CN111008648A - Fault identification method, system and medium for pure optical fiber electronic current transformer - Google Patents
Fault identification method, system and medium for pure optical fiber electronic current transformer Download PDFInfo
- Publication number
- CN111008648A CN111008648A CN201911100607.6A CN201911100607A CN111008648A CN 111008648 A CN111008648 A CN 111008648A CN 201911100607 A CN201911100607 A CN 201911100607A CN 111008648 A CN111008648 A CN 111008648A
- Authority
- CN
- China
- Prior art keywords
- fault
- fault identification
- current transformer
- value
- function
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Testing Of Short-Circuits, Discontinuities, Leakage, Or Incorrect Line Connections (AREA)
Abstract
The invention discloses a method, a system and a medium for identifying faults of a pure optical fiber electronic current transformer, wherein the method for identifying the faults comprises the following steps: acquiring a state signal x (t) of the pure optical fiber type current transformer; carrying out modal decomposition on the state signal x (t) to obtain a plurality of intrinsic mode function components; performing principal component analysis on the plurality of eigenmode function components by using a principal component analysis method to extract a feature vector for fault identification; and inputting the characteristic vector into a pre-trained machine learning classification model to obtain a fault identification result of the pure optical fiber type current transformer, wherein the machine learning classification model is pre-trained to establish mapping between the characteristic vector and the fault identification result. The method can effectively realize the fault identification of the pure optical fiber electronic current transformer, and has the advantages of high fault identification accuracy, simple and quick principle and high identification speed.
Description
Technical Field
The invention relates to a current transformer fault detection technology, in particular to a pure optical fiber electronic current transformer fault identification method, a system and a medium.
Background
In recent years, as all-fiber electronic transformers are gradually applied and popularized in intelligent substations, sensing characteristics and operation state characteristic identification of the all-fiber electronic transformers are very necessary. With the development of large capacity and extra-high voltage of power systems, the requirements on miniaturization, intellectualization and high reliability of power equipment are higher and higher. In recent years, various electronic transformers have been developed rapidly along with the progress of optical fibers and electronic technology. Different from the traditional mutual inductor, the electronic mutual inductor has the advantages of light weight, small size, excellent insulating property, strong anti-electromagnetic interference capability, large dynamic measurement range, wide frequency response range, high sensitivity, convenience in realizing intellectualization and the like. The all-fiber current transformer is one of the main technical lines of an electronic current transformer, has the general advantages of the electronic transformer, and simultaneously can measure direct current as well as alternating current because the measurement principle is that the phase difference of a current induced magnetic field to a light beam propagating in an optical fiber is utilized.
The optical fiber current sensor is based on Faraday magneto-optical effect and digital closed loop detection technology, and has the advantages of high detection precision, good insulation performance, high bandwidth, strong external interference resistance, small volume, flexible structure and the like. There are many electromagnetic current transformers using the electromagnetic induction principle in the power system, but such transformers have many problems, such as: the insulation structure has large volume, high cost and reduced reliability under high voltage; poor linearity, small dynamic range and the like. The pure optical fiber electronic transformer is a novel current and voltage sensor with mature theory, wide spectrum range and excellent response characteristic, is one of the development trends of strong smart grid application equipment in the future, is also an important sensing element ubiquitous in the sensing layer of the power internet of things, is applied more in an extra-high voltage direct current system at present, and is applied to a small number of electronic transformers including the pure optical fiber electronic transformer in the early stage of smart grid construction. The characteristic extraction and the fault identification of the optical fiber sensing signals are researched, and the characteristic extraction has very important influence on the accuracy of the mode identification.
The pure optical fiber electronic current transformer can be well utilized for identifying the faults of the pure optical fiber current transformer, on one hand, environment disturbance signals can be accurately sensed by utilizing the sensitivity of optical fiber sensing, on the other hand, signal processing and characteristic extraction and fault identification can be carried out on mixed signals which are sensed by the optical fiber sensor and contain various noises and disturbance information, so that corresponding measures can be taken by workers, and certain promotion effect is played for the subsequent application of the ubiquitous power internet of things and the smart grid.
Disclosure of Invention
The technical problems to be solved by the invention are as follows: aiming at the problems in the prior art, the invention provides a method, a system and a medium for identifying the fault of the pure optical fiber electronic current transformer.
In order to solve the technical problems, the invention adopts the technical scheme that:
a fault identification method for a pure optical fiber electronic current transformer comprises the following implementation steps:
1) acquiring a state signal x (t) of the pure optical fiber type current transformer;
2) carrying out modal decomposition on the state signal x (t) to obtain a plurality of intrinsic mode function components;
3) performing principal component analysis on the plurality of eigenmode function components by using a principal component analysis method to extract a feature vector for fault identification;
4) and inputting the characteristic vector into a pre-trained machine learning classification model to obtain a fault identification result of the pure optical fiber type current transformer, wherein the machine learning classification model is pre-trained to establish mapping between the characteristic vector and the fault identification result.
Optionally, the detailed steps of step 2) include:
2.1) determining all extreme points of the state signal x (t), including a maximum point and a minimum point;
2.2) interpolating all extreme points by adopting a spline function, solving upper envelope lines and lower envelope lines, and connecting the mean values m (t) of the upper envelope lines and the lower envelope lines in sequence to obtain a mean value line;
2.3) finding extreme values at the left end and the right end according to the waveform characteristics of the state signals x (t), and eliminating an end effect by utilizing mirror continuation;
2.4) subtracting the state signal x (t) from the mean value m (t) of the mean value line to obtain a difference value h (t);
2.5) checking whether the difference h (t) meets the condition of the intrinsic mode function, and if not, taking the difference h (t) as a newly input state signal x (t) to jump to execute the step 2.1); otherwise, calculating residual errors r (t) among the state signals x (t) and the difference values h (t);
2.6) judging whether the residual r (t) meets the termination condition, if the residual r (t) is a non-monotonic function, judging that the residual r (t) does not meet the termination condition, and using the residual r (t) as a newly input state signal x (t) to jump to execute the step 2.1); otherwise, judging that the termination condition is met, ending the modal decomposition process, outputting the intrinsic mode function as the intrinsic mode function component obtained by the iteration of the current round, and skipping to execute the step 3).
Optionally, the detailed steps of step 3) include:
3.1) taking the plurality of intrinsic mode function components as a data sample set, and calculating a mean vector of samples in the sample data set;
3.2) removing the mean value of each sample, and centralizing the sample data;
3.3) constructing a covariance matrix of a data matrix obtained by centralizing sample data;
3.4) carrying out characteristic decomposition on the covariance matrix to obtain an eigenvalue and a corresponding eigenvector;
and 3.5) screening out the feature vectors with the contribution rate larger than the set contribution rate threshold value as the feature vectors for fault identification.
Optionally, the machine learning classification model in step 4) is an SVM identification classifier composed of support vector machines SVM 1-SVMk, the number of the support vector machines SVM 1-SVMk is k in total and respectively corresponds to k fault identification results one by one, the support vector machines SVM 1-SVMk form a hierarchical relationship, and any one-level support vector machine SVMiThe output result of (1) comprises corresponding fault identification result and other faults, and the input of the output result is input into the next-stage support vector machine SVM under the condition of outputting other faultsi+1And step 4) is also preceded by a step of training a machine learning classification model, and the detailed steps comprise:
s1) respectively acquiring corresponding state signal samples for various fault identification results of the pure optical fiber current type mutual inductor, wherein the number of the state signal samples of each fault identification result is the same; adding corresponding fault identification result classification labels to all the state signal samples to form training samples with labels and test samples;
s2) establishing a machine learning classification model;
s3) for each support vector machine SVMi in the machine learning classification model: inputting the training sample into a Support Vector Machine (SVMi) for training to find the optimal parameter of the SVMi; inputting the test sample into the SVMi to be tested to obtain a corresponding fault identification result, and judging that the training of the SVMi is finished if the identification accuracy of the fault identification result of the test sample reaches a preset requirement; otherwise, continuing to train the support vector machine SVMi until the identification accuracy of the fault identification result of the test sample reaches the preset requirement.
Optionally, in step S1), the step of attaching corresponding fault identification result classification labels to all the status signal samples specifically means that five types of fault identification results, namely "1", "2", "3", "4" and "5", are respectively attached to the identification results of the normal status, the fiber link fault, the polarizer fault, the fiber transceiver fault and the power failure, and the five types of labels "1", "2", "3", "4" and "5" output by the machine learning classification model respectively refer to the identification results of the normal status, the fiber link fault, the polarizer fault, the fiber transceiver fault and the power failure.
Optionally, the step S3) of training and finding the optimal parameter of the SVMi refers to finding the optimal value of the kernel function parameter gamma and the penalty parameter C by using a longicorn whisker search algorithm, and the specific steps include:
s3.1) representing left beard sitting by xlThe index xr represents the coordinates of the right whisker, x represents the coordinates of the centroid, d0 represents the distance between the two whiskers, and the orientation between the two whiskers is represented and normalized by the random vector dir ═ rands (n, 1):
expressing the coordinates as
The random function rands (k,1) is used for taking a random number rands between 1 and the iteration number n;
s3.2) establishing an optimizing function by using a model corresponding to the initialized kernel function parameter gamma and the penalty parameter C, and enabling the function H to be equal to<gamma,C>And calculating the fitness values Hleft and Hright of the left and right whiskers, Hleft ═ H (xl), Hright ═ H (xr); fitness value H (x) based on position at time tt) And a fitness value H (x) of the position at time t-1t-1) Calculating the probability p according to the following formula;
if H (x)t) Greater than H (x)t-1) If yes, the value of the acceptance probability p is 1, and the step S3.4) is skipped to update the position; if H (x)t) Less than or equal to H (x)t-1) If yes, judging that the value of the probability p is smaller than 1, further judging whether the probability p is smaller than the random number rands, and if yes, skipping to execute the step S3.4) to update the position; otherwise, the position is not updated and the step S3.5 is directly skipped to execute);
s3.4) updating the position according to the adaptability values of the left and right whiskers of the longicorn at the position: if the value of the left whisker is smaller than the value of the right whisker (Hleft < Hright), in order to find the maximum value of the function H, a search is required to be performed by the longicorn in the direction of the right whisker, the distance step is traveled, and the updated position is
If the left whisker value is greater than the right whisker value (Hleft > Hright), then to find the maximum value of the function H, a search is required of the longicorn in the direction of the left whisker, the distance step is traveled,update the position to
Wherein x istAs the current position, xt+1Sign () is a sign function, step, for the shifted position of the longicorntH (xl) is the current step length, and H (xr) is the values of the left and right whiskers;
s3.5) judging whether the nth optimization function value H is satisfiednAnd the optimization function value H of the (n-1) th iterationn-1The difference between them is less than the preset threshold (the specific determination condition in this embodiment is H)n-Hn-1≤Hn1%), if satisfied, stopping iteration to obtain the optimal values of the kernel function parameter gamma and the penalty parameter C; if not, judging whether the iteration number n is equal to a preset upper limit value or not, if not, adding 1 to the iteration number n, and directly jumping to execute the step S3.1) to continuously search the optimal values of the kernel function parameter gamma and the penalty parameter C; otherwise, judging that the optimization fails, ending and exiting.
Optionally, step S3.4) is followed by a step of improving the step size using the following formula:
stept+1＝ε×stept
in the above formula, stept+1Step for improved step sizetFor the current step size, ε is the improvement factor.
In addition, the invention also provides a fault identification system for the pure optical fiber electronic current transformer, which comprises the following steps:
the signal input program unit is used for acquiring a state signal x (t) of the pure optical fiber type current transformer;
the signal decomposition program unit is used for carrying out modal decomposition on the state signal x (t) to obtain a plurality of intrinsic modal function components;
a principal component analysis program unit for performing principal component analysis on the plurality of eigenmode function components by using a principal component analysis method to extract a feature vector for fault identification;
and the fault identification program unit is used for inputting the characteristic vector into a pre-trained machine learning classification model to obtain a fault identification result of the pure optical fiber current transformer, and the machine learning classification model is pre-trained to establish mapping between the characteristic vector and the fault identification result.
In addition, the invention also provides a system for identifying the fault of the pure optical fiber electronic current transformer, which comprises a computer device, wherein the computer device is programmed or configured to execute the steps of the method for identifying the fault of the pure optical fiber electronic current transformer, or a computer program which is programmed or configured to execute the method for identifying the fault of the pure optical fiber electronic current transformer is stored in a memory of the computer device.
In addition, the present invention also provides a computer readable storage medium having stored thereon a computer program programmed or configured to execute the method for fault identification of a pure fiber electronic current transformer.
Compared with the prior art, the invention has the following advantages: according to the method, modal decomposition is carried out on state signals x (t) of the pure optical fiber type current transformer to obtain a plurality of intrinsic mode function components, Principal Component Analysis (PCA) is utilized to carry out principal component analysis on the plurality of intrinsic mode function components to extract a feature vector for fault identification, the feature vector is input into a machine learning classification model trained in advance to obtain a fault identification result of the pure optical fiber type current transformer, the machine learning classification model is trained in advance to establish mapping between the feature vector and the fault identification result, fault information of a sensor can be extracted more effectively, and follow-up fault identification is facilitated. Human participation can be reduced in the whole process, the analysis accuracy is improved, meanwhile, the principle is simpler and faster based on the fault identification of the SVM, and the identification speed and the accuracy of the pure optical fiber current sensor are high.
Drawings
FIG. 1 is a schematic diagram of a basic flow of a method according to an embodiment of the present invention.
Fig. 2 is a basic flow chart of the EMD algorithm according to the embodiment of the present invention.
Fig. 3 is a schematic diagram of the structural principle of the machine learning classification model in the embodiment of the present invention.
Fig. 4 is a schematic flow chart of the parameter optimization by using the longicorn whisker search algorithm in the embodiment of the present invention.
Detailed Description
As shown in fig. 1, the implementation steps of the method for identifying a fault of a pure optical fiber electronic current transformer in the embodiment include:
1) acquiring a state signal x (t) of the pure optical fiber type current transformer;
2) carrying out modal decomposition on the state signal x (t) to obtain a plurality of intrinsic mode function components;
3) performing principal component analysis on the plurality of eigenmode function components by using a Principal Component Analysis (PCA) method to extract a feature vector for fault identification;
4) and inputting the characteristic vector into a pre-trained machine learning classification model to obtain a fault identification result of the pure optical fiber type current transformer, wherein the machine learning classification model is pre-trained to establish mapping between the characteristic vector and the fault identification result.
As shown in fig. 2, in step 2) of this embodiment, mode decomposition (EMD decomposition) is performed on the state signal x (t) to obtain a plurality of eigen-mode function components (IMF components), which is improved EMD decomposition, and in addition, other existing EMD decomposition methods may be adopted according to needs. The detailed steps of step 2) to improve EMD decomposition in this example include:
2.1) determining all extreme points of the state signal x (t), including a maximum point and a minimum point;
2.2) interpolating all extreme points by adopting a spline function, solving upper envelope lines and lower envelope lines, and connecting the mean values m (t) of the upper envelope lines and the lower envelope lines in sequence to obtain a mean value line; the mean value m (t) can be expressed as:
in the above formula, xmax(t) represents a maximum point, xmin(t) represents a minimum value point;
2.3) finding extreme values at the left end and the right end according to the waveform characteristics of the state signals x (t), and eliminating an end effect by utilizing mirror continuation; the specific point of eliminating the endpoint effect by using mirror continuation is as follows: the mirror is placed at an extreme value position with symmetrical signals by utilizing the characteristic of mirror symmetry mapping, then the signals in the mirror are mapped outwards to obtain a periodic signal with the length being twice as long as the data length of the signals in the mirror, and the periodic signal is connected end to form a closed curve, so that the end point effect is eliminated.
2.4) subtracting the state signal x (t) from the mean value m (t) of the mean line to obtain a difference value h (t), namely: h (t) x (t) -m (t);
2.5) checking whether the difference h (t) meets the condition of the intrinsic mode function, and if not, taking the difference h (t) as a newly input state signal x (t) to jump to execute the step 2.1); otherwise, calculating the residual r (t) between the state signal x (t) and the difference h (t), namely: r (t) x (t) -h (t); the eigenmode function condition may also be referred to as an IMF condition for short, which is specifically defined as follows: (1) in the whole time range of the function, the number of local extreme points and zero-crossing points must be equal, or at most, the difference is one; (2) at any point in time, the envelope of the local maxima (upper envelope) and the envelope of the local minima (lower envelope) must be, on average, zero.
2.6) judging whether the residual r (t) meets the termination condition, if the residual r (t) is a non-monotonic function, judging that the residual r (t) does not meet the termination condition, and using the residual r (t) as a newly input state signal x (t) to jump to execute the step 2.1); otherwise, judging that the termination condition is met, ending the modal decomposition process, outputting the intrinsic mode function as the intrinsic mode function component obtained by the iteration of the current round, and skipping to execute the step 3).
In this embodiment, step 3) is used to extract feature vectors of the obtained IMF by using PCA principal component analysis. The detailed steps of the step 3) comprise:
3.1) taking a plurality of intrinsic mode function components as a data sample set, and calculating a mean vector of samples in the data sample set, wherein the mean vector can be expressed as the following function expression:
in the above formula, μ is the sample data set XIMFMean vector of medium samples, n is the number of samples, XIMFIs a sampleA data set.
3.2) averaging each sample, centralizing the sample data, and expressing the sample data as the following functional expression:
XIMF＝XIMF-μ
in the above formula, XIMFFor data matrices, X, obtained by centralization of sample dataIMFIs a sample data set, mu is a sample data set XIMFMean vector of the medium samples.
3.3) constructing a covariance matrix of a data matrix obtained by centralizing the sample data, wherein the covariance matrix can be expressed as the following functional expression:
in the above formula, V is covariance matrix, XIMFAnd centralizing the obtained data matrix for the sample data.
3.4) carrying out feature decomposition on the covariance matrix to obtain an eigenvalue and a corresponding eigenvector, namely: obtaining an eigenvalue lambda of the covariance matrix ViAnd corresponding feature vector omegai；
And 3.5) screening out the feature vectors with the contribution rate larger than the set contribution rate threshold value as the feature vectors for fault identification. The screening in this embodiment specifically refers to d (d) before selection<Number of original components) to make the cumulative variance contribution rate meet certain requirements (the contribution rate is more than 80%). Before screening, the eigenvalues lambda need to be sorted in descending orderiAccording to the contribution rate, the first d characteristic values are taken, namely lambda, diag [ lambda ]1、λ2、…、λi]And corresponding feature vector omegad＝[ω1、ω2、…、ωi]As a basis for the subspace, then the d principal components to be extracted are
Reconstructing raw data X from the extracted principal components2ω F + μ as a feature vector for fault identification.
In this embodiment, the failure identification result includes a normal state, a failure of the optical fiber link, a failure of the polarizer, a failure of the optical fiber transceiver, and a failure of the power supply.
As shown in fig. 3, the machine learning classification model in step 4) of this embodiment is an SVM identification classifier composed of support vector machines SVM 1-SVMk, the number of support vector machines SVM 1-SVMk is k in total and corresponds to k fault identification results one by one, the support vector machines SVM 1-SVMk form a hierarchical relationship, and any one level of support vector machine SVMiThe output result of (1) comprises corresponding fault identification result and other faults, and the input of the output result is input into the next-stage support vector machine SVM under the condition of outputting other faultsi+1. The training situation of each SVMi in this embodiment is as follows:
in this embodiment, step 4) is preceded by a step of training a machine learning classification model, and the detailed steps include:
s1) respectively acquiring corresponding state signal samples for various fault identification results of the pure optical fiber current type mutual inductor, wherein the number of the state signal samples of each fault identification result is the same; adding corresponding fault identification result classification labels to all the state signal samples to form training samples with labels and test samples;
s2) establishing a machine learning classification model;
s3) for each support vector machine SVMi in the machine learning classification model: inputting the training sample into a Support Vector Machine (SVMi) for training to find the optimal parameter of the SVMi; inputting the test sample into the SVMi to be tested to obtain a corresponding fault identification result, and judging that the training of the SVMi is finished if the identification accuracy of the fault identification result of the test sample reaches a preset requirement; otherwise, continuing to train the support vector machine SVMi until the identification accuracy of the fault identification result of the test sample reaches the preset requirement.
In this embodiment, in step S1), the method adds corresponding fault identification result classification labels to all the status signal samples, specifically, adds five labels "1", "2", "3", "4" and "5" to the identification results of five faults, i.e., normal status, fiber link fault, polarizer fault, fiber transceiver fault, and power failure, respectively, and outputs five labels "1", "2", "3", "4" and "5" by using a machine learning classification model to respectively refer to the identification results of five faults, i.e., normal status, fiber link fault, polarizer fault, fiber transceiver fault, and power failure. In the multiple groups of samples in this embodiment, the training sample setting labels "1 to 5" correspond to each other: no. 1-15 is normal state, No. 16-30 is optical fiber link failure, No. 31-45 is polarizer failure, No. 46-60 is optical fiber transceiver failure, and No. 61-75 is power failure.
In this example, a total of 150 data samples, and in 5 cases, 30 cases in each case, in the above table, feature vectors were calculated. The data samples were divided into two groups: training and test samples, each containing 15 sets of normal status, 15 sets of fiber link failures, 15 sets of polarizer failures, 15 sets of fiber transceiver failures, and 15 sets of power failures, are labeled "1", "2", "3", "4", "5". The kernel function of the support vector machine SVMi adopts No. 2: RBF function exp (-gamma | (x)i-xj)2) And the penalty parameter C, and the soft boundary is C-SVM.
As shown in fig. 4, the step S3) of the embodiment of the present invention to train and find the optimal parameter of the SVMi is to specifically use a skyhook beard search algorithm (BAS) to find the optimal value of the kernel function parameter gamma and the penalty parameter C, and the specific steps include:
s3.1) the left whisker coordinate is denoted by xl, the right whisker coordinate is denoted by xr, the centroid coordinate is denoted by d0, and the distance between the two whiskers is denoted by using a random vector dir ═ rands (n,1) to denote the orientation between the two whiskers and normalize it:
expressing the coordinates as
The random function rands (n,1) is used for taking a random number rands between 1 and the iteration number n;
s3.2) establishing an optimizing function by using a model corresponding to the initialized kernel function parameter gamma and the penalty parameter C, and enabling the function H to be equal to<gamma,C>And calculating the fitness values Hleft and Hright of the left and right whiskers, Hleft ═ H (xl), Hright ═ H (xr); fitness value H (x) based on position at time tt) And a fitness value H (x) of the position at time t-1t-1) Calculating the probability p according to the following formula;
if H (x)t) Greater than H (x)t-1) If yes, the value of the acceptance probability p is 1, and the step S3.4) is skipped to update the position; if H (x)t) Less than or equal to H (x)t-1) If yes, judging that the value of the probability p is smaller than 1, further judging whether the probability p is smaller than the random number rands, and if yes, skipping to execute the step S3.4) to update the position; otherwise, the position is not updated and the step S3.5 is directly skipped to execute);
s3.4) updating the position according to the adaptability values of the left and right whiskers of the longicorn at the position: if the value of the left whisker is smaller than the value of the right whisker (Hleft < Hright), in order to find the maximum value of the function H, a search is required to be performed by the longicorn in the direction of the right whisker, the distance step is traveled, and the updated position is
If the left whisker value is greater than the right whisker value (Hleft > Hright), then to find the maximum value of the function H, a search is made by the longicorn in the direction of the left whisker, the distance step is traveled, and the location is updated to
Wherein x istAs the current position, xt+1Sign () is a sign function, step, for the shifted position of the longicorntH (xl) is the current step length, and H (xr) is the values of the left and right whiskers;
s3.5) judging whether the nth optimization function value H is satisfiednAnd the optimization function value H of the (n-1) th iterationn-1The difference between them is less than the preset threshold (specifically determined in this embodiment)Off condition is Hn-Hn-1≤H n1%), if satisfied, stopping iteration to obtain the optimal values of the kernel function parameter gamma and the penalty parameter C; if not, judging whether the iteration number n is equal to a preset upper limit value or not, if not, adding 1 to the iteration number n, and directly jumping to execute the step S3.1) to continuously search the optimal values of the kernel function parameter gamma and the penalty parameter C; otherwise, judging that the optimization fails, ending and exiting.
And obtaining the required iterative process according to the steps, and finally obtaining the optimization results of the 2 parameter variables.
In order to improve the accuracy of algorithm optimization, step S3.4) in this embodiment further includes a step of improving the step size by using the following formula:
stept+1＝ε×stept
in the above formula, stept+1Step for improved step sizetFor the current step size, ε is the improvement factor, which is between 0 and 1 and is close to 1, and may be equal to 0.95.
In summary, the present embodiment discloses a method for identifying a fault of a pure fiber electronic current transformer based on an improved EMD and a BAS-SVM, in which the method performs empirical mode decomposition on a state signal of the pure fiber current transformer by using the EMD, performs mirror extension on the state signal to improve an end effect, and obtains a set of IMF components of an intrinsic mode function. And carrying out Principal Component Analysis (PCA) on the extracted IMF components, and extracting a feature vector for fault identification according to the contribution rate of each component. And dividing the extracted feature vectors into two groups as samples, wherein the two groups of samples are equal in number and are respectively used as training samples and testing samples, and training out the fault identifier of the SVM pure fiber current type mutual inductor after finding out the optimal parameters of the SVM by utilizing a Tianniu whisker search algorithm, thereby achieving the effect of identifying faults. The method of the embodiment fully reduces human participation, is simple and convenient in principle, improves the identification precision, and has important significance for the development of the future power grid.
In addition, this embodiment still provides a pure fiber electronic current transformer fault identification system, includes:
the signal input program unit is used for acquiring a state signal x (t) of the pure optical fiber type current transformer;
the signal decomposition program unit is used for carrying out modal decomposition on the state signal x (t) to obtain a plurality of intrinsic modal function components;
a principal component analysis program unit for performing principal component analysis on the plurality of eigenmode function components by using a principal component analysis method to extract a feature vector for fault identification;
and the fault identification program unit is used for inputting the characteristic vector into a pre-trained machine learning classification model to obtain a fault identification result of the pure optical fiber current transformer, and the machine learning classification model is pre-trained to establish mapping between the characteristic vector and the fault identification result.
In addition, the present embodiment further provides a system for identifying a fault of a pure fiber electronic current transformer, which includes a computer device programmed or configured to execute the steps of the method for identifying a fault of a pure fiber electronic current transformer, or a memory of the computer device having a computer program stored thereon programmed or configured to execute the method for identifying a fault of a pure fiber electronic current transformer.
In addition, the present embodiment also provides a computer readable storage medium, which stores thereon a computer program programmed or configured to execute the foregoing method for identifying a fault of a pure fiber electronic current transformer.
As will be appreciated by one skilled in the art, embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein. The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks. These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks. These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
The above description is only a preferred embodiment of the present invention, and the protection scope of the present invention is not limited to the above embodiments, and all technical solutions belonging to the idea of the present invention belong to the protection scope of the present invention. It should be noted that modifications and embellishments within the scope of the invention may occur to those skilled in the art without departing from the principle of the invention, and are considered to be within the scope of the invention.
Claims (10)
1. A fault identification method for a pure fiber electronic current transformer is characterized by comprising the following implementation steps:
1) acquiring a state signal x (t) of the pure optical fiber type current transformer;
2) carrying out modal decomposition on the state signal x (t) to obtain a plurality of intrinsic mode function components;
3) performing principal component analysis on the plurality of eigenmode function components by using a principal component analysis method to extract a feature vector for fault identification;
4) and inputting the characteristic vector into a pre-trained machine learning classification model to obtain a fault identification result of the pure optical fiber type current transformer, wherein the machine learning classification model is pre-trained to establish mapping between the characteristic vector and the fault identification result.
2. The method for identifying the fault of the pure fiber electronic current transformer according to claim 1, wherein the detailed steps of the step 2) comprise:
2.1) determining all extreme points of the state signal x (t), including a maximum point and a minimum point;
2.2) interpolating all extreme points by adopting a spline function, solving upper envelope lines and lower envelope lines, and connecting the mean values m (t) of the upper envelope lines and the lower envelope lines in sequence to obtain a mean value line;
2.3) finding extreme values at the left end and the right end according to the waveform characteristics of the state signals x (t), and eliminating an end effect by utilizing mirror continuation;
2.4) subtracting the state signal x (t) from the mean value m (t) of the mean value line to obtain a difference value h (t);
2.5) checking whether the difference h (t) meets the condition of the intrinsic mode function, and if not, taking the difference h (t) as a newly input state signal x (t) to jump to execute the step 2.1); otherwise, calculating residual errors r (t) among the state signals x (t) and the difference values h (t);
2.6) judging whether the residual r (t) meets the termination condition, if the residual r (t) is a non-monotonic function, judging that the residual r (t) does not meet the termination condition, and using the residual r (t) as a newly input state signal x (t) to jump to execute the step 2.1); otherwise, judging that the termination condition is met, ending the modal decomposition process, outputting the intrinsic mode function as the intrinsic mode function component obtained by the iteration of the current round, and skipping to execute the step 3).
3. The method for identifying the fault of the pure fiber electronic current transformer according to claim 1, wherein the detailed steps of the step 3) comprise:
3.1) taking the plurality of intrinsic mode function components as a data sample set, and calculating a mean vector of samples in the sample data set;
3.2) removing the mean value of each sample, and centralizing the sample data;
3.3) constructing a covariance matrix of a data matrix obtained by centralizing sample data;
3.4) carrying out characteristic decomposition on the covariance matrix to obtain an eigenvalue and a corresponding eigenvector;
and 3.5) screening out the feature vectors with the contribution rate larger than the set contribution rate threshold value as the feature vectors for fault identification.
4. The method for identifying faults of pure fiber electronic current transformers according to claim 1, wherein the machine learning classification model in the step 4) is an SVM identification classifier composed of support vector machines SVM 1-SVMk, the number of support vector machines SVM 1-SVMk is k in total and respectively corresponds to k fault identification results one by one, the support vector machines SVM 1-SVMk form a hierarchical relationship, and any one-level support vector machine SVM is selectediThe output result of (1) comprises corresponding fault identification result and other faults, and the input of the output result is input into the next-stage support vector machine SVM under the condition of outputting other faultsi+1(ii) a Step 4) is also preceded by a step of training a machine learning classification model, and the detailed steps comprise:
s1) respectively acquiring corresponding state signal samples for various fault identification results of the pure optical fiber current type mutual inductor, wherein the number of the state signal samples of each fault identification result is the same; adding corresponding fault identification result classification labels to all the state signal samples to form training samples with labels and test samples;
s2) establishing a machine learning classification model;
s3) for each support vector machine SVMi in the machine learning classification model: inputting the training sample into a Support Vector Machine (SVMi) for training to find the optimal parameter of the SVMi; inputting the test sample into the SVMi to be tested to obtain a corresponding fault identification result, and judging that the training of the SVMi is finished if the identification accuracy of the fault identification result of the test sample reaches a preset requirement; otherwise, continuing to train the support vector machine SVMi until the identification accuracy of the fault identification result of the test sample reaches the preset requirement.
5. The method for identifying the faults of the pure fiber electronic current transformer according to claim 4, wherein in step S1), the corresponding fault identification result classification labels are attached to all the status signal samples, specifically, five types of labels "1", "2", "3", "4" and "5" are respectively attached to the identification results of the normal status, the fault of the fiber link, the fault of the polarizer, the fault of the fiber transceiver and the fault of the power supply, and the five types of labels "1", "2", "3", "4" and "5" output by the machine learning classification model respectively refer to five types of fault identification results of the normal status, the fault of the fiber link, the fault of the polarizer, the fault of the fiber transceiver and the fault of the power supply.
6. The method for identifying the fault of the pure fiber electronic current transformer according to claim 4, wherein the step S3) of training and searching for the optimal parameter of the SVMi is to search for the optimal value of the kernel function parameter gamma and the penalty parameter C by using a Tianniu whisker search algorithm, and the specific steps include:
s3.1) the left whisker coordinate is denoted by xl, the right whisker coordinate is denoted by xr, the centroid coordinate is denoted by d0, the distance between the two whiskers is denoted by the random vector dir ═ rands (n,1) for the orientation between the two whiskers and normalized:
expressing the coordinates as
The random function rands (n,1) is used for taking a random number rands between 1 and the iteration number n;
s3.2) establishing an optimization function by using a model corresponding to the initialized kernel function parameter gamma and the penalty parameter CNumber, let function H equal<gamma,C>And calculating the fitness values Hleft and Hright of the left and right whiskers, Hleft ═ H (xl), Hright ═ H (xr); fitness value H (x) based on position at time tt) And a fitness value H (x) of the position at time t-1t-1) Calculating the probability p according to the following formula;
if H (x)t) Greater than H (x)t-1) If yes, the value of the acceptance probability p is 1, and the step S3.4) is skipped to update the position; if H (x)t) Less than or equal to H (x)t-1) If yes, judging that the value of the probability p is smaller than 1, further judging whether the probability p is smaller than the random number rands, and if yes, skipping to execute the step S3.4) to update the position; otherwise, the position is not updated and the step S3.5 is directly skipped to execute);
s3.4) updating the position according to the adaptability values of the left and right whiskers of the longicorn at the position: if the value of the left whisker is smaller than the value of the right whisker (Hleft < Hright), in order to find the maximum value of the function H, a search is required to be performed by the longicorn in the direction of the right whisker, the distance step is traveled, and the updated position is
If the left whisker value is greater than the right whisker value (Hleft > Hright), then to find the maximum value of the function H, a search is made by the longicorn in the direction of the left whisker, the distance step is traveled, and the location is updated to
Wherein x istAs the current position, xt+1Sign () is a sign function, step, for the shifted position of the longicorntH (xl) is the current step length, and H (xr) is the values of the left and right whiskers;
s3.5) judging whether the nth optimization function value H is satisfiednAnd the optimization function value H of the (n-1) th iterationn-1The difference between them is less than the preset threshold (the specific determination condition in this embodiment is H)n-Hn-1≤Hn1%) if satisfied, stopIteration is carried out to obtain the optimal values of the kernel function parameter gamma and the penalty parameter C; if not, judging whether the iteration number n is equal to a preset upper limit value or not, if not, adding 1 to the iteration number n, and directly jumping to execute the step S3.1) to continuously search the optimal values of the kernel function parameter gamma and the penalty parameter C; otherwise, judging that the optimization fails, ending and exiting.
7. The method for identifying faults of the pure fiber electronic current transformer according to claim 6, wherein the step S3.4) is followed by a step of improving the step size by adopting the following formula:
stept+1＝ε×stept
in the above formula, stept+1Step for improved step sizetFor the current step size, ε is the improvement factor.
8. The utility model provides a pure fiber electronic current transformer fault identification system which characterized in that includes:
the signal input program unit is used for acquiring a state signal x (t) of the pure optical fiber type current transformer;
the signal decomposition program unit is used for carrying out modal decomposition on the state signal x (t) to obtain a plurality of intrinsic modal function components;
a principal component analysis program unit for performing principal component analysis on the plurality of eigenmode function components by using a principal component analysis method to extract a feature vector for fault identification;
and the fault identification program unit is used for inputting the characteristic vector into a pre-trained machine learning classification model to obtain a fault identification result of the pure optical fiber current transformer, and the machine learning classification model is pre-trained to establish mapping between the characteristic vector and the fault identification result.
9. A fault identification system for a pure fiber electronic current transformer, comprising a computer device, wherein the computer device is programmed or configured to perform the steps of the method for identifying faults of a pure fiber electronic current transformer according to any one of claims 1 to 7, or a memory of the computer device has a computer program programmed or configured to perform the method for identifying faults of a pure fiber electronic current transformer according to any one of claims 1 to 7 stored thereon.
10. A computer-readable storage medium having stored thereon a computer program programmed or configured to perform the method for fault identification of a pure fiber electronic current transformer according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911100607.6A CN111008648B (en) | 2019-11-12 | 2019-11-12 | Fault identification method, system and medium for pure optical fiber electronic current transformer |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911100607.6A CN111008648B (en) | 2019-11-12 | 2019-11-12 | Fault identification method, system and medium for pure optical fiber electronic current transformer |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111008648A true CN111008648A (en) | 2020-04-14 |
| CN111008648B CN111008648B (en) | 2023-10-27 |
Family
ID=70111926
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911100607.6A Active CN111008648B (en) | 2019-11-12 | 2019-11-12 | Fault identification method, system and medium for pure optical fiber electronic current transformer |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111008648B (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111814954A (en) * | 2020-06-19 | 2020-10-23 | 武汉光迅科技股份有限公司 | Optical fiber quality analysis method and device, electronic equipment and storage medium |
| CN112414713A (en) * | 2020-11-04 | 2021-02-26 | 吉电(滁州)章广风力发电有限公司 | Rolling bearing fault detection method based on measured signals |
| CN112651498A (en) * | 2020-09-22 | 2021-04-13 | 杭州杭越传感科技有限公司 | Method and device for improving temperature stability of self-learning current sensor |
| CN112910567A (en) * | 2021-01-25 | 2021-06-04 | 北京邮电大学 | Interception classification monitoring method based on recurrent neural network and related equipment |
| CN115034337A (en) * | 2022-08-10 | 2022-09-09 | 江西科骏实业有限公司 | Method and device for identifying state of traction motor in rail transit vehicle and medium |
| CN115630847A (en) * | 2022-12-07 | 2023-01-20 | 四川省华盾防务科技股份有限公司 | Transceiving assembly detection method and system based on data prediction and storage medium |
| WO2023115875A1 (en) * | 2021-12-24 | 2023-06-29 | 深圳前海微众银行股份有限公司 | Hardware device maintenance method and apparatus, and electronic device |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102937668A (en) * | 2012-11-08 | 2013-02-20 | 电子科技大学 | Electric system low-frequency oscillation detection method |
| CN106092574A (en) * | 2016-05-30 | 2016-11-09 | 西安工业大学 | The Method for Bearing Fault Diagnosis selected with sensitive features is decomposed based on improving EMD |
| CN107091737A (en) * | 2017-06-06 | 2017-08-25 | 太原理工大学 | A kind of study of typical faults of rotor systems combining diagnostic method based on current signal |
| CN107505133A (en) * | 2017-08-10 | 2017-12-22 | 滁州学院 | The probability intelligent diagnosing method of rolling bearing fault based on adaptive M RVM |
| CN107908863A (en) * | 2017-11-14 | 2018-04-13 | 哈尔滨理工大学 | A kind of hydraulic turbine operating condition decision method based on EMD theories with HHT conversion |
| US20180357542A1 (en) * | 2018-06-08 | 2018-12-13 | University Of Electronic Science And Technology Of China | 1D-CNN-Based Distributed Optical Fiber Sensing Signal Feature Learning and Classification Method |
| CN109026647A (en) * | 2018-08-14 | 2018-12-18 | 东华大学 | A kind of the Hydraulic pump fault detection method and system of BASA optimization GRBF-SVM |
| CN109255490A (en) * | 2018-09-28 | 2019-01-22 | 西安建筑科技大学 | Corrosion rate prediction technique outside a kind of buried pipeline based on KPCA-BAS-GRNN |
| CN109307862A (en) * | 2018-07-05 | 2019-02-05 | 西安电子科技大学 | A kind of target radiation source individual discrimination method |
| CN110044623A (en) * | 2019-04-15 | 2019-07-23 | 中国人民解放军海军工程大学 | The rolling bearing fault intelligent identification Method of empirical mode decomposition residual signal feature |
| CN110146294A (en) * | 2019-04-23 | 2019-08-20 | 莆田学院 | A kind of wind-driven generator vibrating failure diagnosis method and storage medium |
-
2019
- 2019-11-12 CN CN201911100607.6A patent/CN111008648B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102937668A (en) * | 2012-11-08 | 2013-02-20 | 电子科技大学 | Electric system low-frequency oscillation detection method |
| CN106092574A (en) * | 2016-05-30 | 2016-11-09 | 西安工业大学 | The Method for Bearing Fault Diagnosis selected with sensitive features is decomposed based on improving EMD |
| CN107091737A (en) * | 2017-06-06 | 2017-08-25 | 太原理工大学 | A kind of study of typical faults of rotor systems combining diagnostic method based on current signal |
| CN107505133A (en) * | 2017-08-10 | 2017-12-22 | 滁州学院 | The probability intelligent diagnosing method of rolling bearing fault based on adaptive M RVM |
| CN107908863A (en) * | 2017-11-14 | 2018-04-13 | 哈尔滨理工大学 | A kind of hydraulic turbine operating condition decision method based on EMD theories with HHT conversion |
| US20180357542A1 (en) * | 2018-06-08 | 2018-12-13 | University Of Electronic Science And Technology Of China | 1D-CNN-Based Distributed Optical Fiber Sensing Signal Feature Learning and Classification Method |
| CN109307862A (en) * | 2018-07-05 | 2019-02-05 | 西安电子科技大学 | A kind of target radiation source individual discrimination method |
| CN109026647A (en) * | 2018-08-14 | 2018-12-18 | 东华大学 | A kind of the Hydraulic pump fault detection method and system of BASA optimization GRBF-SVM |
| CN109255490A (en) * | 2018-09-28 | 2019-01-22 | 西安建筑科技大学 | Corrosion rate prediction technique outside a kind of buried pipeline based on KPCA-BAS-GRNN |
| CN110044623A (en) * | 2019-04-15 | 2019-07-23 | 中国人民解放军海军工程大学 | The rolling bearing fault intelligent identification Method of empirical mode decomposition residual signal feature |
| CN110146294A (en) * | 2019-04-23 | 2019-08-20 | 莆田学院 | A kind of wind-driven generator vibrating failure diagnosis method and storage medium |
Non-Patent Citations (1)
| Title |
|---|
| 卢光辉等: "基于改进天牛须搜索算法的分布式电源选址定容" * |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111814954A (en) * | 2020-06-19 | 2020-10-23 | 武汉光迅科技股份有限公司 | Optical fiber quality analysis method and device, electronic equipment and storage medium |
| CN111814954B (en) * | 2020-06-19 | 2023-09-08 | 武汉光迅科技股份有限公司 | Optical fiber quality analysis method and device, electronic equipment and storage medium |
| CN112651498A (en) * | 2020-09-22 | 2021-04-13 | 杭州杭越传感科技有限公司 | Method and device for improving temperature stability of self-learning current sensor |
| CN112414713A (en) * | 2020-11-04 | 2021-02-26 | 吉电(滁州)章广风力发电有限公司 | Rolling bearing fault detection method based on measured signals |
| CN112910567A (en) * | 2021-01-25 | 2021-06-04 | 北京邮电大学 | Interception classification monitoring method based on recurrent neural network and related equipment |
| CN112910567B (en) * | 2021-01-25 | 2022-07-01 | 北京邮电大学 | Interception classification monitoring method based on recurrent neural network and related equipment |
| WO2023115875A1 (en) * | 2021-12-24 | 2023-06-29 | 深圳前海微众银行股份有限公司 | Hardware device maintenance method and apparatus, and electronic device |
| CN115034337A (en) * | 2022-08-10 | 2022-09-09 | 江西科骏实业有限公司 | Method and device for identifying state of traction motor in rail transit vehicle and medium |
| CN115034337B (en) * | 2022-08-10 | 2022-11-01 | 江西科骏实业有限公司 | Method, device and medium for identifying state of traction motor in rail transit vehicle |
| CN115630847A (en) * | 2022-12-07 | 2023-01-20 | 四川省华盾防务科技股份有限公司 | Transceiving assembly detection method and system based on data prediction and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111008648B (en) | 2023-10-27 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111008648A (en) | Fault identification method, system and medium for pure optical fiber electronic current transformer | |
| CN109541961A (en) | For the discrete state event-driven simulation method of power electronics hybrid system emulation | |
| KR101889451B1 (en) | Weight benefit evaluator for training data | |
| CN114841257B (en) | Small sample target detection method based on self-supervision comparison constraint | |
| CN108051660A (en) | A kind of transformer fault combined diagnosis method for establishing model and diagnostic method | |
| CN107480386B (en) | Test excitation optimization method based on response aliasing measurement and genetic algorithm | |
| CN111079780A (en) | Training method of space map convolution network, electronic device and storage medium | |
| CN110044525A (en) | A kind of flexible resistive dot matrix pressure detecting system, method and apparatus | |
| CN110489521A (en) | Text categories detection method, device, electronic equipment and computer-readable medium | |
| US9177258B2 (en) | Implementing a learning algorithm and estimation function utilizing a randomly selected prototype function with randomly set parameters | |
| US20220122343A1 (en) | Image feature extraction and network training method, apparatus, and device | |
| CN110119505A (en) | Term vector generation method, device and equipment | |
| CN108629345A (en) | Dimensional images feature matching method and device | |
| CN110309922A (en) | A kind of network model training method and device | |
| CN104463916A (en) | Eye movement fixation point measurement method based on random walk | |
| CN112420125A (en) | Molecular attribute prediction method and device, intelligent equipment and terminal | |
| CN116295752A (en) | Test strength control method and system for SMT (surface mounting technology) feeding equipment | |
| CN107742516A (en) | Intelligent identification Method, robot and computer-readable recording medium | |
| CN110188798A (en) | A kind of object classification method and model training method and device | |
| CN109492401A (en) | A kind of content vector risk checking method, device, equipment and medium | |
| CN105335763A (en) | Fabric defect classification method based on improved extreme learning machine | |
| CN112100509A (en) | Information recommendation method, device, server and storage medium | |
| CN108681490A (en) | For the vector processing method, device and equipment of RPC information | |
| CN112040401B (en) | Indoor positioning method and device, electronic equipment and storage medium | |
| CN115619496A (en) | Accurate pushing method for E-commerce products and server |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |