CN109063902A - A kind of short-term load forecasting method, device, equipment and storage medium - Google Patents
A kind of short-term load forecasting method, device, equipment and storage medium Download PDFInfo
- Publication number
- CN109063902A CN109063902A CN201810784848.6A CN201810784848A CN109063902A CN 109063902 A CN109063902 A CN 109063902A CN 201810784848 A CN201810784848 A CN 201810784848A CN 109063902 A CN109063902 A CN 109063902A
- Authority
- CN
- China
- Prior art keywords
- dbn network
- short
- dbn
- subsequence
- prediction model
- 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
- 238000013277 forecasting method Methods 0.000 title claims abstract description 24
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 23
- 239000002245 particle Substances 0.000 claims description 37
- 230000002354 daily effect Effects 0.000 claims description 20
- 239000011159 matrix material Substances 0.000 claims description 14
- 230000003203 everyday effect Effects 0.000 claims description 9
- 238000013528 artificial neural network Methods 0.000 claims description 8
- 230000006978 adaptation Effects 0.000 claims description 4
- 230000000694 effects Effects 0.000 abstract description 4
- 239000000523 sample Substances 0.000 description 63
- 238000000034 method Methods 0.000 description 12
- 238000010586 diagram Methods 0.000 description 6
- 239000008186 active pharmaceutical agent Substances 0.000 description 5
- 241001269238 Data Species 0.000 description 4
- 240000002853 Nelumbo nucifera Species 0.000 description 4
- 235000006508 Nelumbo nucifera Nutrition 0.000 description 4
- 235000006510 Nelumbo pentapetala Nutrition 0.000 description 4
- 230000008901 benefit Effects 0.000 description 4
- 238000005457 optimization Methods 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 238000004590 computer program Methods 0.000 description 3
- 241000083513 Punctum Species 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 238000003062 neural network model Methods 0.000 description 2
- 230000009471 action Effects 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 238000002360 preparation method Methods 0.000 description 1
- 230000000750 progressive effect Effects 0.000 description 1
- 230000000717 retained effect Effects 0.000 description 1
- 239000012488 sample solution Substances 0.000 description 1
- 230000002123 temporal effect Effects 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
- G06Q10/04—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
- 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
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Economics (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Marketing (AREA)
- General Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Public Health (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Water Supply & Treatment (AREA)
- Development Economics (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
A kind of short-term load forecasting method disclosed by the invention, historical load data is obtained first, integrated empirical mode decomposition is then based on historical load data is decomposed to obtain multiple subsequences, secondly training sample is chosen and using training sample training DBN network to obtain DBN Network Prediction Model, finally short term is predicted using DBN Network Prediction Model and each subsequence, using this programme, subsequence is obtained due to decompose historical load data, the complex characteristics (nonlinearity and non-stationary) of load data are reduced to a certain extent, therefore, short term is predicted using the subsequence of DBN Network Prediction Model and the load data of low complex characteristics, this programme is higher to the precision of prediction of short term, very big guarantee is provided to the normal operation of operation of power networks scheduling.In addition, the invention also discloses a kind of short-term load forecasting device, equipment and storage medium, effect are as above.
Description
Technical field
The present invention relates to technical field of electric power, in particular to a kind of short-term load forecasting method, device, equipment and storage are situated between
Matter.
Background technique
Load prediction be operation of power networks scheduling pith, be to ensure that electric system can accomplish safety but also take into account through
The important evidence of Ji operation, load forecast will not only consider the variation of load, it is also contemplated that temperature, humidity etc. influence load
The factor of variation, therefore, it is vital for establishing perfect load forecasting model.
Currently, the prediction model for short-term load forecasting includes: time series models, artificial intelligence model, hybrid guided mode
Type, BP neural network model and DBN network model, compared with traditional BP neural network model, depth confidence network (DBN)
More hidden layer network structures and parameter can be trained using greedy successively training method, traditional BP neural network is efficiently solved and be difficult to
The problem of more implicit layer networks is established, and compared with conventional DBN network model, using the DBN after the optimization of crossover algorithm in length and breadth
Network model compensates for the defect that network parameter falls into local optimum, improves the generalization ability of DBN network, so can be used for short
Phase load prediction.However, the nonlinearity having due to load sequence and non-stationary complex characteristics, and DBN network mould
Type (DBN network model or using the DBN network model after crossover algorithm optimization in length and breadth) is difficult to handle the load of high complex characteristics
Sequence, therefore (DBN network model uses the DBN network mould after crossover algorithm optimization in length and breadth using single DBN network model
Type) it is not very accurate to the prediction of short term, to influence the normal operation that operation of power networks is dispatched.
Therefore, how to improve the accuracy predicted short term is with the normal operation for guaranteeing operation of power networks scheduling
Those skilled in the art's problem to be solved.
Summary of the invention
It is an object of the invention to disclose a kind of short-term load forecasting method, device, equipment and storage medium, improve pair
The accuracy that short term is predicted ensure that the normal operation of operation of power networks scheduling.
To achieve the above object, the embodiment of the invention discloses following technical solutions:
First, the embodiment of the invention discloses a kind of short-term load forecasting methods, comprising:
Obtain historical load data；
The historical load data is decomposed based on integrated empirical mode decomposition to obtain multiple subsequences；
Choose training sample and using training sample training DBN network to obtain DBN Network Prediction Model；
Short term is predicted using the DBN Network Prediction Model and each subsequence.
Preferably, described that the historical load data is decomposed to obtain sub-series of packets based on integrated empirical mode decomposition
It includes:
Gaussian sequence is added in daily time series corresponding with the historical load data；
It is IMF component by the daily Time Series for adding the Gaussian sequence；
The first difference of the daily time series and the IMF component of adding the Gaussian sequence is calculated to incite somebody to action
First difference is as residual components；
Using the residual components as time series to be decomposed and repeat the above steps until final residual components are less than
Or be equal to preset value when stop decompose, finally obtain multiple IMF components corresponding with the daily time series and finally
Residual components, corresponding subsequence every day in the historical load data, each subsequence include: with it is described every
The corresponding multiple IMF components of it time series and final residual components.
Preferably, described to include: for IMF component by the daily Time Series for adding the Gaussian sequence
The maximum point and minimum point of the daily time series of the Gaussian sequence are added in identification；
It is fitted and corresponding first envelope of the maximum point and the second envelope corresponding with the minimum point；
Utilize first envelope and the second envelope line computation target surplus；
Calculate the second difference of the time series and the target surplus of adding the Gaussian sequence；
Judge whether second difference meets IMF condition；
If so, using second difference as IMF component；
If it is not, then steps be repeated alternatively until that second difference meets the IMF condition.
Preferably, the selection training sample and using the training sample training DBN network to obtain DBN neural network forecast
Model includes:
Training sample is chosen from historical load data.
The training parameter of the DBN network is determined according to subsequence corresponding with the training sample；
The optimal solution of the DBN network is determined using the training parameter and the training sample；
Using the optimal solution as weight corresponding with the DBN network to obtain the DBN Network Prediction Model, sub- sequence
A IMF component and final residual components in column have unique corresponding DBN Network Prediction Model.
Preferably, the optimal solution that the DBN network is determined using the training parameter and the training sample includes:
Determine the initial population scale and maximum number of iterations in the training parameter；
Using subsequence corresponding with the training sample as the input of the DBN network；
Particle to be optimized is encoded and generates initial population corresponding with the initial population scale；
Calculate the fitness of each encoded particles in the initial population；
Lateral cross is carried out to the encoded particles in the initial population based on crossover algorithm in length and breadth and crossed longitudinally is obtained
Kind mass matrix；
Calculate the fitness of the particle in described kind of mass matrix and with each encoded particles in the initial population
Fitness is compared to choose target fitness；
Using intended particle corresponding with the target fitness as the individual in progeny population；
The number of iterations be steps be repeated alternatively until beyond the maximum number of iterations, by finally obtained maximum adaptation degree pair
The particle answered is as the optimal solution.
Preferably, described prediction is carried out to short term using the DBN Network Prediction Model and the subsequence to include:
Forecast sample is chosen from the historical load data；
Using subsequence corresponding with the forecast sample as the input of the DBN Network Prediction Model；
It take the subsequence as the input of the DBN Network Prediction Model, using weight corresponding with the optimal solution as institute
The weight for stating DBN Network Prediction Model determines that the reality output of the DBN Network Prediction Model, the reality output are to measure in advance
The short term arrived.
Second, the embodiment of the invention discloses a kind of short-term load forecasting devices, comprising:
Module is obtained, for obtaining historical load data；
Decomposing module, for being decomposed to obtain multiple sons to the historical load data based on set empirical mode decomposition
Sequence；
Module is chosen, for choosing training sample and training DBN network pre- to obtain DBN network using the training sample
Survey model；
Prediction module, for being predicted using the DBN Network Prediction Model and each subsequence short term.
Preferably, the selection module includes:
Selection unit, for choosing training sample from the historical load data；
First determination unit, for determining the training of the DBN network according to subsequence corresponding with the training sample
Parameter；
Second determination unit, for determining the optimal of the DBN network using the training parameter and the training sample
Solution；
Setup unit, for using the optimal solution as weight corresponding with the DBN network to obtain the DBN network
Prediction model, each IMF component and final residual components in the subsequence have unique corresponding DBN neural network forecast mould
Type.Third, the embodiment of the invention discloses a kind of short-term load forecasting equipment, comprising:
Memory is used for Storage Estimation program；
Processor is realized short-term as described in any of the above for executing the Prediction program stored in the memory
The step of load forecasting method.
4th, the embodiment of the invention discloses a kind of computer readable storage medium, deposited on computer readable storage medium
Computer program is contained, any short-term load forecasting method as above is realized when computer program is executed by processor
Step.
As it can be seen that a kind of short-term load forecasting method disclosed by the embodiments of the present invention, first acquisition historical load data, then
Historical load data is decomposed based on integrated empirical mode decomposition to obtain multiple subsequences, secondly chooses training sample and benefit
With training sample training DBN network to obtain DBN Network Prediction Model, DBN Network Prediction Model and each subsequence are finally utilized
Short term is predicted, it is middle compared with the prior art that highly complex feature can not be handled using single DBN prediction model
Load data and cause the problem for being easy to cause precision of prediction low, using this programme, since historical load data being carried out
Decomposition obtains multiple subsequences, reduces complex characteristics (nonlinearity and the non-stationary of load data to a certain extent
Property), therefore, short term is predicted using the subsequence of DBN Network Prediction Model and the load data of low complex characteristics,
This programme is higher to the precision of prediction of short term, provides very big guarantee to the normal operation of operation of power networks scheduling.In addition,
The embodiment of the invention also discloses a kind of short-term load forecasting device, equipment and storage medium, effect are as above.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is only this
Some embodiments of invention for those of ordinary skill in the art without creative efforts, can be with
It obtains other drawings based on these drawings.
Fig. 1 is a kind of short-term load forecasting method flow diagram disclosed by the embodiments of the present invention；
Fig. 2 is a kind of short-term load forecasting apparatus structure schematic diagram disclosed by the embodiments of the present invention；
Fig. 3 is a kind of short-term load forecasting device structure schematic diagram disclosed by the embodiments of the present invention.
Specific embodiment
Following will be combined with the drawings in the embodiments of the present invention, and technical solution in the embodiment of the present invention carries out clear, complete
Site preparation description, it is clear that described embodiments are only a part of the embodiments of the present invention, instead of all the embodiments.It is based on
Embodiment in the present invention, it is obtained by those of ordinary skill in the art without making creative efforts every other
Embodiment shall fall within the protection scope of the present invention.
The embodiment of the invention discloses a kind of short-term load forecasting method, device, equipment and storage mediums, improve to short
The accuracy that phase load is predicted ensure that the normal operation of operation of power networks scheduling.
Referring to Figure 1, Fig. 1 is a kind of short-term load forecasting method flow diagram disclosed by the embodiments of the present invention, the party
Method includes:
S101, historical load data is obtained.
Specifically, historical load data includes: any a period of time (present invention before current time in the present embodiment
In embodiment be preferably 1 year) Power system load data and weather data.It is small that the temporal resolution of historical load data is set as 1
When, that is, the time series that time interval is 1 hour is formed, i.e., one day includes that (data point includes time point to 24 data points
With electric load corresponding with the time point).
S102, historical load data is decomposed based on integrated empirical mode decomposition to obtain multiple subsequences.
Specifically, the decomposition principle of integrated empirical mode decomposition may refer to the prior art, in the present invention in the present embodiment
In embodiment, adaptively the load sequence of every day in historical load data can be decomposed using integrated empirical mode decomposition
For the intrinsic mode functions (IMF component) and a final surplus of a series of frequencies from high to low.It is wrapped in historical load data
Containing many days load datas, a subsequence will can be corresponded to every day of historical load data, it can in each subsequence
To include multiple IMF components and a final IMF component.In the embodiment of the present invention, preferably by the time series of every day
It is decomposed into 7 IMF components and 1 final IMF component.Certainly, difference according to the actual situation, daily time series may be used also
To be divided into more IMF components.
Wherein, as preferred embodiment, step S102 includes:
Gaussian sequence is added in daily time series corresponding with historical load data.
It is IMF component by the daily Time Series for adding Gaussian sequence.
The daily time series of addition Gaussian sequence and the first difference of IMF component are calculated with by the first difference
As residual components.
Using residual components as time series to be decomposed and repeat the above steps until final residual components are beyond default
Stop decomposing when value, finally obtain multiple IMF components corresponding with daily time series and final residual components, history is negative
Corresponding subsequence every day in lotus data, each subsequence includes multiple IMFs corresponding with daily time series points
Amount and final residual components.
Specifically, in the present embodiment, it is assumed that time series corresponding with historical load data is x (t), the height added at random
This white noise sequence is nm(t), then Gaussian sequence n is addedm(t) time series xm(t) it can be indicated with following formula:
xm(t)=x (t)+nm(t)
Integrated empirical mode decomposition (EMD (decomposition principle may refer to the prior art)) is then based on to time series xm
(t) it is decomposed, in each decompose by time series xm(t) it is decomposed to obtain an IMF component lm1(t), then with adding
The time series x of Gaussian sequence is addedm(t) it subtracts this decomposition and obtains IMF component lm1(t), residual components (are obtained
One difference) rm(t).Specific formula is as follows:
rm(t)=xm(t)-lm1(t)
Then, the residual components r to obtainmIt (t) is new time series (time series to be decomposed), then to residue point
Measure rm(t) repeat to decompose according to above-mentioned steps, until residual components rm(t) when being less than or equal to preset value δ (t), just stop
Decompose and will be less than or equal to the residual components r of preset value δ (t)m(t) as final residual components rn(t), finally obtained
Subsequence includes multiple IMF component ci,m(t) and final residual components rn(t), wherein i=1,2,3...n.
It wherein, is IMF component packet by the Time Series for adding Gaussian sequence as preferred embodiment
It includes:
The maximum point and minimum point of the time series of identification addition white noise sequence.
It is fitted and corresponding first envelope of maximum point and the second envelope corresponding with minimum point.
Utilize the first envelope and the second envelope line computation target surplus.
Calculate the time series of addition Gaussian sequence and the second difference of target surplus.
Judge whether the second difference meets IMF condition.
If so, using the second difference as IMF component.
If it is not, then steps be repeated alternatively until that the second difference meets IMF condition.
Specifically, in the present embodiment, to addition white noise sequence nm(t) time series xm(t) when being decomposed, know
Other time series xm(t) then maximum point and minimum point are fitted the envelope d of maximum pointh(t) (the first envelope)
With the envelope d of fitting minimum pointl(t) (the second envelope).Utilize the first envelope dh(t) and the second envelope dl(t) it counts
Calculate target surplus dav(t), target surplus dav(t) calculating can be calculated using following formula:
Calculate target surplus dav(t) after, time series x is then utilizedm(t) target surplus d is subtractedav(t) second is obtained
Difference, then judge whether the second difference meets IMF condition, if it is satisfied, then using the second difference as IMF component lm1(t).If
It is unsatisfactory for, then repeatedly the step in embodiment meets IMF condition and by final meet IMF condition second until the second difference
Difference is as the IMF component in the embodiment of the present invention, until the second difference meets IMF condition, wherein IMF condition includes: to obtain
The corresponding sequence of the second difference in entire time range, the number of Local Extremum and zero crossing must be equal, or at most
Difference one, and point at any time, the envelope (lower envelope of envelope (the coenvelope line) and local minimum of local maximum
Line) average value must be zero.Certainly, according to the practical operation situation of electric system, IMF condition may be other conditions, in this regard,
The embodiment of the present invention is simultaneously not construed as limiting.
In addition, in order to guarantee that IMF component in finally obtained subsequence and final residual components are more representative,
In the embodiment of the present invention, the history is born by repeating to obtain historical load data and be based on integrated empirical mode decomposition
Lotus data are decomposed the step of obtaining multiple subsequences, and each subsequence has multiple groups value, every sub- sequence in every group of subsequence
Include multiple IMF components and a final residual components in column, then the IMF component of every group of subsequence and final are remained
Remaining component, which is averaged, acquires average IMF component and average residual components.If the son in finally obtained every group of subsequence
The number of sequence is M, and the IMF component in every group of subsequence in each subsequence is 7, and residual components are 1, then can benefit
The average value of the IMF component in every group of subsequence and the average value of residual components are calculated with following formula:
Wherein, i=1,2,3 ... 7；The number M of subsequence in every group of subsequence can be determined according to the actual situation,
Number of the embodiment of the present invention for the number of the IMF component in each subsequence and the number for subsequence number are not
It limits.Therefore, according to above formula, finally obtain 7 it is average after IMF component and 1 it is average after residual components.
It should be noted that being corresponding with one corresponding to every day in historical load data in the embodiment of the present invention
Subsequence, comprising multiple IMF components and a final residual components in each subsequence, in addition, training sample and pre-
Survey sample standard deviation can be chosen from historical load data, e.g., using a few days in historical load data load datas as
Training sample, using remaining load data as forecast sample.Whether DBN neural network forecast mould is obtained using training sample training
Type still uses forecast sample combination DBN Network Prediction Model to predict that short term, the input of model is all to decompose
The subsequence arrived.Corresponding IMF component and a final surplus are corresponding with a DBN neural network forecast in each subsequence
Model.
S103, training sample is chosen and using training sample training DBN network to obtain DBN Network Prediction Model.
Specifically, training sample can be the historical load data and day destiny of prediction day the first four months in the present embodiment
According to.Utilize the optimization and training that training sample training DBN network is based on crossover algorithm in length and breadth to DBN network.It is instructed about obtaining
Practice sample and using training sample training DBN network with obtain the concept of DBN Network Prediction Model by following embodiment into
Row is introduced, and the embodiment of the present invention wouldn't explain herein.Such as the introduction of above-described embodiment, training sample is from historical load data
Middle to choose (be any certain days load datas), after choosing training sample, the time series of every day is corresponding negative
Lotus data are all broken down into 8 components (8 components are only to illustrate, or the component of other numbers), and 8 components include
Then 7 IMF components and 1 final residual components establish DBN net for each IMF component and 1 final residual components
Then network recycles the IMF component of the daily synchronization point of the training sample selected to instruct respective DBN network
Practice (best initial weights for being principally obtaining DBN network), finally obtains 8 DBN Network Prediction Models.
Wherein, as preferred embodiment, step S103 includes:
Training sample is chosen from historical load data.
The training parameter of DBN network is determined according to subsequence corresponding with training sample.
The optimal solution of DBN network is determined using training sample.
Optimal solution is obtained into DBN Network Prediction Model as weight corresponding with DBN network, the IMF in subsequence points
Amount and final residual components have unique corresponding DBN Network Prediction Model.
Specifically, from training sample is chosen in historical load data being chosen in historical load data in the present embodiment
Then a few days load datas establish the corresponding DBN network of each component in subsequence corresponding with these days and utilize instruction
Practice sample to be trained each DBN network.(what certainly, training sample specifically selected the load data in which day can basis
Actual conditions determine, the embodiment of the present invention herein and be not construed as limiting).After having chosen training sample, according to training sample
Size of data determines the training parameter (initial population scale, maximum number of iterations of DBN network etc.) of DBN network.Due to history
Load data, which passes through, has obtained multiple subsequences (including multiple IMF components in subsequence) based on integrated empirical mode decomposition, because
This, establishes a DBN network for each IMF component corresponding with training sample, then in training sample every day same a period of time
The corresponding IMF component of punctum as DBN network input and DBN network is trained until determine the optimal of DBN network
One group of weight (optimal solution), specifically, the training of DBN network may refer to the prior art.When DBN network exports best initial weights
After matrix (optimal solution), best initial weights matrix as the weight of DBN Network Prediction Model and is predicted short term.
S104, short term is predicted using DBN Network Prediction Model and each subsequence.
Specifically, in the present embodiment, it is negative from history after obtaining DBN Network Prediction Model (may refer to the prior art)
Forecast sample is chosen in lotus data, and each subsequence of forecast sample is then input in DBN Network Prediction Model (every sub- sequence
Column are corresponding with multiple components, and the synchronization point in each subsequence is obtained component as its corresponding DBN Network Prediction Model
Input), short-term load prediction is then exported by each DBN Network Prediction Model, then by each DBN Network Prediction Model
Each output valve as prediction day predicted load, by taking the load for predicting some day as an example, forecast sample be prediction a few days ago
One day load data, which, which is equally decomposed, weighs 8 components, this 8 components are brought into and each component respectively
Corresponding DBN Network Prediction Model obtains 8 load prediction results, finally is overlapped to obtain by this 8 load prediction results
Predict the load prediction of day.In addition, the concept of short term also may refer to the prior art.
Based on the above embodiment, as preferred embodiment, step S104 includes:
Forecast sample is chosen from historical load data.
Using subsequence corresponding with forecast sample as the input of DBN Network Prediction Model.
Using weight corresponding with optimal solution as the weight of DBN Network Prediction Model.
It take each subsequence as the input of DBN Network Prediction Model, with weight corresponding with optimal solution for DBN neural network forecast mould
The weight of type determines that the reality output of DBN Network Prediction Model, reality output are the short term that prediction obtains.
Using other data in subsequence in addition to training sample as DBN net corresponding with IMF component each in subsequence
The input of network prediction model is to predict the data of a certain moment point, then, the actual number that each DBN Network Prediction Model is exported
(predict that the data of a certain moment point can be with using DBN Network Prediction Model according to the data for just obtaining a certain moment point are overlapped
Referring to the prior art).
It should be noted that the number of IMF component is how many in subsequence in the embodiment of the present invention, the DBN net established
The number of network prediction model is just how many, the time series in each IMF component be DBN Network Prediction Model training sample and
Forecast sample.The finally result to the load prediction sometime put in a short time are as follows: each DBN Network Prediction Model is in certain a period of time
The superposition value of the reality output of punctum.
As it can be seen that a kind of short-term load forecasting method disclosed by the embodiments of the present invention, first acquisition historical load data, then
Historical load data is decomposed based on integrated empirical mode decomposition to obtain multiple subsequences, secondly chooses training sample and benefit
With training sample training DBN network to obtain DBN Network Prediction Model, DBN Network Prediction Model and each subsequence are finally utilized
Short term is predicted, it is middle compared with the prior art that highly complex feature can not be handled using single DBN prediction model
Load data and cause the problem for being easy to cause precision of prediction low, using this programme, since historical load data being carried out
Decomposition obtains subsequence, reduces the complex characteristics (nonlinearity and non-stationary) of load data to a certain extent, because
This, predicts short term using the subsequence of DBN Network Prediction Model and the load data of low complex characteristics, this programme
It is higher to the precision of prediction of short term, very big guarantee is provided to the normal operation of operation of power networks scheduling.
Based on the above embodiment, as preferred embodiment, DBN network is determined most using training parameter and training sample
Excellent solution includes:
Determine the initial population scale and maximum number of iterations in training parameter.
Using training sample as the input of DBN network.
Particle to be optimized is encoded and generate initial population scale corresponding with initial population scale it is corresponding just
Beginning population.
Calculate the fitness of each encoded particles in initial population.
Lateral cross is carried out to the encoded particles in initial population based on crossover algorithm in length and breadth and crossed longitudinally obtains population
Matrix.
It calculates the fitness of the particle in kind of mass matrix and is carried out pair with the fitness of each encoded particles in initial population
Than to choose target fitness.
Using intended particle corresponding with target fitness as the individual in progeny population.
The number of iterations be steps be repeated alternatively until beyond the maximum number of iterations, by finally obtained maximum adaptation degree pair
The particle answered is as optimal solution.
Specifically, particle to be optimized is the weight of DBN network in the present embodiment, in the solution space of particle coding,
Initial population X=[X is randomly generated1,X2,...,XM]T, the fitness of each encoded particles in initial population is then calculated, each
The fitness of encoded particles can be calculated using following formula:
Wherein, ptIndicate the reality output of DBN network,Indicate the target output of DBN network, N indicates number of training.
The principle of crossover algorithm may refer to the prior art in length and breadth, carry out briefly in the embodiment of the present invention with regard to its application
Bright, the lateral cross in crossover algorithm is the intersection behaviour that counts done in initial population with two one-dimensional particles (weight) in length and breadth
Make, and two particle is randomly generated with one-dimensional.The filial generation that lateral cross operation obtains is stored in matrix MShcThe inside (MShcClaim
For golden mean of the Confucian school solution), then calculate the MShcThe fitness of all particles in matrix, by obtained fitness and parent population (DSvc, the
Except generation initial population) it compares, from MShcFitness particle more better than fitness in parent population is selected in matrix to protect
Stay in progeny population DShcIn.Wherein, lateral cross operates to obtain matrix MShcIt can be indicated with following formula:
MShc(i, d)=r1×X(i,d)+(1-r1)×X(j,d)+c1×(X(i,d)-X(j,d))
MShc(j, d)=r2×X(j,d)+(1-r2)×X(i,d)+c2×(X(j,d)-X(i,d))
I, j ∈ N (1, M), d ∈ N (1, D)
In formula, r1、r2It is the random number between [0,1]；c1、c2It is the random number between [- 1,1]；M is the model of population
It encloses；D is the dimension of variable corresponding with particle；The d that X (i, d), X (j, d) respectively indicate parent particle X (i) and X (j) is tieed up；
MShc(i,d)、MShc(j, d) respectively indicates the filial generation that X (i, d) and X (j, d) is generated by lateral cross in d dimension.
After carrying out lateral operation, crossed longitudinally operation is then carried out again, crossed longitudinally is all particles in different dimensional
The one kind carried out between degree counts intersection, and bidimensional is random combine, and the filial generation that crossover operation obtains is stored in matrix MSvcIn
(MSvcReferred to as golden mean of the Confucian school solution), the adaptive value of each particle in the matrix is then calculated, with its parent population (i.e. DShc) compared
Compared with selecting more excellent particle (intended particle, fitness corresponding with more excellent particle be target fitness) to be retained in DSvc
In (progeny population).Crossed longitudinally operation can carry out according to the following formula:
MSvc(i,d1)=rX (i, d1)+(1-r)·X(i,d2)
I ∈ N (1, M), d1,d2∈ N (1, D), r ∈ [0,1]
In formula: MSvc(i,d1) be particle X (i) d1Peacekeeping d2Tie up the filial generation by generating after crossed longitudinally operation.
At this point, carried out a lateral cross and it is crossed longitudinally after, judge whether current iteration number is more than that maximum changes
Generation number, if exceeding maximum number of iterations, by finally obtained progeny population DSvcIn the corresponding grain of maximum fitness
Son is used as optimal solution (best initial weights and threshold value) if without departing from maximum number of iterations, continues the step more than repeating with again
It is secondary to be iterated, wherein maximum number of iterations can be determined according to training sample or forecast sample, and the embodiment of the present invention exists
This is simultaneously not construed as limiting.
A kind of short-term load forecasting device disclosed by the embodiments of the present invention is introduced below, refers to Fig. 2, Fig. 2 is
A kind of short-term load forecasting apparatus structure schematic diagram disclosed by the embodiments of the present invention, the device include:
Module 201 is obtained, for obtaining historical load data.
Decomposing module 202, for being decomposed to obtain multiple sons to historical load data based on set empirical mode decomposition
Sequence.
Module 203 is chosen, for choosing training sample and training DBN network pre- to obtain DBN network using training sample
Survey model.
Prediction module 204, for being predicted using DBN Network Prediction Model and each subsequence short term.
Based on the above embodiment, as preferred embodiment, choosing module 203 includes:
Selection unit, for choosing training sample from historical load data；
First determination unit, for determining the training parameter of DBN network according to subsequence corresponding with training sample；
Second determination unit, for determining the optimal solution of DBN network using training parameter and training sample；
Setup unit, for optimal solution to be obtained DBN Network Prediction Model as weight corresponding with DBN network, son
Each IMF component and final residual components in sequence have unique corresponding DBN Network Prediction Model.
As it can be seen that a kind of short-term load forecasting device disclosed by the embodiments of the present invention, first acquisition historical load data, then
Historical load data is decomposed based on integrated empirical mode decomposition to obtain multiple subsequences, secondly chooses training sample and benefit
With training sample training DBN network to obtain DBN Network Prediction Model, DBN Network Prediction Model and each subsequence are finally utilized
Short term is predicted, it is middle compared with the prior art that highly complex feature can not be handled using single DBN prediction model
Load data and cause the problem for being easy to cause precision of prediction low, using this programme, since historical load data being carried out
Decomposition obtains subsequence, reduces the complex characteristics (nonlinearity and non-stationary) of load data to a certain extent, because
This, predicts short term using the subsequence of DBN Network Prediction Model and the load data of low complex characteristics, this programme
It is higher to the precision of prediction of short term, very big guarantee is provided to the normal operation of operation of power networks scheduling.
In addition, referring to Fig. 3 the embodiment of the invention also discloses a kind of short-term load forecasting equipment, Fig. 3 is that the present invention is real
Applying a kind of short-term load forecasting device structure schematic diagram, the equipment disclosed in example includes:
Memory 301 is used for Storage Estimation program；
Processor 302, for executing the Prediction program stored in the memory to realize that any of the above embodiment is mentioned
Short-term load forecasting method the step of.
It should be noted that a kind of short-term load forecasting equipment disclosed by the embodiments of the present invention have as it is above-mentioned any one
Technical effect possessed by embodiment, details are not described herein for the embodiment of the present invention.
This programme in order to better understand, a kind of computer readable storage medium disclosed by the embodiments of the present invention, computer
It is stored with Prediction program on readable storage medium storing program for executing, realizes what any embodiment as above was mentioned when computer program is executed by processor
The step of short-term load forecasting method.
It should be noted that a kind of computer readable storage medium disclosed by the embodiments of the present invention has as above-mentioned any one
Technical effect possessed by a embodiment, details are not described herein for the embodiment of the present invention.
A kind of short-term load forecasting method of the disclosure as set forth herein, device, equipment and storage medium have been carried out in detail above
It is thin to introduce.Specific examples are used herein to illustrate the principle and implementation manner of the present application, and above embodiments are said
It is bright to be merely used to help understand the present processes and its core concept.It should be pointed out that for the ordinary skill of the art
For personnel, under the premise of not departing from the application principle, can also to the application, some improvement and modification can also be carried out, these improvement
It is also fallen into the protection scope of the claim of this application with modification.
Each embodiment is described in a progressive manner in specification, the highlights of each of the examples are with other realities
The difference of example is applied, the same or similar parts in each embodiment may refer to each other.For device disclosed in embodiment
Speech, since it is corresponded to the methods disclosed in the examples, so being described relatively simple, related place is referring to method part illustration
?.
It should also be noted that, in the present specification, relational terms such as first and second and the like be used merely to by
One entity or operation are distinguished with another entity or operation, without necessarily requiring or implying these entities or operation
Between there are any actual relationship or orders.Moreover, the terms "include", "comprise" or its any other variant meaning
Covering non-exclusive inclusion, so that the process, method, article or equipment for including a series of elements not only includes that
A little elements, but also including other elements that are not explicitly listed, or further include for this process, method, article or
The intrinsic element of equipment.In the absence of more restrictions, the element limited by sentence "including a ...", is not arranged
Except there is also other identical elements in the process, method, article or apparatus that includes the element.
Claims (10)
1. a kind of short-term load forecasting method characterized by comprising
Obtain historical load data；
The historical load data is decomposed based on integrated empirical mode decomposition to obtain multiple subsequences；
Choose training sample and using training sample training DBN network to obtain DBN Network Prediction Model；
Short term is predicted using the DBN Network Prediction Model and each subsequence.
2. short-term load forecasting method according to claim 1, which is characterized in that described based on integrated empirical mode decomposition
The historical load data is decomposed to obtain multiple subsequences include:
Gaussian sequence is added in daily time series corresponding with the historical load data；
It is IMF component by the daily Time Series for adding the Gaussian sequence；
The first difference of the daily time series and the IMF component of adding the Gaussian sequence is calculated with will be described
First difference is as residual components；
Using the residual components as time series to be decomposed and repeat the above steps until final residual components are less than or wait
Stop decomposing when preset value, finally obtains multiple IMF components corresponding with the daily time series and final residue
Component, corresponding subsequence every day in the historical load data, each subsequence include: with it is described daily
The corresponding multiple IMF components of time series and final residual components.
3. short-term load forecasting method according to claim 2, which is characterized in that described to add the white Gaussian noise
The daily Time Series of sequence are that IMF component includes:
The maximum point and minimum point of the daily time series of the Gaussian sequence are added in identification；
It is fitted and corresponding first envelope of the maximum point and the second envelope corresponding with the minimum point；
Utilize first envelope and the second envelope line computation target surplus；
Calculate the second difference of the time series and the target surplus of adding the Gaussian sequence；
Judge whether second difference meets IMF condition；
If so, using second difference as IMF component；
If it is not, then steps be repeated alternatively until that second difference meets the IMF condition.
4. short-term load forecasting method according to claim 2, which is characterized in that the selection training sample simultaneously utilizes institute
Training sample training DBN network, which is stated, to obtain DBN Network Prediction Model includes:
Training sample is chosen from the historical load data；According to subsequence corresponding with training sample determination
The training parameter of DBN network；
The optimal solution of the DBN network is determined using the training parameter and the training sample；
The optimal solution is obtained into the DBN Network Prediction Model, the sub- sequence as weight corresponding with the DBN network
Each IMF component and final residual components in column have unique corresponding DBN Network Prediction Model.
5. short-term load forecasting method according to claim 4, which is characterized in that described to utilize the training parameter and institute
It states training sample and determines that the optimal solution of the DBN network includes:
Determine the initial population scale and maximum number of iterations in the training parameter；
Using subsequence corresponding with the training sample as the input of the DBN network；
Particle to be optimized is encoded and generates initial population corresponding with the initial population scale；
Calculate the fitness of each encoded particles in the initial population；
Lateral cross is carried out to the encoded particles in the initial population based on crossover algorithm in length and breadth and crossed longitudinally obtains population
Matrix；
Calculate the fitness of the particle in described kind of mass matrix and the adaptation with each encoded particles in the initial population
Degree is compared to choose target fitness；
Using intended particle corresponding with the target fitness as the individual in progeny population；
It steps be repeated alternatively until that the number of iterations exceeds the maximum number of iterations, finally obtained maximum adaptation degree is corresponding
Particle is as the optimal solution.
6. according to claim 1 to short-term load forecasting method described in 5 any one, which is characterized in that described in the utilization
DBN Network Prediction Model and each subsequence carry out prediction to short term and include:
Forecast sample is chosen from the historical load data；
Using subsequence corresponding with the forecast sample as the input of the DBN Network Prediction Model；
Using weight corresponding with the optimal solution as the weight of the DBN Network Prediction Model；
It take the subsequence as the input of the DBN Network Prediction Model, with weight corresponding with the optimal solution for the DBN
The weight of Network Prediction Model determines the reality output of the DBN Network Prediction Model, and the reality output prediction obtains
Short term.
7. a kind of short-term load forecasting device characterized by comprising
Module is obtained, for obtaining historical load data；
Decomposing module, for being decomposed to obtain multiple sub- sequences to the historical load data based on set empirical mode decomposition
Column；
Module is chosen, for choosing training sample and using training sample training DBN network to obtain DBN neural network forecast mould
Type；
Prediction module, for being predicted using the DBN Network Prediction Model and each subsequence short term.
8. short-term load forecasting device according to claim 7, which is characterized in that the selection module includes:
Selection unit, for choosing training sample from the historical load data；
First determination unit, for determining the training parameter of the DBN network according to subsequence corresponding with the training sample；
Second determination unit, for determining the optimal solution of the DBN network using the training parameter and the training sample；
Setup unit, for using the optimal solution as weight corresponding with the DBN network to obtain the DBN neural network forecast
Model, each IMF component and final residual components in the subsequence have unique corresponding DBN Network Prediction Model.
9. a kind of short-term load forecasting equipment characterized by comprising
Memory is used for Storage Estimation program；
Processor, for executing the Prediction program stored in the memory to realize as claimed in any one of claims 1 to 6
The step of short-term load forecasting method.
10. a kind of computer readable storage medium, it is stored with Prediction program on the computer readable storage medium, feature exists
In the Prediction program is executed by processor to realize such as short-term load forecasting method as claimed in any one of claims 1 to 6
Step.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810784848.6A CN109063902A (en) | 2018-07-17 | 2018-07-17 | A kind of short-term load forecasting method, device, equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810784848.6A CN109063902A (en) | 2018-07-17 | 2018-07-17 | A kind of short-term load forecasting method, device, equipment and storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109063902A true CN109063902A (en) | 2018-12-21 |
Family
ID=64816999
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810784848.6A Pending CN109063902A (en) | 2018-07-17 | 2018-07-17 | A kind of short-term load forecasting method, device, equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109063902A (en) |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109636059A (en) * | 2018-12-24 | 2019-04-16 | 国网北京市电力公司 | Electric heating distribution transformer load forecasting method and device |
| CN111382906A (en) * | 2020-03-06 | 2020-07-07 | 南京工程学院 | Power load prediction method, system, equipment and computer readable storage medium |
| CN112952814A (en) * | 2021-03-04 | 2021-06-11 | 四川云起老和科技有限公司 | Regional energy Internet evolution simulation method considering town growth characteristics |
| CN116663863A (en) * | 2023-07-28 | 2023-08-29 | 石家庄科林电气股份有限公司 | Virtual power plant load prediction method based on scheduling plan |
| CN116755641A (en) * | 2023-08-22 | 2023-09-15 | 山东凌远机电科技有限公司 | Distribution box operation data optimization acquisition and storage method |
| CN116979531A (en) * | 2023-09-25 | 2023-10-31 | 山西京能售电有限责任公司 | Novel energy data monitoring method and method for monitoring auxiliary power market |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20140337002A1 (en) * | 2013-05-08 | 2014-11-13 | Instant Access Networks, Llc | Method and Instrumentation for Sustainable Energy Load Flow Management System (SelfMaster(TM)) |
| CN107256439A (en) * | 2017-06-01 | 2017-10-17 | 常州英集动力科技有限公司 | Joint EEMD and neutral net short-term load forecasting method and system |
| CN107292453A (en) * | 2017-07-24 | 2017-10-24 | 国网江苏省电力公司电力科学研究院 | A kind of short-term wind power prediction method based on integrated empirical mode decomposition Yu depth belief network |
-
2018
- 2018-07-17 CN CN201810784848.6A patent/CN109063902A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20140337002A1 (en) * | 2013-05-08 | 2014-11-13 | Instant Access Networks, Llc | Method and Instrumentation for Sustainable Energy Load Flow Management System (SelfMaster(TM)) |
| CN107256439A (en) * | 2017-06-01 | 2017-10-17 | 常州英集动力科技有限公司 | Joint EEMD and neutral net short-term load forecasting method and system |
| CN107292453A (en) * | 2017-07-24 | 2017-10-24 | 国网江苏省电力公司电力科学研究院 | A kind of short-term wind power prediction method based on integrated empirical mode decomposition Yu depth belief network |
Non-Patent Citations (2)
| Title |
|---|
| XUEHENG QIU ETAL: ""Empirical Mode Decomposition based ensemble deep learning for load demand time series forecasting"", 《APPLIED SOFT COMPUTING》 * |
| 陈冬沣等: ""基于纵横交叉算法与 Elman 神经网络的短期负荷预测研究"", 《贵州电力技术》 * |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109636059A (en) * | 2018-12-24 | 2019-04-16 | 国网北京市电力公司 | Electric heating distribution transformer load forecasting method and device |
| CN111382906A (en) * | 2020-03-06 | 2020-07-07 | 南京工程学院 | Power load prediction method, system, equipment and computer readable storage medium |
| CN111382906B (en) * | 2020-03-06 | 2024-02-27 | 南京工程学院 | Power load prediction method, system, equipment and computer readable storage medium |
| CN112952814A (en) * | 2021-03-04 | 2021-06-11 | 四川云起老和科技有限公司 | Regional energy Internet evolution simulation method considering town growth characteristics |
| CN112952814B (en) * | 2021-03-04 | 2022-12-09 | 四川云起老和科技有限公司 | Regional energy Internet evolution simulation method considering town growth characteristics |
| CN116663863A (en) * | 2023-07-28 | 2023-08-29 | 石家庄科林电气股份有限公司 | Virtual power plant load prediction method based on scheduling plan |
| CN116663863B (en) * | 2023-07-28 | 2023-10-20 | 石家庄科林电气股份有限公司 | Virtual power plant load prediction method based on scheduling plan |
| CN116755641A (en) * | 2023-08-22 | 2023-09-15 | 山东凌远机电科技有限公司 | Distribution box operation data optimization acquisition and storage method |
| CN116755641B (en) * | 2023-08-22 | 2023-10-24 | 山东凌远机电科技有限公司 | Distribution box operation data optimization acquisition and storage method |
| CN116979531A (en) * | 2023-09-25 | 2023-10-31 | 山西京能售电有限责任公司 | Novel energy data monitoring method and method for monitoring auxiliary power market |
| CN116979531B (en) * | 2023-09-25 | 2023-12-12 | 山西京能售电有限责任公司 | Novel energy data monitoring method and method for monitoring auxiliary power market |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109063902A (en) | A kind of short-term load forecasting method, device, equipment and storage medium | |
| Gaur et al. | Real-time wave forecasting using genetic programming | |
| Jiang et al. | A hybrid forecasting approach applied in the electrical power system based on data preprocessing, optimization and artificial intelligence algorithms | |
| CN110738010A (en) | Wind power plant short-term wind speed prediction method integrated with deep learning model | |
| CN108985514A (en) | Load forecasting method, device and equipment based on EEMD and LSTM | |
| Carapellucci et al. | A methodology for the synthetic generation of hourly wind speed time series based on some known aggregate input data | |
| CA2436352A1 (en) | Process and system for developing a predictive model | |
| CN106651023A (en) | Grey correlation analysis-based improved fireworks algorithm mid-long term load prediction method | |
| CN112464566A (en) | Transformer oil temperature prediction method based on genetic algorithm and BP neural network | |
| CN111461445B (en) | Short-term wind speed prediction method and device, computer equipment and storage medium | |
| Tian et al. | A network traffic hybrid prediction model optimized by improved harmony search algorithm | |
| CN116599050A (en) | Photovoltaic prediction method and related device based on self-attention mechanism | |
| CN109635938A (en) | A kind of autonomous learning impulsive neural networks weight quantization method | |
| CN107358059A (en) | Short-term photovoltaic energy Forecasting Methodology and device | |
| CN107301478A (en) | A kind of cable run short-term load forecasting method | |
| CN109543879A (en) | Load forecasting method and device neural network based | |
| CN113111592A (en) | Short-term wind power prediction method based on EMD-LSTM | |
| CN115034422A (en) | Wind power short-term power prediction method and system based on fluctuation identification and error correction | |
| CN105389625B (en) | Active power distribution network ultra-short term load prediction method | |
| CN116845875A (en) | WOA-BP-based short-term photovoltaic output prediction method and device | |
| CN114925940A (en) | Holiday load prediction method and system based on load decomposition | |
| CN116822722A (en) | Water level prediction method, system, device, electronic equipment and medium | |
| CN109840308B (en) | Regional wind power probability forecasting method and system | |
| CN115660893A (en) | Transformer substation bus load prediction method based on load characteristics | |
| CN115907000A (en) | Small sample learning method for optimal power flow prediction of power system |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181221 |
|
| RJ01 | Rejection of invention patent application after publication |