CN110210672A - The method and apparatus of electricity demand forecasting, electronic equipment, computer-readable medium - Google Patents
The method and apparatus of electricity demand forecasting, electronic equipment, computer-readable medium Download PDFInfo
- Publication number
- CN110210672A CN110210672A CN201910482201.2A CN201910482201A CN110210672A CN 110210672 A CN110210672 A CN 110210672A CN 201910482201 A CN201910482201 A CN 201910482201A CN 110210672 A CN110210672 A CN 110210672A
- Authority
- CN
- China
- Prior art keywords
- vector
- information
- future
- neural networks
- convolutional neural
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 230000005611 electricity Effects 0.000 title claims abstract description 92
- 238000000034 method Methods 0.000 title claims abstract description 54
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 78
- 239000013598 vector Substances 0.000 claims description 96
- 208000037170 Delayed Emergence from Anesthesia Diseases 0.000 claims description 26
- 230000008569 process Effects 0.000 claims description 24
- 238000012545 processing Methods 0.000 claims description 18
- 238000003860 storage Methods 0.000 claims description 8
- 238000004590 computer program Methods 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 abstract description 2
- 239000010410 layer Substances 0.000 description 79
- 238000010586 diagram Methods 0.000 description 13
- 238000010606 normalization Methods 0.000 description 12
- 238000012549 training Methods 0.000 description 11
- 238000013528 artificial neural network Methods 0.000 description 7
- 230000005284 excitation Effects 0.000 description 6
- 238000013473 artificial intelligence Methods 0.000 description 4
- 230000000712 assembly Effects 0.000 description 4
- 238000000429 assembly Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 230000000306 recurrent effect Effects 0.000 description 4
- 230000015654 memory Effects 0.000 description 3
- 239000000203 mixture Substances 0.000 description 3
- 238000011176 pooling Methods 0.000 description 3
- 230000033228 biological regulation Effects 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 230000004087 circulation Effects 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 238000007906 compression Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 230000002779 inactivation Effects 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 230000008901 benefit Effects 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 239000011248 coating agent Substances 0.000 description 1
- 238000000576 coating method Methods 0.000 description 1
- 230000006835 compression Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 238000013277 forecasting method Methods 0.000 description 1
- 239000011229 interlayer Substances 0.000 description 1
- 238000003475 lamination Methods 0.000 description 1
- 230000014759 maintenance of location Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000004064 recycling Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
- 239000002699 waste material Substances 0.000 description 1
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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- 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
-
- 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
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- 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
- G06Q10/0631—Resource planning, allocation, distributing or scheduling for enterprises or organisations
- G06Q10/06315—Needs-based resource requirements planning or analysis
-
- 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
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- General Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- General Business, Economics & Management (AREA)
- Entrepreneurship & Innovation (AREA)
- Marketing (AREA)
- General Health & Medical Sciences (AREA)
- Tourism & Hospitality (AREA)
- Biophysics (AREA)
- Operations Research (AREA)
- Development Economics (AREA)
- Quality & Reliability (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Game Theory and Decision Science (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Educational Administration (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- Primary Health Care (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
Present disclose provides a kind of methods of electricity demand forecasting, it is related to field of cloud calculation, this method comprises: obtaining historical information and Future Information, the historical information includes the additional information for respectively corresponding M historical time section and practical electricity consumption, the Future Information includes the additional information for respectively corresponding N number of future time section, and the M and N are positive integer；The historical information and Future Information are inputted into preset convolutional neural networks, obtain the prediction electricity consumption of each future time section of correspondence of the convolutional neural networks output.The disclosure additionally provides the device, electronic equipment, computer-readable medium of a kind of electricity demand forecasting.
Description
Technical field
The embodiment of the present disclosure is related to field of computer technology (specially field of cloud calculation), in particular to electricity demand forecasting
Method and apparatus, electronic equipment, computer-readable medium.
Background technique
The use of a certain range (a such as city, a cell, specific one or more users) in different time period
Electricity is significantly different, since the electric energy that electricity power enterprise issues is difficult to be stored on a large scale, therefore when generated energy and electricity consumption are obvious
When mismatch, it can perhaps cause electricity shortage or will cause waste of energy.Therefore, the electricity consumption in correctly predicted future, to hair
It is all critically important for electric enterprise, sale of electricity enterprise, urban development planning person etc..
Electricity demand forecasting can be carried out by similar date comparison, time series etc., but this simple prediction mode considers
Factor it is few, prediction accuracy is low, poor expandability.
Pass through recurrent neural network (RNN, Recurrent Neural Network), length time gate (Long Short-
Term Memory) etc. prediction electricity consumption be also it is feasible, but such mode equally exist model training it is difficult, to data and
Network parameter requires the problems such as high.
Summary of the invention
The embodiment of the present disclosure provides a kind of method and apparatus of electricity demand forecasting, electronic equipment, computer-readable medium.
In a first aspect, the embodiment of the present disclosure provides a kind of method of electricity demand forecasting comprising:
It obtains historical information and Future Information, the historical information includes the additional letter for respectively corresponding M historical time section
Breath and practical electricity consumption, the Future Information includes the additional information for respectively corresponding N number of future time section, and the M and N are positive
Integer；
The historical information and Future Information are inputted into preset convolutional neural networks, it is defeated to obtain the convolutional neural networks
The prediction electricity consumption of each future time section of correspondence out.
In some embodiments, the additional information of corresponding each period includes:
The Weather information of the period, and/or, the holiday information of the period.
In some embodiments, the M and N is the integer more than or equal to 2.
In some embodiments, the convolutional neural networks include:
First conformable layer, it is each described to go through in the history image for the historical information to be integrated into history image
The history period correspond to a column pixel, the history electricity consumption, the additional information each project correspond to one-row pixels；
Second conformable layer, for the Future Information to be integrated into future image, in the future image, it is each it is described not
Carry out a period corresponding column pixel, each project of the additional information corresponds to one-row pixels.
In some embodiments, the convolutional neural networks further include:
The first convolution process layer being connect with the output end of first conformable layer, for handling the history image to obtain
Element number to one-dimensional primary vector, the primary vector is equal to M；
The second convolution process layer being connect with the output end of second conformable layer, for handling the future image to obtain
Element number to one-dimensional secondary vector, the secondary vector is equal to N；
The merging layer being connect with the output end of the output end of the first convolution process layer and the second convolution process layer,
For the primary vector and secondary vector to be merged into one-dimensional third vector, the element number of the third vector is equal to M+
N。
In some embodiments, the convolutional neural networks further include:
The residual error network layer being connect with the output end for merging layer, for handling the third vector to obtain N number of one
4th vector of dimension；The residual error network layer includes multiple sequentially connected residual blocks, and each residual block includes multiple volumes
Lamination, and direct-connected connection is equipped between the input terminal and output end of each residual block；
The output layer being connect with the output end of the residual error network layer, for N number of 4th vector to be changed into one
The element number of one-dimensional output vector, the output vector is equal to N, wherein each element is that a future time section is corresponding
Predict electricity consumption.
Second aspect, the embodiment of the present disclosure provide a kind of device of electricity demand forecasting comprising:
Module is obtained, for obtaining historical information and Future Information, the historical information includes when respectively corresponding M history
Between section additional information and practical electricity consumption, the Future Information includes the additional information for respectively corresponding N number of future time section, institute
It states M and N is positive integer；
Convolutional neural networks module, for the historical information and Future Information to be inputted preset convolutional neural networks,
Obtain the prediction electricity consumption of each future time section of correspondence of the convolutional neural networks output.
In some embodiments, the additional information of corresponding each period includes:
The Weather information of the period, and/or, the holiday information of the period.
In some embodiments, the M and N is the integer more than or equal to 2.
In some embodiments, the convolutional neural networks module includes:
First integral unit, it is each described in the history image for the historical information to be integrated into history image
Historical time section correspond to a column pixel, the history electricity consumption, the additional information each project correspond to one-row pixels；
Second integral unit, it is each described in the future image for the Future Information to be integrated into future image
Future time section corresponds to a column pixel, and each project of the additional information corresponds to one-row pixels.
In some embodiments, the convolutional neural networks module further include:
The first convolution processing unit being connect with the output end of first integral unit, for handling the history image
To obtain one-dimensional primary vector, the element number of the primary vector is equal to M；
The second convolution processing unit being connect with the output end of second integral unit, for handling the future image
To obtain one-dimensional secondary vector, the element number of the secondary vector is equal to N；
The merging list being connect with the output end of the output end of first convolution processing unit and the second convolution processing unit
Member, for the primary vector and secondary vector to be merged into one-dimensional third vector, the element number etc. of the third vector
In M+N.
In some embodiments, the convolutional neural networks module further include:
The residual error network unit being connect with the output end of the combining unit, for handling the third vector to obtain N
A the 4th one-dimensional vector；The residual error network unit includes multiple sequentially connected residual blocks, and each residual block includes
Multiple convolutional layers, and direct-connected connection is equipped between the input terminal and output end of each residual block；
The output unit being connect with the output end of the residual error network unit, for N number of 4th vector to be changed into
The element number of one one-dimensional output vector, the output vector is equal to N, wherein each element is that a future time section is right
The prediction electricity consumption answered.
The third aspect, the embodiment of the present disclosure provide a kind of electronic equipment comprising:
One or more processors；
Storage device is stored thereon with one or more programs, when one or more of programs are by one or more
A processor executes, so that the method that one or more of processors realize any one of the above electricity demand forecasting.
Fourth aspect, the embodiment of the present disclosure provide a kind of computer-readable medium, are stored thereon with computer program, described
The method of any one of the above electricity demand forecasting is realized when program is executed by processor.
In the embodiment of the present disclosure, electricity demand forecasting is carried out using convolutional neural networks, wherein input convolutional neural networks
Information further includes the additional information (such as Weather information, holiday information) of corresponding each period other than time and electricity consumption,
Consider the correlation between data variant under the same time not only to convolutional neural networks, it is also contemplated that data are at any time
The trend factor (correlation between same data under different time) of variation, i.e., its intersect point from two time, space dimensions
The correlation of data has been analysed, to have stronger ability to express, more accurate prediction can be made.
Further, convolutional neural networks are simple to similar date comparison, time series etc. as artificial intelligence system
Prediction technique is compared, and prediction accuracy is high, scalability is good；And convolutional neural networks and recurrent neural network, length time gate
It is compared Deng others artificial intelligence system, training is easier, and is required data and network parameter lower.
Detailed description of the invention
Attached drawing is used to provide to further understand the embodiment of the present disclosure, and constitutes part of specification, with this public affairs
The embodiment opened is used to explain the disclosure together, does not constitute the limitation to the disclosure.By reference to attached drawing to detailed example reality
It applies example to be described, the above and other feature and advantage will become apparent those skilled in the art, in the accompanying drawings:
Fig. 1 is a kind of flow chart of the method for electricity demand forecasting that the embodiment of the present disclosure provides；
Fig. 2 is the structural schematic diagram of convolutional neural networks used in the embodiment of the present disclosure；
Fig. 3 is the schematic diagram of the stream compression process in convolutional neural networks used in the embodiment of the present disclosure；
Fig. 4 obtains the knot of history image for the first conformable layer in convolutional neural networks used in the embodiment of the present disclosure
Structure schematic diagram；
Fig. 5 obtains the knot of future image for the second conformable layer in convolutional neural networks used in the embodiment of the present disclosure
Structure schematic diagram；
Fig. 6 is the structural schematic diagram of the process of convolution layer in convolutional neural networks used in the embodiment of the present disclosure；
Fig. 7 is the structural representation of the residual block of the residual error network layer in convolutional neural networks used in the embodiment of the present disclosure
Figure；
Fig. 8 is the structural schematic diagram of the output layer in convolutional neural networks used in the embodiment of the present disclosure；
Fig. 9 is the flow chart of the method for another electricity demand forecasting that the embodiment of the present disclosure provides；
Figure 10 is the training method schematic diagram for the convolutional neural networks that the embodiment of the present disclosure provides；
Figure 11 is a kind of composition block diagram of the device for electricity demand forecasting that the embodiment of the present disclosure provides；
Figure 12 is the composition block diagram of the device for another electricity demand forecasting that the embodiment of the present disclosure provides.
Specific embodiment
To make those skilled in the art more fully understand the technical solution of the disclosure, the disclosure is mentioned with reference to the accompanying drawing
The method and apparatus of the electricity demand forecasting of confession, electronic equipment, computer-readable medium are described in detail.
Example embodiment will hereinafter be described more fully hereinafter with reference to the accompanying drawings, but the example embodiment can be with difference
Form come embody and should not be construed as being limited to the disclosure elaboration embodiment.Conversely, the purpose for providing these embodiments exists
It is thoroughly and complete in making the disclosure, and those skilled in the art will be made to fully understand the scope of the present disclosure.
As used in the disclosure, term "and/or" includes any and all groups of one or more associated listed entries
It closes.
Term used in the disclosure is only used for description specific embodiment, and is not intended to limit the disclosure.Such as disclosure institute
It uses, "one" is also intended to "the" including plural form singular, unless in addition context is expressly noted that.
When in the disclosure use term " includes " and/or " by ... be made " when, specify there are the feature, entirety, steps
Suddenly, operation, element and/or component, but do not preclude the presence or addition of one or more of the other feature, entirety, step, operation, member
Part, component and/or its group.
Embodiment described in the disclosure can be by the idealized schematic diagram of the disclosure and reference planes figure and/or sectional view are retouched
It states.It therefore, can be according to manufacturing technology and/or tolerance come modified example diagram.
Embodiment of the disclosure is not limited to embodiment shown in the drawings, but matches including what is formed based on manufacturing process
The modification set.Therefore, the area illustrated in attached drawing has schematic attribute, and the shape in area as shown in the figure instantiates the area of element
Concrete shape, but be not intended to restrictive.
Unless otherwise defined, the otherwise meaning and ability of all terms (including technical and scientific term) used in the disclosure
The normally understood meaning of domain those of ordinary skill is identical.It will also be understood that such as those those of limit term in common dictionary
It should be interpreted as having and its consistent meaning of meaning under the background of the relevant technologies and the disclosure, and will be not interpreted as
With idealization or excessively formal meaning, so limited unless the disclosure is clear.
Technical term explanation
In the embodiments of the present disclosure, unless otherwise specified, then following technical term should be according to understanding explained below:
Electricity consumption refers to a certain range (a such as city, a cell, specific one or more users) at one
The summation of the electricity issued used in period by electricity power enterprise can be the past period that statistics obtains
Electricity consumption (practical electricity consumption), be also possible to the electricity consumption (prediction electricity consumption) for the future time section that prediction obtains.
Convolutional neural networks (CNN, Convolutional Neural Networks) are a kind of with depth structure
Feedforward neural network, can be used for image recognition, image characteristics extraction etc., convolutional neural networks include at least one convolutional layer.
Convolutional layer is used to carry out convolution (Convolution)；Convolution is carried out using convolution collecting image (or vector)
A kind of processing.
(Normalization) layer is normalized, is used to that feature to be normalized, such as carries out batch normalization
(Batch Normalization)。
Excitation layer is also referred to as active coating, for carrying out Nonlinear Mapping, such as progress Rule excitation.
Random inactivation (Dropcut) layer is used for the fractional weight of hidden layer or the random zero of output.
Residual error network (ResNet, Residual Neural Network) is a kind of specific shape of convolutional neural networks
Formula, residual error network include multiple residual blocks (Res Block), and each residual block includes multiple convolutional layers, and each residual block
Direct-connected connection is equipped between input terminal and output end.
Direct-connected connection (Shortcut) is used to across at least one convolutional layer input the data of a scale rear
Layer in.
Pond (Pool) layer is used to carry out pond down-sampling, such as carries out maximum pond (Max Pooling), is averaged
Pond (Average Pooling).
Fig. 1 is the method flow diagram of the electricity demand forecasting of the embodiment of the present disclosure.
In a first aspect, referring to Fig.1, the embodiment of the present disclosure provides a kind of method of electricity demand forecasting comprising:
S101, acquisition historical information and Future Information, historical information include the additional letter for respectively corresponding M historical time section
Breath and practical electricity consumption, Future Information includes the additional information for respectively corresponding N number of future time section, and M and N are positive integer.
The additional information in the past M period (historical time section) and practical electricity consumption are obtained, and is obtained still
The additional information in N number of period (future time section) not arrived, to pass through these information predictions N number of future time section
Electricity consumption.
Wherein, historical time section is pass by, therefore its additional information and practical electricity consumption can be by existing records (as used
Electricity record) it obtains.And future time section does not arrive also, but its additional information can also be obtained by means such as regulation, predictions (after
It is continuous to be described in detail).
Wherein, each period (including historical time section and future time section) refers to continuous a period of time, specific
Length entity can be set as needed, for example, a few hours, one day, three days, one week, two weeks etc..And the variant period
Duration can not wait, but as preferred mode, the duration of each period used in each electricity demand forecasting process is phase
Deng, it is such as one day.
Wherein, multiple periods can be between have spaced, but as preferred mode, multiple historical time sections are to connect
Continuous, multiple future time sections are also continuous；And the historical time section after with respectively can be near preceding future time section
Adjacent (abutment points are current time), be also possible to have it is spaced, for example, historical time Duan Kewei future time section
The same period last year or same period last month etc..
Wherein, the number (i.e. M) of historical time section can also be able to be for one or more, the number (i.e. N) of future time section
It is one or more；Moreover, the selection of the above M and N be it is independent, i.e. M and N not can be not etc., but as preferred mode, M
It can be equal to N.
In some embodiments, M and N is the integer more than or equal to 2.
For the considerations of improving predictablity rate and efficiency, multiple historical time sections and multiple future times are preferably selected
Section.
Wherein, additional information is the information with time correlation, i.e., the additional information in each period is by the period
Concrete condition determine.Therefore, in each period, each project of additional information all has determining value；Without simultaneously
Between in section, the value of the identical items of additional information may be different.
Certainly, during each electricity demand forecasting, variant period (including historical time section and future time section)
The detailed programs of corresponding additional information should be identical.
In some embodiments, the additional information of corresponding each period includes: the Weather information of the period, and/or,
The holiday information of the period.
That is, the additional information of a period, it may include the weather conditions (Weather information) of period can also wrap
Include the festivals or holidays situation (holiday information) of the period.
Wherein, the festivals or holidays situation of each period (including historical time section and future time section) be can be according to regulation
Determining.The Weather information of each historical time section can be determined according to meteorological record, and the weather of each future time section is believed
The predicted value of weather forecast can be used in breath.
Wherein, Weather information may include many specific projects, such as temperature, humidity, wind-force, wind direction, rainfall, snowfall, ice
Hail, frost etc., and the concrete form of these projects, specific division mode are also multiplicity.
For example, about temperature, can use one of maximum temperature in a period of time, minimum temperature, mean temperature as
The temperature project of the period；Alternatively, can also be using maximum temperature, minimum temperature, mean temperature etc. as about temperature
Multiple and different projects.
For another example, about rainfall, it can only use in the period " whether having rainfall " as particular content (the i.e. rainfall of rainfall project
The value of project is only yes/no)；Alternatively, rainfall can also be regard rainfall grade (light rain, moderate rain, heavy rain etc.), rainfall etc. as
The particular content of project (i.e. the value of rainfall project can be heavy rain, 15 millimeters etc.).
Wherein, the concrete form of holiday information, specific division mode are also multiplicity.
For example, can only whether there is particular content (the i.e. festivals or holidays of red-letter day or holiday as holiday information using in the period
Only one project of information, value are only yes/no)；Alternatively, whether can also will have holiday in the period, whether have some tool
Body segment day etc. respectively as multiple and different festivals or holidays project (such as whether the holiday project for being, whether be the project on Christmas Day, be
The no project etc. for the Spring Festival)；Alternatively, can also be specially which using the period as the particular content of holiday information in red-letter day；Or
Person, for longer time section (such as one week), can also including festivals or holidays number as the specific of holiday information
Content.
S102, historical information and Future Information are inputted into preset convolutional neural networks, obtains convolutional neural networks output
Each future time section of correspondence prediction electricity consumption.
Using pre-set convolutional neural networks, historical information and Future Information are handled, processing result is
For the electricity consumption of each future time section, i.e. prediction electricity consumption.
In the embodiment of the present disclosure, electricity demand forecasting is carried out using convolutional neural networks, wherein input convolutional neural networks
Information further includes the additional information (such as Weather information, holiday information) of corresponding each period other than time and electricity consumption,
Consider the correlation between data variant under the same time not only to convolutional neural networks, it is also contemplated that data are at any time
The trend factor (correlation between same data under different time) of variation, i.e., its intersect point from two time, space dimensions
The correlation of data has been analysed, to have stronger ability to express, more accurate prediction can be made.
Further, convolutional neural networks are simple to similar date comparison, time series etc. as artificial intelligence system
Prediction technique is compared, and prediction accuracy is high, scalability is good；And convolutional neural networks and recurrent neural network, length time gate
It is compared Deng others artificial intelligence system, training is easier, and is required data and network parameter lower.
Below in the embodiment of the present disclosure using convolutional neural networks in the form of be specifically introduced.
Referring to Fig. 2, Fig. 3, in some embodiments, the above convolutional neural networks include:
First conformable layer, for historical information to be integrated into history image, in history image, each historical time section is corresponding
One column pixel, history electricity consumption, additional information each project correspond to one-row pixels；
Second conformable layer, for Future Information to be integrated into future image, in future image, each future time section is corresponding
Each project of one column pixel, additional information corresponds to one-row pixels.
The above historical information, Future Information substantially comprise two dimensions, and a dimension is " time ", another dimension is
" feature (space in other words) ", i.e., it is two-dimensional information, therefore it can be converted into two-dimensional image.
Specifically, referring to Fig. 4, the first conformable layer is a column with each period, using history electricity consumption as a line, additional letter
Each project of breath is a line, and historical information is integrated into the format of " image ".The width of history image is that M (has M as a result,
Column), wherein the coordinate of each pixel is (period, characteristic item), and the pixel value of each pixel is its corresponding period
The value of characteristic item (project of history electricity consumption or additional information): for example, the pixel value of the pixel of (May 8, temperature project)
It can be " 28 degrees Celsius " that the pixel value of (the 3rd week, history electricity consumption) pixel can be " 50000 kilowatt hour " etc..
Similar, referring to Fig. 5, the second conformable layer is also a column (therefore shared N column) with each period, with additional information
Each project is a line, and Future Information is integrated into the format of " image ".The width of future image is N (having N column) as a result,
And height (line number) fewer than the height of history image 1, that is, lack the row of electricity consumption.
Certainly, in the above history image and future image, in addition to the row of electricity consumption, the row of the row of projects of additional information
Column mode, sequence etc. are preferably identical.
Convolutional neural networks are usually for being handled image, and above procedure is equivalent to historical information and not
Carry out information to change respectively for two images, and wherein future image a line fewer than history image；Therefore convolutional neural networks
Work is equivalent to according to history image (row including electricity consumption) and future image (row of no electricity consumption), " prediction " following figure
The row of the electricity consumption of picture predicts the following electricity consumption.
Referring to Fig. 2, Fig. 3, in some embodiments, convolutional neural networks further include:
The first convolution process layer connecting with the output end of the first conformable layer, obtains one-dimensional for handling history image
The element number of primary vector, primary vector is equal to M；
The second convolution process layer connecting with the output end of the second conformable layer, obtains one-dimensional for handling future image
The element number of secondary vector, secondary vector is equal to N.
After obtaining history image and future image, respectively they can be carried out to include convolution by two process of convolution layers
Processing, and guarantee that final process obtained is the one-dimensional vector that length is respectively M and N (width of image before being equal to).
Certainly, one-dimensional vector actually also can be considered that width is respectively M and N, highly for 1 image, therefore it can also regard as by original image
It is obtained along short transverse " compression ".
The above process of convolution considers the information of time and feature (space) two dimensions simultaneously, therefore can preferably extract letter
Feature in breath improves prediction accuracy.
It certainly, is realization convolution, it is inevitable in process of convolution layer (including the first convolution process layer and second convolution process layer)
Concrete form including convolutional layer, but process of convolution layer can be multiplicity.For example, referring to Fig. 6, each process of convolution layer may be used also
Including the normalization layer for carrying out batch normalization (Batch Normalization), for carrying out the excitation of Relu excitation
Layer etc..Wherein, batch normalization layer can be such that convolutional neural networks stablize faster in training.
Fig. 2, Fig. 3, in some embodiments, convolutional neural networks further include:
The merging layer being connect with the output end of the output end of the first convolution process layer and the second convolution process layer, for by the
One vector and secondary vector merge into one-dimensional third vector, and the element number of third vector is equal to M+N.
After process of convolution layer, historical information and Future Information have been compressed to primary vector and secondary vector respectively,
Therefore the two vectors can be merged into the one-dimensional vector (third vector) of a length longer (M+N) by merging layer.
Certainly, it is various for merging layer to carry out the concrete mode of vector merging, such as two vector first places can connect, or
Person can also upset the element in two vectors and cross arrangement.
Fig. 2, Fig. 3, in some embodiments, convolutional neural networks further include:
With the residual error network layer that connect of output end for merging layer, the N number of one-dimensional 4th is obtained for handling third vector
Vector；Residual error network layer includes multiple sequentially connected residual blocks, and each residual block includes multiple convolutional layers, and each residual block
Input terminal and output end between be equipped with direct-connected connection.
To the one-dimensional third vector that merging obtains, can continue to handle it with residual error network, and final output is N number of
One-dimensional vector (the 4th vector).
Wherein, residual error network layer includes multiple residual blocks, has multiple convolutional layers in each residual block, and be inputted
Hold the direct-connected connection connected with output end.By direct-connected connection, the data for having across part convolutional layer in residual error network can be made
Connection, so that the problems such as gradient caused by avoiding because of network depth increase disappears, gradient explosion, makes network depth that can increase
Add, improves feature extraction effect.
Certainly, the concrete form of the concrete form of residual error network layer and each residual block is various.For example, referring to
Fig. 7 may include two convolutional layers in each residual block, and be connected to two convolution interlayers for carrying out batch normalization
Normalization layer, the excitation layer for carrying out Relu excitation, random inactivation (Dropcut) layer of (Batch Normalization)
Deng.Wherein, the effect of random deactivating layer mainly reduces the interdependency between node, prevents overfitting；And direct-connected connection
Concrete form be also multiplicity, may be, for example, maximum pond (Max Pooling).
Fig. 2, Fig. 3, in some embodiments, convolutional neural networks further include:
The output layer being connect with the output end of residual error network layer, for by N number of 4th vector be changed into one it is one-dimensional defeated
The element number of outgoing vector, output vector is equal to N, wherein each element is the corresponding prediction electricity consumption of a future time section.
To P the 4th one-dimensional vectors of residual error network layer output, being processed into a length by output layer is N's
One-dimensional vector (output vector), each element of the output vector as predict the electricity consumption of obtained one future time section of correspondence
Amount (therefore being altogether N number of).
Certainly, the concrete form of output layer is various.For example, referring to Fig. 8, output layer may include a normalization layer
(as carried out a batch normalization) and pond layer (such as carrying out average pond).
Certainly, referring to Fig. 2, Fig. 3, convolutional neural networks may also include other layers, such as connecing in addition to above each layer
Receive historical information and the input layer of Future Information etc..
Referring to Fig. 9, in some embodiments, the step of obtaining historical information and Future Information above (S101 step) it
Before, further includes:
S100, convolutional neural networks are trained.
That is, can be trained in advance to convolutional neural networks, with the convolutional neural networks being had excellent performance.When
So, if the convolutional neural networks in step S101 are obtained and feasible by other ways such as duplications.
The training of convolutional neural networks may include multiple training circulations, and it is refreshing to enter information into convolution in recycling for each training
It is handled through network and by convolutional neural networks, processing result is evaluated later, and convolution is adjusted according to evaluation result
The parameter of neural network.By largely training circulation, constantly convolutional neural networks can be adjusted, improve its performance until
It restrains or reaches expected.
Due to needing to evaluate processing result, therefore the historical information that different historical time sections can be selected carries out convolution mind
Training through network.
It, can will be defeated as historical information in the practical electricity consumption of preceding historical time section, additional information specifically, referring to Fig.1 0
Enter convolutional neural networks, and convolutional neural networks will be inputted as Future Information in the additional information of rear historical time section, thus
The output of convolutional neural networks is its prediction in rear historical time section electricity consumption；Later by the prediction in rear historical time section
Electricity consumption (prediction result) is compared with the practical electricity consumption (standard results) in rear historical time section, and according to comparison result
The parameter for adjusting convolutional neural networks, completes the training of convolutional neural networks.
Figure 11 is the composition block diagram of the device of the electricity demand forecasting of the embodiment of the present disclosure.
Second aspect, referring to Fig.1 1, the embodiment of the present disclosure provides a kind of device of electricity demand forecasting comprising:
Module is obtained, for obtaining historical information and Future Information, historical information includes respectively corresponding M historical time section
Additional information and practical electricity consumption, Future Information includes the additional information for respectively corresponding N number of future time section, and M and N are positive
Integer；
Convolutional neural networks module is obtained for historical information and Future Information to be inputted preset convolutional neural networks
The prediction electricity consumption of each future time section of correspondence of convolutional neural networks output.
In some embodiments, the additional information of corresponding each period includes:
The Weather information of the period, and/or, the holiday information of the period.
In some embodiments, M and N is the integer more than or equal to 2.
Referring to Fig.1 2, in some embodiments, convolutional neural networks module includes:
First integral unit, for historical information to be integrated into history image, in history image, each historical time section is right
Ying Yilie pixel, history electricity consumption, additional information each project correspond to one-row pixels；
Second integral unit, for Future Information to be integrated into future image, in future image, each future time section is right
Each project of Ying Yilie pixel, additional information corresponds to one-row pixels.
Referring to Fig.1 2, in some embodiments, convolutional neural networks module further include:
The first convolution processing unit being connect with the output end of the first integral unit, for handling history image to obtain one
The element number of the primary vector of dimension, primary vector is equal to M；
The second convolution processing unit being connect with the output end of the second integral unit, for handling future image to obtain one
The element number of the secondary vector of dimension, secondary vector is equal to N；
The combining unit connecting with the output end of the output end of the first convolution processing unit and the second convolution processing unit is used
In primary vector and secondary vector are merged into one-dimensional third vector, the element number of third vector is equal to M+N.
Referring to Fig.1 2, in some embodiments, convolutional neural networks module further include:
The residual error network unit connecting with the output end of combining unit obtains N number of one-dimensional for handling third vector
4th vector；Residual error network unit includes multiple sequentially connected residual blocks, and each residual block includes multiple convolutional layers, and each
Direct-connected connection is equipped between the input terminal and output end of residual block；
The output unit being connect with the output end of residual error network unit, for by N number of 4th vector be changed into one it is one-dimensional
Output vector, the element number of output vector is equal to N, wherein each element is the corresponding prediction electricity consumption of a future time section
Amount.
The third aspect, the embodiment of the present disclosure provide a kind of electronic equipment comprising:
One or more processors；
Storage device is stored thereon with one or more programs, when one or more programs are by one or more processors
It executes, so that the method that one or more processors realize any one of the above electricity demand forecasting.
Fourth aspect, the embodiment of the present disclosure provide a kind of computer-readable medium, are stored thereon with computer program, program
The method of any one of the above electricity demand forecasting is realized when being executed by processor.
It will appreciated by the skilled person that whole or certain steps, system, dress in method disclosed hereinabove
Functional module/unit in setting may be implemented as software, firmware, hardware and its combination appropriate.In hardware embodiment,
Division between the functional module/unit referred in the above description not necessarily corresponds to the division of physical assemblies；For example, one
Physical assemblies can have multiple functions or a function or step and can be executed by several physical assemblies cooperations.Certain objects
Reason component or all physical assemblies may be implemented as by processor, such as central processing unit, digital signal processor or micro process
The software that device executes, is perhaps implemented as hardware or is implemented as integrated circuit, such as specific integrated circuit.Such software
Can be distributed on a computer-readable medium, computer-readable medium may include computer storage medium (or non-transitory be situated between
Matter) and communication media (or fugitive medium).As known to a person of ordinary skill in the art, term computer storage medium includes
In any method or skill for storing information (such as computer readable instructions, data structure, program module or other data)
The volatile and non-volatile implemented in art, removable and nonremovable medium.Computer storage medium includes but is not limited to
RAM, ROM, EEPROM, flash memory or other memory technologies, CD-ROM, digital versatile disc (DVD) or other optical disc storages, magnetic
Box, tape, disk storage or other magnetic memory apparatus or it can be used for storing desired information and can be visited by computer
Any other medium asked.In addition, known to a person of ordinary skill in the art be, communication media generally comprises computer-readable
Other numbers in the modulated data signal of instruction, data structure, program module or such as carrier wave or other transmission mechanisms etc
According to, and may include any information delivery media.
Example embodiment has been disclosed in the disclosure, although and use concrete term, they are only used for simultaneously only answering
When being interpreted general remark meaning, and it is not used in the purpose of limitation.In some instances, aobvious to those skilled in the art
And be clear to, unless otherwise expressly stated, otherwise can be used alone the feature that description is combined with specific embodiment, characteristic
And/or element, or the feature, characteristic and/or element of description can be combined with other embodiments and be applied in combination.Therefore, this field
The skilled person will understand that can be carried out various in the case where not departing from the scope of the present disclosure illustrated by the attached claims
Change in form and details.
Claims (14)
1. a kind of method of electricity demand forecasting comprising:
Obtain historical information and Future Information, the historical information include respectively correspond the additional information of M historical time section with
Practical electricity consumption, the Future Information include the additional information for respectively corresponding N number of future time section, and the M and N are positive whole
Number；
The historical information and Future Information are inputted into preset convolutional neural networks, obtain the convolutional neural networks output
The prediction electricity consumption of corresponding each future time section.
2. according to the method described in claim 1, wherein, the additional information of corresponding each period includes:
The Weather information of the period, and/or, the holiday information of the period.
3. according to the method described in claim 1, wherein,
The M and N is the integer more than or equal to 2.
4. according to the method described in claim 1, wherein, the convolutional neural networks include:
First conformable layer, for the historical information to be integrated into history image, in the history image, when each history
Between the corresponding column pixel of section, the history electricity consumption, the additional information each project correspond to one-row pixels；
Second conformable layer, for the Future Information to be integrated into future image, in the future image, when each described following
Between the corresponding column pixel of section, each project of the additional information corresponds to one-row pixels.
5. according to the method described in claim 4, wherein, the convolutional neural networks further include:
The first convolution process layer being connect with the output end of first conformable layer, for handling the history image to obtain one
The element number of the primary vector of dimension, the primary vector is equal to M；
The second convolution process layer being connect with the output end of second conformable layer, for handling the future image to obtain one
The element number of the secondary vector of dimension, the secondary vector is equal to N；
The merging layer connecting with the output end of the output end of the first convolution process layer and the second convolution process layer, is used for
The primary vector and secondary vector are merged into one-dimensional third vector, the element number of the third vector is equal to M+N.
6. according to the method described in claim 5, wherein, the convolutional neural networks further include:
The residual error network layer being connect with the output end for merging layer, it is N number of one-dimensional to obtain for handling the third vector
4th vector；The residual error network layer includes multiple sequentially connected residual blocks, and each residual block includes multiple convolutional layers,
And direct-connected connection is equipped between the input terminal and output end of each residual block；
The output layer being connect with the output end of the residual error network layer, for by N number of 4th vector be changed into one it is one-dimensional
Output vector, the element number of the output vector is equal to N, wherein each element is the corresponding prediction of a future time section
Electricity consumption.
7. a kind of device of electricity demand forecasting comprising:
Module is obtained, for obtaining historical information and Future Information, the historical information includes respectively corresponding M historical time section
Additional information and practical electricity consumption, the Future Information includes the additional information for respectively corresponding N number of future time section, the M
It is positive integer with N；
Convolutional neural networks module is obtained for the historical information and Future Information to be inputted preset convolutional neural networks
The prediction electricity consumption of each future time section of correspondence of the convolutional neural networks output.
8. device according to claim 7, wherein the additional information of corresponding each period includes:
The Weather information of the period, and/or, the holiday information of the period.
9. device according to claim 7, wherein
The M and N is the integer more than or equal to 2.
10. device according to claim 7, wherein the convolutional neural networks module includes:
First integral unit, for the historical information to be integrated into history image, in the history image, each history
Period correspond to a column pixel, the history electricity consumption, the additional information each project correspond to one-row pixels；
Second integral unit, for the Future Information to be integrated into future image, in the future image, each future
Period corresponds to a column pixel, and each project of the additional information corresponds to one-row pixels.
11. device according to claim 10, wherein the convolutional neural networks module further include:
The first convolution processing unit being connect with the output end of first integral unit, for handling the history image to obtain
Element number to one-dimensional primary vector, the primary vector is equal to M；
The second convolution processing unit being connect with the output end of second integral unit, for handling the future image to obtain
Element number to one-dimensional secondary vector, the secondary vector is equal to N；
The combining unit connecting with the output end of the output end of first convolution processing unit and the second convolution processing unit is used
In the primary vector and secondary vector are merged into one-dimensional third vector, the element number of the third vector is equal to M+N.
12. device according to claim 11, wherein the convolutional neural networks module further include:
The residual error network unit being connect with the output end of the combining unit, for handling the third vector to obtain N number of one
4th vector of dimension；The residual error network unit includes multiple sequentially connected residual blocks, and each residual block includes multiple
Convolutional layer, and direct-connected connection is equipped between the input terminal and output end of each residual block；
The output unit being connect with the output end of the residual error network unit, for N number of 4th vector to be changed into one
The element number of one-dimensional output vector, the output vector is equal to N, wherein each element is that a future time section is corresponding
Predict electricity consumption.
13. a kind of electronic equipment comprising:
One or more processors；
Storage device is stored thereon with one or more programs, when one or more of programs are by one or more of places
It manages device to execute, so that one or more of processors are realized according to claim 1 to electricity demand forecasting described in 6 any one
Method.
14. a kind of computer-readable medium is stored thereon with computer program, basis is realized when described program is executed by processor
The method of electricity demand forecasting described in claim 1 to 6 any one.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910482201.2A CN110210672A (en) | 2019-06-04 | 2019-06-04 | The method and apparatus of electricity demand forecasting, electronic equipment, computer-readable medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910482201.2A CN110210672A (en) | 2019-06-04 | 2019-06-04 | The method and apparatus of electricity demand forecasting, electronic equipment, computer-readable medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110210672A true CN110210672A (en) | 2019-09-06 |
Family
ID=67790733
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910482201.2A Pending CN110210672A (en) | 2019-06-04 | 2019-06-04 | The method and apparatus of electricity demand forecasting, electronic equipment, computer-readable medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110210672A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111539841A (en) * | 2019-12-31 | 2020-08-14 | 远景智能国际私人投资有限公司 | Electric quantity prediction method, device, equipment and readable storage medium |
| CN112132211A (en) * | 2020-09-20 | 2020-12-25 | 郑州精铖能源技术有限公司 | Efficient cascade energy management method and system |
| CN113469396A (en) * | 2020-03-30 | 2021-10-01 | 海南博川电力设计工程有限公司 | Method for predicting short-term power load of power system |
| CN114978292A (en) * | 2022-06-24 | 2022-08-30 | 广州爱浦路网络技术有限公司 | Satellite network connection method, system, electronic equipment and storage medium |
| CN117495434A (en) * | 2023-12-25 | 2024-02-02 | 天津大学 | Electric energy demand prediction method, model training method, device and electronic equipment |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102930354A (en) * | 2012-11-06 | 2013-02-13 | 北京国电通网络技术有限公司 | Method and device for predicating electricity consumption of residential area |
| CN105260803A (en) * | 2015-11-06 | 2016-01-20 | 国家电网公司 | Power consumption prediction method for system |
| US20160321750A1 (en) * | 2015-04-30 | 2016-11-03 | Fujitsu Limited | Commodity price forecasting |
| CN108876070A (en) * | 2018-09-25 | 2018-11-23 | 新智数字科技有限公司 | A kind of method and apparatus that Load Prediction In Power Systems are carried out based on neural network |
| CN108921225A (en) * | 2018-07-10 | 2018-11-30 | 深圳市商汤科技有限公司 | A kind of image processing method and device, computer equipment and storage medium |
| CN109543901A (en) * | 2018-11-20 | 2019-03-29 | 国网辽宁省电力有限公司经济技术研究院 | Short-Term Load Forecasting Method based on information fusion convolutional neural networks model |
-
2019
- 2019-06-04 CN CN201910482201.2A patent/CN110210672A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102930354A (en) * | 2012-11-06 | 2013-02-13 | 北京国电通网络技术有限公司 | Method and device for predicating electricity consumption of residential area |
| US20160321750A1 (en) * | 2015-04-30 | 2016-11-03 | Fujitsu Limited | Commodity price forecasting |
| CN105260803A (en) * | 2015-11-06 | 2016-01-20 | 国家电网公司 | Power consumption prediction method for system |
| CN108921225A (en) * | 2018-07-10 | 2018-11-30 | 深圳市商汤科技有限公司 | A kind of image processing method and device, computer equipment and storage medium |
| CN108876070A (en) * | 2018-09-25 | 2018-11-23 | 新智数字科技有限公司 | A kind of method and apparatus that Load Prediction In Power Systems are carried out based on neural network |
| CN109543901A (en) * | 2018-11-20 | 2019-03-29 | 国网辽宁省电力有限公司经济技术研究院 | Short-Term Load Forecasting Method based on information fusion convolutional neural networks model |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111539841A (en) * | 2019-12-31 | 2020-08-14 | 远景智能国际私人投资有限公司 | Electric quantity prediction method, device, equipment and readable storage medium |
| EP4085387A4 (en) * | 2019-12-31 | 2023-06-07 | Envision Digital International Pte. Ltd. | Method and apparatus for predicting power consumption, device and readiable storage medium |
| CN113469396A (en) * | 2020-03-30 | 2021-10-01 | 海南博川电力设计工程有限公司 | Method for predicting short-term power load of power system |
| CN112132211A (en) * | 2020-09-20 | 2020-12-25 | 郑州精铖能源技术有限公司 | Efficient cascade energy management method and system |
| CN114978292A (en) * | 2022-06-24 | 2022-08-30 | 广州爱浦路网络技术有限公司 | Satellite network connection method, system, electronic equipment and storage medium |
| CN117495434A (en) * | 2023-12-25 | 2024-02-02 | 天津大学 | Electric energy demand prediction method, model training method, device and electronic equipment |
| CN117495434B (en) * | 2023-12-25 | 2024-04-05 | 天津大学 | Electric energy demand prediction method, model training method, device and electronic equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110210672A (en) | The method and apparatus of electricity demand forecasting, electronic equipment, computer-readable medium | |
| Li et al. | Prediction for tourism flow based on LSTM neural network | |
| CN110866190B (en) | Method and device for training neural network model for representing knowledge graph | |
| US11227190B1 (en) | Graph neural network training methods and systems | |
| Saeed et al. | Hybrid bidirectional LSTM model for short-term wind speed interval prediction | |
| CN109345130B (en) | Method and device for commercial site selection, computer equipment and storage medium | |
| CN111242292B (en) | OD data prediction method and system based on deep space-time network | |
| WO2018222308A1 (en) | Time-based features and moving windows sampling for machine learning | |
| CN111539841A (en) | Electric quantity prediction method, device, equipment and readable storage medium | |
| CN111985622A (en) | Graph neural network training method and system | |
| CN113657607B (en) | Continuous learning method for federal learning | |
| CN115170565B (en) | Image fraud detection method and device based on automatic neural network architecture search | |
| CN110059066A (en) | The method of spark combination tensorflow progress remote sensing image information extraction | |
| CN108897757A (en) | A kind of photo storage method, storage medium and server | |
| CN113870422A (en) | Pyramid Transformer-based point cloud reconstruction method, device, equipment and medium | |
| CN111242395A (en) | Method and device for constructing prediction model for OD (origin-destination) data | |
| CN115810149A (en) | High-resolution remote sensing image building extraction method based on superpixel and image convolution | |
| CN117688984A (en) | Neural network structure searching method, device and storage medium | |
| CN117689731B (en) | Lightweight new energy heavy-duty battery pack identification method based on improved YOLOv model | |
| CN113569960B (en) | Small sample image classification method and system based on domain adaptation | |
| CN111726592B (en) | Method and apparatus for obtaining architecture of image signal processor | |
| CN115238134A (en) | Method and apparatus for generating a graph vector representation of a graph data structure | |
| Xi et al. | Continual learning for scene classification of high resolution remote sensing images | |
| CN111242394B (en) | Method and system for extracting spatial correlation characteristics | |
| CN116310391B (en) | Identification method for tea diseases |
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 |