CN108805366A - Multifactor adaptive neural network Methods of electric load forecasting based on decision formal context and system - Google Patents
Multifactor adaptive neural network Methods of electric load forecasting based on decision formal context and system Download PDFInfo
- Publication number
- CN108805366A CN108805366A CN201810747257.1A CN201810747257A CN108805366A CN 108805366 A CN108805366 A CN 108805366A CN 201810747257 A CN201810747257 A CN 201810747257A CN 108805366 A CN108805366 A CN 108805366A
- Authority
- CN
- China
- Prior art keywords
- attribute
- decision
- value
- load
- neural network
- 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
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 34
- 230000003044 adaptive effect Effects 0.000 title claims abstract description 29
- 238000000034 method Methods 0.000 title claims abstract description 26
- 238000012549 training Methods 0.000 claims abstract description 22
- 238000000605 extraction Methods 0.000 claims abstract description 20
- 238000005457 optimization Methods 0.000 claims abstract description 10
- 239000008186 active pharmaceutical agent Substances 0.000 claims description 18
- 239000013598 vector Substances 0.000 claims description 7
- 230000006870 function Effects 0.000 claims description 6
- 238000013507 mapping Methods 0.000 claims description 4
- 239000011159 matrix material Substances 0.000 claims description 4
- 238000003860 storage Methods 0.000 claims description 4
- 238000012217 deletion Methods 0.000 claims description 3
- 230000037430 deletion Effects 0.000 claims description 3
- 230000006978 adaptation Effects 0.000 claims description 2
- 238000004590 computer program Methods 0.000 claims description 2
- 241000208340 Araliaceae Species 0.000 claims 1
- 235000005035 Panax pseudoginseng ssp. pseudoginseng Nutrition 0.000 claims 1
- 235000003140 Panax quinquefolius Nutrition 0.000 claims 1
- 230000014509 gene expression Effects 0.000 claims 1
- 235000008434 ginseng Nutrition 0.000 claims 1
- 230000005611 electricity Effects 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 229910052799 carbon Inorganic materials 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000013277 forecasting method Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 239000013589 supplement Substances 0.000 description 1
- 230000001360 synchronised 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
- 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
- G06Q10/0637—Strategic management or analysis, e.g. setting a goal or target of an organisation; Planning actions based on goals; Analysis or evaluation of effectiveness of goals
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
Landscapes
- Business, Economics & Management (AREA)
- Engineering & Computer Science (AREA)
- Human Resources & Organizations (AREA)
- Economics (AREA)
- Strategic Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Entrepreneurship & Innovation (AREA)
- Marketing (AREA)
- General Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- Tourism & Hospitality (AREA)
- Quality & Reliability (AREA)
- Game Theory and Decision Science (AREA)
- Operations Research (AREA)
- Development Economics (AREA)
- Health & Medical Sciences (AREA)
- Educational Administration (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses multifactor adaptive neural network Methods of electric load forecasting and system based on decision formal context, including：History power load charge values are obtained, and influence the factor of history power load charge values, the input data as system prediction；Discretization is carried out to input data；Yojan optimization is carried out to input data using the old attribute reduction algorithms of decision formal context, extraction influences the crucial historical load data value of load prediction and crucial external influence factor value；The crucial historical load data value of extraction and crucial external influence factor value are input to adaptive neural network system, Electric Load Forecasting measured value is obtained by training.The present invention has fully considered influence of the extraneous factor for electric load, enables to Electric Load Forecasting geodesic structure more accurate.
Description
Technical field
The present invention relates to the Techniques for Prediction of Electric Loads fields of power industry distribution network planning, more particularly to one kind is based on certainly
The multifactor adaptive neural network Methods of electric load forecasting and system of plan Formal Context.
Background technology
During the managed operation of power industry system, load forecast is not only electric system and formulates production, battalion
The important foundation of decision is sold, and meets the important guarantee of power supply and demand balance, and is power grid, the planning construction and electricity of power supply
Net enterprise, the business decision of power grid user provides information and foundation.With economic growth, electricity consumption is also synchronous therewith to be increased
It is long, if the accuracy and timeliness of load forecast are extremely difficult to ensure, thus may cause it is serious economical and
Social concern.On the one hand, the operating cost that too low load forecast can bring electric system additional, it is necessary to be adopted by the energy
It purchases to supplement the workload demand that electric system itself cannot be satisfied, once there is power supply shortage, it will to national economy and the people
Life generates tremendous influence, increases social cost；On the other hand, it will cause resource waves for excessively high load forecast result
Take.Electric system cost is not only increased, or even can also increase Environmental costs.In short supply in current international energy, countries in the world are big
It strives vigorously to advocate under the overall situation for leading low-carbon and economize on electricity economy, improves the accuracy of load forecast, national economy and society will be sent out
Exhibition has great significance.
However, the factor for influencing electric load often has very much, such as season, temperature, festivals or holidays etc..Due to influencing electricity
There is non-linear relation between power load factor, there are redundancies between factor, moreover, data volume is big, need by multiple
Following power load charge values are predicted in miscellaneous analysis and calculating, so, how to remove redundancy, extract it is main it is crucial because
Element, it is main problem to be solved to improve the efficiency of load forecast and accuracy.
Invention content
The purpose of the present invention is exactly to solve the above-mentioned problems, to propose a kind of multifactor adaptive based on decision formal context
It answers neural network for forecasting power load method and system, this method that can catch the key factor for influencing electric load, improves pre-
Operation efficiency and accuracy are surveyed, to preferably Electric Power Network Planning be instructed to work.
The disclosed multifactor adaptive neural network electric power based on decision formal context in one or more embodiments
Load forecasting method, including：
History power load charge values are obtained, and influence the factor of history power load charge values, the input number as system prediction
According to；
Discretization is carried out to input data；
Yojan optimization is carried out to input data using the old attribute reduction algorithms of decision formal context, extraction influences load prediction
Crucial historical load data value and crucial external influence factor value；
The crucial historical load data value of extraction and crucial external influence factor value are input to adaptive neural network system
System obtains Electric Load Forecasting measured value by training.
Further, described to carry out discretization to input data, specific method is：
Step 2.1：Calculate a Candidate point set；
S=(U, A ∪ { d }) is a decision table, U={ x1,…,xnIt is limited object set i.e. domain, A=
{a1,…,anIt is conditional attribute set, d is decision attribute；It is arbitrary for a ∈ A, there is information function U → Va, VaIt is on attribute a
Codomain；For d, there is Vd={ 1 ..., r (d) }, r (d) are the number of decision type, VdIt is the codomain on attribute d；For appointing
(a, c) of meaning, wherein a ∈ A, c ∈ R, c be called be a a Candidate point；A Candidate point set is constituted by multiple c；
The history electric load Value Data of extraction is placed in U, the factor for influencing history power load charge values is placed in A；
Step 2.2：A subset, referred to as Result Cuts subset are selected from Candidate point set；
Step 2.3：By following mapping mode, directly using the given data of Result Cuts subset discretization；
Given DS=(U, V, f, A ∪ { d }), Data Discretization are to determine the cluster dividing P={ P of Aa| a ∈ A } and it is corresponding
Breakpoint set C={ Ca| a ∈ A }, convert DS to DSp=(U, Vp,fp,Ap∪ { d }) process；
Wherein DSpThe P of referred to as DS is divided；Ap={ ap| a ∈ A } it is known as the P divisions of A；apIt is the P divisions of a；fp：U→Ap∪
{ d }, ifThen fp(ap, x) and=i；It is Va's
It divides；It is a breakpoint set；It is codomain VaOne value,That is attribute a
On i-th of break value,That is (i-1)-th break value on attribute a；Value notes of the example x ∈ U on attribute a ∈ A ∪ { d }
For a (x), i.e. f (x, a)=a (x)；Example x ∈ U are in attribute aP∈APValue on ∪ { d } is denoted as aP(x), i.e. fP(x, a)=aP
(x)。
Further, if | Pa|=1, then VaAll values will all be divided into same section, it is same to be correspondingly discretized
One as a result, still, if apAny example cannot be differentiated, then from DSpMiddle deletion.
Further, the old attribute reduction algorithms using decision formal context carry out yojan optimization, tool to input data
Body is：
Step 3.1：According to the data after discretization, strong Consistent Decision Formal Context (U, M, I, N, J) is defined, wherein M is
Conditional attribute collection, N are decision kind sets,m∈M；
Step 3.2：For conditional attribute m, its inclusion relation E (m, n) with decision attribute n, all E (m, n) packets are asked
Matrix is constituted containing attribute；
Step 3.3：A Consistent Sets D is defined, if there is property setIt is the Consistent Sets of (U, M, I) then to claim D；
Step 3.4：One of them untreated attribute a is deleted from Consistent Sets D；
Step 3.5：Judge whether attribute a is reducible, if reducible then follow the steps 3.6, adds back a if irreducible
Consistent Sets D continues to execute step 3.6；
Step 3.6：Judge whether also have untreated attribute in Consistent Sets D at this time；If also untreated category
Property then continues to execute step 3.4, otherwise terminates algorithm, D is a yojan of the Formal Context at this time.
Further, judge that the whether reducible methods of attribute a are in the step 3.5：
When attribute a does not have sub- attribute, attribute a is reducible；
When attribute a has sub- attribute, for the sub- attribute b of attribute a, if there is sub- attribute b Consistent Sets D all fathers
The collection of attribute is combined into B, and a is deleted from D at this time, is hadIf meeting g (b) ≠ g (B), attribute a is irreducible, such as
Such sub- attribute b is not present in fruit, then attribute a is reducible；Wherein, g (b) indicates the set of the object with attribute b；G (B) table
Show the set of the object for all properties for possessing set B.
Further, the crucial historical load data value of extraction and crucial external influence factor value are input to adaptive god
Through network system, Electric Load Forecasting measured value is obtained by training, specially：
Step 4.1：Randomly generate comprising u individual initial population and set evolution number k=1, individual for a real value to
Amount is to (λi,ei),Wherein, λ is feedback term constant；E is parameter vector；
Step 4.2：Each individual (λi,ei) filial generation of oneself is generated as the following formula：
In formula：N (0,1) is the random number being distributed in One-Dimensional Normal；Nj(0,1) it is random with the relevant normal state of component of a vector
Number；Parameter
Step 4.3：Parent and the fitness of filial generation are determined according to network training error；
Step 4.4：Q individual is randomly selected from parent and filial generation, by them by fitness and parent and filial generation
Each individual is compared, and calculates individual amount W (j) wherein poorer than the individual adaptation degree, and as the individual
Then score presses the sequence that score declines and sorts to individual, selects the u individuals with higher score as follow-on parent
Group；
Step 4.5：Judge whether to meet evolution target, if satisfied, stopping evolving, goes to step 4.6；Otherwise, k=k+1 turns
Step 4.2 continues to evolve；
Step 4.6：Data after input simplification are obtained required as forecast sample by the operation of neural network
Predicted value.
Further, fitness function is defined as in the step 4.3：
In formula, ei=| oi(j)-oi|, oi(j) the neural network output valve of i-th of training sample, j-th of individual is indicated；oi
Indicate the actual value of i-th of training sample；K is training sample sum.
Further, the evolution target in the step 4.5 refers to：The parent of continuous T time and filial generation are unchanged or repeatedly
Generation number is more than the number of setting, and T is setting value.
The disclosed multifactor adaptive neural network electric power based on decision formal context in one or more embodiments
Load prediction system, including server, the server include memory, processor and storage on a memory and can handle
The computer program run on device, the processor realize following steps when executing described program：
History power load charge values are obtained, and influence the factor of history power load charge values, the input number as system prediction
According to；
Discretization is carried out to input data；
Yojan optimization is carried out to input data using the old attribute reduction algorithms of decision formal context, extraction influences load prediction
Crucial historical load data value and crucial external influence factor value；
The crucial historical load data value of extraction and crucial external influence factor value are input to adaptive neural network system
System obtains Electric Load Forecasting measured value by training.
Disclosed a kind of computer readable storage medium in one or more embodiments, is stored thereon with computer journey
Sequence, the program execute following steps when being executed by processor：
History power load charge values are obtained, and influence the factor of history power load charge values, the input number as system prediction
According to；
Discretization is carried out to input data；
Yojan optimization is carried out to input data using the old attribute reduction algorithms of decision formal context, extraction influences load prediction
Crucial historical load data value and crucial external influence factor value；
The crucial historical load data value of extraction and crucial external influence factor value are input to adaptive neural network system
System obtains Electric Load Forecasting measured value by training.
Advantageous effect of the present invention：
The present invention has fully considered influence of the extraneous factor for electric load, enables to Electric Load Forecasting geodesic structure more
It is accurate to add.
The present invention extracts crucial historical load data value by brief algorithm and crucial external influence factor is used as input sample
This predicts electric load, can improve prediction operation efficiency and accuracy, to preferably Electric Power Network Planning be instructed to work.
Description of the drawings
Fig. 1 is the multifactor adaptive neural network Methods of electric load forecasting flow chart based on decision formal context.
Specific implementation mode
The present invention is further illustrated with specific implementation mode below in conjunction with the accompanying drawings.
It is noted that following detailed description is all illustrative, it is intended to provide further instruction to the application.Unless another
It indicates, all technical and scientific terms used herein has usual with the application person of an ordinary skill in the technical field
The identical meanings of understanding.
It should be noted that term used herein above is merely to describe specific implementation mode, and be not intended to restricted root
According to the illustrative embodiments of the application.As used herein, unless the context clearly indicates otherwise, otherwise singulative
It is also intended to include plural form, additionally, it should be understood that, when in the present specification using term "comprising" and/or " packet
Include " when, indicate existing characteristics, step, operation, device, component and/or combination thereof.
As background technology is introduced, in order to solve due to being influenced to lead to load forecast result by extraneous factor
Inaccurate problem, present applicant proposes a kind of multifactor adaptive neural network Electric Load Forecasting based on decision formal context
Survey method, as shown in Figure 1, specifically comprising the following steps：
Step 1：Obtain history power load charge values, and influence history power load charge values factor, such as season, weather,
The impact factors such as festivals or holidays, the input data as system prediction, wherein exist between history power load charge values and influence factor
Corresponding mapping relations；The factor data form for influencing history power load charge values is as shown in the table：
The influence factor collection of 1 daily peak load of table
Step 2：Discretization input data；It is as follows：
Step 2.1：Calculate a Candidate point set；
S=(U, A ∪ { d }) is a decision table, U={ x1,…,xnIt is limited object set i.e. domain, A=
{a1,…,anIt is conditional attribute set, d is decision attribute.It is arbitrary for a ∈ A, there is information function U → Va, VaIt is attribute a
Codomain.For d, there is Vd={ 1 ..., r (d) }, r (d) are the number of decision type.Arbitrarily (a, c), wherein a ∈ A, c ∈ R
(R is set of real numbers), c be called be a a Candidate point；A Candidate point set is constituted by multiple c；
For the data in step 1, history Power system load data is placed in U, and influence factor is placed in A.
Step 2.2：A subset, referred to as Result Cuts subset are selected from Candidate point set；
Step 2.3：By following mapping mode, directly using the given data of Result Cuts subset discretization.
Given DS=(U, V, f, A ∪ { d }), Data Discretization are to determine the cluster dividing P={ P of Aa| a ∈ A } and it is corresponding
Breakpoint set C={ Ca| a ∈ A }, convert DS to DSp=(U, Vp,fp,Ap∪ { d }) process.The P that wherein DSp is known as DS is drawn
Point；Ap={ ap| a ∈ A } it is known as the P divisions of A；fp:U→Ap∪ { d }, ifThen fp(ap, x) and=i.
VdIt is the codomain of attribute d；VaIt is the codomain of attribute a,It is codomain VaOne value,It is VaDivision；It is one
A breakpoint set；apIt is the P divisions of a；That is i-th of break value on attribute a,That is (i-1)-th breakpoint on attribute a
Value；Values of the example x ∈ U on attribute a ∈ A ∪ { d } is denoted as a (x), i.e. f (x, a)=a (x)；Example x ∈ U are in attribute aP∈AP
Value on ∪ { d } is denoted as aP(x), i.e. fP(x, a)=aP(x)。
If | Pa|=1, then VaAll values will all be divided into same section, be correspondingly discretized as same result.But
It is, if apAny example cannot be differentiated, it can be from DSpMiddle deletion.
Step 3：Yojan optimization, extraction are carried out to the data that will be inputted using the old attribute reduction algorithms of decision formal context
Influence the crucial historical load data value of load prediction and crucial external influence factor value；
(U, M, I, N, J) is one strong Consistent Decision Formal Context (decision formal context herein and discretization in step 2
DS laterpIt is corresponding), wherein M is conditional attribute collection, and N is decision kind set,M ∈ M, if meetingIt is the parent attribute of n then to claim m, and corresponding n is the sub- attribute of m.
ClaimFor the relationship between n and m inclusion relations；
Claim Mr=(E (m, n)), m ∈ M, n ∈ N are the attribute inclusion relation matrix of strong Consistent Decision Formal Context.In the square
The value of m rows n row is 1 in battle array, then attribute m is the parent attribute of attribute n.
The detailed process of old attribute reduction algorithms is as follows：
Step 3.1：For conditional attribute m, its inclusion relation E (m, n) with decision attribute n, all E (m, n) packets are asked
Matrix is constituted containing attribute；
Step 3.2：A Consistent Sets D is defined, if there is property setIt is the Consistent Sets of (U, M, I) then to claim D,
One of them untreated attribute a is deleted from set D；
Step 3.3：Judge whether attribute a is reducible, if reducible then follow the steps 3.4, a is added into meeting if irreducible
Set D continues to execute step 3.4
Step 3.4：Judge whether also have untreated attribute in D at this time.If also untreated attribute after
It is continuous to execute step 3.2, otherwise terminate algorithm, D is a yojan of the Formal Context at this time.
Judge that whether reducible attribute a methods be as follows：
When attribute a does not have sub- attribute, attribute a is reducible.When attribute a has sub- attribute, for the sub- attribute b as attribute a,
It is combined into B in the collection of all parent attributes of D if there is sub- attribute b, a is deleted from D at this time, is hadIf meeting g (b)
≠ g (B), then attribute a is irreducible, and if there is no such sub- attribute b, then attribute a is reducible.
Wherein, g (b) indicates the set of the object with attribute b；G (B) indicates the object for possessing all properties of set B
Set.
Step 4：The key data values of extraction are input to adaptive neural network system, electric load is obtained by training
Predicted value.
Based on this sentences the data of yojan, load forecast is carried out, algorithm steps are as follows：
Step 4.1：Randomly generate comprising u individual initial population and set evolution number k=1, individual for a real value to
Amount is to (λi,ei),Wherein, λ is feedback term constant (including a and b)；E is parameter vector (i.e. adaptive neural network
Policing parameter in network system).
Step 4.2：Each individual (λi,ei) filial generation of oneself is generated as the following formula：
In formula：N (0,1) is the random number being distributed in One-Dimensional Normal；Nj(0,1) it is random with the relevant normal state of component of a vector
Number；
Parameter
Step 4.3：Determine that parent and the fitness of filial generation, fitness function are defined as according to network training error
In formula：ei=| oi(j)-oi|, oi(j) the neural network output valve of i-th of training sample, j-th of individual is indicated；oi
Indicate the actual value of i-th of training sample；K is training sample sum.
Step 4.4：Q individual is randomly selected from parent and filial generation, by them by fitness and parent and filial generation
Each individual is compared, and calculates the individual amount W (j) wherein than the individual difference, and as the score of the individual, so
The sequence that score declines is pressed afterwards to sort to individual, selects the u individuals with higher score as follow-on parent group.
Step 4.5：Judge whether to meet evolution target (continuous 10 parents and filial generation is unchanged or iteration is more than setting time
Number), if satisfied, stopping evolving, go to step 4.6；Otherwise, k=k+1 goes to step 4.2 and continues to evolve.
Step 4.6：Input prediction sample (passes through the data after step 3 yojan), by the operation of neural network, obtains
Required predicted value.
Above-mentioned, although the foregoing specific embodiments of the present invention is described with reference to the accompanying drawings, not protects model to the present invention
The limitation enclosed, those skilled in the art should understand that, based on the technical solutions of the present invention, those skilled in the art are not
Need to make the creative labor the various modifications or changes that can be made still within protection scope of the present invention.
Claims (10)
1. the multifactor adaptive neural network Methods of electric load forecasting based on decision formal context, which is characterized in that including：
History power load charge values are obtained, and influence the factor of history power load charge values, the input data as system prediction；
Discretization is carried out to input data；
Yojan optimization is carried out to input data using the old attribute reduction algorithms of decision formal context, extraction influences the pass of load prediction
Key historical load data value and crucial external influence factor value；
The crucial historical load data value of extraction and crucial external influence factor value are input to adaptive neural network system, led to
It crosses training and obtains Electric Load Forecasting measured value.
2. the multifactor adaptive neural network load forecast side based on decision formal context as described in claim 1
Method, which is characterized in that described to carry out discretization to input data, specific method is：
Step 2.1：Calculate a Candidate point set；
S=(U, A ∪ { d }) is a decision table, U={ x1,…,xnIt is limited object set i.e. domain, A={ a1,…,an}
For conditional attribute set, d is decision attribute；It is arbitrary for a ∈ A, there is information function U → Va, VaIt is the codomain on attribute a；To d
For, there is Vd={ 1 ..., r (d) }, r (d) are the number of decision type, VdIt is the codomain on attribute d；For arbitrary (a, c),
Wherein a ∈ A, c ∈ R, c be called be a a Candidate point；A Candidate point set is constituted by multiple c；
The history electric load Value Data of extraction is placed in U, the factor for influencing history power load charge values is placed in A；
Step 2.2：A subset, referred to as Result Cuts subset are selected from Candidate point set；
Step 2.3：By following mapping mode, directly using the given data of Result Cuts subset discretization；
Given DS=(U, V, f, A ∪ { d }), Data Discretization are to determine the cluster dividing P={ P of Aa| a ∈ A } and corresponding break point set
Close C={ Ca| a ∈ A }, convert DS to DSp=(U, Vp,fp,Ap∪ { d }) process；
Wherein DSpThe P of referred to as DS is divided；Ap={ ap| a ∈ A } it is known as the P divisions of A；apIt is the P divisions of a；fp：U→Ap∪ { d } is such as
FruitThen fp(ap, x) and=i；It is VaDivision；It is a breakpoint set；It is codomain VaOne value,That is on attribute a
I-th of break value,That is (i-1)-th break value on attribute a；Values of the example x ∈ U on attribute a ∈ A ∪ { d } is denoted as a
(x), i.e. f (x, a)=a (x)；Example x ∈ U are in attribute aP∈APValue on ∪ { d } is denoted as aP(x), i.e. fP(x, a)=aP(x)。
3. the multifactor adaptive neural network load forecast side based on decision formal context as claimed in claim 2
Method, which is characterized in that if | Pa|=1, then VaAll values will all be divided into same section, it is same to be correspondingly discretized
As a result, still, if apAny example cannot be differentiated, then from DSpMiddle deletion.
4. the multifactor adaptive neural network load forecast side based on decision formal context as described in claim 1
Method, which is characterized in that the old attribute reduction algorithms using decision formal context carry out yojan optimization to input data, specifically
For：
Step 3.1：According to the data after discretization, strong Consistent Decision Formal Context (U, M, I, N, J) is defined, wherein M is condition
Property set, N are decision kind sets,m∈M；
Step 3.2：For conditional attribute m, it includes to belong to ask its inclusion relation E (m, n) with decision attribute n, all E (m, n)
Property constitute matrix；
Step 3.3：A Consistent Sets D is defined, if there is property setIt is the Consistent Sets of (U, M, I) then to claim D；
Step 3.4：One of them untreated attribute a is deleted from Consistent Sets D；
Step 3.5：Judge whether attribute a is reducible, if reducible then follow the steps 3.6, a is added into back coordination if irreducible
Collection D continues to execute step 3.6；
Step 3.6：Judge whether also have untreated attribute in Consistent Sets D at this time；If also untreated attribute
Step 3.4 is continued to execute, algorithm is otherwise terminated, D is a yojan of the Formal Context at this time.
5. the multifactor adaptive neural network load forecast side based on decision formal context as claimed in claim 4
Method, which is characterized in that judge that the whether reducible methods of attribute a are in the step 3.5：
When attribute a does not have sub- attribute, attribute a is reducible；
When attribute a has sub- attribute, for the sub- attribute b of attribute a, if there is sub- attribute b Consistent Sets D all parent attributes
Collection be combined into B, a is deleted from D at this time, is hadIf meeting g (b) ≠ g (B), attribute a is irreducible, if not
There are such sub- attribute b, then attribute a is reducible；Wherein, g (b) indicates the set of the object with attribute b；G (B) expressions are gathered around
There is the set of the object of all properties of set B.
6. the multifactor adaptive neural network load forecast side based on decision formal context as described in claim 1
Method, which is characterized in that the crucial historical load data value of extraction and crucial external influence factor value are input to adaptive neural network
Network system obtains Electric Load Forecasting measured value, specially by training：
Step 4.1：It randomly generates the initial population comprising u individual and sets evolution number k=1, individual is a real-valued vectors pair
(λi,ei),Wherein, λ is feedback term constant；E is parameter vector；
Step 4.2：Each individual (λi,ei) filial generation of oneself is generated as the following formula：
In formula：N (0,1) is the random number being distributed in One-Dimensional Normal；Nj(0,1) it is and the relevant normal random number of component of a vector；Ginseng
Number
Step 4.3：Parent and the fitness of filial generation are determined according to network training error；
Step 4.4：Q individual is randomly selected from parent and filial generation, they are pressed into fitness and each of parent and filial generation
Individual is compared, and calculates individual amount W (j) wherein poorer than the individual adaptation degree, and as the score of the individual,
Then it presses the sequence that score declines to sort to individual, selects the u individuals with higher score as follow-on parent group；
Step 4.5：Judge whether to meet evolution target, if satisfied, stopping evolving, goes to step 4.6；Otherwise, k=k+1 is gone to step
4.2 continue to evolve；
Step 4.6：Data after input simplification obtain required prediction as forecast sample by the operation of neural network
Value.
7. the multifactor adaptive neural network load forecast side based on decision formal context as claimed in claim 6
Method, which is characterized in that fitness function is defined as in the step 4.3：
In formula, ei=| oi(j)-oi|, oi(j) the neural network output valve of i-th of training sample, j-th of individual is indicated；oiIt indicates
The actual value of i-th of training sample；K is training sample sum.
8. the multifactor adaptive neural network load forecast side based on decision formal context as claimed in claim 6
Method, which is characterized in that the evolution target in the step 4.5 refers to：The parent of continuous T time and filial generation is unchanged or iteration time
For number more than the number of setting, T is setting value.
9. the multifactor adaptive neural network Electric Load Prediction System based on decision formal context, which is characterized in that including
Server, the server include memory, processor and storage on a memory and the computer that can run on a processor
Program, the processor realize following steps when executing described program：
History power load charge values are obtained, and influence the factor of history power load charge values, the input data as system prediction；
Discretization is carried out to input data；
Yojan optimization is carried out to input data using the old attribute reduction algorithms of decision formal context, extraction influences the pass of load prediction
Key historical load data value and crucial external influence factor value；
The crucial historical load data value of extraction and crucial external influence factor value are input to adaptive neural network system, led to
It crosses training and obtains Electric Load Forecasting measured value.
10. a kind of computer readable storage medium, is stored thereon with computer program, which is characterized in that the program is by processor
Following steps are executed when execution：
History power load charge values are obtained, and influence the factor of history power load charge values, the input data as system prediction；
Discretization is carried out to input data；
Yojan optimization is carried out to input data using the old attribute reduction algorithms of decision formal context, extraction influences the pass of load prediction
Key historical load data value and crucial external influence factor value；
The crucial historical load data value of extraction and crucial external influence factor value are input to adaptive neural network system, led to
It crosses training and obtains Electric Load Forecasting measured value.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810747257.1A CN108805366A (en) | 2018-07-09 | 2018-07-09 | Multifactor adaptive neural network Methods of electric load forecasting based on decision formal context and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810747257.1A CN108805366A (en) | 2018-07-09 | 2018-07-09 | Multifactor adaptive neural network Methods of electric load forecasting based on decision formal context and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108805366A true CN108805366A (en) | 2018-11-13 |
Family
ID=64076024
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810747257.1A Pending CN108805366A (en) | 2018-07-09 | 2018-07-09 | Multifactor adaptive neural network Methods of electric load forecasting based on decision formal context and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108805366A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113591362A (en) * | 2021-04-26 | 2021-11-02 | 湖南师范大学 | Clinker proportion optimization and regulation method based on big data intelligent control algorithm |
-
2018
- 2018-07-09 CN CN201810747257.1A patent/CN108805366A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113591362A (en) * | 2021-04-26 | 2021-11-02 | 湖南师范大学 | Clinker proportion optimization and regulation method based on big data intelligent control algorithm |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Maghouli et al. | A scenario-based multi-objective model for multi-stage transmission expansion planning | |
| Fahrni et al. | An application‐oriented guide to R&D project selection and evaluation methods | |
| CN102792542A (en) | Device for prediction of electricity generation amount, method for same, and program for same | |
| CN106022614A (en) | Data mining method of neural network based on nearest neighbor clustering | |
| CN104933505A (en) | Decision and evaluation method for intelligent power distribution network group based on fuzzy assessment | |
| CN105023070A (en) | Output power prediction method of photovoltaic system | |
| Wang et al. | Combined probability density model for medium term load forecasting based on quantile regression and kernel density estimation | |
| CN106570250A (en) | Power big data oriented microgrid short-period load prediction method | |
| CN109242169A (en) | A kind of short-term load forecasting method towards large power customers | |
| Jadidoleslam et al. | Application of shuffled frog leaping algorithm to long term generation expansion planning | |
| Han et al. | An evaluation and optimization methodology for efficient power plant programs | |
| Liu et al. | Dynamic economic dispatch with wind power penetration based on non-parametric kernel density estimation | |
| Hu et al. | Optimal operation of power systems with energy storage under uncertainty: A scenario-based method with strategic sampling | |
| CN112446175B (en) | Distributed power supply site selection optimization method based on power distribution network probability load flow calculation | |
| CN108805366A (en) | Multifactor adaptive neural network Methods of electric load forecasting based on decision formal context and system | |
| Guo et al. | Skewed normal cloud modified whale optimization algorithm for degree reduction of S-λ curves | |
| CN104392317A (en) | Project scheduling method based on genetic culture gene algorithm | |
| CN104679988B (en) | A kind of multiple attributive decision making method based on cloud TOPSIS | |
| Zhang et al. | Application of analytic network process in agricultural products logistics performance evaluation | |
| Das et al. | Economic load dispatch using PSO and TLBO | |
| CN110490488B (en) | Power enterprise main network planning data analysis system based on big data analysis technology | |
| Amjady et al. | Optimal reliable operation of hydrothermal power systems with random unit outages | |
| Yuan et al. | Hydrothermal systems generation scheduling using cultural algorithm | |
| Zheng et al. | Probabilistic Load Forecasting with High Penetration of Renewable Energy Based on Variable Selection and Residual Modeling | |
| Bai et al. | Correlation analysis and prediction of power network loss based on mutual information and artificial neural network |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181113 |