CN109447362A - A kind of Spatial Load Forecasting method based on Fuzzy Information Granulation and support vector machines - Google Patents
A kind of Spatial Load Forecasting method based on Fuzzy Information Granulation and support vector machines Download PDFInfo
- Publication number
- CN109447362A CN109447362A CN201811323801.6A CN201811323801A CN109447362A CN 109447362 A CN109447362 A CN 109447362A CN 201811323801 A CN201811323801 A CN 201811323801A CN 109447362 A CN109447362 A CN 109447362A
- Authority
- CN
- China
- Prior art keywords
- class
- load
- cellular
- load density
- unit interval
- 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
- 238000005469 granulation Methods 0.000 title claims abstract description 36
- 230000003179 granulation Effects 0.000 title claims abstract description 36
- 238000012706 support-vector machine Methods 0.000 title claims abstract description 20
- 238000013277 forecasting method Methods 0.000 title claims abstract description 11
- 230000001413 cellular effect Effects 0.000 claims abstract description 199
- 230000003834 intracellular effect Effects 0.000 claims description 56
- 239000011159 matrix material Substances 0.000 claims description 43
- 238000000034 method Methods 0.000 claims description 38
- 239000002245 particle Substances 0.000 claims description 26
- NAWXUBYGYWOOIX-SFHVURJKSA-N (2s)-2-[[4-[2-(2,4-diaminoquinazolin-6-yl)ethyl]benzoyl]amino]-4-methylidenepentanedioic acid Chemical compound C1=CC2=NC(N)=NC(N)=C2C=C1CCC1=CC=C(C(=O)N[C@@H](CC(=C)C(O)=O)C(O)=O)C=C1 NAWXUBYGYWOOIX-SFHVURJKSA-N 0.000 claims description 6
- 230000008859 change Effects 0.000 claims description 5
- 238000005259 measurement Methods 0.000 claims description 5
- 240000002853 Nelumbo nucifera Species 0.000 claims description 4
- 235000006508 Nelumbo nucifera Nutrition 0.000 claims description 4
- 235000006510 Nelumbo pentapetala Nutrition 0.000 claims description 4
- 238000002790 cross-validation Methods 0.000 claims description 4
- 238000011161 development Methods 0.000 claims description 4
- 230000002159 abnormal effect Effects 0.000 claims description 3
- 239000008187 granular material Substances 0.000 claims description 3
- 238000012549 training Methods 0.000 claims description 3
- 239000000284 extract Substances 0.000 claims description 2
- 230000008569 process Effects 0.000 claims description 2
- 230000000694 effects Effects 0.000 abstract description 4
- 230000008901 benefit Effects 0.000 abstract description 3
- PEDCQBHIVMGVHV-UHFFFAOYSA-N Glycerine Chemical compound OCC(O)CO PEDCQBHIVMGVHV-UHFFFAOYSA-N 0.000 description 51
- 230000000875 corresponding effect Effects 0.000 description 20
- 230000018109 developmental process Effects 0.000 description 3
- 230000005611 electricity Effects 0.000 description 3
- 238000011156 evaluation Methods 0.000 description 2
- 238000009499 grossing Methods 0.000 description 2
- 238000012417 linear regression Methods 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 238000009412 basement excavation Methods 0.000 description 1
- 238000012512 characterization method Methods 0.000 description 1
- 239000004020 conductor Substances 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 230000005259 style development Effects 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—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
The present invention is a kind of Spatial Load Forecasting method based on Fuzzy Information Granulation and support vector machines, characterized in that is comprised the step of: building GIS for electric power first, and generates two class cellulars respectively wherein;Secondly according to the thickness of the length differentiating I class cellular duty granular degree of time scale; by dividing fuzzy granulation window; it establishes reasonable fuzzy set and Fuzzy Information Granulation is carried out to the historical load data under I class cellular fine granularity, and then determine the reasonable maximum value of the historical load under I class cellular coarse-grain;Then supporting vector machine model is used, I class cellular load under coarse-grain is predicted;It finally determines I class cellular load density equalizing coefficient, seeks classed load density index, acquire each II class cellular predicted load in conjunction with land used information, to realize the gridding to Spatial Load Forecasting result.It is reasonable with methodological science, strong applicability, prediction effect more preferably the advantages that.
Description
Technical field
The present invention relates to Spatial Load Forecasting field in distribution network planning, be it is a kind of based on Fuzzy Information Granulation and support to
The Spatial Load Forecasting method of amount machine.
Background technique
Basic work of the Spatial Load Forecasting (spatial load forecasting, SLF) as Power System Planning
Make, result instructs the investment planning of the following power grid, and the accuracy of Spatial Load Forecasting not only influences transformer capacity, electricity
The selection of web frame, voltage class, conductor cross-section, to the reasonability of entire electric power networks layout, power grid construction and its operation
Economy and safety have profound influence.
Most common SLF method has four classes: polytomy variable method, landuse emulation, tendency method, load density target
Method.Polytomy variable method establishes corresponding Extrapolating model on this basis using the factor of influence load as correlated variables to predict
The load in the cellular non-coming year.In view of the factor and its complexity for influencing load, the non-linear relation between whole influence factors is considered
Relatively difficult, in addition more demanding to the quantity of data and precision, prediction effect is poor, studies both at home and abroad it very few.With
Simulation method usual way in ground is to be evaluated in the form of marking land-use style development degree, and core is land used decision.This
Class method is classed load density as known conditions, but the determination of its size is not easy to.Tendency method is in region to be predicted
Generate cellular, according to cellular historical data extrapolate its plan year Load Development Trend, this method it is although easy to operate and relatively hold
It easily realizes, but cellular develops the complicated factors such as the continuous variation of unbalanced, load transfer and cellular and this method is had centainly
Limitation.District load density index method is domestic because it has the advantage of stronger adaptability to the variation of urban planning scheme
Outer extensive use.Such method is studied, is needed most based on demand history data and relevant Decision variable, existing method
It is the improvement and application to relevant Decision variable, and is also the pass for improving reliability forecasting to the effective use of historical load data
Key.Such method is insufficient to the utilization rate of historical load data, ignores the deep excavation to actual measurement historical load data.
If directly using actual measurement load data maximum value progress Spatial Load Forecasting in practical projects, in cellular load
Abnormal data will lead to prediction result precision reduction, by determine and can be obviously improved using the reasonable maximum value of cellular load
Precision of prediction.
Summary of the invention
It is an object of the present invention to provide a kind of scientific and reasonable, strong applicability, prediction effect is more preferably based on fuzzy information granule
Change the Spatial Load Forecasting method with support vector machines.
Realize the object of the invention the technical scheme adopted is that a kind of sky based on Fuzzy Information Granulation and support vector machines
Between load forecasting method, which is characterized in that it the following steps are included:
1) GIS for electric power is constructed
1. being registrated base figure;
2. establishing the 10kV feeder line supply district figure layer in region to be predicted;
3. establishing the land used information figure layer in region to be predicted;
4. establishing I class, II class cellular figure layer respectively;
I class cellular is generated with each 10kV feeder line supply district in region to be predicted, establishes I class cellular figure layer;It chooses suitable
When spatial resolution, in region to be predicted generate II class cellular, establish II class cellular figure layer;
2) Fuzzy Information Granulation and SVM prediction model are established
Fuzzy Information Granulation window is divided first, and fuzzy letter then is carried out to the load data under I class cellular fine granularity
Breath granulation, extracts the effective information of obscure particle in each window, the reasonable maximum value of I class cellular historical load is determined, finally to I
The reasonable maximum value of class cellular is supported vector machine prediction, obtains the Spatial Load Forecasting result based on I class cellular;
Granularity is divided according to time scale, and the thickness of granularity is an opposite concept, the selection of coarse-grain
It is determined according to the time scale of prediction target, fine granularity is closed according to the sample frequency and forecast demand of historical load data
Reason is chosen;
1. dividing fuzzy granulation window
Complicated information data is divided into several set according to certain feature by Information Granulating, and each set is exactly one
A information;Using the Information Granulating model based on fuzzy set;Fuzzy granulation window is divided by I class cellular historical load data sequence
Column are divided into several load data subsequences, each load data subsequence is an Information Granulating window;
2. the blurring of window information
The key of the blurring of window information is to establish reasonable fuzzy set, is blurred to window information, uses mould
Paste the information that information replaces the load data of parent window;Window information blurring is built in the load data sequence X of division
An obscure particle h is found, i.e., one the fuzzy concept G, G that can rationally describe X is the fuzzy set using X as domain, it is determined that G
Also obscure particle h has been determined that, relationship such as formula (1):
Wherein, x is the variable in domain;Fuzzy concept G is the fuzzy set using X as domain;The essence of fuzzification process
It is determining function A, A is the membership function of fuzzy concept G;First have to determine that the form of obscure particle determines tool again when fuzzy granulation
The membership function A of body selects triangle obscure particle, membership function A such as formula (2):
Wherein, x is the variable in domain;A, l, b are the parameter of membership function, correspond respectively to the fuzzy granulation of each window
Three parameters afterwards: LOW, R and UP;For single obscure particle, LOW parameter describes the corresponding initial data of the particle
Minimum value, R parameter describes the average value of the corresponding initial data of the particle, and it is original accordingly that UP parameter describes the particle
The maximum value of data;
3. determining the reasonable maximum value of I class cellular historical load
To any one I class cellular, each window information is the historical load data under I class cellular fine granularity, passes through mould
Useful information is excavated and extracted to paste Information Granulating technology during being blurred to it, overcomes abnormal data bring to interfere, obtains
To the UP particle of each window;It is empty that UP particle accurately describes the maximum that I class cellular historical load data under coarse-grain changes
Between, so that the UP particle of the window to be determined as to the reasonable maximum value of I class cellular historical load under coarse-grain;
4. Support vector regression is predicted
Kernel function using gaussian kernel function as support vector machines (support vector machine, SVM) uses
K walks cross-validation method (k-fold cross validation, K-CV) method and carries out optimizing to penalty parameter c and nuclear parameter g to mention
The accuracy rate of high regression forecasting;It is propped up using the reasonable maximum value of I class cellular historical load as the training sample of SVM model
Vector machine prediction is held, the Spatial Load Forecasting result based on I class cellular is obtained;
3) gridding of space electric load
1. generating II class cellular
In the GIS for electric power established according to step 1), with etc. sizes square net generate II class cellular,
Establish II class cellular figure layer;
2. determining load density equalizing coefficient
Firstly, the load density of each I class cellular is found out, the load density of the I class cellular in each unit interval
It is acquired by formula (3),
dik=pik/Sik (3)
Wherein, dikFor the load density of i-th of I class cellular in k-th of unit interval, i=1,2 ..., n, n is I class member
The number of born of the same parents;K=1,2 ..., q, q are the number of unit period;pikFor i-th I class cellular in k-th of unit interval
Load maximum value;SikFor the area of i-th of I class cellular in k-th of unit interval;
In view of load density is also variant in the different I class cellular of development degree for same type load, here, introducing load
Density equalization coefficient, is denoted as β,
βik=dik/dk.max (4)
Wherein, βikFor the load density equalizing coefficient of i-th of I class cellular in k-th of unit interval;I=1,2 ..., n,
N is the number of I class cellular;K=1,2 ..., q, q are the number of unit period;dikFor i-th of I in k-th of unit interval
The load density of class cellular;dk.maxFor the maximum value of load density in all I class cellulars in k-th of unit interval;
3. seeking each classed load density index
The maximum value of each I class member jth class land used load density intracellular can be obtained multiplied by corresponding load density equalizing coefficient
To each I class member jth class land used load density intracellular, formula (5) are seen,
Dijk=βikDjk (5)
Wherein, DijkFor i-th of I class member jth class land used load density intracellular in k-th of unit interval, i=1,2 ...,
N, n are the number of I class cellular;J=1,2 ..., m, m are the class number of land-use style;K=1,2 ..., q, q are the unit period
Number;βikFor the load density equalizing coefficient of i-th of I class cellular in k-th of unit interval;DjkFor k-th of unit interval
The maximum value of interior each I class member jth class land used load density intracellular;
For any one I class cellular, load is equal to the first all kinds of land areas intracellular of I class and corresponding classed load is close
The sum for spending product, is shown in formula (6),
Wherein, pikFor the load maximum value of i-th of I class cellular in k-th of unit interval, i=1,2 ..., n, n is I class
The number of cellular;K=1,2 ..., q, q are the number of unit period;SijkFor i-th of I class cellular in k-th of unit interval
Interior jth class land area;J=1,2 ..., m, m are the class number of land-use style;DijkFor i-th of I class in k-th of unit interval
Member jth class land used load density intracellular;
In conjunction with formula (5) and formula (6), formula (7) can be obtained,
Wherein, pikFor the load maximum value of i-th of I class cellular in k-th of unit interval, i=1,2 ..., n, n is I class
The number of cellular;K=1,2 ..., q, q are the number of unit period;βikFor i-th of I class cellular in k-th of unit interval
Load density equalizing coefficient;SijkFor i-th of I class member jth class land area intracellular in k-th of unit interval, j=1,
2 ..., m, m are the class number of land-use style;DjkFor I class member each in k-th of unit interval jth class land used load density intracellular
Maximum value;
The matrix representation forms such as formula (8) of formula (7),
P=B S D=C D (8)
Wherein, P is the unit interval internal loading maximum value matrix of I class cellular;B is the equalizing coefficient square of each load density
Battle array;S is I class member land area matrix intracellular;D is each I class member classed load density maxima matrix intracellular;C is that I class member is intracellular
The matrix of the corresponding load density equalizing coefficient product of the area of all kinds of lands used;
Relationship between the unit interval internal loading maximum estimated value and classed load density maximum estimated value of I class cellular
Formula (9) are expressed as,
Wherein,For the unit interval internal loading maximum estimated value matrix of I class cellular;C is I class member all kinds of lands used intracellular
The corresponding load density equalizing coefficient product of area matrix;Estimate for each I class member classed load density maximum intracellular
Evaluation matrix;
In view of error in measurement have just have it is negative, the unit interval internal loading maximum value and estimated value of all I class cellulars
The summation of quadratic sum of difference be denoted as Q, see formula (10),
Wherein, PiFor the unit interval internal loading maximum value matrix of i-th of I class cellular, i=1,2 ..., n, n is I class
The number of cellular;For the unit interval internal loading maximum estimated value matrix of i-th of I class cellular, i=1,2 ..., n, n I
The number of class cellular;Q is the total of the quadratic sum of the unit interval internal loading maximum value of all I class cellulars and the difference of estimated value
With;
Using principle of least square method to formula (10) classed load density maximum estimated value matrixIt is solved, is obtained
Classed load density maximum estimated value, is shown in formula (11) and (12),
Wherein, P is the unit interval internal loading maximum value matrix of I class cellular;C is the face of I class member all kinds of lands used intracellular
The matrix of the corresponding load density equalizing coefficient product of product;For each I class member classed load density maximum estimated value intracellular
Matrix;
The average load density value-acquiring method of land-use style of the same race is shown in formula (13),
Wherein,Indicate the average load density value of jth class land-use style in k-th of unit interval;It is k-th
Each I class member jth class land used load density maximum estimated value intracellular in unit interval, j=1,2 ..., m, m are land-use style
Class number;βikFor the load density equalizing coefficient of i-th of I class cellular in k-th of unit interval;SijkFor k-th of unit interval
Interior i-th of I class member jth class land area intracellular;
The average load density value for the various land-use styles being calculated is each classed load density index;
4. calculating each II class cellular predicted load
In the GIS for electric power established according to step 1), in conjunction with the use 3. established according to step 1) sub-step
Ground information figure layer determines each area according to step 3) sub-step 1. II class generated member various land-use styles intracellular,
And by it multiplied by the average load density value of the corresponding various land-use styles according to calculated by formula (13), multiplied by according to public affairs
2. corresponding load density equalizing coefficient that formula step 3) sub-step determines, calculates the load of each II class cellular target time section
Predicted value, to realize the gridding to Spatial Load Forecasting result.
A kind of Spatial Load Forecasting method based on Fuzzy Information Granulation and support vector machines of the invention, first building electricity
Power GIS-Geographic Information System, and generate two class cellulars respectively wherein.Secondly according to the length differentiating I class cellular load of time scale
It is negative to the history under I class cellular fine granularity to establish reasonable fuzzy set by dividing fuzzy granulation window for the thickness of granularity
Lotus data carry out Fuzzy Information Granulation, and then determine the reasonable maximum value of the historical load under I class cellular coarse-grain;Then
Using supporting vector machine model, I class cellular load under coarse-grain is predicted.Finally determine I class cellular load density
Equalizing coefficient seeks classed load density index, acquires each II class cellular predicted load in conjunction with land used information, thus realization pair
The gridding of Spatial Load Forecasting result.With scientific and reasonable, strong applicability, prediction effect more preferably the advantages that.
Detailed description of the invention
Fig. 1 is the basic of Spatial Load Forecasting method of the utilization based on Fuzzy Information Granulation and support vector machines of the invention
Schematic illustration.
Fig. 2 is region 10kV feeder line supply district figure layer to be predicted.
Fig. 3 is land used information figure layer in region to be predicted.
Fig. 4 is I class cellular figure layer.
Fig. 5 is II class cellular figure layer.
Fig. 6 is the reasonable maximum value of each moon historical load of literary 28 lines of I class cellular.
Fig. 7 is SVM parameter optimization result figure.
Fig. 8 is the quasi- measured value of each II class cellular load.
Fig. 9 is each II class cellular predicted load.
Specific embodiment
Below with drawings and examples, invention is further explained.
The art usually by: English based on Fuzzy Information Granulation and is abbreviated as (fuzzy information
Granulation, FIG);It the English of support vector machines and is abbreviated as (support vector machine, SVM);Space is negative
The English and be abbreviated as (spatial load forecasting, SLF) that lotus is predicted;Therefore, of the invention based on fuzzy message
The Spatial Load Forecasting method of granulation and support vector machines is referred to as FIG-SVM SLF method.
- Fig. 9 referring to Fig.1, the Spatial Load Forecasting method of the invention based on Fuzzy Information Granulation and support vector machines,
In, comprising the following steps:
By each I class cellular history on the November 30,1 day to 2015 January in 2013 in an administrative area in the city of Henan
Load data is as modeling domain, using each I class cellular historical load data in December, 2015 as prediction domain.
1) GIS for electric power is constructed
1. being registrated base figure
Obtain Henan city satellite photo, and in GIS-Geographic Information System environment according to actual longitude and latitude to its into
Row registration is used as base figure.
2. establishing the 10kV feeder line supply district figure layer in region to be predicted
The base figure that the sub-step of step 1) is 1. registrated determines region to be predicted as background, and establishes area to be predicted
10kV feeder line supply district figure layer in domain is shown in Fig. 2 wherein sharing 28 10kV feeder lines.
3. establishing the land used information figure layer in region to be predicted
The base figure that the sub-step of step 1) is 1. registrated determines region to be predicted as background, and establishes area to be predicted
Land used information figure layer in domain is shown in Fig. 3 wherein sharing 8 kinds of land-use styles.
4. establishing I class, II class cellular cellular figure layer respectively
In the GIS for electric power established according to step 1), power according to 10kV feeder line each in region to be predicted
Range generates I class cellular, each 10kV feeder line supply district is exactly an I class cellular, and one shares 28 I in region to be predicted
Class cellular, is shown in Fig. 4.In the GIS for electric power established according to step 1), according to side length be 0.3km etc. sizes just
Square net generates II class cellular, each grid is exactly an II class cellular, and one shares 148 II classes members in region to be predicted
Born of the same parents see Fig. 5.
2) Fuzzy Information Granulation and SVM prediction model are established
1. dividing fuzzy granulation window
In region endosymbiosis to be predicted at 28 I class cellulars.Unit interval to be predicted is the moon, therefore I class cellular is thick
Load data under granularity selects matching moon load maximum value, according to the demand of prediction, under I class cellular fine granularity
Load data be chosen for daily load maximum value.By taking literary 28 lines of I class cellular as an example, in order to unit interval to be predicted
Match, with 30 days for a window, therefore models domain and be divided into 35 windows.
2. the blurring of window information
Triangle obscure particle is selected to carry out Fuzzy Information Granulation;Fuzzy Information Granulation is carried out to modeling domain using formula (2)
Afterwards, each window obtains tri- fuzzy information granules of LOW, R and UP, 3 parameters of a, l, b of as membership function A, corresponding description text
28 the line moon load minimum value, average value and maximum value;
3. determining the reasonable maximum value of I class cellular historical load data
The UP particle of corresponding 35 windows, forms UP array.The load of UP array Efficient Characterization literary 28 line history moons
Maximum value, determine it as the reasonable maximum value of the load of the literary 28 line history moons.The load of the I class cellular literary 28 line history moons
Reasonable maximum value is shown in Fig. 6.Similarly determine that the reasonable maximum value of the load of each I class cellular history moon, relevant specific value are shown in Table
1。
The reasonable maximum value of the load of each I class cellular history moon of table 1
4. Support vector regression is predicted
Establish SVM model, be normalized first, using text 28 the line history moon load reasonable maximum value as training set into
The building of row model.Kernel functional parameter g and penalty factor c is chosen using K-CV method.Rough searching is carried out first, and observation is rough
It is finely selected again after the result of searching, obtains kernel functional parameter g=0.044, penalty factor c=256, SVM parameter optimization
As a result see Fig. 7.
SVM is trained using obtained optimized parameter c and g, literary 28 lines in December, 2015 (i.e. the target moon) are born
Lotus maximum value is predicted.
Similarly predict the target moon load maximum value of each I class cellular.As a comparison, in January, 2013 to 2015 11 is utilized
Month actual moon load maximum value of I class cellular, using support vector machines method, gray theory method, exponential smoothing and linear regression
Method predicts I class cellular of the target moon, is shown in Table 2.
Each I class cellular target moon load maximum value of table 2
3) gridding of space electric load
1. generating II class cellular
In the GIS for electric power established according to step 1), with side length for 0.3km etc. sizes square net
Lattice generate II class cellular, each grid is exactly an II class cellular, and one shares 148 II class cellulars in region to be predicted, see
Fig. 5.
2. determining load density equalizing coefficient
Firstly, the load density of each I class cellular is found out, in the GIS for electric power established according to step 1)
In, determine the area of I class member various land-use styles intracellular.According to the target moon load maximum value of I class cellular in table 2, in target
The load density of month each I class cellular is acquired by formula (3),
dik=pik/Sik (3)
Wherein, dikFor the load density of i-th of I class cellular of the target moon, i=1,2 ..., 28, k=1;pikFor the target moon
The load maximum value of i I class cellular;SikFor the area of i-th of I class cellular of the target moon;In the present invention, according to step 1) sub-step
Suddenly the 10kV feeder line supply district figure layer sub-step 2. established 3. established and land used information figure layer be fixed, therefore I class
The area of cellular does not change over.The calculated target moon, the load density of each I class cellular was as shown in table 3:
The 3 target moon of table each I class cellular load density
In view of load density is also variant in the different I class cellular of development degree for same type load, here, introducing load
Density equalization coefficient, is denoted as β;It determines load density equalizing coefficient, sees formula (4),
βik=dik/dk.max (4)
In formula: βikFor the load density equalizing coefficient of i-th of I class cellular of the target moon;I=1,2 ..., 28, k=1;dikFor
The target moon i-th of I class cellular load density;dk.maxFor the maximum value of load density in the target moon all I class cellulars;It finds out
The target moon, the load density equalizing coefficient of each I class cellular was as shown in table 3:
The 4 target moon of table each I class cellular load density equalizing coefficient
3. seeking each classed load density index
The maximum value of each I class member jth class land used load density intracellular can be obtained multiplied by corresponding load density equalizing coefficient
To each I class member jth class land used load density intracellular, formula (5) are seen,
Dijk=βikDjk (5)
Wherein, DijkFor i-th of target moon I class member jth class land used load density intracellular, i=1,2 ..., 28, j=1,
2,…,8;βikFor the load density equalizing coefficient of i-th of I class cellular of the target moon, i=1,2 ..., 28, k=1;DjkFor the target moon
The maximum value of each I class member jth class land used load density intracellular;
For any one I class cellular, load is equal to the first all kinds of land areas intracellular of I class and corresponding classed load is close
The sum for spending product, is shown in formula (6),
Wherein, pikFor the load maximum value of i-th of I class cellular of the target moon, i=1,2 ..., 28, k=1;SijkFor the target moon
I-th of I class member jth class land area intracellular, j=1,2 ..., 8;In the present invention, it is 2. established according to step 1) sub-step
3. 10kV feeder line supply district figure layer sub-step is established and land used information figure layer is fixed, therefore each I class member is intracellular
Various land areas do not change over.DijkFor i-th of I class member jth class land used load density intracellular of the target moon;
In conjunction with formula (5) and formula (6), formula (7) can be obtained,
Wherein, pikFor the load maximum value of i-th of I class cellular of the target moon, i=1,2 ..., 28, k=1;βikFor the target moon
The load density equalizing coefficient of i-th of I class cellular;SijkFor i-th of target moon I class member jth class land area intracellular, j=1,
2,…,8;DjkFor the maximum value of the target moon each I class member jth class land used load density intracellular;
The matrix representation forms such as formula (8) of formula (7),
P=B S D=C D (8)
Wherein, P is the load maximum value matrix of target moon I class cellular;B is the equalizing coefficient matrix of each load density;S is
I class member land area matrix intracellular;D is each I class member classed load density maxima matrix intracellular;C is I class member all kinds of use intracellular
The matrix of the corresponding load density equalizing coefficient product of the area on ground;
I class cellular the target moon load maximum estimated value and classed load density maximum estimated value between relationship can indicate
For formula (9),
Wherein,For the load maximum estimated value matrix of target moon I class cellular;C is the area of I class member all kinds of lands used intracellular
The matrix of corresponding load density equalizing coefficient product;For each I class member classed load density maximum estimated value square intracellular
Battle array;
In view of error in measurement have just have it is negative, the difference of the load maximum value and estimated value of the target moon all I class cellulars
The summation of quadratic sum is denoted as Q, sees formula (10),
Wherein, PiFor the load maximum value matrix of i-th of I class cellular of the target moon, i=1,2 ..., 28;For i-th of I class
The load maximum estimated value matrix of the cellular target moon, i=1,2 ..., 28;Q is the load maximum value of the target moon all I class cellulars
And the summation of the quadratic sum of the difference of estimated value;
Using principle of least square method to formula (10) classed load density maximum estimated value matrixIt is solved, is obtained
Classed load density maximum estimated value is shown in formula (11) (12),
Wherein, P is the load maximum value matrix of target moon I class cellular;C is the area of I class member all kinds of lands used intracellular and its
The matrix of corresponding load density equalizing coefficient product;For each I class member classed load density maximum estimated value matrix intracellular;
The average load density value-acquiring method of land-use style of the same race is shown in formula (13),
Wherein,Indicate the average load density value of target moon jth class land-use style;For the target moon each I class cellular
Interior jth class land used load density maximum estimated value, j=1,2 ..., 8;βikFor the target moon, the load density of i-th of I class cellular is equal
Weigh coefficient, i=1, and 2 ..., 28, k=1;SijkFor i-th of target moon I class member jth class land area intracellular, j=1,2 ..., 8;
The classed load density index of the target moon in calculated region to be predicted, as shown in table 5:
The classed load density index of the 5 target moon of table
4. seeking each II class cellular predicted load
In the GIS for electric power established according to step 1), in conjunction with what is 3. established according to step 1) sub-step
Land used information figure layer determines each face according to step 3) sub-step 1. II class generated member various land-use styles intracellular
Product, and by it multiplied by the average load density value of the corresponding various land-use styles according to calculated by formula (13), multiplied by root
According to 2. corresponding load density equalizing coefficient that formula step 3) sub-step determines, the load of each II class cellular target moon is calculated
Predicted value, to realize the gridding to Spatial Load Forecasting result.
As a comparison, formula is utilized to the actual moon load maximum value of I class cellular November in 2015 using in January, 2013
(4)-formula (14) seeks the classed load density index of each moon, calculates monthly each II class cellular predicted load.Using support
Vector machine method, gray theory method, exponential smoothing and linear regression method predict target moon II class cellular.Each method it is pre-
Survey the corresponding related specific value of result and be shown in Table 6, as space is limited, cannot whole 148 II class cellulars prediction result, only list
8 II class cellulars, the location of these cellulars are as shown in Figure 5.
6 II class cellular load prediction results of table
The relative error of each method prediction result is sought, and counts the II class cellular number in relative error section and accounts for cellular
The ratio of sum, the results are shown in Table 7.
7 error evaluation table of table
As can be seen from Table 7, the SLF based on Fuzzy Information Granulation and support vector machines misses the prediction of all II class cellulars
Difference both less than 40%, and in the burst error of 0%-10% II class cellular number account for cellular sum ratio be 71.62%, it is bright
Aobvious 34.46%, 21.62%, 26.35% and 27.03% better than other 4 kinds of methods.It can be seen that being based on Fuzzy Information Granulation
SLF with support vector machines has higher precision compared with other 4 kinds of prediction techniques.
To keep prediction result more intuitive, by the quasi- measured value of in December, 2015 each II class cellular load and predicted value in electricity
It is shown in power GIS, sees Fig. 8, Fig. 9 respectively.
The particular embodiment of the present invention is made that detailed explanation to the contents of the present invention, but does not limit to the present embodiment,
Those skilled in the art are according to the present invention to enlighten any obvious change done, and belongs to rights protection of the present invention
Range.
Claims (1)
1. a kind of Spatial Load Forecasting method based on Fuzzy Information Granulation and support vector machines, which is characterized in that it include with
Lower step:
1) GIS for electric power is constructed
1. being registrated base figure;
2. establishing the 10kV feeder line supply district figure layer in region to be predicted;
3. establishing the land used information figure layer in region to be predicted;
4. establishing I class, II class cellular figure layer respectively;
I class cellular is generated with each 10kV feeder line supply district in region to be predicted, establishes I class cellular figure layer;It chooses appropriate
Spatial resolution generates II class cellular in region to be predicted, establishes II class cellular figure layer;
2) Fuzzy Information Granulation and SVM prediction model are established
Fuzzy Information Granulation window is divided first, and fuzzy information granule then is carried out to the load data under I class cellular fine granularity
Change, extract the effective information of obscure particle in each window, determine the reasonable maximum value of I class cellular historical load, finally to I class member
The reasonable maximum value of born of the same parents is supported vector machine prediction, obtains the Spatial Load Forecasting result based on I class cellular;
Granularity is divided according to time scale, and the thickness of granularity is an opposite concept, the selection of coarse-grain according to
The time scale of target is predicted to determine, fine granularity is rationally selected according to the sample frequency and forecast demand of historical load data
It takes;
1. dividing fuzzy granulation window
Complicated information data is divided into several set according to certain feature by Information Granulating, and each set is exactly a letter
Cease grain;Using the Information Granulating model based on fuzzy set;Fuzzy granulation window is divided to draw I class cellular historical load data sequence
It is divided into several load data subsequences, each load data subsequence is an Information Granulating window;
2. the blurring of window information
The key of the blurring of window information is to establish reasonable fuzzy set, be blurred to window information, with fuzzy letter
Cease the information that grain replaces the load data of parent window;Window information blurring is to establish one in the load data sequence X of division
A obscure particle h, i.e., one the fuzzy concept G, G that can rationally describe X are the fuzzy sets using X as domain, it is determined that G is also
Obscure particle h is determined, relationship such as formula (1):
Wherein, x is the variable in domain;Fuzzy concept G is the fuzzy set using X as domain;The essence of fuzzification process is true
Determine function A, A is the membership function of fuzzy concept G;First have to determine that the form of obscure particle determines specifically again when fuzzy granulation
Membership function A selects triangle obscure particle, membership function A such as formula (2):
Wherein, x is the variable in domain;A, l, b are the parameter of membership function, after corresponding respectively to the fuzzy granulation of each window
Three parameters: LOW, R and UP;For single obscure particle, LOW parameter describes the corresponding initial data of the particle most
Small value, R parameter describe the average value of the corresponding initial data of the particle, and UP parameter describes the corresponding initial data of the particle
Maximum value;
3. determining the reasonable maximum value of I class cellular historical load
To any one I class cellular, each window information is the historical load data under I class cellular fine granularity, passes through fuzzy letter
Useful information is excavated and extracted to breath granulation technique during being blurred to it, and abnormal data bring is overcome to interfere, and obtains each
The UP particle of window;UP particle accurately describes the maximum space that I class cellular historical load data changes under coarse-grain, from
And the UP particle of the window is determined as to the reasonable maximum value of I class cellular historical load under coarse-grain;
4. Support vector regression is predicted
Kernel function using gaussian kernel function as support vector machines (support vector machine, SVM), is walked using K
Cross-validation method (k-fold cross validation, K-CV) method carries out optimizing to penalty parameter c and nuclear parameter g to improve
The accuracy rate of regression forecasting;It is supported using the reasonable maximum value of I class cellular historical load as the training sample of SVM model
Vector machine prediction, obtains the Spatial Load Forecasting result based on I class cellular;
3) gridding of space electric load
1. generating II class cellular
In the GIS for electric power established according to step 1), with etc. sizes square net generate II class cellular, establish
II class cellular figure layer;
2. determining load density equalizing coefficient
Firstly, finding out the load density of each I class cellular, the load density of the I class cellular in each unit interval is by public affairs
Formula (3) acquires,
dik=pik/Sik (3)
Wherein, dikFor the load density of i-th of I class cellular in k-th of unit interval, i=1,2 ..., n, n is I class cellular
Number;K=1,2 ..., q, q are the number of unit period;pikFor the load of i-th of I class cellular in k-th of unit interval
Maximum value;SikFor the area of i-th of I class cellular in k-th of unit interval;
In view of load density is also variant in the different I class cellular of development degree for same type load, here, introducing load density
Equalizing coefficient is denoted as β,
βik=dik/dk.max (4)
Wherein, βikFor the load density equalizing coefficient of i-th of I class cellular in k-th of unit interval;I=1,2 ..., n, n I
The number of class cellular;K=1,2 ..., q, q are the number of unit period;dikFor i-th of I class member in k-th of unit interval
The load density of born of the same parents;dk.maxFor the maximum value of load density in all I class cellulars in k-th of unit interval;
3. seeking each classed load density index
The maximum value of each I class member jth class land used load density intracellular can be obtained each multiplied by corresponding load density equalizing coefficient
I class member jth class land used load density intracellular, is shown in formula (5),
Dijk=βikDjk (5)
Wherein, DijkFor i-th of I class member jth class land used load density intracellular, i=1,2 ..., n, n in k-th of unit interval
For the number of I class cellular;J=1,2 ..., m, m are the class number of land-use style;K=1,2 ..., q, q are of unit period
Number;βikFor the load density equalizing coefficient of i-th of I class cellular in k-th of unit interval;DjkFor in k-th of unit interval
The maximum value of each I class member jth class land used load density intracellular;
For any one I class cellular, load is equal to the I class member all kinds of land areas intracellular and multiplies with corresponding classed load density
Long-pending sum is shown in formula (6),
Wherein, pikFor the load maximum value of i-th of I class cellular in k-th of unit interval, i=1,2 ..., n, n is I class cellular
Number;K=1,2 ..., q, q are the number of unit period;SijkFor i-th of I class member in k-th of unit interval intracellular the
J class land area;J=1,2 ..., m, m are the class number of land-use style;DijkFor i-th of I class cellular in k-th of unit interval
Interior jth class land used load density;
In conjunction with formula (5) and formula (6), formula (7) can be obtained,
Wherein, pikFor the load maximum value of i-th of I class cellular in k-th of unit interval, i=1,2 ..., n, n is I class cellular
Number;K=1,2 ..., q, q are the number of unit period;βikFor in k-th of unit interval i-th I class cellular it is negative
Lotus density equalization coefficient;SijkFor i-th of I class member jth class land area intracellular in k-th of unit interval, j=1,2 ..., m,
M is the class number of land-use style;DjkFor the maximum value of I class member each in k-th of unit interval jth class land used load density intracellular;
The matrix representation forms such as formula (8) of formula (7),
P=BSD=CD (8)
Wherein, P is the unit interval internal loading maximum value matrix of I class cellular;B is the equalizing coefficient matrix of each load density;S
For I class member land area matrix intracellular;D is each I class member classed load density maxima matrix intracellular;C is that I class member is intracellular all kinds of
The matrix of the corresponding load density equalizing coefficient product of the area of land used;
Relationship between the unit interval internal loading maximum estimated value and classed load density maximum estimated value of I class cellular indicates
For formula (9),
Wherein,For the unit interval internal loading maximum estimated value matrix of I class cellular;C is the face of I class member all kinds of lands used intracellular
The matrix of the corresponding load density equalizing coefficient product of product;For each I class member classed load density maximum estimated value intracellular
Matrix;
In view of error in measurement have just have it is negative, the unit interval internal loading maximum value of all I class cellulars and the difference of estimated value
The summation of quadratic sum be denoted as Q, see formula (10),
Wherein, PiFor the unit interval internal loading maximum value matrix of i-th of I class cellular, i=1,2 ..., n, n is I class cellular
Number;For the unit interval internal loading maximum estimated value matrix of i-th of I class cellular, i=1,2 ..., n, n is I class cellular
Number;Q is the summation of the quadratic sum of the unit interval internal loading maximum value of all I class cellulars and the difference of estimated value;
Using principle of least square method to formula (10) classed load density maximum estimated value matrixIt is solved, is classified
Load density maximum estimated value is shown in formula (11) and (12),
Wherein, P is the unit interval internal loading maximum value matrix of I class cellular;C be I class member all kinds of lands used intracellular area with
The matrix of its corresponding load density equalizing coefficient product;For each I class member classed load density maximum estimated value matrix intracellular;
The average load density value-acquiring method of land-use style of the same race is shown in formula (13),
Wherein,Indicate the average load density value of jth class land-use style in k-th of unit interval;For k-th of unit
Each I class member jth class land used load density maximum estimated value intracellular in period, j=1,2 ..., m, m are the class number of land-use style;
βikFor the load density equalizing coefficient of i-th of I class cellular in k-th of unit interval;SijkIt is in k-th of unit interval i-th
A I class member jth class land area intracellular;
The average load density value for the various land-use styles being calculated is each classed load density index;
4. calculating each II class cellular predicted load
In the GIS for electric power established according to step 1), in conjunction with the land used letter 3. established according to step 1) sub-step
Figure layer is ceased, determines each area according to step 3) sub-step 1. II class generated member various land-use styles intracellular, and will
It is walked multiplied by the average load density value of the corresponding various land-use styles according to calculated by formula (13) multiplied by according to formula
2. corresponding load density equalizing coefficient that rapid 3) sub-step determines, calculates the load prediction of each II class cellular target time section
Value, to realize the gridding to Spatial Load Forecasting result.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811323801.6A CN109447362A (en) | 2018-11-08 | 2018-11-08 | A kind of Spatial Load Forecasting method based on Fuzzy Information Granulation and support vector machines |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811323801.6A CN109447362A (en) | 2018-11-08 | 2018-11-08 | A kind of Spatial Load Forecasting method based on Fuzzy Information Granulation and support vector machines |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109447362A true CN109447362A (en) | 2019-03-08 |
Family
ID=65551305
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811323801.6A Pending CN109447362A (en) | 2018-11-08 | 2018-11-08 | A kind of Spatial Load Forecasting method based on Fuzzy Information Granulation and support vector machines |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109447362A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111062538A (en) * | 2019-12-21 | 2020-04-24 | 东北电力大学 | CEEMD method for determining reasonable maximum value of cellular load in space load prediction |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103258246A (en) * | 2013-05-16 | 2013-08-21 | 东北电力大学 | Method for obtaining load density index based on cellular historical data |
CN107910864A (en) * | 2017-10-11 | 2018-04-13 | 长沙理工大学 | A kind of 220kV busbar equivalent load volatility forecast methods based on Information Granulating and support vector machines |
-
2018
- 2018-11-08 CN CN201811323801.6A patent/CN109447362A/en active Pending
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103258246A (en) * | 2013-05-16 | 2013-08-21 | 东北电力大学 | Method for obtaining load density index based on cellular historical data |
CN107910864A (en) * | 2017-10-11 | 2018-04-13 | 长沙理工大学 | A kind of 220kV busbar equivalent load volatility forecast methods based on Information Granulating and support vector machines |
Non-Patent Citations (7)
Title |
---|
侯聪 等: ""基于模糊信息粒化和支持向量机的空调负荷预测"", 《建筑热能通风空调》 * |
孔平 等: ""基于模糊信息粒化支持向量机的短期电力负荷预测"", 《电力信息与通信技术》 * |
聂鹏: ""基于支持向量机的城市电网空间负荷预测方法"", 《东北电力大学学报》 * |
肖白 等: ""基于元胞历史负荷数据的负荷密度指标法"", 《电网技术》 * |
肖白 等: ""基于模糊粗糙集理论和时空信息的空间负荷预测"", 《电力建设》 * |
肖白: ""基于多级聚类分析和支持向量机的空间负荷预测方法"", 《电力系统自动化》 * |
肖白: ""空间电力负荷预测方法综述与展望"", 《中国电机工程学报》 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111062538A (en) * | 2019-12-21 | 2020-04-24 | 东北电力大学 | CEEMD method for determining reasonable maximum value of cellular load in space load prediction |
CN111062538B (en) * | 2019-12-21 | 2022-09-20 | 东北电力大学 | CEEMD method for determining reasonable maximum value of cellular load in space load prediction |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN103679263B (en) | Forecasting Methodology is closed on based on the thunder and lightning of particle swarm support vector machine | |
Liu et al. | Optimal allocation of water quantity and waste load in the Northwest Pearl River Delta, China | |
CN108182484A (en) | Spatial Load Forecasting method based on gridding technology and BP neural network | |
Wu et al. | Prediction and analysis of water resources demand in Taiyuan City based on principal component analysis and BP neural network | |
CN110119590A (en) | A kind of water quality model particle filter assimilation method based on multi-source observation data | |
CN109086951A (en) | It is a kind of meter and urban development degree multistage Spatial Load Forecasting method | |
CN104504280A (en) | Planning-demand-considered comprehensive evaluation method for communication mode of cluster management system of charging piles of electric automobile | |
Yi et al. | Intelligent prediction of transmission line project cost based on least squares support vector machine optimized by particle swarm optimization | |
Miettinen et al. | Simulating wind power forecast error distributions for spatially aggregated wind power plants | |
CN103839113B (en) | Microscopic simulation method based on house selecting models of house-renting selectors and house selecting models of house-purchasing selectors | |
CN101599142A (en) | Land evaluation index classification quantitative method based on spatial data field | |
Cao | 1.17 Spatial optimization for sustainable land use planning | |
Sun et al. | Exploring the effects of population growth on future land use change in the Las Vegas Wash watershed: an integrated approach of geospatial modeling and analytics | |
Chen et al. | Flood control operation of reservoir group using Yin-Yang Firefly Algorithm | |
Giustolisi et al. | An evolutionary multiobjective strategy for the effective management of groundwater resources | |
Tian et al. | Combined Sewer Overflow and Flooding Mitigation Through a Reliable Real‐Time Control Based on Multi‐Reinforcement Learning and Model Predictive Control | |
CN105825295A (en) | Space load predication method with consideration of cellular development degree | |
Lin et al. | Intelligent prediction of the construction cost of substation projects using support vector machine optimized by particle swarm optimization | |
CN109447362A (en) | A kind of Spatial Load Forecasting method based on Fuzzy Information Granulation and support vector machines | |
CN109615119A (en) | A kind of Spatial Load Forecasting method based on rank Set Pair Analysis Theory | |
Xu et al. | A novel intelligent deep learning-based uncertainty-guided network training in market price | |
Suh et al. | A water demand forecasting model using BPNN for residential building | |
CN105243446A (en) | Electricity consumption combined forecasting method based on particle swarm optimization | |
CN108053055A (en) | Large size city Middle-long Electric Power Load Forecast method based on support vector machines | |
Gu et al. | Multi‐scenario simulation of land use change based on MCR‐SD‐FLUS model: A case study of Nanchang, China |
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 | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190308 |
|
WD01 | Invention patent application deemed withdrawn after publication |