CN109993424A - A kind of non-interfering formula load decomposition method based on width learning algorithm - Google Patents
A kind of non-interfering formula load decomposition method based on width learning algorithm Download PDFInfo
- Publication number
- CN109993424A CN109993424A CN201910230729.0A CN201910230729A CN109993424A CN 109993424 A CN109993424 A CN 109993424A CN 201910230729 A CN201910230729 A CN 201910230729A CN 109993424 A CN109993424 A CN 109993424A
- Authority
- CN
- China
- Prior art keywords
- matrix
- vector
- load decomposition
- spliced
- augmented
- 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
- 238000000354 decomposition reaction Methods 0.000 title claims abstract description 53
- 238000000034 method Methods 0.000 title claims abstract description 27
- 230000002452 interceptive effect Effects 0.000 title claims abstract description 22
- 238000012549 training Methods 0.000 claims abstract description 27
- 239000011159 matrix material Substances 0.000 claims description 202
- 239000013598 vector Substances 0.000 claims description 96
- 230000003190 augmentative effect Effects 0.000 claims description 49
- 238000013507 mapping Methods 0.000 claims description 28
- 230000002708 enhancing effect Effects 0.000 claims description 18
- 238000012360 testing method Methods 0.000 claims description 18
- 230000004913 activation Effects 0.000 claims description 17
- 230000008859 change Effects 0.000 claims description 12
- 238000010606 normalization Methods 0.000 claims description 11
- 230000009466 transformation Effects 0.000 claims description 6
- 230000003416 augmentation Effects 0.000 claims description 4
- 230000006870 function Effects 0.000 claims 1
- 230000005611 electricity Effects 0.000 abstract description 7
- 238000013473 artificial intelligence Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 230000008901 benefit Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000013011 mating Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012806 monitoring device Methods 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0639—Performance analysis of employees; Performance analysis of enterprise or organisation operations
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
Landscapes
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Engineering & Computer Science (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- Tourism & Hospitality (AREA)
- General Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Educational Administration (AREA)
- Marketing (AREA)
- Entrepreneurship & Innovation (AREA)
- Theoretical Computer Science (AREA)
- Development Economics (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Game Theory and Decision Science (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Supply And Distribution Of Alternating Current (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
Non-interfering formula load decomposition method provided by the invention based on width learning algorithm, it acquires the operation data before and after electric operation as input data, in conjunction in conjunction with width learning algorithm, by data prediction, load decomposition model initial training, the study of load decomposition model incremental, switch state changes identification model initial training and switch state changes identification model incremental learning, fully consider multiple factors of electrical equipment information, the electric load operation conditions of more electrical equipment scenes is obtained in non-interfering mode, effectively realize the load decomposition of electrical equipment, more accurately electricity consumption decision is provided, and there are preferable versatility and replicability, greatly reduce application cost.
Description
Technical field
The present invention relates to load decomposition technical field, specially a kind of non-interfering formula load based on width learning algorithm point
Solution method.
Background technique
Electric load, which decomposes, usually requires mating interference formula or non-interfering formula device, and traditional load decomposition method needs are matched
Standby dedicated sensor is at high cost to be unfavorable for scale popularization to monitor, acquire electric equipment operation state and related power information,
And deployment cycle is long, slow effect.A large amount of intelligent electric meter class monitoring device has been emerged in large numbers on Vehicles Collected from Market, has been had to electrical equipment
Voltage (U), electric current (I), active power (P), reactive power (Q), power factor (PF), frequency (f), active energy (kWh)
Equal electrical parameters are effectively monitored and information collection.In addition, traditional electrical equipment load decomposition method, consideration it is electrical because
Element is generally power, and the accuracy rate of data sheet one, decomposition is lower.For disadvantages mentioned above, it is proposed that improve.
Summary of the invention
A kind of non-interfering formula load decomposition method based on width learning algorithm provided by the invention is not necessarily to dedicated sensing
Device device fully considers multiple factors of electrical equipment information, using the relevant technologies of big data and artificial intelligence, effectively realizes
The load decomposition of electrical equipment provides more accurately electricity consumption decision.
The present invention provides a kind of non-interfering formula load decomposition method based on width learning algorithm, is used to monitor analysis electricity
The power operating state of device and related power information to realize the load decomposition of the electric appliance, method includes the following steps:
Step 1: a kind of data acquisition chooses the electric switch state unconverted multiple periods, with K1 using frequency
Rate acquire in electric operation data, which includes voltage, electric current, instantaneous active power, instantaneous reactive power, will be described
Operation data is spliced to form input vectorAnd input vector is spliced to form after making Fourier transformation to the operation data
By the input vectorSplicing becomes input vectorMeanwhile the switch state data of electric appliance are acquired, it opens
It is denoted as 1, closing is denoted as 0, and switch state is spliced to form label vectorWherein, i=1,2 ..., I are denoted as data number,
I is total amount of data；Two class data acquisition, with electric appliance fortune in the T period after the sample frequency acquisition electric switch state change of K2
The operation data is spliced to form input vector by row dataWherein, remember that j=1,2 ..., J are data number, J is that data are total
Amount, meanwhile, construct label vectorElectric switch state is then acquired without change with the sample frequency of K2 for complete 1 vector of j dimension
The operation data is spliced to form input vector by electric operation data in the T period of changeMeanwhile constructing label vector
For the full 0 vector of j dimension；
Step 2: data prediction, a kind of data prediction, by the input vector
Normalization is spliced into input matrixOther input vectorsIt normalizes and is spliced into test input matrixBy the label vectorIt is spliced into label matrixOther marks
Sign vectorIt is spliced into test label matrixTwo class data predictions, by the input vectorNormalization is spliced into input matrixOther input vectorsNormalization is spliced into test matrixBy the label vector
It is spliced into label vectorOther label vectorsIt is spliced into test label vector
Step 3: load decomposition model initial training is based on the input matrixUtilize the first random initializtion square
Battle arrayFirst activation primitiveAnd first bias vectorConstruct mappings characteristics node matrix equation
Based on the mappings characteristics node matrix equationUtilize the second random initializtion matrixSecond activation primitiveAnd second bias vectorBuilding enhancing node matrix equationUtilize the mappings characteristics node matrix equationAnd enhancing node matrix equationConstruct the first augmented matrixPass through first augmented matrixAnd
The label matrixAcquire the first weight matrix
Step 4: using test input matrixIt is tested, if meeting the first training error, output load point
Model is solved, and enters step 6；If training error is unsatisfactory for the first training error, 5 are entered step；
Step 5: the study of load decomposition model incremental is based on the mappings characteristics node matrix equationIt is random using third
Initialize matrixSecond activation primitiveAnd third bias vectorConstructing increment enhances node square
Battle arrayEnhance node matrix equation using the incrementAnd first augmented matrixConstruct the second augmented matrixPass through the second augmented matrixAnd label matrixAcquire the second weight matrixM+1 is assigned
It is worth to M, and m+1 is assigned to m, return step 4；
Step 6: switch state changes identification model initial training, is based on the input matrixUsing the 4th with
Machine initializes matrixThird activation primitiveAnd the 4th bias vectorConstruct mappings characteristics section
Dot matrixBased on the mappings characteristics node matrix equationUtilize the 5th random initializtion matrix
4th activation primitiveAnd the 5th bias vectorBuilding enhancing node matrix equationUtilize the mapping
Characteristic node matrixAnd enhancing node matrix equationConstruct third augmented matrixPass through the third
Augmentation squareBattle array and the label matrixAcquire third weight matrix
Step 7: using test input matrixIt is tested, if meeting the second training error, exports switch shape
State changes identification model, and enters step 9；If training error is unsatisfactory for the second training error, 8 are entered step；
Step 8: switch state changes identification model incremental learning, is based on the mappings characteristics node matrix equationBenefit
With the 6th random initializtion matrix4th activation primitiveAnd the 6th bias vectorBuilding increases
Amount enhancing node matrix equationEnhance node matrix equation using the incrementAnd third augmented matrixStructure
Build the 4th augmented matrixPass through the 4th augmented matrixAnd label matrixAcquire the 4th weight
MatrixM+1 is assigned to M, and m+1 is assigned to m, return step 7；
Step 9: switch state variation identification is continued the electric operation data for being acquired K period with the sample frequency of K2, spelled
It connects and normalizes and constitute input vector Xswitch, by XswitchThe switch state variation identification model is inputted, identifies electric switch
Whether state changes, if so, 10 are entered step after the T2 period of delay, if it is not, then executing step 10 with Fixed Time Interval；
And
Step 10: electric appliance load decomposition acquires the electric operation information of signal period, splicing is simultaneously with the sample frequency of K1
Normalization constitutes input vector X1, and Fourier transformation is made to the operation information, splice and normalizes composition input vector X2, will
X1、X2It is spliced to form input vector Xcycle, by input vector XcycleThe load decomposition model is inputted, obtains electric appliance load point
Solve result Ycycle。
Preferably, the mapping node matrix is constructed based on following formulaAndNote Then Note Then Wherein, k
=1,2 ..., n.
Preferably, the enhancing node matrix equation is constructed based on following formulaAndNote ThenNoteThen Wherein, l=1,2 ..., m.
Preferably, step 3 is based on following formula and acquires first weight matrixFirst augmented matrixSolve the first augmented matrixPseudoinverse Obtain the first weight matrix
Preferably, step 5, which is based on following formula building increment, enhances node matrix equation
Preferably, step 5 is based on following formula and acquires the second weight matrixConstruct the second augmented matrixWherein, it enables
Solve the second augmented matrixPseudoinverse
Then solve the second weight matrix
Preferably, step 6 is based on following formula and acquires the third weight matrixThe third augmentation square
Battle arraySolve third augmented matrixPseudoinverse Obtain third weight matrix
Preferably, step 8, which is based on following formula building increment, enhances node matrix equation
Preferably, step 8 is based on following formula and acquires the 4th weight matrixConstruct the 4th augmented matrixWherein, it enables
Solve the 4th augmented matrixPseudoinverse
Then solve the 4th weight matrix
Preferably, the frequency K1 can selection range be 1kHz-10kHz, the frequency K2 can selection range be 1kHz-
10kHz。
Non-interfering formula load decomposition method provided by the invention based on width learning algorithm, when realizing load decomposition without
Dedicated sensor device is needed, the most power information acquisition devices (such as intelligent electric meter) of reusable in the market are based on
Width learning algorithm fully considers that multiple factors of electrical equipment information are had using the relevant technologies of big data and artificial intelligence
Effect realizes the load decomposition of electrical equipment, has preferable versatility and replicability, greatly reduces application cost, more accurately
Ground recognizes the electric load operation conditions of more device contexts, provides more accurately electricity consumption decision, effectively promotion economic benefit.
Detailed description of the invention
Fig. 1 is a kind of non-interfering formula load decomposition method based on width learning algorithm that first embodiment of the invention provides
Flow chart；
Fig. 2 is width Learning work load decomposition model schematic diagram of the present invention；
Fig. 3 is the voltage for the electric heater switch state variation front and back that first embodiment of the invention provides, current graph；
Fig. 4 is the electric appliance load decomposition result figure that first embodiment of the invention provides.
Specific embodiment
With reference to the accompanying drawing to a kind of non-interfering formula load decomposition side based on width learning algorithm provided by the present invention
Method is described further, it should be pointed out that below only with a kind of technical solution of optimization to technical solution of the present invention with
And design principle is described in detail.
A kind of non-interfering formula load decomposition method based on width learning algorithm that first embodiment of the invention provides is used
Power operating state and the correlation power information of electric appliance are analyzed in monitoring to realize the load decomposition of the electric appliance.
Refering to fig. 1 and Fig. 2, a kind of non-interfering formula load based on width learning algorithm that first embodiment of the invention provides
Decomposition method comprising following steps:
Step 1: a kind of data acquisition chooses the electric switch state unconverted multiple periods, with the use of 10kHz
Electric operation data in frequency collection, which includes voltage, electric current, instantaneous active power, instantaneous reactive power, by institute
It states operation data and is spliced to form input vectorAnd input vector is spliced to form after making Fourier transformation to the operation dataBy the input vectorSplicing becomes input vectorMeanwhile the switch state data of electric appliance are acquired,
Unlatching is denoted as 1, and closing is denoted as 0, and switch state is spliced to form label vectorWherein, i=1,2 ..., I are denoted as data
Number, I is total amount of data；Two class data acquisition, in the T period after the sample frequency acquisition electric switch state change of 1kHz
The operation data is spliced to form input vector by electric operation dataWherein, remember j=1,2 ..., J is data number, and J is
Total amount of data, meanwhile, construct label vectorElectric switch shape is then acquired with the sample frequency of K2 for complete 1 vector of j dimension
Electric operation data in the state unconverted T period, are spliced to form input vector for the operation dataMeanwhile constructing label
VectorFor the full 0 vector of j dimension；
Step 2: data prediction, a kind of data prediction, by the input vector
Normalization is spliced into input matrixOther input vectorsIt normalizes and is spliced into test input matrixBy the label vectorIt is spliced into label matrixIts
His label vectorIt is spliced into test label matrixTwo class data predictions, by the input vectorNormalization is spliced into input matrixOther input vectorsNormalization is spliced into test matrixBy the label vector
It is spliced into label vectorOther label vectorsIt is spliced into test label vector
Step 3: load decomposition model initial training is based on the input matrixUtilize the first random initializtion square
Battle arrayFirst activation primitiveAnd first bias vectorConstruct mappings characteristics node matrix equation
Based on the mappings characteristics node matrix equationUtilize the second random initializtion matrixSecond activation primitiveAnd second bias vectorBuilding enhancing node matrix equationUtilize the mappings characteristics node matrix equationAnd enhancing node matrix equationConstruct the first augmented matrixPass through first augmented matrixAnd
The label matrixAcquire the first weight matrix
Specifically, above-mentioned steps 3 acquire the first weight matrix based on following formulaFirstly, building the mapping section
Dot matrixNoteThen Secondly, constructing the enhancing node matrix equationNoteThenThen, first is constructed
Augmented matrixSolve the first augmented matrixPseudoinverse Finally, acquiring the first weight matrix
Step 4: using test input matrixIt is tested, if meeting the first training error, output load point
Model is solved, and enters step 6；If training error is unsatisfactory for the first training error, 5 are entered step；
Step 5: the study of load decomposition model incremental is based on the mappings characteristics node matrix equationIt is random using third
Initialize matrixSecond activation primitiveAnd third bias vectorConstructing increment enhances node square
Battle arrayEnhance node matrix equation using the incrementAnd first augmented matrixConstruct the second augmented matrixPass through the second augmented matrixAnd label matrixAcquire the second weight matrixM+1 is assigned
It is worth to M, and m+1 is assigned to m, return step 4；
Specifically, step 5, which is based on following formula, acquires the second weight matrixFirstly, building increment enhances node
MatrixThen, the second augmented matrix is constructedIt enables
Solve the second augmented matrixPseudoinverse
Finally, acquiring the second weight matrix
Step 6: switch state changes identification model initial training, is based on the input matrixUsing the 4th with
Machine initializes matrixThird activation primitiveAnd the 4th bias vectorConstruct mappings characteristics section
Dot matrixBased on the mappings characteristics node matrix equationUtilize the 5th random initializtion matrix
4th activation primitiveAnd the 5th bias vectorBuilding enhancing node matrix equationUtilize the mapping
Characteristic node matrixAnd enhancing node matrix equationConstruct third augmented matrixPass through the third
Augmentation squareBattle array and the label matrixAcquire third weight matrix
Specifically, step 6, which is based on following formula, acquires third weight matrixFirstly, constructing the mapping node
Matrix
NoteThen Secondly, constructing the enhancing node matrix equationNoteThen Then, third augmented matrix is constructed
Solve third augmented matrixPseudoinverse
Finally, acquiring third power
Value matrix
Step 7: using test input matrixIt is tested, if meeting the second training error, exports switch shape
State changes identification model, and enters step 9；If training error is unsatisfactory for the second training error, 8 are entered step；
Step 8: switch state changes identification model incremental learning, is based on the mappings characteristics node matrix equationBenefit
With the 6th random initializtion matrix4th activation primitiveAnd the 6th bias vectorBuilding increases
Amount enhancing node matrix equationEnhance node matrix equation using the incrementAnd third augmented matrixStructure
Build the 4th augmented matrixPass through the 4th augmented matrixAnd label matrixAcquire the 4th weight
MatrixM+1 is assigned to M, and m+1 is assigned to m, return step 7；
Specifically, step 8, which is based on following formula, acquires the 4th weight matrixFirstly, building increment enhances node
MatrixThen, building the 4th increases
Wide matrixIt enables
Solve the 4th augmented matrixPseudoinverse
Finally, acquiring the 4th weight matrix
Step 9: switch state variation identification, as shown in figure 3, continuing the electricity for acquiring K period with the sample frequency of 1kHz
Device operation data splices and normalizes composition input vector Xswitch, by XswitchThe switch state variation identification model is inputted,
Whether identification electric switch state changes, if so, 10 are entered step after 25 periods of delay, if it is not, then between the set time
Every execution step 10；
Step 10: electric appliance load decomposition, as shown in figure 4, acquiring the electric appliance fortune of signal period with the sample frequency of 10kHz
Row information splices and normalizes composition input vector X1, and Fourier transformation is made to the operation information, splice and normalizes composition
Input vector X2, by X1、X2It is spliced to form input vector Xcycle, by input vector XcycleThe load decomposition model is inputted,
Obtain electric appliance load decomposition result Ycycle。
Specifically, in the present invention, the electric appliance of institute's identification switch state change is electric heater refering to Fig. 3 and Fig. 4, divide
The electric appliance for solving load is electric heater, television set and notebook.
Non-interfering formula load decomposition method provided by the invention based on width learning algorithm, before and after acquisition electric operation
Operation data as input data, in conjunction with width learning algorithm is combined, initially instructed by data prediction, load decomposition model
Practice, the study of load decomposition model incremental, switch state changes identification model initial training and switch state changes identification model
Incremental learning fully considers multiple factors of electrical equipment information, obtains the electricity of more electrical equipment scenes in non-interfering mode
Power load operation condition effectively realizes the load decomposition of electrical equipment, provides more accurately electricity consumption decision, and has preferable logical
With property and replicability, application cost is greatly reduced.
The above is only the preferred embodiment of the present invention, it is noted that above-mentioned preferred embodiment is not construed as pair
Limitation of the invention, protection scope of the present invention should be defined by the scope defined by the claims..For the art
For those of ordinary skill, without departing from the spirit and scope of the present invention, several improvements and modifications can also be made, these change
It also should be regarded as protection scope of the present invention into retouching.
Claims (10)
1. a kind of non-interfering formula load decomposition method based on width learning algorithm is used to monitor the operation power of analysis electric appliance
State and related power information are to realize the load decomposition of the electric appliance, which is characterized in that method includes the following steps:
Step 1: a kind of data acquisition is chosen the electric switch state unconverted multiple periods, is adopted with the use frequency of K1
Electric operation data in collecting, which includes voltage, electric current, instantaneous active power, instantaneous reactive power, by the operation
Data are spliced to form input vectorAnd input vector is spliced to form after making Fourier transformation to the operation dataBy institute
State input vectorSplicing becomes input vectorMeanwhile the switch state data of electric appliance are acquired, unlatching is denoted as
1, closing is denoted as 0, and switch state is spliced to form label vectorWherein, i=1,2 ..., I are denoted as data number, and I is
Total amount of data；Two class data acquisition, with electric operation number in the T period after the sample frequency acquisition electric switch state change of K2
According to the operation data is spliced to form input vectorWherein, remembering j=1,2 ..., J is data number, and J is total amount of data,
Meanwhile constructing label vectorIt is then, unchanged with the sample frequency acquisition electric switch state of K2 for complete 1 vector of j dimension
T period in electric operation data, which is spliced to form input vectorMeanwhile constructing label vectorFor
The full 0 vector of j dimension；
Step 2: data prediction, a kind of data prediction, by the input vector
Normalization is spliced into input matrixOther input vectorsIt normalizes and is spliced into test input matrixBy the label vector It is spliced into label matrixOther marks
Sign vectorIt is spliced into test label matrixTwo class data predictions, by the input vectorNormalization is spliced into input matrixOther input vectorsNormalization is spliced into test matrixBy the label vectorIt is spliced into label vectorOther label vectorsIt is spliced into
Test label vector
Step 3: load decomposition model initial training is based on the input matrixUtilize the first random initializtion matrixFirst activation primitiveAnd first bias vectorConstruct mappings characteristics node matrix equationBase
In the mappings characteristics node matrix equationUtilize the second random initializtion matrixSecond activation primitive
And second bias vectorBuilding enhancing node matrix equationUtilize the mappings characteristics node matrix equationWith
And enhancing node matrix equationConstruct the first augmented matrixPass through first augmented matrixAnd the label
MatrixAcquire the first weight matrix
Step 4: using test input matrixIt is tested, if meeting the first training error, output load decomposes mould
Type, and enter step 6；If training error is unsatisfactory for the first training error, 5 are entered step；
Step 5: the study of load decomposition model incremental is based on the mappings characteristics node matrix equationIt is initial at random using third
Change matrixSecond activation primitiveAnd third bias vectorConstructing increment enhances node matrix equationEnhance node matrix equation using the incrementAnd first augmented matrixConstruct the second augmented matrixPass through the second augmented matrixAnd label matrixAcquire the second weight matrixBy M+1 assignment
To M, and m+1 is assigned to m, return step 4；
Step 6: switch state changes identification model initial training, is based on the input matrixIt is random initial using the 4th
Change matrixThird activation primitiveAnd the 4th bias vectorConstruct mappings characteristics node matrix equationBased on the mappings characteristics node matrix equationUtilize the 5th random initializtion matrix4th swashs
Function livingAnd the 5th bias vectorBuilding enhancing node matrix equationUtilize the mappings characteristics section
Dot matrixAnd enhancing node matrix equationConstruct third augmented matrixPass through the third augmentation square
Battle arrayAnd the label matrixAcquire third weight matrix
Step 7: using test input matrixIt is tested, if meeting the second training error, exports switch state change
Change identification model, and enters step 9；If training error is unsatisfactory for the second training error, 8 are entered step；
Step 8: switch state changes identification model incremental learning, is based on the mappings characteristics node matrix equationUtilize
Six random initializtion matrixes4th activation primitiveAnd the 6th bias vectorIncrement is constructed to increase
Strong node matrix equationEnhance node matrix equation using the incrementAnd third augmented matrixBuilding the
Four augmented matrixesPass through the 4th augmented matrixAnd label matrixAcquire the 4th weight matrixM+1 is assigned to M, and m+1 is assigned to m, return step 7；
Step 9: switch state variation identification continues the electric operation data that K period is acquired with the sample frequency of K2, splicing is simultaneously
Normalization constitutes input vector Xswitch, by XswitchThe switch state variation identification model is inputted, identifies electric switch state
Whether change, if so, 10 are entered step after the T2 period of delay, if it is not, then executing step 10 with Fixed Time Interval；And
Step 10: electric appliance load decomposition acquires the electric operation information of signal period with the sample frequency of K1, splices simultaneously normalizing
Change and constitutes input vector X1, and Fourier transformation is made to the operation information, splice and normalizes composition input vector X2, by X1、X2
It is spliced to form input vector Xcycle, by input vector XcycleThe load decomposition model is inputted, obtains electric appliance load decomposition knot
Fruit Ycycle。
2. a kind of non-interfering formula load decomposition method based on width learning algorithm according to claim 1, feature exist
In based on the following formula building mapping node matrixAndNote
Then Note
Then Wherein, k=1,2 ..., n.
3. a kind of non-interfering formula load decomposition method based on width learning algorithm according to claim 2, which is characterized in that base
The enhancing node matrix equation is constructed in following formulaAndNoteThenNote
Then Wherein, l=1,2 ..., m.
4. a kind of non-interfering formula load decomposition method based on width learning algorithm according to claim 3, feature exist
In step 3 is based on following formula and acquires first weight matrixFirst augmented matrixSolve the first augmented matrixPseudoinverse Obtain the first weight matrix
5. a kind of non-interfering formula load decomposition method based on width learning algorithm according to claim 4, feature exist
In step 5, which is based on following formula building increment, enhances node matrix equation
6. a kind of non-interfering formula load decomposition method based on width learning algorithm according to claim 5, which is characterized in that step
Rapid 5 acquire the second weight matrix based on following formulaConstruct the second augmented matrixIts
In, it enables
Solve the second augmented matrixPseudoinverse
Then solve the second weight matrix
7. a kind of non-interfering formula load decomposition method based on width learning algorithm according to claim 3,
It is characterized in that, step 6 is based on following formula and acquires the third weight matrixThe third augmented matrixSolve third augmented matrixPseudoinverseObtain third weight matrix
8. a kind of non-interfering formula load decomposition method based on width learning algorithm, feature according to claim exist
In step 8, which is based on following formula building increment, enhances node matrix equation
9. a kind of non-interfering formula load decomposition method based on width learning algorithm according to claim, which is characterized in that step
8 acquire the 4th weight matrix based on following formulaConstruct the 4th augmented matrix
Wherein, it enables
Solve the 4th augmented matrixPseudoinverse
Then solve the 4th weight matrix
10. a kind of non-interfering formula load decomposition side based on width learning algorithm described in any one according to claim 1~9
Method, which is characterized in that the frequency K1 can selection range be 1kHz-10kHz, the frequency K2 can selection range be 1kHz-
10kHz。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910230729.0A CN109993424B (en) | 2019-03-26 | 2019-03-26 | Non-interference type load decomposition method based on width learning algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910230729.0A CN109993424B (en) | 2019-03-26 | 2019-03-26 | Non-interference type load decomposition method based on width learning algorithm |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109993424A true CN109993424A (en) | 2019-07-09 |
| CN109993424B CN109993424B (en) | 2023-06-23 |
Family
ID=67131501
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910230729.0A Active CN109993424B (en) | 2019-03-26 | 2019-03-26 | Non-interference type load decomposition method based on width learning algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109993424B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112256123A (en) * | 2020-09-25 | 2021-01-22 | 北京师范大学 | Brain load-based control work efficiency analysis method, equipment and system |
| CN116304762A (en) * | 2023-05-17 | 2023-06-23 | 杭州致成电子科技有限公司 | Method and device for decomposing load |
| CN116610922A (en) * | 2023-07-13 | 2023-08-18 | 浙江大学滨江研究院 | Non-invasive load identification method and system based on multi-strategy learning |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20130110756A1 (en) * | 2011-10-31 | 2013-05-02 | Siemens Corporation | Short-term Load Forecast Using Support Vector Regression and Feature Learning |
| CN105260803A (en) * | 2015-11-06 | 2016-01-20 | 国家电网公司 | Power consumption prediction method for system |
| CN107330517A (en) * | 2017-06-14 | 2017-11-07 | 华北电力大学 | One kind is based on S_Kohonen non-intrusion type resident load recognition methods |
| CN108960339A (en) * | 2018-07-20 | 2018-12-07 | 吉林大学珠海学院 | A kind of electric car induction conductivity method for diagnosing faults based on width study |
| CN109444757A (en) * | 2018-10-09 | 2019-03-08 | 杭州中恒云能源互联网技术有限公司 | A kind of residual capacity of power battery of electric automobile evaluation method |
| CN109508908A (en) * | 2018-12-25 | 2019-03-22 | 深圳市城市公共安全技术研究院有限公司 | Non-invasive load recognition model training method, load monitoring method and device |
-
2019
- 2019-03-26 CN CN201910230729.0A patent/CN109993424B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20130110756A1 (en) * | 2011-10-31 | 2013-05-02 | Siemens Corporation | Short-term Load Forecast Using Support Vector Regression and Feature Learning |
| CN105260803A (en) * | 2015-11-06 | 2016-01-20 | 国家电网公司 | Power consumption prediction method for system |
| CN107330517A (en) * | 2017-06-14 | 2017-11-07 | 华北电力大学 | One kind is based on S_Kohonen non-intrusion type resident load recognition methods |
| CN108960339A (en) * | 2018-07-20 | 2018-12-07 | 吉林大学珠海学院 | A kind of electric car induction conductivity method for diagnosing faults based on width study |
| CN109444757A (en) * | 2018-10-09 | 2019-03-08 | 杭州中恒云能源互联网技术有限公司 | A kind of residual capacity of power battery of electric automobile evaluation method |
| CN109508908A (en) * | 2018-12-25 | 2019-03-22 | 深圳市城市公共安全技术研究院有限公司 | Non-invasive load recognition model training method, load monitoring method and device |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112256123A (en) * | 2020-09-25 | 2021-01-22 | 北京师范大学 | Brain load-based control work efficiency analysis method, equipment and system |
| CN112256123B (en) * | 2020-09-25 | 2022-08-23 | 北京师范大学 | Brain load-based control work efficiency analysis method, equipment and system |
| CN116304762A (en) * | 2023-05-17 | 2023-06-23 | 杭州致成电子科技有限公司 | Method and device for decomposing load |
| CN116610922A (en) * | 2023-07-13 | 2023-08-18 | 浙江大学滨江研究院 | Non-invasive load identification method and system based on multi-strategy learning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109993424B (en) | 2023-06-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109993424A (en) | A kind of non-interfering formula load decomposition method based on width learning algorithm | |
| CN103020459B (en) | A kind of cognitive method of various dimensions electricity consumption behavior and system | |
| CN109145949A (en) | Non-intrusive electrical load monitoring and decomposition method and system based on integrated study | |
| CN106483449B (en) | Analog-circuit fault diagnosis method based on deep learning and Complex eigenvalues | |
| Zufferey et al. | Machine learning approaches for electric appliance classification | |
| CN109165604A (en) | The recognition methods of non-intrusion type load and its test macro based on coorinated training | |
| CN111209976B (en) | Non-invasive load monitoring method and system | |
| CN112434799B (en) | Non-invasive load identification method based on full convolution neural network | |
| CN111639586B (en) | Non-invasive load identification model construction method, load identification method and system | |
| CN102224425A (en) | Method and device for the non-intrusive determination of the electrical power consumed by an installation, by analysing load transients | |
| CN112598303A (en) | Non-invasive load decomposition method based on combination of 1D convolutional neural network and LSTM | |
| CN109697548A (en) | Electrical energy consumption analysis server and its electrical energy consumption analysis method | |
| CN113408341B (en) | Load identification method and device, computer equipment and storage medium | |
| CN102262211A (en) | Analog circuit failure diagnosis method based on dynamic classification of echo state network | |
| CN105807231A (en) | Method and system for storage battery residual capacity detection | |
| CN108763847A (en) | Non-invasive electric signal analysis tracking method and device and electronic equipment | |
| CN110633870A (en) | Harmonic early warning method, harmonic early warning device and terminal equipment | |
| CN114091282B (en) | Lithium ion battery state estimation method and system based on fractional order model | |
| CN106707118B (en) | Partial Discharge Pattern Recognition Method and device | |
| CN109782176A (en) | Battery remaining power On-line Estimation method based on NARX model | |
| CN117272230A (en) | Non-invasive load monitoring method and system based on multi-task learning model | |
| Stošović et al. | Classification of nonlinear loads based on artificial neural networks | |
| CN107024975A (en) | Exchange method and equipment | |
| CN209265691U (en) | A kind of test device and test macro | |
| CN112130075B (en) | Method and system for estimating OCV (open Circuit Voltage) by coupling of offline and online gas-liquid battery models |
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 |