the trained deep forest model.

9. A readable storage medium storing a computer program, wherein when the computer program is executed by a processor, the processor is caused to execute the steps of the method according to any one of claims 1 to 7.

10. A computer device comprising a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of the method according to any one of claims 1 to 7.

Images