CN110059357A - A kind of intelligent electric energy meter failure modes detection method and system based on autoencoder network - Google Patents
A kind of intelligent electric energy meter failure modes detection method and system based on autoencoder network Download PDFInfo
- Publication number
- CN110059357A CN110059357A CN201910208116.7A CN201910208116A CN110059357A CN 110059357 A CN110059357 A CN 110059357A CN 201910208116 A CN201910208116 A CN 201910208116A CN 110059357 A CN110059357 A CN 110059357A
- Authority
- CN
- China
- Prior art keywords
- autoencoder network
- network model
- electric energy
- energy meter
- intelligent electric
- 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.)
- Pending
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 70
- 238000012549 training Methods 0.000 claims abstract description 46
- 238000012360 testing method Methods 0.000 claims abstract description 18
- 238000009434 installation Methods 0.000 claims abstract description 8
- 238000000034 method Methods 0.000 claims description 43
- 230000003252 repetitive effect Effects 0.000 claims description 11
- 238000004458 analytical method Methods 0.000 claims description 7
- 238000005457 optimization Methods 0.000 claims description 7
- 230000007423 decrease Effects 0.000 claims description 6
- 238000007781 pre-processing Methods 0.000 claims description 3
- 230000005611 electricity Effects 0.000 claims description 2
- 230000009467 reduction Effects 0.000 abstract description 20
- 238000000605 extraction Methods 0.000 abstract description 4
- 238000007796 conventional method Methods 0.000 abstract description 3
- 230000036039 immunity Effects 0.000 abstract description 3
- 230000001737 promoting effect Effects 0.000 abstract description 2
- 230000006870 function Effects 0.000 description 21
- 238000010586 diagram Methods 0.000 description 7
- 230000004913 activation Effects 0.000 description 5
- 238000004891 communication Methods 0.000 description 4
- 230000003247 decreasing effect Effects 0.000 description 4
- 238000007689 inspection Methods 0.000 description 4
- 230000005856 abnormality Effects 0.000 description 3
- 230000005540 biological transmission Effects 0.000 description 3
- 238000003745 diagnosis Methods 0.000 description 3
- 239000000284 extract Substances 0.000 description 3
- 241000208340 Araliaceae Species 0.000 description 2
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 description 2
- 235000003140 Panax quinquefolius Nutrition 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 235000008434 ginseng Nutrition 0.000 description 2
- 239000013307 optical fiber Substances 0.000 description 2
- 238000000513 principal component analysis Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 230000000644 propagated effect Effects 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 230000004888 barrier function Effects 0.000 description 1
- 230000033228 biological regulation Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000012217 deletion Methods 0.000 description 1
- 230000037430 deletion Effects 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 210000004218 nerve net Anatomy 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 230000000630 rising effect Effects 0.000 description 1
- 238000003786 synthesis reaction Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R35/00—Testing or calibrating of apparatus covered by the other groups of this subclass
- G01R35/04—Testing or calibrating of apparatus covered by the other groups of this subclass of instruments for measuring time integral of power or current
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a kind of intelligent electric energy meter failure modes detection method and system based on autoencoder network, comprising: be divided into training set and test set after the history detection data of the intelligent electric energy meter of acquisition is normalized；The parameter of Initialize installation autoencoder network model；Sample data is chosen in training set to be input in the autoencoder network model, is classified with obtaining signal characteristic and being input in classifier, and be iterated training according to classification results；The autoencoder network model parameter is constantly adjusted according to the classification results of test set, to determine the optimized parameter of autoencoder network model；Classification and Detection is carried out using failure of the corresponding autoencoder network model of the optimized parameter to intelligent electric energy meter.The present invention carries out the feature extraction of unsupervised ground using signal of the depth noise reduction autoencoder network to acquisition, can be realized the quick and precisely classification of fault-signal, facilitates the fault identification ability for promoting intelligent electric energy meter, has stronger noise immunity compared to conventional method.
Description
Technical field
The present invention relates to intelligent electric energy meter failure analysis techniques fields, and more particularly, to one kind based on coding certainly
The intelligent electric energy meter failure modes detection method and system of network.
Background technique
The external interference etc. being subject to during being passed with the use of nonlinear loads all kinds of in electric system and power line
Problem, all kinds of failure problems increase in power grid, and failure mode tends to be complicated.Quick, accurate detection to failures all kinds of in route
And classification, electric system can be effectively reduced by the issuable damaging influence of failure.
Feature is extracted from voltage and current signals, is facilitated researcher and is more fully understood the property of failure, feature inspection
Survey and classification task, and be conducive to complete these tasks using more consistent and effective mode.
In traditional fault detection method, time-domain signal is converted into the key step that frequency-region signal is feature extraction.
Common method has wavelet transform (Discrete Fourier transform, DFT), S-transformation (S-transform,
The Time-Frequency Analysis Methods such as ST).In addition, the validity feature for classification, such as principal component analysis also can be generated in Dimension Reduction Analysis
(principal component analysis, PCA).Although features described above extractive technique has been applied to different types of electric power
System failure identification, but certain priori knowledge is the premise of failure Accurate classification, and some details of the process of implementation often
It needs to modify and adjust repeatedly.Moreover, the accuracy that DFT and ST is converted by remaining factor constraint with influence.Therefore, real
These existing technologies can be very difficult, time-consuming and lack versatility.
Application No. is the 201810522073.5 electric energy metering device method for detecting abnormality based on long memory models in short-term
Deep learning theory is applied in digitalized electrical energy meter abnormality detection, it is big suitable for data volume, learn under easy deletion condition each
The data variation feature of class failure can effectively improve electric energy metering device abnormality detection precision；But this method is time-consuming long, calculates
Amount is big, for a large amount of redundancy electric energy meter acquires signal, needs higher hardware support to grid responsive in time.
Summary of the invention
The present invention proposes a kind of intelligent electric energy meter failure modes detection method and system based on autoencoder network, to solve
The problem of how classification and Detection efficiently and accurately being carried out to intelligent electric energy meter failure.
To solve the above-mentioned problems, according to an aspect of the invention, there is provided a kind of intelligence based on autoencoder network
Electrical energy meter fault classification and Detection method, which is characterized in that the described method includes:
The history detection data of the intelligent electric energy meter of acquisition is normalized, and going through for normalized will be passed through
History detection data is divided into training set and test set；Wherein, the history detection data is voltage signal sample or current signal sample
This, comprising: normal signal data and fault-signal data；
The parameter of Initialize installation autoencoder network model；Wherein, the parameter includes: the layer of autoencoder network model
Several, every layer of number of nodes, the weight of coding structure and biasing；
The sample data that the first-level nodes number is chosen in the training set is input in the autoencoder network model, with
It obtains signal characteristic and is input in classifier and classify, and be iterated training according to classification results；
Measuring accuracy is calculated according to the classification results obtained using test set, and fortune is iterated according to the measuring accuracy
It calculates, adjusts the autoencoder network model parameter, constantly with the optimized parameter of the determination autoencoder network model；
Classification and Detection is carried out using failure of the corresponding autoencoder network model of the optimized parameter to intelligent electric energy meter.
Preferably, wherein the history detection data of the intelligent electric energy meter of described pair of acquisition is normalized, comprising:
Wherein, x'(n) be intelligent electric energy meter by normalized voltage signal or current signal, E () is letter
The mean value of number x (n), Std () are the standard deviation of signal x (n).
Preferably, wherein the method also includes:
Before the history detection data to the intelligent electric energy meter of acquisition is normalized, to the intelligent electric energy of acquisition
Rejecting redundant data and wrong data in the history detection data of table are rejected, and are filled to missing data.
Preferably, wherein the autoencoder network model include: coding structure and decoding structure,
The coding structure is expressed as following formula:
ht=f (Wx'(n)+b)t,
The decoding representation is following formula:
xt=f (W'ht+b')t,
Wherein, x'(n) it is the voltage signal of intelligent electric energy meter by normalized or current signal f () be to activate
Function, W are the weight of coding structure, and b is the biasing of coding structure, and t=1,2 ..., N are the number of plies of autoencoder network model, n
To input dimension, the dimension of different layers is gradually decreased, effectively to extract signal characteristic；W ' is the weight for decoding structure, and b ' is solution
The biasing of code structure, xtIt is equal with the input of coding structure for the output for decoding structure.
Preferably, wherein being initialized using transversal normal distribution to the weight and deviation of coding structure, comprising:
Wherein, f (x) is density function, and u and δ are respectively the mean value and standard deviation of normal distribution, the weight W of coding structure
Value range be (u-2 δ, u+2 δ).
Preferably, wherein the method also includes:
When the penalty values in the cycle of training in continuous first predetermined number threshold value no longer reduce or measuring accuracy even
When starting to reduce after the cycle of training of the second continuous predetermined number threshold value, stop iteration.
Preferably, wherein the method also includes:
When being iterated trained using the autoencoder network model, input signal is added the noise of preset ratio,
And determine the optimization object function of the autoencoder network model are as follows:
Wherein, L (W, b) is penalty values, and σ is the noise being added, and W is the weight of coding structure, and W ' is the power for decoding structure
Weight, first item are quadratic loss function, and Section 2 increases the dilute of model for L2 regularization expression formula to prevent model over-fitting
Dredge property.
Preferably, wherein the method also includes:
In upper one layer of Dropout layers of addition of classifier, for giving up the feature of 1-r probability quantity at random, to mitigate
Fitting；
Wherein, r meets Bernoulli Jacob's distribution, are as follows: r~Bernoulli (p), p are Bernoulli Jacob's distribution parameter, for classification
Actual node number is rht, htFor the output of autoencoder network.
Preferably, wherein adjusting the number of plies and every layer of number of nodes of the autoencoder network model using gridding method, according to
Measuring accuracy adjusts p.
According to another aspect of the present invention, a kind of intelligent electric energy meter failure modes inspection based on autoencoder network is provided
Examining system, which is characterized in that the system comprises:
Normalized module, the history detection data for the intelligent electric energy meter to acquisition are normalized, and
History detection data Jing Guo normalized is divided into training set and test set；Wherein, the history detection data is voltage
Sample of signal or current signal sample, comprising: normal signal data and fault-signal data；
Parameter initialization setup module, the parameter for Initialize installation autoencoder network model；Wherein, the parameter packet
It includes: the weight and biasing of the number of plies of autoencoder network model, every layer of number of nodes, coding structure；
Autoencoder network model repetitive exercise module, for choosing the sample number of the first-level nodes number in the training set
According to being input in the autoencoder network model, classified with obtaining signal characteristic and being input in classifier, and according to classification
As a result it is iterated training；
Optimized parameter determining module, the classification results calculating measuring accuracy for being obtained according to utilization test set, and according to
The measuring accuracy is iterated operation, constantly adjusts the autoencoder network model parameter, with the determination autoencoder network
The optimized parameter of model；
Failure analysis module, for the event using the corresponding autoencoder network model of the optimized parameter to intelligent electric energy meter
Barrier carries out classification and Detection.
Preferably, wherein the normalized module, returns the history detection data of the intelligent electric energy meter of acquisition
One change processing, comprising:
Wherein, x'(n) be intelligent electric energy meter by normalized voltage signal or current signal, E () is letter
The mean value of number x (n), Std () are the standard deviation of signal x (n).
Preferably, wherein the system also includes:
It is normalized for the history detection data in the intelligent electric energy meter to acquisition in data preprocessing module
Before, in the history detection data of the intelligent electric energy meter of acquisition rejecting redundant data and wrong data reject, and to lack
Data are lost to be filled.
Preferably, wherein the autoencoder network model include: coding structure and decoding structure,
The coding structure is expressed as following formula:
ht=f (Wx'(n)+b)t,
The decoding representation is following formula:
xt=f (W'ht+b')t,
Wherein, x'(n) it is the voltage signal of intelligent electric energy meter by normalized or current signal f () be to activate
Function, W are the weight of coding structure, and b is the biasing of coding structure, and t=1,2 ..., N are the number of plies of autoencoder network model, n
To input dimension, the dimension of different layers is gradually decreased, effectively to extract signal characteristic；W ' is the weight for decoding structure, and b ' is solution
The biasing of code structure, xtIt is equal with the input of coding structure for the output for decoding structure.
Preferably, wherein in the parameter initialization setup module, using transversal normal distribution to the weight of coding structure
And deviation is initialized, comprising:
Wherein, f (x) is density function, and u and δ are respectively the mean value and standard deviation of normal distribution, the weight W of coding structure
Value range be (u-2 δ, u+2 δ).
Preferably, wherein the autoencoder network model repetitive exercise module, further includes:
When the penalty values in the cycle of training in continuous first predetermined number threshold value no longer reduce or measuring accuracy even
When starting to reduce after the cycle of training of the second continuous predetermined number threshold value, deconditioning.
Preferably, wherein the autoencoder network model repetitive exercise module, further includes:
When being iterated trained using the autoencoder network model, input signal is added the noise of preset ratio,
And determine the optimization object function of the autoencoder network model are as follows:
Wherein, L (W, b) is penalty values, and σ is the noise being added, and W is the weight of coding structure, and W ' is the power for decoding structure
Weight, first item are quadratic loss function, and Section 2 increases the dilute of model for L2 regularization expression formula to prevent model over-fitting
Dredge property.
Preferably, wherein the system also includes:
In upper one layer of Dropout layers of addition of classifier, for giving up the feature of 1-r probability quantity at random, to mitigate
Fitting；
Wherein, r meets Bernoulli Jacob's distribution, are as follows: r~Bernoulli (p), p are Bernoulli Jacob's distribution parameter, for classification
Actual node number is rht, htFor the output of autoencoder network.
Preferably, wherein adjusting the number of plies and every layer of number of nodes of the autoencoder network model using gridding method, according to
Measuring accuracy adjusts p.
The present invention provides a kind of intelligent electric energy meter failure modes detection method and system based on autoencoder network, packet
It includes: being divided into training set and test set after the history detection data of the intelligent electric energy meter of acquisition is normalized；Initialization
The parameter of autoencoder network model is set；Sample data is chosen in the training set is input to the autoencoder network model
In, classified with obtaining signal characteristic and being input in classifier, and be iterated training according to classification results；According to test set
Classification results constantly adjust the autoencoder network model parameter, with the optimized parameter of the determination autoencoder network model；
Classification and Detection is carried out using failure of the corresponding autoencoder network model of the optimized parameter to intelligent electric energy meter.The present invention is first
It is compressed for the data feeding autoencoder network of largely non-label, net is extracted by multilevel encoder and decoder constitutive characteristic
Network extracts the feature of fault-signal and normal signal；Finally, using there is the fault sample of label to divide classifier on a small quantity
Class.The present invention carries out the feature extraction of unsupervised ground using signal of the depth noise reduction autoencoder network to acquisition, is suitable for label and believes
Insufficient condition is ceased, in conjunction with unsupervised learning and Supervised classification, the quick and precisely classification of fault-signal is realized, helps to mention
The fault identification ability for rising intelligent electric energy meter has stronger noise immunity compared to conventional method.
Detailed description of the invention
By reference to the following drawings, exemplary embodiments of the present invention can be more fully understood by:
Fig. 1 is the intelligent electric energy meter failure modes detection method based on autoencoder network according to embodiment of the present invention
100 flow chart；
Fig. 2 is the schematic diagram according to the intelligent electric energy meter data transmission system of embodiment of the present invention；
Fig. 3 is the schematic diagram according to the autoencoder network model of embodiment of the present invention；
Fig. 4 is the schematic diagram using autoencoder network model depth noise reduction according to embodiment of the present invention；And
Fig. 5 is the intelligent electric energy meter failure modes detection system based on autoencoder network according to embodiment of the present invention
500 structural schematic diagram.
Specific embodiment
Exemplary embodiments of the present invention are introduced referring now to the drawings, however, the present invention can use many different shapes
Formula is implemented, and is not limited to the embodiment described herein, and to provide these embodiments be at large and fully disclose
The present invention, and the scope of the present invention is sufficiently conveyed to person of ordinary skill in the field.Show for what is be illustrated in the accompanying drawings
Term in example property embodiment is not limitation of the invention.In the accompanying drawings, identical cells/elements use identical attached
Icon note.
Unless otherwise indicated, term (including scientific and technical terminology) used herein has person of ordinary skill in the field
It is common to understand meaning.Further it will be understood that with the term that usually used dictionary limits, should be understood as and its
The context of related fields has consistent meaning, and is not construed as Utopian or too formal meaning.
Fig. 1 is the intelligent electric energy meter failure modes detection method based on autoencoder network according to embodiment of the present invention
100 flow chart.As shown in Figure 1, the intelligent electric energy meter failure based on autoencoder network point that embodiments of the present invention provide
Class detection method is compressed first against the data feeding autoencoder network of largely non-label, by multilevel encoder and decoding
Device constitutive characteristic extracts network, extracts the feature of fault-signal and normal signal；Finally, using the fault sample for having label on a small quantity
Classify to classifier, by this way, can be realized and the fault-signal under label data small sample is effectively monitored.This hair
Bright embodiment carries out the feature extraction of unsupervised ground using signal of the depth noise reduction autoencoder network to acquisition, is suitable for label
The insufficient condition of information is realized the quick and precisely classification of fault-signal, is facilitated in conjunction with unsupervised learning and Supervised classification
The fault identification ability for promoting intelligent electric energy meter has stronger noise immunity compared to conventional method.Embodiments of the present invention mention
The intelligent electric energy meter failure modes detection method 100 based on autoencoder network supplied is right in step 101 since step 101 place
The history detection data of the intelligent electric energy meter of acquisition is normalized, and by the history detection data Jing Guo normalized
It is divided into training set and test set；Wherein, the history detection data is voltage signal sample or current signal sample, comprising: just
Regular signal data and fault-signal data.
Preferably, wherein the history detection data of the intelligent electric energy meter of described pair of acquisition is normalized, comprising:
Wherein, x'(n) be intelligent electric energy meter by normalized voltage signal or current signal, E () is letter
The mean value of number x (n), Std () are the standard deviation of signal x (n).
Preferably, wherein the method also includes:
Before the history detection data to the intelligent electric energy meter of acquisition is normalized, to the intelligent electric energy of acquisition
Rejecting redundant data and wrong data in the history detection data of table are rejected, and are filled to missing data.
In embodiments of the present invention, the intelligence in certain typical province is obtained using data transmission system as shown in Figure 2
Electric energy meter historical data, comprising: signal and fault-signal when normal work, the signal can be voltage signal or current signal
At least one of.Intelligent electric energy meter will be sent to base station server host by concentrator after voltage and current signal sampling,
Wherein optical fiber or GPRS communication can be used in communication modes.
After getting the history detection data of intelligent electric energy meter, firstly for missing data using Spline Interpolation Method into
Row interpolation, and for digital signal value be greater than some threshold value when value, redundant data and wrong data directly rejected.
Then, it is all made of following method for normalizing for all data to be normalized, and by the history Jing Guo normalized
Detection data is divided into training set and test set.Wherein, place is normalized to the history detection data of the intelligent electric energy meter of acquisition
Reason, comprising:
Wherein, x'(n) be intelligent electric energy meter by normalized voltage signal or current signal, E () is letter
The mean value of number x (n), Std () are the standard deviation of signal x (n).
In step 102, the parameter of Initialize installation autoencoder network model；Wherein, the parameter includes: autoencoder network
The weight and biasing of the number of plies of model, every layer of number of nodes, coding structure.
Preferably, wherein the autoencoder network model include: coding structure and decoding structure,
The coding structure is expressed as following formula:
ht=f (Wx'(n)+b)t,
The decoding representation is following formula:
xt=f (W'ht+b')t,
Wherein, x'(n) it is the voltage signal of intelligent electric energy meter by normalized or current signal f () be to activate
Function, W are the weight of coding structure, and b is the biasing of coding structure, and t=1,2 ..., N are the number of plies of autoencoder network model, n
To input dimension, the dimension of different layers is gradually decreased, effectively to extract signal characteristic；W ' is the weight for decoding structure, and b ' is solution
The biasing of code structure, xtIt is equal with the input of coding structure for the output for decoding structure.
Preferably, wherein being initialized using transversal normal distribution to the weight and deviation of coding structure, comprising:
Wherein, f (x) is density function, and u and δ are respectively the mean value and standard deviation of normal distribution, the weight W of coding structure
Value range be (u-2 δ, u+2 δ).
In embodiments of the present invention, autoencoder network model is the nerve net being made of the sparse self-encoding encoder of multilayer
Network, each layer of output are connected to next layer of input.Autoencoder network model is made of encoder and decoder two parts, right
For one N layers of depth encoder, the output of coding structure is indicated with following equation are as follows: ht=f (Wx'(n)+b)t；Its
In, f () is activation primitive, and W presentation code structure ratio, the biasing of b presentation code structure, t=1,2 ..., N indicate the number of plies；
Wherein, input dimension is n, and the dimension of different layers gradually decreases, effectively to extract signal characteristic.
Fig. 3 is the schematic diagram according to the autoencoder network model of embodiment of the present invention.As shown in figure 3, for one 3 layers
Autoencoder network model.
It can be line rectification (RELU) activation primitive, expression formula for activation primitive are as follows: f (x)=max (0, x),
Gradient decline solution can be effectively ensured in RELU and direction is propagated, and avoid the problem of gradient disappears, and calculating is easier,
Reduce the calculation amount of network.
If the output node number of encoder is set as m, decoder architecture may be expressed as: xt=f (W'ht+b')t, wherein
W ' indicates that decoding structure ratio, b ' indicate the biasing of decoding structure, and exports input of the xt for decoder output, with encoder
It is equal.
In embodiments of the present invention, weights initialisation, the table of the initial method are carried out using transversal normal distribution
Up to formula are as follows:
Wherein, u and δ is the mean value and standard deviation of normal distribution, and the value range of weight W is (u-2 δ, u+2 δ).
In step 103, the sample data that the first-level nodes number is chosen in the training set is input to the coding net certainly
In network model, classified with obtaining signal characteristic and being input in classifier, and be iterated training according to classification results.
Preferably, wherein the method also includes:
When being iterated trained using the autoencoder network model, input signal is added the noise of preset ratio,
And determine the optimization object function of the autoencoder network model are as follows:
Wherein, L (W, b) is penalty values, and σ is the noise being added, and W is the weight of coding structure, and W ' is the power for decoding structure
Weight, first item are quadratic loss function, and Section 2 increases the dilute of model for L2 regularization expression formula to prevent model over-fitting
Dredge property.
Preferably, wherein the method also includes:
When the penalty values in the cycle of training in continuous first predetermined number threshold value no longer reduce or measuring accuracy even
When starting to reduce after the cycle of training of the second continuous predetermined number threshold value, stop iteration.
Preferably, wherein the method also includes:
In upper one layer of Dropout layers of addition of classifier, for giving up the feature of 1-r probability quantity at random, to mitigate
Fitting；
Wherein, r meets Bernoulli Jacob's distribution, are as follows: r~Bernoulli (p), p are Bernoulli Jacob's distribution parameter, for classification
Actual node number is rht, htFor the output of autoencoder network.
Fig. 4 is the schematic diagram using autoencoder network model depth noise reduction according to embodiment of the present invention.Such as Fig. 4 institute
Show, in order to improve the noiseproof feature of model, model is trained from coding using noise reduction in embodiments of the present invention, i.e.,
A certain proportion of noise is added to input signal, input signal becomes x (n)+σ at this time.To solve depth noise reduction from encoding model
Parameter, the optimization object function of model are as follows:
Wherein, first item is quadratic loss function, and Section 2 is L2 regularization expression formula, it is therefore intended that prevents model excessively quasi-
It closes, increases the sparsity of model.
The sample data that the first-level nodes number is chosen in the training set is input in the autoencoder network model, with
Signal characteristic is obtained, and the characteristic signal that will acquire is input in classifier and classifies to obtain classification results, when continuous
The first predetermined number threshold value cycle of training in penalty values when no longer reducing, stop repetitive exercise.Wherein, first default
Number threshold value can be 5,8,10 etc..
In embodiments of the present invention, classifier uses Softmax classifier.The expression formula of the classifier are as follows:
Wherein, hlIt is depth noise reduction from the output of the coding structure encoded, K indicates intelligent electric energy meter failure kind collected
Class number.The purpose of Softmax be for the output valve of model to be tied to (0,1] in, the corresponding position of output most probable value
As corresponding fault category.
In addition, classifier upper one layer of Dropout layers of addition, for giving up the feature of 1-r probability quantity at random, with
Mitigate over-fitting；Wherein, r meets Bernoulli Jacob's distribution, are as follows: r~Bernoulli (p), p are Bernoulli Jacob's distribution parameter, for classifying
Actual node number be rht, htFor the output of autoencoder network.
In step 104, measuring accuracy is calculated according to the classification results obtained using test set, and according to the measuring accuracy
It is iterated operation, constantly adjusts the autoencoder network model parameter, with the optimal ginseng of the determination autoencoder network model
Number.
Preferably, wherein adjusting the number of plies and every layer of number of nodes of the autoencoder network model using gridding method, according to
Measuring accuracy adjusts p.
When being tested using test set, measuring accuracy is calculated, when measuring accuracy is in continuous second predetermined number threshold
When starting to reduce after the cycle of training of value, stop iteration；Otherwise, the number of plies of the autoencoder network model is adjusted using gridding method
With every layer of number of nodes, p is adjusted according to measuring accuracy, until measuring accuracy meets the requirement of setting.
In step 105, carried out using failure of the corresponding autoencoder network model of the optimized parameter to intelligent electric energy meter
Classification and Detection.
Trained depth noise reduction is passed through into intelligent electric energy meter from the host that encoding model is placed in base station server
The data of acquisition analyze electric network fault type in real time.
The embodiment illustrated the present invention in detail below
With the intelligent electric energy meter data instance in certain typical province, the intelligent electric energy meter failure modes inspection based on autoencoder network
Survey method, comprising the following steps:
S1: obtaining the intelligent electric energy meter historical data in certain typical province using data transmission system as shown in Figure 2, including
Voltage and current signal.Intelligent electric energy meter will be sent to base station server master by concentrator after voltage and current signal sampling
Machine.Wherein optical fiber or GPRS communication can be used in communication modes.
S2: it after carrying out rejecting and filling processing to the intelligent electric energy meter historical data of acquisition, is normalized, to subtract
Influence of the small dimension to result.It specifically includes:
Wherein, x ' (n) indicates that the voltage and current signal of intelligent electric energy meter acquisition, E () indicate the mean value of signal x (n),
Std indicates the standard deviation of signal x (n).
In the present embodiment, it for the signal after normalized, chooses 70% and is used as training set, choose 30% as survey
Examination collection.
To classify in the present embodiment for 6 kinds of fault types, every kind of fault type takes 1000 samples, i.e., and totally 6000
Fault sample.In addition 1000 groups of normal samples are taken to be analyzed, i.e., total number of training is 7000.In addition, being directed to different samples
Length it is different, the signal of all samples is uniformly downsampled to 600 data points.I.e. each sample length is 600, be ensure that
The uniformity of data and the consistency of model structure.
S3: Initialize installation is carried out to the parameter of autoencoder network model, and all data are carried out using self-encoding encoder
Training.
In order to improve the noiseproof feature of model, model is trained from coding using noise reduction, i.e., input signal is added
A certain proportion of noise, input signal becomes x (n)+σ at this time.
In the present embodiment, the white Gaussian noise that signal-to-noise ratio is respectively 20dB, 40dB is added for input signal.To optimize depth
Noise reduction encodes model parameter, the majorized function of model certainly are as follows:
Wherein, first item is quadratic loss function, and Section 2 is L2 regularization expression formula, it is therefore intended that prevents model excessively quasi-
It closes, increases the sparsity of model.
In the present embodiment, selected activation primitive is line rectification (RELU) activation primitive, expression formula are as follows: f (x)=
max(0,x)；Gradient decline solution can be effectively ensured in RELU and direction is propagated, and avoid the problem of gradient disappears, and calculate
It is easier, reduce the calculation amount of network.
In the present embodiment, the initialization of weight and offset parameter is carried out using transversal normal distribution, the initial method
Expression formula are as follows:
Wherein, u and δ is the mean value and standard deviation of normal distribution, then the value range of respective weights W is (u-2 δ, u+2 δ).
Using it is shown in Fig. 4 model is trained from coding structure when, which contains one 5 layers of depth
Noise reduction autoencoder network and one layer of classification layer.In the present embodiment according to data volume size, the super ginseng of suitable network is set
Number.By 600 nodes that are dimensioned to of input layer, i.e. first layer size is 600, and second layer number of nodes is 300, and third layer is special
The number of nodes of sign output layer is set as 100, i.e. m=100.The number of nodes that the 4th layer of decoder architecture part is 300, layer 5
Node is 600.
In training, when penalty values are kept for no longer to reduce continuous 10 cycles of training；Or 10, measuring accuracy interval instruction
Practice when starting to reduce after the period, then deconditioning.
Wherein, classifier uses Softmax classifier, the expression formula of the classifier are as follows:
Wherein, hl is the output of depth noise reduction from the coding structure encoded, and K indicates intelligent electric energy meter failure kind collected
Class number.The purpose of Softmax be for target function value to be tied to (0,1] in.Then export the corresponding position of most probable value
As corresponding fault category.
Model over-fitting in order to prevent, one layer of addition, one layer of Dropout is excessively quasi- to reduce model on Softmax classifier
It closes.After the layer is added in network, the number of features of the certain probability 1-r of random drop is then used for the reality of classification to mitigate over-fitting
Number of nodes is rht.
In the present embodiment, 0.4 is set as the parameter r that the number of nodes of Softmax classifier is 100, Dropout.
S4: according to the performance of model classifiers, the hyper parameter of continuous percentage regulation noise reduction autoencoder network needs to adjust
Hyper parameter mainly includes the Probability p of the model number of plies, the number of nodes of every layer of model and Dropout, repeats above step until full
The default measuring accuracy threshold value of foot.
S5: trained depth noise reduction is passed through into intelligent electricity from the host that encoding model is placed in base station server
The data of energy table acquisition, analyze electric network fault type in real time.
In the present embodiment, three-phase or single-phase voltage and current signal are inputted respectively according to sample order multiple from coding
Model.Single-phase voltage and current signal are used in the present embodiment, i.e., 2 depth noise reductions is needed to be instructed from encoding model altogether
Practice.
The present embodiment is using the optimal hyper parameter of gridding method search model, by taking the network number of plies as an example, by the number of plies from 3 to 10 according to
The secondary increase network number of plies, according to network fault diagnosis as a result, determining the optimal number of plies after, then model node number carried out similar
Search.
The present embodiment is by taking the number of plies as an example, with the output result of Softmax classifier maximum probability for final fault category knot
Fruit.The precision of depth noise reduction autoencoder network is tested, table 1 gives the relationship between the number of plies and accuracy.It is anti-
The output node number of the only interference of number of nodes, middle layer characteristic layer is both configured to 100.
The intelligent electric energy meter fault-signal diagnostic classification result that table 1 is encoded certainly based on depth noise reduction
As known from Table 1, as the depth noise reduction autoencoder network number of plies increases, precision has a larger increase, but when the number of plies is greater than
When certain amount, precision is begun to decline, and shows that the number of plies is to influence one of the principal element of intelligent electric energy meter fault diagnosis precision.When
When the number of plies increases, model parameter increases, and the training time is consequently increased.Show that more model parameters are presented with fault diagnosis
It is certain to influence.In addition, being directed to 6 kinds of fault-signals, synthesis precision and training time can be shown that 5 layers are that a preferably number of plies is selected
It selects.
Fig. 5 is the intelligent electric energy meter failure modes detection system based on autoencoder network according to embodiment of the present invention
500 structural schematic diagram.As shown in figure 5, the intelligent electric energy meter event based on autoencoder network that embodiments of the present invention provide
Hinder classified detection system 500, comprising: normalized module 501, parameter initialization setup module 502, autoencoder network model
Repetitive exercise module 503, optimized parameter determining module 504 and failure analysis module 505.
Preferably, the normalized module 501, the history detection data for the intelligent electric energy meter to acquisition carry out
Normalized, and the history detection data Jing Guo normalized is divided into training set and test set；Wherein, the history inspection
Measured data is voltage signal sample or current signal sample, comprising: normal signal data and fault-signal data.
Preferably, wherein the normalized module 501, carries out the history detection data of the intelligent electric energy meter of acquisition
Normalized, comprising:
Wherein, x'(n) be intelligent electric energy meter by normalized voltage signal or current signal, E () is letter
The mean value of number x (n), Std () are the standard deviation of signal x (n).
Preferably, wherein the system also includes data preprocessing modules, for going through in the intelligent electric energy meter to acquisition
Before history detection data is normalized, to the rejecting redundant data in the history detection data of the intelligent electric energy meter of acquisition
It is rejected with wrong data, and missing data is filled.
Preferably, the parameter initialization setup module 502, the parameter for Initialize installation autoencoder network model；
Wherein, the parameter includes: the weight and biasing of the number of plies of autoencoder network model, every layer of number of nodes, coding structure.
Preferably, wherein the autoencoder network model includes: coding structure and decoding structure, the coding structure is indicated
For following formula:
ht=f (Wx'(n)+b)t,
The decoding representation is following formula:
xt=f (W'ht+b')t,
Wherein, x'(n) it is the voltage signal of intelligent electric energy meter by normalized or current signal f () be to activate
Function, W are the weight of coding structure, and b is the biasing of coding structure, and t=1,2 ..., N are the number of plies of autoencoder network model, n
To input dimension, the dimension of different layers is gradually decreased, effectively to extract signal characteristic；W ' is the weight for decoding structure, and b ' is solution
The biasing of code structure, xtIt is equal with the input of coding structure for the output for decoding structure.
Preferably, wherein in the parameter initialization setup module 502, using transversal normal distribution to the power of coding structure
Weight and deviation are initialized, comprising:
Wherein, f (x) is density function, and u and δ are respectively the mean value and standard deviation of normal distribution, the weight W of coding structure
Value range be (u-2 δ, u+2 δ).
Preferably, the autoencoder network model repetitive exercise module 503, for choosing first layer in the training set
The sample data of number of nodes is input in the autoencoder network model, is divided with obtaining signal characteristic and being input in classifier
Class, and training is iterated according to classification results.
Preferably, wherein the autoencoder network model repetitive exercise module 503, further includes: when pre- continuous first
If the penalty values in the cycle of training of number threshold value no longer reduce or measuring accuracy continuous second predetermined number threshold value instruction
When starting to reduce after the white silk period, deconditioning.
Preferably, wherein the autoencoder network model repetitive exercise module 503, further includes: encoded certainly using described
When network model is iterated trained, input signal is added the noise of preset ratio, and determine the autoencoder network model
Optimization object function are as follows:
Wherein, L (W, b) is penalty values, and σ is the noise being added, and W is the weight of coding structure, and W ' is the power for decoding structure
Weight, first item are quadratic loss function, and Section 2 increases the dilute of model for L2 regularization expression formula to prevent model over-fitting
Dredge property.
Preferably, wherein the system also includes: in upper one layer of Dropout layers of addition of classifier, for giving up at random
The feature of 1-r probability quantity, to mitigate over-fitting；Wherein, r meets Bernoulli Jacob's distribution, are as follows: r~Bernoulli (p), p primary
Sharp distribution parameter is exerted, the actual node number for classification is rht, htFor the output of autoencoder network.
Preferably, the optimized parameter determining module 504 is surveyed for being calculated according to the classification results obtained using test set
Precision is tried, and operation is iterated according to the measuring accuracy, adjusts the autoencoder network model parameter, constantly to determine
State the optimized parameter of autoencoder network model.
Preferably, wherein adjusting the number of plies and every layer of number of nodes of the autoencoder network model using gridding method, according to
Measuring accuracy adjusts p.
Preferably, the failure analysis module 505, for utilizing the corresponding autoencoder network model pair of the optimized parameter
The failure of intelligent electric energy meter carries out classification and Detection.
The intelligent electric energy meter failure modes detection system 500 and the present invention based on autoencoder network of the embodiment of the present invention
Another embodiment the intelligent electric energy meter failure modes detection method 100 based on autoencoder network it is corresponding, herein no longer
It repeats.
The present invention is described by reference to a small amount of embodiment.However, it is known in those skilled in the art, as
Defined by subsidiary Patent right requirement, in addition to the present invention other embodiments disclosed above equally fall in it is of the invention
In range.
Normally, all terms used in the claims are all solved according to them in the common meaning of technical field
It releases, unless in addition clearly being defined wherein.All references " one/described/be somebody's turn to do [device, component etc.] " are all opened ground
At least one example being construed in described device, component etc., unless otherwise expressly specified.Any method disclosed herein
Step need not all be run with disclosed accurate sequence, unless explicitly stated otherwise.
Claims (18)
1. a kind of intelligent electric energy meter failure modes detection method based on autoencoder network, which is characterized in that the described method includes:
The history detection data of the intelligent electric energy meter of acquisition is normalized, and the history Jing Guo normalized is examined
Measured data is divided into training set and test set；Wherein, the history detection data is voltage signal sample or current signal sample, packet
It includes: normal signal data and fault-signal data；
The parameter of Initialize installation autoencoder network model；Wherein, the parameter includes: the number of plies, every of autoencoder network model
Layer number of nodes, coding structure weight and biasing；
The sample data that the first-level nodes number is chosen in the training set is input in the autoencoder network model, to obtain
Signal characteristic, which is input in classifier, classifies, and is iterated training according to classification results；
Measuring accuracy is calculated according to the classification results obtained using test set, and operation is iterated according to the measuring accuracy,
The autoencoder network model parameter is adjusted, constantly with the optimized parameter of the determination autoencoder network model；
Classification and Detection is carried out using failure of the corresponding autoencoder network model of the optimized parameter to intelligent electric energy meter.
2. the method according to claim 1, wherein the history detection data of the intelligent electric energy meter of described pair of acquisition
It is normalized, comprising:
Wherein, x'(n) be intelligent electric energy meter by normalized voltage signal or current signal, E () is signal x
(n) mean value, Std () are the standard deviation of signal x (n).
3. the method according to claim 1, wherein the method also includes:
Before the history detection data to the intelligent electric energy meter of acquisition is normalized, to the intelligent electric energy meter of acquisition
Rejecting redundant data and wrong data in history detection data are rejected, and are filled to missing data.
4. the method according to claim 1, wherein the autoencoder network model includes: coding structure reconciliation
Code structure,
The coding structure is expressed as following formula:
ht=f (Wx'(n)+b)t,
The decoding representation is following formula:
xt=f (W'ht+b')t,
Wherein, x'(n) it is the voltage signal of intelligent electric energy meter by normalized or current signal f () be to activate letter
Number, W are the weight of coding structure, and b is the biasing of coding structure, and t=1,2 ..., N are the number of plies of autoencoder network model, and n is
Dimension is inputted, the dimension of different layers gradually decreases, effectively to extract signal characteristic；W ' is the weight for decoding structure, and b ' is decoding
The biasing of structure, xtIt is equal with the input of coding structure for the output for decoding structure.
5. according to the method described in claim 4, it is characterized in that, using transversal normal distribution to the weight of coding structure and partially
Difference is initialized, comprising:
Wherein, f (x) is density function, and u and δ are respectively the mean value and standard deviation of normal distribution, and the weight W's of coding structure takes
Being worth range is (u-2 δ, u+2 δ).
6. the method according to claim 1, wherein the method also includes:
When the penalty values in the cycle of training in continuous first predetermined number threshold value no longer reduce or measuring accuracy is continuous
When starting to reduce after the cycle of training of the second predetermined number threshold value, stop iteration.
7. the method according to claim 1, wherein the method also includes:
When being iterated trained using the autoencoder network model, input signal is added the noise of preset ratio, and really
The optimization object function of the fixed autoencoder network model are as follows:
Wherein, L (W, b) is penalty values, and σ is the noise being added, and W is the weight of coding structure, and W ' is the weight for decoding structure, the
One is quadratic loss function, and Section 2 increases the sparsity of model for L2 regularization expression formula to prevent model over-fitting.
8. the method according to claim 1, wherein the method also includes:
In upper one layer of Dropout layers of addition of classifier, for giving up the feature of 1-r probability quantity at random, to mitigate over-fitting；
Wherein, r meets Bernoulli Jacob's distribution, are as follows: r~Bernoulli (p), p are Bernoulli Jacob's distribution parameter, the reality for classification
Number of nodes is rht, htFor the output of autoencoder network.
9. according to the method described in claim 8, it is characterized in that, adjusting the layer of the autoencoder network model using gridding method
Several and every layer of number of nodes adjusts p according to measuring accuracy.
10. a kind of intelligent electric energy meter failure modes detection system based on autoencoder network, which is characterized in that the system packet
It includes:
Normalized module, the history detection data for the intelligent electric energy meter to acquisition are normalized, and will be through
The history detection data for crossing normalized is divided into training set and test set；Wherein, the history detection data is voltage signal
Sample or current signal sample, comprising: normal signal data and fault-signal data；
Parameter initialization setup module, the parameter for Initialize installation autoencoder network model；Wherein, the parameter includes:
The weight and biasing of the number of plies of autoencoder network model, every layer of number of nodes, coding structure；
Autoencoder network model repetitive exercise module, the sample data for choosing the first-level nodes number in the training set are defeated
Enter into the autoencoder network model, is classified with obtaining signal characteristic and being input in classifier, and according to classification results
It is iterated training；
Optimized parameter determining module, for calculating measuring accuracy according to the classification results obtained using test set, and according to described
Measuring accuracy is iterated operation, constantly adjusts the autoencoder network model parameter, with the determination autoencoder network model
Optimized parameter；
Failure analysis module, for using the corresponding autoencoder network model of the optimized parameter to the failure of intelligent electric energy meter into
Row classification and Detection.
11. system according to claim 10, which is characterized in that the normalized module, to the intelligence electricity of acquisition
The history detection data of energy table is normalized, comprising:
Wherein, x'(n) be intelligent electric energy meter by normalized voltage signal or current signal, E () is signal x
(n) mean value, Std () are the standard deviation of signal x (n).
12. system according to claim 10, which is characterized in that the system also includes:
Data preprocessing module, for before the history detection data to the intelligent electric energy meter of acquisition is normalized,
To in the history detection data of the intelligent electric energy meter of acquisition rejecting redundant data and wrong data reject, and to missing number
According to being filled.
13. system according to claim 10, which is characterized in that the autoencoder network model include: coding structure and
Structure is decoded,
The coding structure is expressed as following formula:
ht=f (Wx'(n)+b)t,
The decoding representation is following formula:
xt=f (W'ht+b')t,
Wherein, x'(n) it is the voltage signal of intelligent electric energy meter by normalized or current signal f () be to activate letter
Number, W are the weight of coding structure, and b is the biasing of coding structure, and t=1,2 ..., N are the number of plies of autoencoder network model, and n is
Dimension is inputted, the dimension of different layers gradually decreases, effectively to extract signal characteristic；W ' is the weight for decoding structure, and b ' is decoding
The biasing of structure, xtIt is equal with the input of coding structure for the output for decoding structure.
14. system according to claim 13, which is characterized in that in the parameter initialization setup module, using truncation
Normal distribution initializes the weight and deviation of coding structure, comprising:
Wherein, f (x) is density function, and u and δ are respectively the mean value and standard deviation of normal distribution, and the weight W's of coding structure takes
Being worth range is (u-2 δ, u+2 δ).
15. system according to claim 10, which is characterized in that the autoencoder network model repetitive exercise module, also
Include:
When the penalty values in the cycle of training in continuous first predetermined number threshold value no longer reduce or measuring accuracy is continuous
When starting to reduce after the cycle of training of the second predetermined number threshold value, deconditioning.
16. system according to claim 10, which is characterized in that the autoencoder network model repetitive exercise module, also
Include:
When being iterated trained using the autoencoder network model, input signal is added the noise of preset ratio, and really
The optimization object function of the fixed autoencoder network model are as follows:
Wherein, L (W, b) is penalty values, and σ is the noise being added, and W is the weight of coding structure, and W ' is the weight for decoding structure, the
One is quadratic loss function, and Section 2 increases the sparsity of model for L2 regularization expression formula to prevent model over-fitting.
17. system according to claim 10, which is characterized in that the system also includes:
In upper one layer of Dropout layers of addition of classifier, for giving up the feature of 1-r probability quantity at random, to mitigate over-fitting；
Wherein, r meets Bernoulli Jacob's distribution, are as follows: r~Bernoulli (p), p are Bernoulli Jacob's distribution parameter, the reality for classification
Number of nodes is rht, htFor the output of autoencoder network.
18. system according to claim 17, which is characterized in that adjust the autoencoder network model using gridding method
The number of plies and every layer of number of nodes adjust p according to measuring accuracy.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910208116.7A CN110059357A (en) | 2019-03-19 | 2019-03-19 | A kind of intelligent electric energy meter failure modes detection method and system based on autoencoder network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910208116.7A CN110059357A (en) | 2019-03-19 | 2019-03-19 | A kind of intelligent electric energy meter failure modes detection method and system based on autoencoder network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110059357A true CN110059357A (en) | 2019-07-26 |
Family
ID=67317196
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910208116.7A Pending CN110059357A (en) | 2019-03-19 | 2019-03-19 | A kind of intelligent electric energy meter failure modes detection method and system based on autoencoder network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110059357A (en) |
Cited By (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110458240A (en) * | 2019-08-16 | 2019-11-15 | 集美大学 | A kind of three-phase bridge rectifier method for diagnosing faults, terminal device and storage medium |
| CN110658373A (en) * | 2019-09-30 | 2020-01-07 | 国网湖南省电力有限公司 | Power distribution network electricity stealing monitoring method based on multi-algorithm fusion, model and readable storage medium |
| CN110689203A (en) * | 2019-09-30 | 2020-01-14 | 国网山东省电力公司电力科学研究院 | Self-encoder-based primary fan fault early warning method for thermal power plant |
| CN110779746A (en) * | 2019-10-24 | 2020-02-11 | 西安理工大学 | Diagnosis method for improving composite fault of deep sparse self-encoder network rotating machinery |
| CN110794360A (en) * | 2019-10-22 | 2020-02-14 | 中国电力科学研究院有限公司 | Method and system for predicting fault of intelligent electric energy meter based on machine learning |
| CN110910982A (en) * | 2019-11-04 | 2020-03-24 | 广州金域医学检验中心有限公司 | Self-coding model training method, device, equipment and storage medium |
| CN110956091A (en) * | 2019-11-05 | 2020-04-03 | 广州供电局有限公司 | Electric energy meter identification method and device, computer equipment and storage medium |
| CN110995396A (en) * | 2019-12-19 | 2020-04-10 | 中国电力科学研究院有限公司 | Compression method of communication messages of electricity consumption information acquisition system based on hierarchical structure |
| CN111076934A (en) * | 2019-12-24 | 2020-04-28 | 江苏大学 | Method for diagnosing potential fault of bearing based on S transformation |
| CN111404130A (en) * | 2020-04-14 | 2020-07-10 | 国网内蒙古东部电力有限公司赤峰供电公司 | Novel power distribution network fault detection method and fault self-healing system based on quick switch |
| CN111506598A (en) * | 2020-04-13 | 2020-08-07 | 中国科学院自动化研究所 | Fault discrimination method, system and device based on small sample self-learning fault migration |
| CN111860684A (en) * | 2020-07-30 | 2020-10-30 | 元神科技(杭州)有限公司 | Power plant equipment fault early warning method and system based on dual networks |
| CN112861625A (en) * | 2021-01-05 | 2021-05-28 | 深圳技术大学 | Method for determining stacking denoising autoencoder model |
| CN113533995A (en) * | 2021-07-05 | 2021-10-22 | 上海电享信息科技有限公司 | Power battery consistency detection method |
| CN113642244A (en) * | 2021-08-23 | 2021-11-12 | 广东电网有限责任公司 | Power metering equipment fault prediction method based on artificial intelligence |
| CN113674846A (en) * | 2021-09-18 | 2021-11-19 | 浙江远图互联科技股份有限公司 | Hospital intelligent service public opinion monitoring platform based on LSTM network |
| CN114062812A (en) * | 2021-11-15 | 2022-02-18 | 国网四川省电力公司营销服务中心 | Fault diagnosis method and system for metering cabinet |
| CN114189825A (en) * | 2021-12-09 | 2022-03-15 | 苏州捷布森智能科技有限公司 | Data processing method and system based on industrial Internet and intelligent manufacturing |
| CN115545191A (en) * | 2022-12-01 | 2022-12-30 | 中国南方电网有限责任公司超高压输电公司广州局 | Current noise reduction network model training method and fault current limiter current noise reduction method |
| CN117272055A (en) * | 2023-11-23 | 2023-12-22 | 国网山西省电力公司营销服务中心 | Electric energy meter abnormality detection method and device based on filtering enhancement self-encoder |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108255656A (en) * | 2018-02-28 | 2018-07-06 | 湖州师范学院 | A kind of fault detection method applied to batch process |
| CN108921230A (en) * | 2018-07-25 | 2018-11-30 | 浙江浙能嘉华发电有限公司 | Method for diagnosing faults based on class mean value core pivot element analysis and BP neural network |
| 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 |
-
2019
- 2019-03-19 CN CN201910208116.7A patent/CN110059357A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108255656A (en) * | 2018-02-28 | 2018-07-06 | 湖州师范学院 | A kind of fault detection method applied to batch process |
| 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 |
| CN108921230A (en) * | 2018-07-25 | 2018-11-30 | 浙江浙能嘉华发电有限公司 | Method for diagnosing faults based on class mean value core pivot element analysis and BP neural network |
Cited By (31)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110458240A (en) * | 2019-08-16 | 2019-11-15 | 集美大学 | A kind of three-phase bridge rectifier method for diagnosing faults, terminal device and storage medium |
| CN110658373A (en) * | 2019-09-30 | 2020-01-07 | 国网湖南省电力有限公司 | Power distribution network electricity stealing monitoring method based on multi-algorithm fusion, model and readable storage medium |
| CN110689203A (en) * | 2019-09-30 | 2020-01-14 | 国网山东省电力公司电力科学研究院 | Self-encoder-based primary fan fault early warning method for thermal power plant |
| CN110794360A (en) * | 2019-10-22 | 2020-02-14 | 中国电力科学研究院有限公司 | Method and system for predicting fault of intelligent electric energy meter based on machine learning |
| CN110779746B (en) * | 2019-10-24 | 2021-08-10 | 西安理工大学 | Diagnosis method for improving composite fault of deep sparse self-encoder network rotating machinery |
| CN110779746A (en) * | 2019-10-24 | 2020-02-11 | 西安理工大学 | Diagnosis method for improving composite fault of deep sparse self-encoder network rotating machinery |
| CN110910982A (en) * | 2019-11-04 | 2020-03-24 | 广州金域医学检验中心有限公司 | Self-coding model training method, device, equipment and storage medium |
| CN110956091A (en) * | 2019-11-05 | 2020-04-03 | 广州供电局有限公司 | Electric energy meter identification method and device, computer equipment and storage medium |
| CN110956091B (en) * | 2019-11-05 | 2022-07-01 | 广东电网有限责任公司广州供电局 | Electric energy meter identification method and device, computer equipment and storage medium |
| CN110995396A (en) * | 2019-12-19 | 2020-04-10 | 中国电力科学研究院有限公司 | Compression method of communication messages of electricity consumption information acquisition system based on hierarchical structure |
| CN110995396B (en) * | 2019-12-19 | 2022-01-11 | 中国电力科学研究院有限公司 | Compression method of communication messages of electricity consumption information acquisition system based on hierarchical structure |
| CN111076934A (en) * | 2019-12-24 | 2020-04-28 | 江苏大学 | Method for diagnosing potential fault of bearing based on S transformation |
| CN111506598B (en) * | 2020-04-13 | 2021-10-15 | 中国科学院自动化研究所 | Fault discrimination method, system and device based on small sample self-learning fault migration |
| CN111506598A (en) * | 2020-04-13 | 2020-08-07 | 中国科学院自动化研究所 | Fault discrimination method, system and device based on small sample self-learning fault migration |
| CN111404130A (en) * | 2020-04-14 | 2020-07-10 | 国网内蒙古东部电力有限公司赤峰供电公司 | Novel power distribution network fault detection method and fault self-healing system based on quick switch |
| CN111404130B (en) * | 2020-04-14 | 2022-03-22 | 国网内蒙古东部电力有限公司赤峰供电公司 | Novel power distribution network fault detection method and fault self-healing system based on quick switch |
| CN111860684A (en) * | 2020-07-30 | 2020-10-30 | 元神科技(杭州)有限公司 | Power plant equipment fault early warning method and system based on dual networks |
| CN111860684B (en) * | 2020-07-30 | 2023-10-27 | 元神科技(杭州)有限公司 | Power plant equipment fault early warning method and system based on double networks |
| CN112861625B (en) * | 2021-01-05 | 2023-07-04 | 深圳技术大学 | Determination method for stacked denoising self-encoder model |
| CN112861625A (en) * | 2021-01-05 | 2021-05-28 | 深圳技术大学 | Method for determining stacking denoising autoencoder model |
| CN113533995B (en) * | 2021-07-05 | 2023-10-20 | 上海电享信息科技有限公司 | Consistency detection method for power battery |
| CN113533995A (en) * | 2021-07-05 | 2021-10-22 | 上海电享信息科技有限公司 | Power battery consistency detection method |
| CN113642244A (en) * | 2021-08-23 | 2021-11-12 | 广东电网有限责任公司 | Power metering equipment fault prediction method based on artificial intelligence |
| CN113674846A (en) * | 2021-09-18 | 2021-11-19 | 浙江远图互联科技股份有限公司 | Hospital intelligent service public opinion monitoring platform based on LSTM network |
| CN114062812A (en) * | 2021-11-15 | 2022-02-18 | 国网四川省电力公司营销服务中心 | Fault diagnosis method and system for metering cabinet |
| CN114062812B (en) * | 2021-11-15 | 2024-05-07 | 国网四川省电力公司营销服务中心 | Metering cabinet fault diagnosis method and system |
| CN114189825A (en) * | 2021-12-09 | 2022-03-15 | 苏州捷布森智能科技有限公司 | Data processing method and system based on industrial Internet and intelligent manufacturing |
| CN114189825B (en) * | 2021-12-09 | 2022-10-18 | 苏州捷布森智能科技有限公司 | Data processing method and system based on industrial Internet and intelligent manufacturing |
| CN115545191A (en) * | 2022-12-01 | 2022-12-30 | 中国南方电网有限责任公司超高压输电公司广州局 | Current noise reduction network model training method and fault current limiter current noise reduction method |
| CN117272055A (en) * | 2023-11-23 | 2023-12-22 | 国网山西省电力公司营销服务中心 | Electric energy meter abnormality detection method and device based on filtering enhancement self-encoder |
| CN117272055B (en) * | 2023-11-23 | 2024-02-06 | 国网山西省电力公司营销服务中心 | Electric energy meter abnormality detection method and device based on filtering enhancement self-encoder |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110059357A (en) | A kind of intelligent electric energy meter failure modes detection method and system based on autoencoder network | |
| CN110533631B (en) | SAR image change detection method based on pyramid pooling twin network | |
| CN109765333A (en) | A kind of Diagnosis Method of Transformer Faults based on GoogleNet model | |
| CN108870091B (en) | Line leakage system and method based on low-and high-frequency hybrid detection | |
| CN112098714A (en) | ResNet-LSTM-based electricity stealing detection method and system | |
| CN108414923A (en) | A kind of analog-circuit fault diagnosis method based on the extraction of depth confidence network characterization | |
| CN109145706A (en) | A kind of sensitive features selection and dimension reduction method for analysis of vibration signal | |
| CN112036042B (en) | Power equipment abnormality detection method and system based on variational modal decomposition | |
| CN106017876A (en) | Wheel set bearing fault diagnosis method based on equally-weighted local feature sparse filter network | |
| CN109446812A (en) | A kind of embedded system firmware safety analytical method and system | |
| CN108399248A (en) | A kind of time series data prediction technique, device and equipment | |
| CN110070102B (en) | Method for establishing sequence-to-sequence model for identifying power quality disturbance type | |
| CN109858352A (en) | A kind of method for diagnosing faults based on compressed sensing and the multiple dimensioned network of improvement | |
| CN103033362A (en) | Gear fault diagnosis method based on improving multivariable predictive models | |
| CN110348150A (en) | A kind of fault detection method based on dependent probability model | |
| CN103103570B (en) | Based on the aluminium cell condition diagnostic method of pivot similarity measure | |
| CN109767054A (en) | Efficiency cloud appraisal procedure and edge efficiency gateway based on deep neural network algorithm | |
| CN108805195A (en) | A kind of motor group method for diagnosing faults based on two-value deep-neural-network | |
| CN109298225B (en) | Automatic identification model system and method for abnormal state of voltage measurement data | |
| CN105606914A (en) | IWO-ELM-based Aviation power converter fault diagnosis method | |
| CN105844055B (en) | It is fissioned-is polymerize the damage monitoring method of probabilistic model based on guided wave dynamic contract-enhanced | |
| CN111680875A (en) | Unmanned aerial vehicle state risk fuzzy comprehensive evaluation method based on probability baseline model | |
| CN103822786A (en) | Wind turbine generator set mechanical equipment state diagnosis method based on multivariate statistical analysis | |
| CN113538353B (en) | Five-phase asynchronous motor rolling bearing fault diagnosis method based on single-channel diagram data enhancement and migration training residual error network | |
| CN112507479B (en) | Oil drilling machine health state assessment method based on manifold learning and softmax |
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 |