CN107330517A - One kind is based on S_Kohonen non-intrusion type resident load recognition methods - Google Patents
One kind is based on S_Kohonen non-intrusion type resident load recognition methods Download PDFInfo
- Publication number
- CN107330517A CN107330517A CN201710448867.7A CN201710448867A CN107330517A CN 107330517 A CN107330517 A CN 107330517A CN 201710448867 A CN201710448867 A CN 201710448867A CN 107330517 A CN107330517 A CN 107330517A
- Authority
- CN
- China
- Prior art keywords
- mrow
- msub
- kohonen
- neuron
- layer
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/088—Non-supervised learning, e.g. competitive learning
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R19/00—Arrangements for measuring currents or voltages or for indicating presence or sign thereof
- G01R19/0092—Arrangements for measuring currents or voltages or for indicating presence or sign thereof measuring current only
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R23/00—Arrangements for measuring frequencies; Arrangements for analysing frequency spectra
- G01R23/16—Spectrum analysis; Fourier analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
Landscapes
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a kind of non-intrusion type resident load recognition methods based on S_Kohonen, the non-intrusion type resident load recognition methods comprises the following steps:Step one:Switching event is judged according to the change of home electrical porch active power, when occurring switching event, the electric current sample of switching event occurs for collection at family's power portal;Step 2:Frequency-domain analysis is carried out to the electric current sample collected, its frequency domain harmonic amplitude is extracted as the load characteristic of each electrical equipment, forms load characteristic storehouse;Step 3:Design is applied to the S_Kohonen neutral nets that resident load is recognized, the neuron number and the scale of competition layer of S_Kohonen neural network input layers and output layer is determined, to determine the selection mode of triumph neuron and the learning algorithm of weighed value adjusting;Step 4:Initiation parameter;Step 5:Training set is trained to S_Kohonen networks, and test set is tested;Step 6:Adjustment network parameter realizes optimum network performance.
Description
Technical field
It is more particularly to negative based on S_Kohonen non-intrusion types resident the present invention relates to network load monitoring technical field
Lotus recognition methods.
Background technology
The development of intelligent power grid technology enables power consumer to be coordinated with power network.Wherein intelligent power is intelligent electricity
One of important step of net, is the core of interactive service system.Intelligent power is realized, power consumer needs to be best understood from certainly
The power consumption characteristics of body, obtain the consumption information of electrical equipment in time.Load monitoring is the key technology for realizing intelligent power, is passed through
Load monitoring can carry out the analysis of power information to various electrical equipments, so as to guide power consumer to change consumption habit and excellent
Change electricity consumption behavior, so as to reach the purpose of energy-conservation.For power network, load monitoring can help power network to understand load composition simultaneously
The electricity consumption behavior of power consumer is grasped, planning and designing and power generation dispatching for power network provide guidance, and support two-way interaction service
It can be serviced with intelligent use.
Load monitoring includes intrusive mood load monitoring and non-intrusion type load monitoring.Wherein intrusive mood load monitoring is each
Increase checking device in electrical equipment, although this method accuracy is high, cost is big, Maintenance Difficulty.Non-intrusion type load monitoring is
At the power portal that monitoring device is arranged on to power consumer, the power information of power consumer is analyzed by monitoring algorithm, so that
The electricity consumption situation of each equipment inside power consumer is learned, hardware configuration is enormously simplify, reduces financial cost, it is adaptable to be a large amount of
Isolated user Installation Modes.
To there is recognition accuracy in the non-intrusion type load recognizer used at present not high in load monitoring, to small-power
The problem of electrical equipment, multimode the electrical equipment electrical equipment close with feature are difficult to correct identification, this has a strong impact on non-intrusion type load monitoring
Implementation and intelligent power realization.
It is desirable to have a kind of non-intrusion type resident load recognition methods to overcome or at least mitigate non-in the prior art
Intrusive mood load recognizer is the problem of recognition accuracy is low in load monitoring.
The content of the invention
It is an object of the invention to provide a kind of non-intrusion type resident load recognition methods to solve in non-intrusion type load
Household electrical appliance recognition accuracy is not high in monitoring, and low power electric appliance, multimode the electrical equipment electrical equipment close with feature are difficult to correct knowledge
Other problem.
The present invention provides a kind of non-intrusion type resident load recognition methods, comprises the following steps:
Step one:Switching event is judged according to the change of home electrical porch active power, when occurring switching event,
The electric current sample of switching event occurs for collection at family's power portal;
Step 2:Frequency-domain analysis is carried out to the electric current sample collected, its frequency domain harmonic amplitude is extracted as each electricity
The load characteristic of device, forms load characteristic storehouse;
Step 3:Design is applied to the S_Kohonen neutral nets that resident load is recognized, determines S_Kohonen nerve nets
The scale of the neuron number and competition layer of network input layer and output layer, to determine the selection mode and weights of triumph neuron
The learning algorithm of adjustment;
Step 4:Initiation parameter, wherein parameter include:Connection weight ω between input layer and competition layerij, output layer with
Connection weight ω between competition layerjk, the radius of neighbourhood r, ωijLearning rate η1And ωjkLearning rate η2;
Step 5:The load characteristic vector of each electrical equipment is as network inputs, and electrical equipment classification is exported as network, passes through training
Set pair S_Kohonen networks are trained, and training terminates to test test set sample using network, is identified result,
The recognition accuracy and overall recognition accuracy of each electrical equipment are calculated with testing characteristics of network;
Step 6:Adjust the competition layer scale of S_Kohonen neutral nets, terminate threshold value or maximum iteration, research
The relation of network parameter and network performance, selection suitable parameters realize optimum network performance.
Preferably, in the step one, voltage, electricity are carried out using the harvester installed in the home electrical porch
The collection of stream and active power value.
Preferably, in the step one, according to rule judgment method, the change of active power value surpasses at family's power portal
When crossing 20W, judge occur switching event, the steady-state current value before occurring with switching event and the stable state electricity after the generation of switching event
The difference of flow valuve as occur switching event load current sample, and record occur switching event electrical equipment species as event
Label.
Preferably, the step 2 also carries out frequency-domain analysis to each electrical equipment steady-state current for collecting, and by from
Dissipate Fourier expansion and extract each electrical equipment frequency domain character, each harmonic in a current cycle is projected in frequency domain, width is taken
It is worth larger harmonic wave as the load characteristic of each electrical equipment.
Preferably, in the step 3, the S_Kohonen neutral nets include input layer, competition layer and output layer;Institute
The dimension that input layer number is the load characteristic is stated, the competition layer neuron is arranged for two-dimensional array, according to negative
Lotus clusters number determines the scale of the S_Kohonen neutral nets, and the output layer neuron number is to wait to know in resident family
Other Overload Class number;The selection mode of the competition layer triumph neuron is Euclidean distance method, that is, calculates input vector X=
(x1,x2,…xm) with the distance between competition layer neuron j dj, m is the dimension of the load characteristic vector of input, distance meter
Calculating formula is:
The selection competition layer neuron c minimum with input vector X distances realize input vector X as triumph neuron
With the mapping between competition layer neuron c;In addition, the weights of adjustment triumph neuron and its surrounding neighbors neuron, neighborhood rule
It is set to:
Nc(j)=(j | find (norm (posj,posc) < r))
J=1,2 ..., l
posj、poscRespectively neuron j and c position;Norm is to calculate the Euclidean distance between two neurons;R is
The radius of neighbourhood of selection, l is the number of competition layer neuron;Optimum Matching neuron c and its neighborhood Nc(j) in neuron with it is defeated
Enter the weights ω between layerijAccording to XiAdjustment, gradually tends to Xi, and with the weights ω of output layerjkThen according to load desired output
YkAdjustment:
ωij=ωij+y(j)*η1(Xi-ωij)
ωjk=ωjk+y(j)*η2(Yk-ωjk)
Wherein y (j) is the learning algorithm of weighed value adjusting, herein using cook's cap learning function:
Neuron i.e. in radius r neighborhoods adjusts weights with triumph neuron using same way, and beyond neighborhood
Neuron does not adjust weights, and neuron c and its neighborhood N is adjusted respectivelyc(j) power between the node and input layer that are included in
Coefficient ωijAnd the weight coefficient ω between output node layerjk;η1And η2It is weights ω respectivelyijAnd ωjkLearning rate;r
And η1Decline, η with the increase of evolution number of times2Increase with the increase of evolution number of times, wherein n is iterations, the change of three
Rule difference is as follows:
Preferably, in the step 4, ωijIt is initialized as random number, ωjkIt is initialized as 0, η1max, η1min, η2maxWith
η2minValue 0~1, radius of neighbourhood r is selected according to the competition layer scale, rminValue 0~1.
Preferably, in the step 5, the network of design is trained with training set sample, when overall recognition accuracy
Deconditioning when variable quantity is less than threshold value or reaches iterations;Survey is identified to test set sample with the network trained
Examination, examines each electrical equipment recognition accuracy and overall recognition accuracy, and accuracy rate is defined as follows:
The single electrical equipment recognition accuracy=electric appliances correctly recognize the number/electric appliances number of samples,
Overall recognition accuracy=all electrical equipment correctly recognize number/total number of samples of all electrical equipment.
Preferably, in the step 6, the relation of the network parameter and network performance is network under different parameters
Training time, individual event recognition time and overall recognition accuracy, to improve operating efficiency, realize load on-line monitoring selection
Suitable parameter setting.
The non-intrusion type resident load recognition methods of the present invention carries out discrete Fourier transform to electrical equipment steady state operating current,
Frequency domain harmonic current amplitude is extracted as load characteristic, the requirement to harvester sample frequency is reduced, and then reduce hardware
The investment of equipment.Load characteristic is projected to two dimension competition layer plane by the non-intrusion type resident load recognition methods of the present invention, and
According to weighed value adjusting algorithm the class inner region that load characteristic projects to competition layer is constantly reduced, while gradually expanding class and class
The distance between so that the load characteristics clustering region with similar features is gradually distance from.The present invention can improve the knowledge of similar electrical equipment
Other accuracy rate, while making neuron in each type load region have identical output by adjusting output layer weights, makes load
Classification has fault-tolerance, that is, allows that of a sort load characteristic has slight fluctuations, therefore just can occur small echo in line voltage
It can still realize that load is accurately identified, and strengthens the practicality of the algorithm when dynamic.The present invention can be by the different work shape of electrical equipment
State projects to different regions, the correspondence identical output of these regions, can realize the identification to multimode electrical equipment.
In summary, non-intrusion type resident load recognition methods of the present invention has good stability, fault-tolerance and practicality
Property, can solve the problem that household electrical appliance recognition accuracy is not high in non-intrusion type load monitoring, low power electric appliance, multimode electrical equipment and
The problem of close electrical equipment of feature is difficult to correct identification, effective identification is realized to resident's electrical equipment, residential power is can be widely applied to
In load identification.
Brief description of the drawings
Fig. 1 is S_Kohonen neural network model figures.
Fig. 2 is the flow chart of non-intrusion type load recognition methods proposed by the present invention.
Fig. 3 is the competition layer cluster schematic diagram of the close electrical equipment of feature in the embodiment of the present invention.
Fig. 4 is training set sample competition layer cluster schematic diagram in the embodiment of the present invention.
Fig. 5 is test set sample competition layer cluster schematic diagram in the embodiment of the present invention.
Embodiment
To make the purpose, technical scheme and advantage of the invention implemented clearer, below in conjunction with the embodiment of the present invention
Accompanying drawing, the technical scheme in the embodiment of the present invention is further described in more detail.In the accompanying drawings, identical from beginning to end or class
As label represent same or similar element or the element with same or like function.Described embodiment is the present invention
A part of embodiment, rather than whole embodiments.The embodiments described below with reference to the accompanying drawings are exemplary, it is intended to uses
It is of the invention in explaining, and be not considered as limiting the invention.Based on the embodiment in the present invention, ordinary skill people
The every other embodiment that member is obtained under the premise of creative work is not made, belongs to the scope of protection of the invention.Under
Embodiments of the invention are described in detail with reference to accompanying drawing for face.
As shown in Figure 1-2, the present invention provides a kind of non-intrusion type resident load recognition methods based on S_Kohonen, bag
Include following steps:
Step one:Switching event is judged according to the change of home electrical porch active power, when occurring switching event,
The electric current sample of switching event occurs for collection at family's power portal, using installed in the home electrical porch
Harvester carries out the collection of voltage, electric current and active power value, according to rule judgment method, the wattful power at family's power portal
When the change of rate value is more than 20W, judge occur switching event, the steady-state current value before occurring with switching event is sent out with switching event
The difference of steady-state current value after life as occur switching event load current sample, and record occur switching event electrical equipment
Species is used as event tag;
Step 2:Frequency-domain analysis is carried out to each electric current sample collected, while steady to each electrical equipment collected
State electric current carries out frequency-domain analysis, each electrical equipment frequency domain character is extracted by discrete Fourier series expansion, by a current cycle
Each harmonic is projected in frequency domain, is taken the larger harmonic wave of amplitude as load characteristic, is formed the load characteristic storehouse of each electrical equipment;
Step 3:Design is applied to the S_Kohonen neutral nets that resident load is recognized, determines S_Kohonen nerve nets
The scale of the neuron number and competition layer of network input layer and output layer, to determine the selection mode and weights of triumph neuron
The learning algorithm of adjustment, the S_Kohonen neutral nets include input layer, competition layer and output layer;The input layer nerve
First number is the dimension of the load characteristic, and the competition layer neuron is arranged for two-dimensional array, true according to load characteristics clustering number
The scale of the fixed S_Kohonen neutral nets, the output layer neuron number is Overload Class to be identified in resident family
Number;The selection mode of the competition layer triumph neuron is Euclidean distance method, that is, calculates input vector X=(x1,x2,…xm)
With the distance between competition layer neuron j dj, m is the dimension of the load characteristic vector of input, is apart from calculation formula:
The selection competition layer neuron c minimum with input vector X distances realize input vector X as triumph neuron
With the mapping between competition layer neuron c;In addition, the weights of adjustment triumph neuron and its surrounding neighbors neuron, neighborhood rule
It is set to:
Nc(j)=(j | find (norm (posj,posc) < r))
J=1,2 ..., l
posj、poscRespectively neuron j and c position;Norm is to calculate the Euclidean distance between two neurons;R is
The radius of neighbourhood of selection, l is the number of competition layer neuron;Optimum Matching neuron c and its neighborhood Nc(j) in neuron with it is defeated
Enter the weights ω between layerijAccording to XiAdjustment, gradually tends to Xi, and with the weights ω of output layerjkThen according to load desired output
YkAdjustment:
ωij=ωij+y(j)*η1(Xi-ωij)
ωjk=ωjk+y(j)*η2(Yk-ωjk)
Wherein y (j) is the learning algorithm of weighed value adjusting, herein using cook's cap learning function:
Neuron i.e. in radius r neighborhoods adjusts weights with triumph neuron using same way, and beyond neighborhood
Neuron does not adjust weights, and neuron c and its neighborhood N is adjusted respectivelyc(j) power between the node and input layer that are included in
Coefficient ωijAnd the weight coefficient ω between output node layerjk;η1And η2It is weights ω respectivelyijAnd ωjkLearning rate;r
And η1Decline, η with the increase of evolution number of times2Increase with the increase of evolution number of times, wherein n is iterations, the change of three
Rule difference is as follows:
Step 4:Initiation parameter, wherein parameter include:Connection weight ω between input layer and competition layerij, output layer with
Connection weight ω between competition layerjk, the radius of neighbourhood r, ωijLearning rate η1And ωjkLearning rate η2, ωijIt is initialized as
Random number, ωjkIt is initialized as 0, η1max, η1min, η2maxAnd η2minValue 0~1, radius of neighbourhood r enters according to the competition layer scale
Row selection, rminValue 0~1;
Step 5:The load characteristic vector of each electrical equipment is as network inputs, and electrical equipment classification is exported as network, passes through training
Set pair S_Kohonen networks are trained, and training terminates to test test set sample using network, is identified result,
The recognition accuracy and overall recognition accuracy of each electrical equipment are calculated with testing characteristics of network, with net of the training set sample to design
Network is trained, the deconditioning when overall recognition accuracy variable quantity is less than threshold value or reaches iterations;With training
Network test set sample is identified test, examine each electrical equipment recognition accuracy and overall recognition accuracy, accurate calibration
Justice is as follows:
The single electrical equipment recognition accuracy=electric appliances correctly recognize the number/electric appliances number of samples,
Overall recognition accuracy=all electrical equipment correctly recognize number/total number of samples of all electrical equipment;
Step 6:Adjust the competition layer scale of S_Kohonen neutral nets, terminate threshold value or maximum iteration, research
The relation of network parameter and network performance, selection suitable parameters realize optimum network performance, the network parameter and network performance
Relation be training time of the network under different parameters, individual event recognition time and overall recognition accuracy, to improve work
Make efficiency, realize the suitable parameter setting of load on-line monitoring selection.
One embodiment of the invention:
As in Figure 3-5, emulation testing, the database are carried out using BLUED databases disclosed in Carnegie Mellon University
Using 12kHz sample frequency, continuous acquisition one week voltage, electric current, power etc. at the electricity port of 1 American family
Data, and marked the switching moment of each electric equipment.Using the difference of steady-state current at each switching event front and back end portses as
Switching load current sample, each switching event extracts 100 load current samples, and according to voltage correction current waveform phase
Position.Training data is used as training set using 500 load switching events, altogether 50,000 load current samples;Test data is used
There is other 232 loads switching event, 2.32 ten thousand load current samples are used as test set altogether in database.Training set and
Test set is comprising 7 kinds of electrical equipment classifications.Refrigerator, backyard lamp, bathroom ceiling light and bedroom lamp are low power electric appliance in 7 kinds of electrical equipment;
Kitchen chopper, refrigerator etc. have a variety of working conditions;Certain state and bathroom of kitchen chopper and hair-dryer and refrigerator
Ceiling light is the close electrical equipment of two groups of features, with close current waveform and load characteristic.
The step of according to above-mentioned non-intrusion type resident load recognition methods based on S_Kohonen, to all types of in BLUED
Load switching event is identified, in S_Kohonen network training process, given parameters η1min=0.01, η1max=0.1;
η2min=0.5, η2max=1;rmax=5, rmin=0.2;It is 50 × 50 to take competition layer scale;Iteration 10 times, iteration is right each time
50000 samples are trained;After the completion of training 23200 load samples in test set are identified with test, identification is calculated
Accuracy rate, test result is as shown in table 1:
The S_Kohonen neutral nets of table 1 are to household electrical appliance recognition result
As it can be seen from table 1 the inventive method is all very high to the recognition accuracy of all types of loads, can be to small-power
Electrical equipment, multimode the electrical equipment electrical equipment close with feature realize high accuracy identification.
In addition, sample iterations and competition layer scale during the training of adjustment S_Kohonen networks, studies it to S_
The influence of Kohonen network performances.
The relation of the overall accuracy of table 2 and iterations
Iterations | 1 | 2 | 3 | 4 | 5 |
Overall accuracy | 16.20% | 99.62% | 99.78% | 99.86% | 99.92% |
Iterations | 6 | 7 | 8 | 9 | 10 |
Overall accuracy | 99.89% | 99.96% | 99.92% | 99.89% | 99.97% |
From Table 2, it can be seen that overall recognition accuracy is just very high in the 2nd iteration, it is total after the 5th iteration
Body accuracy rate is just basically stable at 99.9% or so, and network stabilization performance is good.
The network performance of table 3 and the relation of competition layer scale
Competition layer scale is studied from 10 × 10 to 80 × 80 two-dimensional arrays, from table 3 it can be seen that with competition layer
Scale increases, and net training time, individual event recognition time and overall accuracy are all in increase.Competition layer scale 40 ×
After 40, overall recognition accuracy is basically stable near 99.97%, and the load recognition capability of network basically reaches stabilization, and
Training time and individual event recognition time continue to increase.By studying network performance and iterations and the pass of competition layer scale
System, adjust optimal parameter, can make network combination property reach it is best.
High accuracy identification can be realized to resident load using the present invention, solved at present to low power electric appliance, many work
The problem of state electrical equipment and the close electrical equipment of feature can not be accurately identified.Appropriate adjustment network parameter, is applied to network performance negative
Lotus is monitored on-line.
It is last it is to be noted that:The above embodiments are merely illustrative of the technical solutions of the present invention, rather than its limitations.To the greatest extent
The present invention is described in detail with reference to the foregoing embodiments for pipe, it will be understood by those within the art that:It is still
Technical scheme described in foregoing embodiments can be modified, or which part technical characteristic is equally replaced
Change;And these modifications or replacement, the essence of appropriate technical solution is departed from the essence of various embodiments of the present invention technical scheme
God and scope.
Claims (8)
1. a kind of non-intrusion type resident load recognition methods based on S_Kohonen, it is characterised in that comprise the following steps:
Step one:Switching event is judged according to the change of home electrical porch active power, when occurring switching event, is in
The electric current sample of switching event occurs for collection at the power portal of front yard;
Step 2:Frequency-domain analysis is carried out to the electric current sample collected, its frequency domain harmonic amplitude is extracted as each electrical equipment
Load characteristic, forms load characteristic storehouse;
Step 3:Design is applied to the S_Kohonen neutral nets that resident load is recognized, determines that S_Kohonen neutral nets are defeated
Enter the neuron number and the scale of competition layer of layer and output layer, to determine the selection mode and weighed value adjusting of triumph neuron
Learning algorithm;
Step 4:Initiation parameter, wherein parameter include:Connection weight ω between input layer and competition layerij, output layer and competition
Connection weight ω between layerjk, the radius of neighbourhood r, ωijLearning rate η1And ωjkLearning rate η2;
Step 5:The load characteristic vector of each electrical equipment is as network inputs, and electrical equipment classification is exported as network, passes through training set pair
S_Kohonen networks are trained, and training terminates to test test set sample using network, is identified result, calculates
The recognition accuracy of each electrical equipment and overall recognition accuracy are with testing characteristics of network;
Step 6:Adjust the competition layer scale of S_Kohonen neutral nets, terminate threshold value or maximum iteration, study network
The relation of parameter and network performance, selection suitable parameters realize optimum network performance.
2. the non-intrusion type resident load recognition methods as claimed in claim 1 based on S_Kohonen, it is characterised in that:Institute
State in step one, voltage, electric current and active power value are carried out using the harvester installed in the home electrical porch
Collection.
3. the non-intrusion type resident load recognition methods as claimed in claim 2 based on S_Kohonen, it is characterised in that:Institute
State in step one, according to rule judgment method, when the change of active power value is more than 20W at family's power portal, judge occur
Switching event, the difference of the steady-state current value after steady-state current value and the generation of switching event before occurring using switching event is used as generation
The current sample of the load of switching event, and the electrical equipment species for occurring switching event is recorded as event tag.
4. the non-intrusion type resident load recognition methods as claimed in claim 1 based on S_Kohonen, it is characterised in that:Institute
State step 2 and frequency-domain analysis is also carried out to each electrical equipment steady-state current collected, and carried by discrete Fourier series expansion
Each electrical equipment frequency domain character is taken, each harmonic in a current cycle is projected in frequency domain, the larger harmonic wave of amplitude is taken as institute
State the load characteristic of each electrical equipment.
5. the non-intrusion type resident load recognition methods as claimed in claim 1 based on S_Kohonen, it is characterised in that:Institute
State in step 3, the S_Kohonen neutral nets include input layer, competition layer and output layer;The input layer
Number is the dimension of the load characteristic, and the competition layer neuron is two-dimensional array arrangement, and institute is determined according to load characteristics clustering number
The scale of S_Kohonen neutral nets is stated, the output layer neuron number is Overload Class number to be identified in resident family;
The selection mode of the competition layer triumph neuron is Euclidean distance method, that is, calculates input vector X=(x1,x2,…xm) with competing
The distance between layer neuron j dj, m is the dimension of the load characteristic vector of input, is apart from calculation formula:
<mfenced open = "" close = "">
<mtable>
<mtr>
<mtd>
<mrow>
<msub>
<mi>d</mi>
<mi>j</mi>
</msub>
<mo>=</mo>
<msqrt>
<mrow>
<munderover>
<mo>&Sigma;</mo>
<mrow>
<mi>i</mi>
<mo>=</mo>
<mn>1</mn>
</mrow>
<mi>m</mi>
</munderover>
<msup>
<mrow>
<mo>(</mo>
<msub>
<mi>x</mi>
<mi>i</mi>
</msub>
<mo>-</mo>
<msub>
<mi>&omega;</mi>
<mrow>
<mi>i</mi>
<mi>j</mi>
</mrow>
</msub>
<mo>)</mo>
</mrow>
<mn>2</mn>
</msup>
</mrow>
</msqrt>
</mrow>
</mtd>
<mtd>
<mrow>
<mi>j</mi>
<mo>=</mo>
<mn>1</mn>
<mo>,</mo>
<mn>2</mn>
<mo>,</mo>
<mn>...</mn>
<mo>,</mo>
<mi>l</mi>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
The selection competition layer neuron c minimum with input vector X distances as triumph neuron, that is, realize input vector X with it is competing
The mapping striven between layer neuron c;In addition, the weights of adjustment triumph neuron and its surrounding neighbors neuron, neighborhood is defined as:
Nc(j)=(j | find (norm (posj,posc) < r))
J=1,2 ..., l
posj、poscRespectively neuron j and c position;Norm is to calculate the Euclidean distance between two neurons;R is selection
The radius of neighbourhood, l be competition layer neuron number;Optimum Matching neuron c and its neighborhood Nc(j) neuron and input layer in
Between weights ωijAccording to XiAdjustment, gradually tends to Xi, and with the weights ω of output layerjkThen according to load desired output YkAdjust
It is whole:
ωij=ωij+y(j)*η1(Xi-ωij)
ωjk=ωjk+y(j)*η2(Yk-ωjk)
Wherein y (j) is the learning algorithm of weighed value adjusting, herein using cook's cap learning function:
<mrow>
<mi>y</mi>
<mrow>
<mo>(</mo>
<mi>j</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<mfenced open = "{" close = "">
<mtable>
<mtr>
<mtd>
<mrow>
<mn>1</mn>
<mo>,</mo>
</mrow>
</mtd>
<mtd>
<mrow>
<mi>j</mi>
<mo>&Element;</mo>
<msub>
<mi>N</mi>
<mi>c</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>j</mi>
<mo>)</mo>
</mrow>
</mrow>
</mtd>
</mtr>
<mtr>
<mtd>
<mrow>
<mn>0</mn>
<mo>,</mo>
</mrow>
</mtd>
<mtd>
<mrow>
<mi>j</mi>
<mo>&NotElement;</mo>
<msub>
<mi>N</mi>
<mi>c</mi>
</msub>
<mrow>
<mo>(</mo>
<mi>j</mi>
<mo>)</mo>
</mrow>
</mrow>
</mtd>
</mtr>
</mtable>
</mfenced>
</mrow>
Neuron i.e. in radius r neighborhoods and triumph neuron adjust weights using same way, and the nerve beyond neighborhood
Member does not adjust weights, and neuron c and its neighborhood N is adjusted respectivelyc(j) weight coefficient between the node and input layer that are included in
ωijAnd the weight coefficient ω between output node layerjk;η1And η2It is weights ω respectivelyijAnd ωjkLearning rate;R and η1With
The increase of evolution number of times and decline, η2Increase with the increase of evolution number of times, wherein n is iterations, the changing rule of three
It is as follows respectively:
<mrow>
<msub>
<mi>&eta;</mi>
<mn>1</mn>
</msub>
<mo>=</mo>
<msub>
<mi>&eta;</mi>
<mrow>
<mn>1</mn>
<mi>m</mi>
<mi>a</mi>
<mi>x</mi>
</mrow>
</msub>
<mo>-</mo>
<mfrac>
<mrow>
<mi>n</mi>
<mo>*</mo>
<mrow>
<mo>(</mo>
<msub>
<mi>&eta;</mi>
<mrow>
<mn>1</mn>
<mi>m</mi>
<mi>a</mi>
<mi>x</mi>
</mrow>
</msub>
<mo>-</mo>
<msub>
<mi>&eta;</mi>
<mrow>
<mn>1</mn>
<mi>m</mi>
<mi>i</mi>
<mi>n</mi>
</mrow>
</msub>
<mo>)</mo>
</mrow>
</mrow>
<mi>N</mi>
</mfrac>
</mrow>
<mrow>
<msub>
<mi>&eta;</mi>
<mn>2</mn>
</msub>
<mo>=</mo>
<msub>
<mi>&eta;</mi>
<mrow>
<mn>2</mn>
<mi>m</mi>
<mi>i</mi>
<mi>n</mi>
</mrow>
</msub>
<mo>+</mo>
<mfrac>
<mrow>
<mi>n</mi>
<mo>*</mo>
<mrow>
<mo>(</mo>
<msub>
<mi>&eta;</mi>
<mrow>
<mn>2</mn>
<mi>m</mi>
<mi>a</mi>
<mi>x</mi>
</mrow>
</msub>
<mo>-</mo>
<msub>
<mi>&eta;</mi>
<mrow>
<mn>2</mn>
<mi>m</mi>
<mi>i</mi>
<mi>n</mi>
</mrow>
</msub>
<mo>)</mo>
</mrow>
</mrow>
<mi>N</mi>
</mfrac>
</mrow>
<mrow>
<mi>r</mi>
<mo>=</mo>
<msub>
<mi>r</mi>
<mrow>
<mi>m</mi>
<mi>a</mi>
<mi>x</mi>
</mrow>
</msub>
<mo>-</mo>
<mfrac>
<mrow>
<mi>n</mi>
<mo>*</mo>
<mrow>
<mo>(</mo>
<msub>
<mi>r</mi>
<mrow>
<mi>m</mi>
<mi>a</mi>
<mi>x</mi>
</mrow>
</msub>
<mo>-</mo>
<msub>
<mi>r</mi>
<mrow>
<mi>m</mi>
<mi>i</mi>
<mi>n</mi>
</mrow>
</msub>
<mo>)</mo>
</mrow>
</mrow>
<mi>N</mi>
</mfrac>
</mrow>
。
6. the non-intrusion type resident load recognition methods as claimed in claim 1 based on S_Kohonen, it is characterised in that:Institute
State in step 4, ωijIt is initialized as random number, ωjkIt is initialized as 0, η1max, η1min, η2maxAnd η2minValue 0~1, neighborhood half
Footpath r is selected according to the competition layer scale, rminValue 0~1.
7. the non-intrusion type resident load recognition methods as claimed in claim 1 based on S_Kohonen, it is characterised in that:Institute
State in step 5, the network of design be trained with training set sample, when overall recognition accuracy variable quantity be less than threshold value or
Deconditioning when person reaches iterations;Test is identified to test set sample with the network trained, examines each electrical equipment to know
Other accuracy rate and overall recognition accuracy, accuracy rate are defined as follows:
The single electrical equipment recognition accuracy=electric appliances correctly recognize the number/electric appliances number of samples,
Overall recognition accuracy=all electrical equipment correctly recognize number/total number of samples of all electrical equipment.
8. the non-intrusion type resident load recognition methods as claimed in claim 1 based on S_Kohonen, it is characterised in that:Institute
State in step 6, training time that the relation of the network parameter and network performance is network under different parameters, individual event
Recognition time and overall recognition accuracy, to improve operating efficiency, realize the suitable parameter setting of load on-line monitoring selection.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710448867.7A CN107330517B (en) | 2017-06-14 | 2017-06-14 | Non-invasive resident load identification method based on S _ Kohonen |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710448867.7A CN107330517B (en) | 2017-06-14 | 2017-06-14 | Non-invasive resident load identification method based on S _ Kohonen |
Publications (2)
Publication Number | Publication Date |
---|---|
CN107330517A true CN107330517A (en) | 2017-11-07 |
CN107330517B CN107330517B (en) | 2021-08-06 |
Family
ID=60194732
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710448867.7A Active CN107330517B (en) | 2017-06-14 | 2017-06-14 | Non-invasive resident load identification method based on S _ Kohonen |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107330517B (en) |
Cited By (29)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108333423A (en) * | 2017-12-27 | 2018-07-27 | 安徽机电职业技术学院 | Non-intrusion type residential power load testing method |
CN108399221A (en) * | 2018-02-11 | 2018-08-14 | 山东建筑大学 | Indoor electric equipment classifying identification method and system based on big data association analysis |
CN108429254A (en) * | 2018-03-07 | 2018-08-21 | 国网四川省电力公司电力科学研究院 | A kind of non-intrusion type power load recognition methods |
CN108537385A (en) * | 2018-04-12 | 2018-09-14 | 广东电网有限责任公司 | A kind of non-intrusion type residential electricity consumption load recognition methods |
CN108664876A (en) * | 2018-03-05 | 2018-10-16 | 佛山科学技术学院 | A kind of recognition methods of electrical appliance |
CN109165604A (en) * | 2018-08-28 | 2019-01-08 | 四川大学 | The recognition methods of non-intrusion type load and its test macro based on coorinated training |
CN109617067A (en) * | 2019-01-16 | 2019-04-12 | 江苏圣通电力新能源科技有限公司 | Non-intrusive electrical load recognition methods neural network based |
CN109934303A (en) * | 2019-03-25 | 2019-06-25 | 宁夏隆基宁光仪表股份有限公司 | A kind of non-invasive household electrical appliance load recognition methods, device and storage medium |
CN109993424A (en) * | 2019-03-26 | 2019-07-09 | 广东艾胜物联网科技有限公司 | A kind of non-interfering formula load decomposition method based on width learning algorithm |
CN110119816A (en) * | 2019-04-15 | 2019-08-13 | 东南大学 | A kind of load characteristic self-learning method suitable for non-intrusion type electric power monitoring |
CN110569876A (en) * | 2019-08-07 | 2019-12-13 | 武汉中原电子信息有限公司 | Non-invasive load identification method and device and computing equipment |
CN110751385A (en) * | 2019-10-08 | 2020-02-04 | 威胜集团有限公司 | Non-invasive load identification method, terminal device and storage medium |
CN110889465A (en) * | 2019-12-12 | 2020-03-17 | 武汉大学 | Power demand side equipment identification method and system based on self-adaptive resonant network |
CN110956220A (en) * | 2019-12-11 | 2020-04-03 | 深圳市活力天汇科技股份有限公司 | Non-invasive household appliance load identification method |
CN110991263A (en) * | 2019-11-12 | 2020-04-10 | 华中科技大学 | Non-invasive load identification method and system for resisting background load interference |
CN111027408A (en) * | 2019-11-19 | 2020-04-17 | 广西电网有限责任公司电力科学研究院 | Load identification method based on support vector machine and V-I curve characteristics |
CN111158239A (en) * | 2020-01-09 | 2020-05-15 | 上海电力大学 | Association rule algorithm and neural network-based pulverizing system performance optimization method |
CN111461189A (en) * | 2020-03-23 | 2020-07-28 | 东南大学 | Resident rhythm detection method based on non-invasive measurement technology |
CN111476298A (en) * | 2020-04-07 | 2020-07-31 | 杭州国彪超声设备有限公司 | Power load state identification method in home and office environment |
CN111563827A (en) * | 2020-04-10 | 2020-08-21 | 杭州电子科技大学 | Load decomposition method based on electrical appliance physical characteristics and residential electricity consumption behaviors |
CN111614159A (en) * | 2020-05-12 | 2020-09-01 | 国网河北省电力有限公司电力科学研究院 | Non-invasive household load identification method for demand side management |
CN112016587A (en) * | 2020-07-09 | 2020-12-01 | 冯燕钧 | Energy consumption monitoring cloud collaborative non-invasive identification method based on master station feature library technology |
CN112180193A (en) * | 2020-09-28 | 2021-01-05 | 华中科技大学 | Non-invasive load identification system and method based on track image identification |
CN112396098A (en) * | 2020-11-12 | 2021-02-23 | 天津天大求实电力新技术股份有限公司 | Non-embedded apartment electrical appliance load identification and analysis method, system and application |
CN113033775A (en) * | 2021-03-10 | 2021-06-25 | 南方电网数字电网研究院有限公司 | Non-invasive load identification network architecture based on supervised learning |
CN113554483A (en) * | 2021-07-19 | 2021-10-26 | 浙江工企信息技术股份有限公司 | Order full life cycle management method and system |
CN113723479A (en) * | 2021-08-18 | 2021-11-30 | 南京工程学院 | Non-invasive load identification method based on GRNN and mean shift algorithm |
CN114184870A (en) * | 2021-12-14 | 2022-03-15 | 河北科技大学 | Non-invasive load identification method and equipment |
CN115456034A (en) * | 2022-11-09 | 2022-12-09 | 广东浩迪创新科技有限公司 | Automatic identification and monitoring method and system for electric bicycle charging |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2016079229A1 (en) * | 2014-11-21 | 2016-05-26 | Universite Libre De Bruxelles | Improved non-intrusive appliance load monitoring method and device |
CN106093630A (en) * | 2016-06-02 | 2016-11-09 | 华北电力大学 | A kind of non-intrusion type household electrical appliance discrimination method |
-
2017
- 2017-06-14 CN CN201710448867.7A patent/CN107330517B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2016079229A1 (en) * | 2014-11-21 | 2016-05-26 | Universite Libre De Bruxelles | Improved non-intrusive appliance load monitoring method and device |
CN106093630A (en) * | 2016-06-02 | 2016-11-09 | 华北电力大学 | A kind of non-intrusion type household electrical appliance discrimination method |
Non-Patent Citations (3)
Title |
---|
李秋硕 等: "《用于NILM的电器设备谐波特征研究》", 《南方电网技术》 * |
祁兵 等: "《基于Fisher有监督判别的非侵入式居民负荷辨识方法》", 《电网技术》 * |
郭欣 等: "《基于有监督Kohonen神经网络的步态识别》", 《自动化学报》 * |
Cited By (38)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108333423A (en) * | 2017-12-27 | 2018-07-27 | 安徽机电职业技术学院 | Non-intrusion type residential power load testing method |
CN108399221A (en) * | 2018-02-11 | 2018-08-14 | 山东建筑大学 | Indoor electric equipment classifying identification method and system based on big data association analysis |
CN108664876A (en) * | 2018-03-05 | 2018-10-16 | 佛山科学技术学院 | A kind of recognition methods of electrical appliance |
CN108429254A (en) * | 2018-03-07 | 2018-08-21 | 国网四川省电力公司电力科学研究院 | A kind of non-intrusion type power load recognition methods |
CN108537385A (en) * | 2018-04-12 | 2018-09-14 | 广东电网有限责任公司 | A kind of non-intrusion type residential electricity consumption load recognition methods |
CN109165604A (en) * | 2018-08-28 | 2019-01-08 | 四川大学 | The recognition methods of non-intrusion type load and its test macro based on coorinated training |
CN109617067A (en) * | 2019-01-16 | 2019-04-12 | 江苏圣通电力新能源科技有限公司 | Non-intrusive electrical load recognition methods neural network based |
CN109934303A (en) * | 2019-03-25 | 2019-06-25 | 宁夏隆基宁光仪表股份有限公司 | A kind of non-invasive household electrical appliance load recognition methods, device and storage medium |
CN109993424A (en) * | 2019-03-26 | 2019-07-09 | 广东艾胜物联网科技有限公司 | A kind of non-interfering formula load decomposition method based on width learning algorithm |
CN109993424B (en) * | 2019-03-26 | 2023-06-23 | 广东艾胜物联网科技有限公司 | Non-interference type load decomposition method based on width learning algorithm |
CN110119816A (en) * | 2019-04-15 | 2019-08-13 | 东南大学 | A kind of load characteristic self-learning method suitable for non-intrusion type electric power monitoring |
CN110569876A (en) * | 2019-08-07 | 2019-12-13 | 武汉中原电子信息有限公司 | Non-invasive load identification method and device and computing equipment |
CN110751385A (en) * | 2019-10-08 | 2020-02-04 | 威胜集团有限公司 | Non-invasive load identification method, terminal device and storage medium |
CN110991263B (en) * | 2019-11-12 | 2022-03-18 | 华中科技大学 | Non-invasive load identification method and system for resisting background load interference |
CN110991263A (en) * | 2019-11-12 | 2020-04-10 | 华中科技大学 | Non-invasive load identification method and system for resisting background load interference |
CN111027408A (en) * | 2019-11-19 | 2020-04-17 | 广西电网有限责任公司电力科学研究院 | Load identification method based on support vector machine and V-I curve characteristics |
CN110956220A (en) * | 2019-12-11 | 2020-04-03 | 深圳市活力天汇科技股份有限公司 | Non-invasive household appliance load identification method |
CN110956220B (en) * | 2019-12-11 | 2023-10-31 | 深圳市活力天汇科技股份有限公司 | Non-invasive household appliance load identification method |
CN110889465A (en) * | 2019-12-12 | 2020-03-17 | 武汉大学 | Power demand side equipment identification method and system based on self-adaptive resonant network |
CN110889465B (en) * | 2019-12-12 | 2020-11-03 | 武汉大学 | Power demand side equipment identification method and system based on self-adaptive resonant network |
CN111158239A (en) * | 2020-01-09 | 2020-05-15 | 上海电力大学 | Association rule algorithm and neural network-based pulverizing system performance optimization method |
CN111158239B (en) * | 2020-01-09 | 2022-06-21 | 上海电力大学 | Association rule algorithm and neural network-based pulverizing system performance optimization method |
CN111461189A (en) * | 2020-03-23 | 2020-07-28 | 东南大学 | Resident rhythm detection method based on non-invasive measurement technology |
CN111461189B (en) * | 2020-03-23 | 2022-11-15 | 东南大学 | Resident rhythm detection method based on non-invasive measurement technology |
CN111476298A (en) * | 2020-04-07 | 2020-07-31 | 杭州国彪超声设备有限公司 | Power load state identification method in home and office environment |
CN111563827A (en) * | 2020-04-10 | 2020-08-21 | 杭州电子科技大学 | Load decomposition method based on electrical appliance physical characteristics and residential electricity consumption behaviors |
CN111563827B (en) * | 2020-04-10 | 2023-08-04 | 杭州电子科技大学 | Load decomposition method based on physical characteristics of electrical appliance and electricity utilization behaviors of residents |
CN111614159A (en) * | 2020-05-12 | 2020-09-01 | 国网河北省电力有限公司电力科学研究院 | Non-invasive household load identification method for demand side management |
CN112016587A (en) * | 2020-07-09 | 2020-12-01 | 冯燕钧 | Energy consumption monitoring cloud collaborative non-invasive identification method based on master station feature library technology |
CN112180193A (en) * | 2020-09-28 | 2021-01-05 | 华中科技大学 | Non-invasive load identification system and method based on track image identification |
CN112180193B (en) * | 2020-09-28 | 2021-11-02 | 华中科技大学 | Non-invasive load identification system and method based on track image identification |
CN112396098A (en) * | 2020-11-12 | 2021-02-23 | 天津天大求实电力新技术股份有限公司 | Non-embedded apartment electrical appliance load identification and analysis method, system and application |
CN113033775A (en) * | 2021-03-10 | 2021-06-25 | 南方电网数字电网研究院有限公司 | Non-invasive load identification network architecture based on supervised learning |
CN113033775B (en) * | 2021-03-10 | 2023-08-18 | 南方电网数字电网研究院有限公司 | Non-invasive load identification network architecture based on supervised learning |
CN113554483A (en) * | 2021-07-19 | 2021-10-26 | 浙江工企信息技术股份有限公司 | Order full life cycle management method and system |
CN113723479A (en) * | 2021-08-18 | 2021-11-30 | 南京工程学院 | Non-invasive load identification method based on GRNN and mean shift algorithm |
CN114184870A (en) * | 2021-12-14 | 2022-03-15 | 河北科技大学 | Non-invasive load identification method and equipment |
CN115456034A (en) * | 2022-11-09 | 2022-12-09 | 广东浩迪创新科技有限公司 | Automatic identification and monitoring method and system for electric bicycle charging |
Also Published As
Publication number | Publication date |
---|---|
CN107330517B (en) | 2021-08-06 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107330517A (en) | One kind is based on S_Kohonen non-intrusion type resident load recognition methods | |
WO2022067562A1 (en) | Method and device for diagnosing fault arc, and computer-readable storage medium | |
CN110956220B (en) | Non-invasive household appliance load identification method | |
Ray et al. | Optimal feature and decision tree-based classification of power quality disturbances in distributed generation systems | |
CN109948597B (en) | High-voltage circuit breaker mechanical fault diagnosis method | |
CN109598451B (en) | Non-invasive load identification method based on PCA (principal component analysis) and LSTM (least Square TM) neural network | |
CN103177188A (en) | Electric system load dynamic characteristic classifying method based on characteristic mapping | |
CN103400009B (en) | Based on the wind energy turbine set Dynamic Equivalence of the semi-supervised spectral clustering of division level | |
CN110991786A (en) | 10kV static load model parameter identification method based on similar daily load curve | |
CN101975910A (en) | Intelligent fault classification and location method for ultra-high voltage direct current transmission line | |
CN103323755A (en) | Method and system for recognition of GIS ultrahigh frequency partial discharge signal | |
CN103886405B (en) | Boiler combustion condition identification method based on information entropy characteristics and probability nerve network | |
CN111242161B (en) | Non-invasive non-resident user load identification method based on intelligent learning | |
CN111612056A (en) | Low-pressure customer variation relation identification method based on fuzzy clustering and zero-crossing offset | |
CN111209976A (en) | Non-invasive load monitoring method and system | |
CN109902340A (en) | It is a kind of meter and complicated meteorology coupled characteristic multi-source-lotus combine scene generating method | |
CN109242136A (en) | A kind of micro-capacitance sensor wind power Chaos-Genetic-BP neural network prediction technique | |
Akarslan et al. | A novel approach based on a feature selection procedure for residential load identification | |
Li et al. | Adaptive S transform for feature extraction in voltage sags | |
CN114943367A (en) | Non-invasive load identification method based on BP neural network model | |
Hu et al. | An electromagnetic environment situation assessment and abnormal detection technology | |
CN113902136A (en) | Load identification method based on electric power fingerprint features and integrated learning mechanism | |
CN112993993A (en) | Non-invasive power load perception identification method based on improved differential evolution algorithm | |
Parija et al. | Faults classification in a microgrid using decision tree technique and support vector machine | |
Massucco et al. | A markov chain load modeling approach through a stream clustering algorithm |
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 |