CN108320046A - Short-term electric load prediction modeling method - Google Patents
Short-term electric load prediction modeling method Download PDFInfo
- Publication number
- CN108320046A CN108320046A CN201711445864.4A CN201711445864A CN108320046A CN 108320046 A CN108320046 A CN 108320046A CN 201711445864 A CN201711445864 A CN 201711445864A CN 108320046 A CN108320046 A CN 108320046A
- Authority
- CN
- China
- Prior art keywords
- electric load
- formula
- neural network
- matrix
- eigenvector
- 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
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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
- G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- 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—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/06—Electricity, gas or water supply
Abstract
The invention discloses short-term electric load prediction modeling methods, the characteristics of presented for electric load, Nonlinear feature extraction is carried out to the influence factor for influencing electric load using core principle component analysis, redundancy, colinearity information between elimination variable, the pivot characteristic component for obtaining influencing electric load, then predicts electric load using the strong BP neural network of non-linear classification；Using the technical program, core principle component analysis has effectively carried out Feature Dimension Reduction under the premise of being sufficiently reserved initial data more information to Power system load data, optimizes the structure of BP neural network, greatly improves predetermined speed and precision.
Description
Technical field
The invention belongs to load forecast fields, it is more particularly related to which short-term electric load prediction models
Method.
Background technology
With the development of economy, the raising of electric appliance utilization rate, energy management system have become regulation and control resident and list
One of the factor that must take into consideration of position electricity consumption, short-term electric load prediction can in a planned way carry out the planning of electric system, marketing,
The work such as marketing, scheduling, the methods of traditional prediction method having time serial method, regression analysis, fuzzy theory, but pass
Unite prediction technique it is affected by many factors, redundancy and conllinear category information cause network structure sufficiently complex, network operation speed compared with
Slowly, to limit the application of model.
Invention content
Technical problem to be solved by the invention is to provide a kind of based on core principle component, eliminates redundancy between variable, conllinear
The short-term electric load prediction modeling method of information.
To achieve the goals above, the technical solution that the present invention takes is：Short-term electric load prediction modeling method, electric power
Stressor obtains its pivot characteristic component after core principle component feature extraction, and prediction model replaces electric power with pivot component
The training and prediction that are originally inputted variable and carry out BP neural network in load sample.
First by input data vector xiIt is mapped to higher-dimension Mercer feature space φ (xi), then in high-dimensional feature space
φ(xi) the linear principal component analysis of middle progress, therefore feature space φ (xi) in linear PCA correspond to the non-thread of the input space
Property PCA.
Short-term electric load prediction modeling method disclosed by the invention, if sample set is X={ x1, x2..., xm, wherein xk
∈ R, l are input vector sum, if φ is a Nonlinear Mapping, corresponding space is denoted as F, and meets following formula：
Then its corresponding covariance matrix is represented by：
Wherein, φ (xj) it is input variable { xjCentralization Nonlinear Mapping, diagonalization by indicate by input vector turn
It is changed to the coordinate of eigenvector v institutes Special composition, for this purpose, finding the eigenvector v ∈ F of characteristic value λ >=0 and non-zero, is met
Following relationship：
λv=Cu (3)
Since the solution of all λ ≠ 0 is located at φ (x1), φ (x2) ... φ (xm) space that is turned into, thus can be pushed away by formula (3)
Export following formula：
N λ a=Ka (4)
In formula, a system of representatives ordered series of numbers vectors a1, a2..., an, meet：
K is the Gram matrixes of a symmetrical n × n simultaneously, and element is
Kij=(φ (xi), φ (xj))=K (xi, xj) (6)
Normalize the finite eigenvalues of Gram matrix KsCorresponding eigenvector vk, makeFinally, eigenvector v is arrived by calculating function phi (x)kOn projection, obtain k-th of variable x
Non-linear principal component：
With V indicate withEigenvectorFor the matrix of column vector,It indicates with the eigenvector of KFor the matrix of column vector, ∧ indicates the diagonal matrix that corresponding characteristic value is formed
Then to training data pointIt is projected, then formula (7) can be rewritten as following expression matrix form：
In formula, φiIndicate the mapping of test sample pointThe n constitutedi× Metzler matrix, KtIndicate ni
× n matrix, element are：
(Kt)ij=(φ (xi), φ (xj))=K (xi, xj) (10)
In formula,WithTraining sample set and test sample collection are indicated respectively.
Short-term electric load prediction modeling method disclosed by the invention, the BP neural network include input layer, imply
Layer, output layer, BP neural network can be arranged as required to one or more hidden layers, be in connect entirely between the neuron of each level
Mode is connect, but does not have any connection between the neuron of same layer, can be indicated as follows：
Y=F2[Wn×m×F1(WM×L×X)] (13)
Error function is defined as the quadratic sum of the difference of desired output and reality output, can be represented by the formula：
Wherein：N is network output layer neuron number；P is training set sample number；yjPFor the real output value of neural network；
djPFor the desired output of neural network.
Using the technical program, short-term electric load prediction is carried out using KPCA-BP neural network models, first to short-term
The impact factor of electric load input sample carries out KPCA processing, and projective transformation is carried out to relative influence factor data, will be a large amount of
Load data projects to the less model space of dimension by higher dimensional space, eliminates the correlation between former input vector, obtains shadow
The pivot characteristic component for ringing the load forecast factor, then as the input variable of BP neural network, and is learnt
Optimal BP neural network Short-term Load Forecasting is obtained, finally carries out load forecast using multigroup prediction number.
Below with reference to drawings and examples, the present invention is described in detail.
Description of the drawings
The content expressed by each width attached drawing of this specification and the label in figure are briefly described below：
Fig. 1 is the schematic diagram of short-term electric load prediction modeling method of the present invention；
Fig. 2 is the system flow chart of short-term electric load prediction modeling method of the present invention.
Specific implementation mode
Below against attached drawing, by the description of the embodiment, each structure for example involved to the specific implementation mode of the present invention
Mutual alignment and connection relation, the effect of each section and operation principle, manufacturing process between the shape of part, construction, each section
And operate with method etc., be described in further detail, with help those skilled in the art to the present invention inventive concept,
Technical solution has more complete, accurate and deep understanding.
Fig. 1 is the schematic diagram of short-term electric load prediction modeling method of the present invention, and short-term electric load as shown in the figure is pre-
Modeling method is surveyed, the electric load factor obtains its pivot characteristic component, prediction model is with master after core principle component feature extraction
First component replaces the training and prediction that are originally inputted variable and carry out BP neural network in electric load sample.
First by input data vector xiIt is mapped to higher-dimension Mercer feature space φ (xi), then in high-dimensional feature space
φ(xi) the linear principal component analysis of middle progress, therefore feature space φ (xi) in linear PCA correspond to the non-thread of the input space
Property PCA.
Fig. 2 is the system flow chart of short-term electric load prediction modeling method of the present invention, and as shown in the figure sets sample set as X
={ x1, x2..., xm, wherein xk∈ R, l are input vector sum, if φ is a Nonlinear Mapping, corresponding space is denoted as
F, and meet following formula：
Then its corresponding covariance matrix is represented by：
Wherein, φ (xj) it is input variable { xjCentralization Nonlinear Mapping, diagonalization by indicate by input vector turn
It is changed to the coordinate of eigenvector v institutes Special composition, for this purpose, finding the eigenvector v ∈ F of characteristic value λ >=0 and non-zero, is met
Following relationship：
λ v=Cv (3)
Since the solution of all λ ≠ 0 is located at φ (x1), φ (x2) ... φ (xm) space that is turned into, thus can be pushed away by formula (3)
Export following formula：
N λ a=Ka (4)
In formula, a system of representatives ordered series of numbers vectors a1, a2..., an, meet：
K is the Gram matrixes of a symmetrical n × n simultaneously, and element is
Kij=(φ (xi), φ (xj))=K (xi, xj) (6)
Normalize the finite eigenvalues of Gram matrix KsCorresponding eigenvector vk, makeFinally, eigenvector v is arrived by calculating function phi (x)kOn projection, obtain k-th of variable x
Non-linear principal component：
With V indicate withEigenvectorFor the matrix of column vector,It indicates with the eigenvector of KFor the matrix of column vector, Λ indicates the diagonal matrix that corresponding characteristic value is formed
Then to training data pointIt is projected, then formula (7) can be rewritten as following expression matrix form：
In formula, φiIndicate the mapping of test sample pointThe n constitutedi× Metzler matrix, KtIndicate ni
× n matrix, element are：
(Kt)ij=(φ (xi), φ (xj))=K (xi, xj) (10)
In formula,WithTraining sample set and test sample collection are indicated respectively.
BP neural network includes input layer, hidden layer, output layer, and BP neural network can be arranged as required to one or more
A hidden layer is in full connection type between the neuron of each level, but does not have any connection between the neuron of same layer, can be indicated
It is as follows：
Y=F2[Wn×m×F1(WM×L×X)] (13)
Error function is defined as the quadratic sum of the difference of desired output and reality output, can be represented by the formula：
Wherein：N is network output layer neuron number；P is training set sample number；yjpFor the real output value of neural network；
djpFor the desired output of neural network.
BP neural network carries out operation using the backpropagation of forward calculation and error, in learning process, if error
More than the minimum error values of setting, then error is subjected to backpropagation, the threshold value and weights of network is adjusted, until its error precision
Less than the error amount of setting, the training learning process of BP neural network terminates at this time, preserves network structure, then passes through network
Reasoning and calculation, and then short-term electric load is predicted.
Using the technical program, short-term electric load prediction is carried out using KPCA-BP neural network models, first to short-term
The impact factor of electric load input sample carries out KPCA processing, and projective transformation is carried out to relative influence factor data, will be a large amount of
Load data projects to the less model space of dimension by higher dimensional space, eliminates the correlation between former input vector, obtains shadow
The pivot characteristic component for ringing the load forecast factor, then as the input variable of BP neural network, and is learnt
Optimal BP neural network Short-term Load Forecasting is obtained, finally carries out load forecast using multigroup prediction number.
The present invention is exemplarily described above in conjunction with attached drawing, it is clear that the present invention implements not by aforesaid way
Limitation, as long as the improvement of the various unsubstantialities of inventive concept and technical scheme of the present invention progress is used, or without changing
Other occasions are directly applied to by the design of the present invention and technical solution, within protection scope of the present invention.
Claims (3)
1. short-term electric load prediction modeling method, it is characterised in that：The electric load factor after core principle component feature extraction,
Its pivot characteristic component is obtained, prediction model replaces the variable that is originally inputted in electric load sample to carry out BP god with pivot component
Training through network and prediction.
First by input data vector xiIt is mapped to higher-dimension Mercer feature space φ (xi), then in high-dimensional feature space φ
(xi) the linear principal component analysis of middle progress, therefore feature space φ (xi) in linear PCA correspond to the non-linear of the input space
PCA。
2. short-term electric load prediction modeling method described in accordance with the claim 1, it is characterised in that：If sample set is X={ x1,
x2..., xm, wherein xk∈ R, l are input vector sum, if φ is a Nonlinear Mapping, corresponding space is denoted as F, and full
Foot formula：
Then its corresponding covariance matrix is represented by：
Wherein, φ (xj) it is input variable { xjCentralization Nonlinear Mapping, diagonalization will indicate to be converted to input vector
The coordinate of eigenvector υ institutes Special composition meets as follows for this purpose, finding the eigenvector υ ∈ F of characteristic value λ >=0 and non-zero
Relationship：
λ v=Cv (3)
Since the solution of all λ ≠ 0 is located at φ (x1), φ (x2) ... φ (xm) space that is turned into, thus can be derived by formula (3)
Following formula：
N λ a=Ka (4)
In formula, a system of representatives ordered series of numbers vectors a1, a2..., an, meet：
K is the Gram matrixes of a symmetrical n × n simultaneously, and element is
Kij=(φ (xi), φ (xj))=K (xi, xj)
Normalize the finite eigenvalues of Gram matrix KsCorresponding eigenvector vk, makeMost
Afterwards, eigenvector v is arrived by calculating function phi (x)kOn projection, obtain k-th of non-linear principal component of variable x：
With V indicate withEigenvectorFor the matrix of column vector,It indicates with the eigenvector of KFor
The matrix of column vector, ∧ indicate the diagonal matrix that corresponding characteristic value is formedThen to training
Data pointIt is projected, then formula (7) can be rewritten as following expression matrix form：
In formula, φiIndicate the mapping of test sample pointThe n constitutedi× Metzler matrix, KiIndicate ni× n squares
Battle array, element are：
(Ki)ij=(φ (xi), φ (xj))=K (xi, xj) (10)
In formula,WithTraining sample set and test sample collection are indicated respectively.
3. short-term electric load prediction modeling method described in accordance with the claim 1, it is characterised in that：The BP neural network
Including input layer, hidden layer, output layer, BP neural network can be arranged as required to one or more hidden layers, each level
It is in full connection type between neuron, but does not have any connection between the neuron of same layer, can indicates as follows：
Y=F2[Wn×m×F1(WM×L×X)] (13)
Error function is defined as the quadratic sum of the difference of desired output and reality output, can be represented by the formula：
Wherein：N is network output layer neuron number；P is training set sample number；yjpFor the real output value of neural network；djpFor
The desired output of neural network.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711445864.4A CN108320046A (en) | 2017-12-27 | 2017-12-27 | Short-term electric load prediction modeling method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711445864.4A CN108320046A (en) | 2017-12-27 | 2017-12-27 | Short-term electric load prediction modeling method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108320046A true CN108320046A (en) | 2018-07-24 |
Family
ID=62893412
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711445864.4A Pending CN108320046A (en) | 2017-12-27 | 2017-12-27 | Short-term electric load prediction modeling method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108320046A (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109473985A (en) * | 2019-01-16 | 2019-03-15 | 江苏圣通电力新能源科技有限公司 | One kind being based on BP neural network smart grid distribution method |
| CN109492748A (en) * | 2018-09-26 | 2019-03-19 | 广东工业大学 | A kind of Mid-long term load forecasting method for establishing model of the electric system based on convolutional neural networks |
| CN109767054A (en) * | 2018-11-22 | 2019-05-17 | 福建网能科技开发有限责任公司 | Efficiency cloud appraisal procedure and edge efficiency gateway based on deep neural network algorithm |
| CN110348630A (en) * | 2019-07-09 | 2019-10-18 | 武汉四创自动控制技术有限责任公司 | A kind of isolated island region Methods of electric load forecasting and system |
| CN117239739A (en) * | 2023-11-13 | 2023-12-15 | 国网冀北电力有限公司 | Method, device and equipment for predicting user side load by knowledge big model |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106600050A (en) * | 2016-12-10 | 2017-04-26 | 国网辽宁省电力有限公司锦州供电公司 | BP neural network-based ultra-short load prediction method |
| CN106651020A (en) * | 2016-12-16 | 2017-05-10 | 燕山大学 | Short-term power load prediction method based on big data reduction |
-
2017
- 2017-12-27 CN CN201711445864.4A patent/CN108320046A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106600050A (en) * | 2016-12-10 | 2017-04-26 | 国网辽宁省电力有限公司锦州供电公司 | BP neural network-based ultra-short load prediction method |
| CN106651020A (en) * | 2016-12-16 | 2017-05-10 | 燕山大学 | Short-term power load prediction method based on big data reduction |
Non-Patent Citations (1)
| Title |
|---|
| 李永斌: "短期电力负荷预测模型的建立与应用", 《计算机仿真》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109492748A (en) * | 2018-09-26 | 2019-03-19 | 广东工业大学 | A kind of Mid-long term load forecasting method for establishing model of the electric system based on convolutional neural networks |
| CN109767054A (en) * | 2018-11-22 | 2019-05-17 | 福建网能科技开发有限责任公司 | Efficiency cloud appraisal procedure and edge efficiency gateway based on deep neural network algorithm |
| CN109473985A (en) * | 2019-01-16 | 2019-03-15 | 江苏圣通电力新能源科技有限公司 | One kind being based on BP neural network smart grid distribution method |
| CN110348630A (en) * | 2019-07-09 | 2019-10-18 | 武汉四创自动控制技术有限责任公司 | A kind of isolated island region Methods of electric load forecasting and system |
| CN117239739A (en) * | 2023-11-13 | 2023-12-15 | 国网冀北电力有限公司 | Method, device and equipment for predicting user side load by knowledge big model |
| CN117239739B (en) * | 2023-11-13 | 2024-02-02 | 国网冀北电力有限公司 | Method, device and equipment for predicting user side load by knowledge big model |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108320046A (en) | Short-term electric load prediction modeling method | |
| Ye et al. | A novel transfer learning framework for time series forecasting | |
| CN106600059B (en) | Intelligent power grid short-term load prediction method based on improved RBF neural network | |
| Han | Deep learning approximation for stochastic control problems | |
| Sun et al. | Solving multiple objective programming problems using feed-forward artificial neural networks: The interactive FFANN procedure | |
| CN107480777A (en) | Sparse self-encoding encoder Fast Training method based on pseudo- reversal learning | |
| CN106503802A (en) | A kind of method of utilization genetic algorithm optimization BP neural network system | |
| CN104636985A (en) | Method for predicting radio disturbance of electric transmission line by using improved BP (back propagation) neural network | |
| Wu et al. | A performance comparison of neural networks in forecasting stock price trend | |
| CN110083125A (en) | A kind of machine tool thermal error modeling method based on deep learning | |
| CN110738363B (en) | Photovoltaic power generation power prediction method | |
| Parasuraman et al. | Cluster-based hydrologic prediction using genetic algorithm-trained neural networks | |
| CN109102021A (en) | The mutual polishing multicore k- mean cluster machine learning method of core under deletion condition | |
| Xu et al. | Multi-cluster based equilibrium optimizer algorithm with compact approach for power system network | |
| CN115393632A (en) | Image classification method based on evolutionary multi-target neural network architecture structure | |
| CN114897451A (en) | Double-layer clustering correction method and device considering key features of demand response user | |
| Zjavka | Recognition of generalized patterns by a differential polynomial neural network | |
| CN113688960A (en) | Grey wolf optimization GHFCM-based residential power data clustering method and device | |
| Motlagh et al. | A new architecture for modeling and prediction of dynamic systems using neural networks: Application in Tehran stock exchange | |
| CN107491841A (en) | Nonlinear optimization method and storage medium | |
| Maciel et al. | Exchange rate forecasting using echo state networks for trading strategies | |
| Cong et al. | Self-paced weight consolidation for continual learning | |
| Bi et al. | Critical direction projection networks for few-shot learning | |
| Jani et al. | 3D affine registration using teaching-learning based optimization | |
| Zhang et al. | A memetic algorithm based extreme learning machine for classification |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20180724 |
|
| RJ01 | Rejection of invention patent application after publication |