CN115456312A - Short-term power load prediction method and system based on octyl geometric modal decomposition - Google Patents
Short-term power load prediction method and system based on octyl geometric modal decomposition Download PDFInfo
- Publication number
- CN115456312A CN115456312A CN202211395000.7A CN202211395000A CN115456312A CN 115456312 A CN115456312 A CN 115456312A CN 202211395000 A CN202211395000 A CN 202211395000A CN 115456312 A CN115456312 A CN 115456312A
- Authority
- CN
- China
- Prior art keywords
- matrix
- power load
- dimension
- value
- formula
- 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
- 238000000034 method Methods 0.000 title claims abstract description 48
- 238000000354 decomposition reaction Methods 0.000 title claims abstract description 27
- 125000002347 octyl group Chemical group [H]C([*])([H])C([H])([H])C([H])([H])C([H])([H])C([H])([H])C([H])([H])C([H])([H])C([H])([H])[H] 0.000 title claims description 12
- 238000012545 processing Methods 0.000 claims abstract description 10
- 239000011159 matrix material Substances 0.000 claims description 146
- 239000013598 vector Substances 0.000 claims description 52
- 238000012549 training Methods 0.000 claims description 28
- 230000009466 transformation Effects 0.000 claims description 19
- 230000008569 process Effects 0.000 claims description 18
- 238000004364 calculation method Methods 0.000 claims description 15
- 241000894006 Bacteria Species 0.000 claims description 14
- 238000012935 Averaging Methods 0.000 claims description 12
- 238000012360 testing method Methods 0.000 claims description 12
- 238000005457 optimization Methods 0.000 claims description 10
- 241001467460 Myxogastria Species 0.000 claims description 9
- 238000006243 chemical reaction Methods 0.000 claims description 9
- 210000002569 neuron Anatomy 0.000 claims description 9
- 210000004027 cell Anatomy 0.000 claims description 6
- 230000003247 decreasing effect Effects 0.000 claims description 6
- 230000035772 mutation Effects 0.000 claims description 6
- 239000002245 particle Substances 0.000 claims description 6
- 238000004458 analytical method Methods 0.000 claims description 5
- 241000863434 Myxococcales Species 0.000 claims description 3
- 230000004913 activation Effects 0.000 claims description 3
- 238000013459 approach Methods 0.000 claims description 3
- 230000001174 ascending effect Effects 0.000 claims description 3
- 238000010276 construction Methods 0.000 claims description 3
- 238000006073 displacement reaction Methods 0.000 claims description 3
- 238000001228 spectrum Methods 0.000 claims description 3
- 229940088594 vitamin Drugs 0.000 claims description 3
- 229930003231 vitamin Natural products 0.000 claims description 3
- 235000013343 vitamin Nutrition 0.000 claims description 3
- 239000011782 vitamin Substances 0.000 claims description 3
- 150000003722 vitamin derivatives Chemical class 0.000 claims description 3
- 238000013277 forecasting method Methods 0.000 claims 1
- 230000006870 function Effects 0.000 description 8
- 230000000737 periodic effect Effects 0.000 description 4
- 238000010248 power generation Methods 0.000 description 4
- 230000008878 coupling Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 230000007774 longterm Effects 0.000 description 3
- 230000008901 benefit Effects 0.000 description 2
- 238000004891 communication Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000012804 iterative process Methods 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 238000007726 management method Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 240000004760 Pimpinella anisum Species 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 238000011217 control strategy Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 239000000446 fuel Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/086—Learning methods using evolutionary algorithms, e.g. genetic algorithms or genetic programming
-
- 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
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J3/00—Circuit arrangements for ac mains or ac distribution networks
- H02J3/003—Load forecast, e.g. methods or systems for forecasting future load demand
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J2203/00—Indexing scheme relating to details of circuit arrangements for AC mains or AC distribution networks
- H02J2203/20—Simulating, e g planning, reliability check, modelling or computer assisted design [CAD]
-
- 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
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Economics (AREA)
- Health & Medical Sciences (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- Life Sciences & Earth Sciences (AREA)
- Marketing (AREA)
- General Health & Medical Sciences (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Operations Research (AREA)
- Data Mining & Analysis (AREA)
- Power Engineering (AREA)
- Entrepreneurship & Innovation (AREA)
- Game Theory and Decision Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Physiology (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Quality & Reliability (AREA)
- Evolutionary Computation (AREA)
- Development Economics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- Primary Health Care (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention relates to the technical field of power load processing, and discloses a short-term power load prediction method and a system based on octagon modal decomposition.
Description
Technical Field
The invention relates to the technical field of power load processing, in particular to a short-term power load prediction method and system based on octyl geometric modal decomposition.
Background
The load prediction is divided into three categories, namely short-term, medium-term and long-term, wherein the short term is several minutes, and the long term can reach several months or even years. The medium and long-term load prediction is commonly used for new station operation, power grid capacity increase and reconstruction, equipment overhaul plan, reservoir optimization scheduling plan, fuel supply plan and the like. The short-term load prediction plays a crucial role in determining the optimal unit combination, reducing the rotating reserve capacity and avoiding safety accidents, and is a key component for guaranteeing the economic operation of the power grid. With the development of the power market, the power load is accurately predicted in a short period, the safe operation of a power grid can be effectively guaranteed, the power generation cost is reduced, the user requirements are met, and the social and economic benefits are improved.
In the past, power load prediction is mainly applied to a scheduling department to make planning arrangement or control strategies of power generation and power supply, research is mainly concentrated on a power generation side, and research on a single user side is rarely related, while with the arrival of a smart grid and the gradual development of distributed renewable energy, user-side power generation plays an increasingly important role in a modern grid structure, how to effectively improve the consumption of renewable energy and the efficiency of a user-side household energy management system becomes increasingly important, analysis on power utilization behaviors of the user side can more effectively promote demand-side management, reasonably inhibit load peaks and improve the utilization rate of grid assets so as to meet the arrival of big data and an intelligent era, on the other hand, a power load prediction method has a certain universality for regional loads and personal loads, compared with regional loads, an individual user load generally has stronger random characteristics, a model with stronger high-frequency random characteristic processing capability can obtain a better effect, and a model capable of predicting the individual user load generally can obtain a more ideal result for regional load prediction with a smoother rule.
Because the load of the power system has a certain periodic characteristic and the factors influencing the load are complex (weather, economy, holidays, observation errors and the like), the load of the power system presents stronger randomness and non-periodic components, and great difficulty is brought to short-term prediction.
At present, short-term load prediction methods can be divided into traditional statistical prediction methods and machine learning prediction methods, statistics comprise generalized autoregressive conditional variance, time series and the like, and load prediction is realized by learning load recursion relations at different moments. With the wide application of machine learning, many researchers apply it to load prediction. Such as a neural Network, an extreme learning machine, a long-time and short-time memory Network, an Echo State Network (ESN), etc., the load sequence has a certain time sequence correlation with the input characteristics of the load prediction model, and if the prediction model can learn the hidden relationship, the prediction accuracy of the short-time load is improved.
However, the ESN parameters are selected only by experience, so that the uncertainty is very high, the prediction accuracy is poor, meanwhile, the basic SMA algorithm is prone to the problems of unstable optimization result, low convergence speed, local optimization and the like when a high-dimensional complex function and a function with an optimal solution not at the origin are optimized, and in addition, the randomness and non-periodic components of the power load on the user side are difficult to process by a single prediction method, so that the modal aliasing problem exists, and the prediction accuracy of the power load is influenced.
Disclosure of Invention
The invention provides a short-term power load prediction method and system based on octyl geometric modal decomposition, and solves the technical problem of poor prediction accuracy of power loads.
In view of this, the first aspect of the present invention provides a short-term power load prediction method based on simmer geometry decomposition, including the following steps:
s1, obtaining historical power loads of a user side, and constructing a historical power load time sequence;
s2, decomposing the historical power load time sequence by adopting a sine geometric mode to obtain a plurality of corresponding load components;
s3, dividing all load components into a training data set and a testing data set;
s4, constructing an echo state network, and optimizing the echo state network through a dimension competition myxomycete algorithm to obtain an initial power load prediction model;
s5, inputting the load components in the training data set to the initial power load prediction model one by one for training to obtain power load prediction submodels corresponding to the load components;
s6, inputting the test data set into the power load forecasting submodel for forecasting to obtain a corresponding load forecasting value;
and S7, performing superposition processing on all the load predicted values to obtain a final user side power load prediction result.
Preferably, step S2 specifically includes:
s201, setting a historical power load time sequence asIn whichFor the data length, the time sequence of the power load is reconstructed by adopting a time sequence delay topological equivalent method to obtain a track matrixComprises the following steps:
in the formula,in order to embed the dimension number of the dimension,in order to delay the time of the delay,;
s202, performing autocorrelation analysis on the track matrix X to obtain a covariance symmetric matrixComprises the following steps:
in the formula, T is a transposed symbol;
s203, a covariance symmetric matrix is matchedConstructing a Hamilton matrixComprises the following steps:
s204, aiming at the Hamilton matrixSquaring to obtain another Hamilton matrixComprises the following steps:
s205, constructing a sine orthogonal matrix through the following formulaFor Hamilton matrixAnd (3) carrying out octyl orthogonal transformation:
in the formula,in the form of a real matrix,for upper triangular matrices, the matrix elements of upper triangular matricesWherein i and j are respectively the number of rows and columns of the matrix、Andis expressed as:
s206, calculating an upper triangular matrix by using QR decompositionCharacteristic value of (D) is notedFrom the upper triangular matrixEigenvalue calculation matrix ofThe characteristic values of (A) are:
for matrixThe characteristic value of the vector is vectorized to obtain a corresponding characteristic vector which is recorded as;
S207, passing matrixFeature vector ofAnd calculating the sum track matrix X to obtain a conversion coefficient matrixComprises the following steps:
using feature vector matricesAnd a conversion coefficient matrixCalculating to obtain a reconstruction matrix:
Representing an initial one-component matrix of the image,,wherein the matrix is reconstructedZIs composed ofMatrix ofAn initial single component matrixIs prepared by the following steps of;
s208, for any initial single-component matrixDefining each element of the matrix asIn whichLet us orderAnd, andwherein, in the process,representing a single component matrixThe minimum of the row and the column of (c),representing a single component matrixOf rows and columns, wherein whenWhen it is takenWhen is coming into contact withTaking outWherein, in the process,the diagonal average transformation matrix calculation element values are represented, and the diagonal average transformation matrix is represented as follows:
in the formula,representing the diagonal average transformation matrix calculation element values,andrespectively representing the values of the elements of the diagonal averaging transformation matrix calculationAnd row and columnThe columns of the image data are,representing the kth element value of the diagonal average transformation matrix;
s209, converting the initial single component matrix by diagonal averagingConversion to a one-dimensional time series of single-component signals, denoted,Representing the data length, i is more than or equal to 1 and less than or equal to d, and sequentially performing matrix comparison on all initial single componentsIs subjected to diagonal averaging to obtainA one-dimensional time series, pairAnd superposing the one-dimensional time sequences to obtain a single-component total signal matrix as follows:
s210, initial single componentForm a group in sequenceVector of vitamin the vector is a vector of a number of,
s211, defining vectorAndin betweenDistance between two adjacent devicesThe element with the largest difference is the corresponding element:
in the formula,representing the original single componentThe j point in the figure begins continuouslyAnThe value of (a) is set to (b),;,;
s212, mixingComparing with a preset threshold value, wherein the preset threshold value is 0.1 to 0.25Wherein, in the process,determining standard deviation for single component signalThe number of the distance is less than the preset threshold value, and the sum of the number and the distance isThe ratio was calculated as:to find out their pairsVector of vitamin of vectors average values are:
s213, dimension ofPlus 1, the sequence of the serial number arrangement forms a new groupThe + 1-dimensional vector is,
s214, defining a vectorAnddistance between themThe element with the largest difference is the one with the largest difference between the two corresponding elements, namely:
s215, mixingComparing with a preset threshold value to determineThe number of the distance is less than the preset threshold value, and the sum of the number and the distance isThe ratio was calculated as:
the formula for calculating the sample entropy value is as follows:
s216, differentiating the entropy values of the samplesInitial single component ofAdding to obtain new componentsNamely:
in the formula,the component obtained by the octyl geometric modal decomposition is represented, and a represents the a-th component sequence.
Preferably, step S4 specifically includes:
s401, establishing an echo state network as follows:
wherein,in order to input the dimension number, the dimension number is input,the number of internal neurons, l is the output dimension, and u (v), x (v) and y (v) are the input vector, the state vector and the output vector of the echo state network respectively;
s402, according to the input vector, the state vector and the output vector of the echo state network, training the echo state network by the following formula to obtain:
wherein f () is an internal neuron activation function Sigmoid, f out () As a function of the output layer(s),Wa connection weight matrix for internal states to internal states,for randomly generated input layers to a reserve pool×The order of the connection weight matrix is,feeding back to reserve tank for randomly generated output layersA connection weight matrix of order x 1,is from reserve pool to output layer++ l) order output weight matrix;
s403, optimizing the echo state network through a dimension competition slime algorithm, which specifically comprises the following steps:
1) Setting the total number U of individual slime organisms and the maximum iteration number of the dimension competition slime organism algorithmProportional parameter z, decreasing parameterRandom numberDetermining parameters of the updating method of the slime locationDimension D of individual slime bacteria, dimension competition probability Pv, and Gaussian variation probability;
2) Randomly generating a group of solutions as initial parameters to fit a dimension competition myxomycete algorithm to optimize an echo state network:
wherein i =1,2.., U;in order to store the scale of the neurons in the pool,in order to be the radius of the spectrum,in order to be sparse in degree,in order to input the dimensions of the cell,in order to displace the input unit, the displacement of the input unit,in order to output the dimensions of the cell,displacing the output unit;
3) Virtually exploring a target space through initial parameters, and in t +1 iterations of the target space, updating the positions of the slime individuals in the following ways:
in the formula,andin order to search the upper and lower boundaries of the range,is composed ofZ is a ratio parameter for determining the ratio of randomly distributed slime bacteria individuals to slime bacteria,the position of the highest food odor concentration currently found, namely the optimal solution position;the current position of the slime mold;andthe positions of two individuals randomly selected from the group are respectively;is the current iteration number;is a coefficient, the value of which isAnd gradually approaches 0 as the number of iterations increases, wherein,,is the maximum iteration number;for a decreasing parameter from 1 to 0,is composed ofA random number in between, and a random number,in order to determine the parameters of the updating method of the slime location,,is shown asThe fitness value of each individual slime mold,representing the optimal fitness value of the slime mold under the current iteration times;
the fitness value calculation formula is as follows:
in the formula,、respectively predicting an actual value of the user side short-term power load and a predicted output value of the user side short-term power load; g is the number of training samples;
in the formula,is composed ofA random number in between, and a random number,indicating the individual position indexes after the fitness values are arranged in ascending order,representThe population of the first half of the middle rank,represents the optimal fitness value obtained by the current iterative process,representing the worst fitness value obtained by the current iteration process;
4) To slime bacteria individualAll dimensions are randomly paired without repeating pairwise pairs, the total is D/2 pairs, and any pair of dimensions is paired, if rand<Pv, performing a dimension crossover operator on the pair of dimensions according to the following formula;
in the formula,is a slime mold individualTo (1) aAnd the firstDimension generation by dimension crossing;
5) Calculating progeny according to the formulaWith parent MyxomycetesAnd updating the individual positions of the slime bacteria, and recording the current global optimal solution;
6) If it is,Then entering a Gaussian mutation operator to carry out optimization on the optimal individualsPerforming Gaussian mutation operation, further performing local search, and updatingThe position of (2):
in the formula,the particles are the optimal particles after Gaussian variation, N (0, 1) is a Gaussian distribution random quantity with the mean value of 0 and the variance of 1;
7) Judging the current iteration numberWhether or not the maximum number of iterations has been reachedIf so, the iteration is finished, the optimal solution is output, otherwise,+1, returning to the step 2) to continue searching until the current iteration timesTo maximum number of iterationsAnd after iteration stops, outputting a current global optimal solution, and updating initial parameters of the echo state network by using the global optimal solution to obtain an initial power load prediction model.
In a second aspect, the present invention further provides a short-term power load prediction system based on simmer geometry decomposition, including:
the load acquisition module is used for acquiring historical power loads of a user side and constructing a historical power load time sequence;
the decomposition module is used for decomposing the historical power load time sequence by adopting a sine-shaped geometric mode to obtain a plurality of corresponding load components;
the dividing module is used for dividing all load components into a training data set and a testing data set;
the network construction module is used for constructing an echo state network, and optimizing the echo state network through a dimension competition myxobacteria algorithm to obtain an initial power load prediction model;
the training module is used for inputting the load components in the training data set to the initial power load prediction model one by one for training to obtain a power load prediction submodel corresponding to each load component;
the prediction module is used for inputting the test data set into the power load prediction submodel for prediction to obtain a corresponding load prediction value;
and the superposition module is used for carrying out superposition processing on all the load predicted values to obtain a final user side power load prediction result.
According to the technical scheme, the invention has the following advantages:
according to the method, the historical power load time sequence is decomposed by adopting a octave geometric mode to obtain a plurality of corresponding load components, the influence of the volatility of the user side power load time sequence on a prediction result is reduced, a prediction model of a dimension competition slime algorithm optimized echo state network is respectively established for each component, the stability and the generalization capability of the prediction model are improved, the prediction values of all the components are superposed to obtain an actual user side power load prediction result, and the prediction accuracy of the power load is improved.
Drawings
Fig. 1 is a flowchart of a short-term power load prediction method based on simmer geometric mode decomposition according to an embodiment of the present invention;
fig. 2 is a schematic structural diagram of a short-term power load prediction system based on simmer geometry mode decomposition according to an embodiment of the present invention.
Detailed Description
In order to make the technical solutions of the present invention better understood, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
For easy understanding, referring to fig. 1, the method for predicting short-term power load based on simmer geometry mode decomposition according to the present invention includes the following steps:
s1, obtaining historical power loads of a user side, and constructing a historical power load time sequence;
s2, decomposing the historical power load time sequence by adopting a sine geometric mode to obtain a plurality of corresponding load components;
s3, dividing all load components into a training data set and a testing data set;
s4, constructing an echo state network, and optimizing the echo state network through a dimension competition myxomycete algorithm to obtain an initial power load prediction model;
s5, inputting the load components in the training data set to the initial power load prediction model one by one for training to obtain a power load prediction submodel corresponding to each load component;
s6, inputting the test data set into a power load forecasting sub-model for forecasting to obtain a corresponding load forecasting value;
and S7, performing superposition processing on all the load predicted values to obtain a final user side power load prediction result.
It can be understood that each load predicted value is a predicted result of the decomposed component, and the predicted values of all the components need to be superimposed to obtain a final load predicted value.
It should be noted that, this embodiment provides a short-term power load prediction method based on simmerical geometry mode decomposition, which decomposes a historical power load time sequence by using a simmerical geometry mode to obtain a plurality of corresponding load components, reduces the influence of the volatility of the user-side power load time sequence on a prediction result, establishes a prediction model for optimizing an echo state network by using a dimension competition slime mold algorithm for each component, improves the stability and generalization capability of the prediction model, and superimposes the prediction values of all the components to obtain an actual user-side power load prediction result, thereby improving the prediction accuracy of the power load.
In a specific embodiment, step S2 specifically includes:
s201, setting a historical power load time sequence asWhereinFor the data length, the time sequence of the power load is reconstructed by adopting a time sequence delay topological equivalent method to obtain a track matrixComprises the following steps:
in the formula,for embedding dimensionThe number of the first and second groups is,in order to delay the time of the process,;
wherein a suitable embedding dimension is selectedAnd a delay timeThe corresponding reconstruction matrix is obtained. Since different embedding dimensions can produce different effects, the idea of determining the embedding dimensions is adopted to calculate the initial time seriesThe frequency corresponding to the maximum peak value of the PSD is the Power Spectral Density (PSD) of. If the normalized frequency is less than a given thresholdThen is provided withIs composed ofWhereinIs the data length. Otherwise the embedding dimension is set to,Is the sampling frequency. Delay timeIs usually taken。
S202, performing autocorrelation analysis on the track matrix X to obtain a covariance symmetric matrixComprises the following steps:
in the formula, T is a transposed symbol;
s203, a covariance symmetric matrix is matchedConstructing a Hamilton matrixComprises the following steps:
s204, aiming at Hamilton matrixSquaring to obtain another Hamilton matrixComprises the following steps:
S205constructing a sinc-orthogonal matrix by the following formulaFor Hamilton matrixAnd (3) carrying out octyl orthogonal transformation:
in the formula,is a real matrix and is characterized in that,for upper triangular matrices, the matrix elements of an upper triangular matrixWherein i and j are respectively the number of rows and columns of the matrix、Andis expressed as:
wherein, the sine orthogonal matrixHHas the property of sine matrix, so that Hamilton matrixThe structure of (2) is not destroyed in the matrix transformation process.
Meanwhile, based on the mathematical theorem: suppose thatIs a matrix of tracks, anIs a symmetric matrix. By matrixConstructing a Hamiltonian matrixI.e. by。
Having a Householder matrixHThen it can pass throughTransformation structure upper triangular Hessenberg matrixNamely:
s206, calculating an upper triangular matrix by QR decompositionCharacteristic value of (D) is recorded asFrom the upper triangular matrixEigenvalue calculation matrix ofThe characteristic values of (A) are:
for matrixThe characteristic value of the vector is vectorized to obtain a corresponding characteristic vector which is recorded as;
S207, passing matrixFeature vector ofAnd calculating the sum track matrix X to obtain a conversion coefficient matrixComprises the following steps:
using eigenvector matricesAnd a conversion coefficient matrixCalculating to obtain a reconstruction matrix:
The initial single-component matrix is represented,,wherein the matrix is reconstructedZIs composed ofMatrix ofAn initial single component matrixIs prepared from (A) and (B);
s208, for any initial single-component matrixDefining each element of the matrix asWhereinLet us orderAnd anWherein, in the process,representing a single component matrixThe minimum of the row and the column of (c),representing a single component matrixIs maximum of row and column, whereinWhen it is takenWhen is coming into contact withGet itWhereinrepresenting calculated values of elements of a diagonal averaging transformation matrix, diagonal averagingThe equalization transformation matrix is represented as follows:
in the formula,representing the diagonal average transformation matrix calculation element values,andrespectively representing the values of the elements of the diagonal averaging transformation matrix calculationAnd row and columnThe columns of the image data are,representing the kth element value of a diagonal averaging transformation matrix;
s209, converting the initial single component matrix by diagonal averagingConversion to a one-dimensional time series of single-component signals, denoted,Representing the data length, i is more than or equal to 1 and less than or equal to d, and sequentially performing matrix comparison on all initial single componentsIs obtained by carrying out diagonal averagingA one-dimensional time series, pairSuperposing the one-dimensional time sequences to obtain a single-component total signal matrix as follows:
it should be noted that, unlike the periodic similarity comparison and reconstruction method of the conventional octyl geometric modal decomposition, the present embodiment introduces the sample entropy into the reconstruction process of the single-component total signal, and aims to adaptively reconstruct the single-component total signal and improve the accuracy of the analysis result.
S211, defining vectorAnddistance between themThe element with the largest difference is the one with the largest difference between the two corresponding elements, namely:
in the formula,representing a single component from the originThe j point of the medium is continuousAnThe value of (a) is,;,;
s212, mixingComparing with a preset threshold value, wherein the preset threshold value is 0.1 to 0.25Whereindetermining standard deviation for single component signalThe number of the distance is less than the preset threshold value, and the sum of the number and the distance isThe ratio was calculated as:to find out their pairsOf vector of dimension average values are:
s213. DimensionPlus 1, the sequence of the serial number arrangement forms a new groupThe + 1-dimensional vector is,
s214, defining vectorAnddistance between themThe element with the largest difference is the one with the largest difference between the two corresponding elements, namely:
s215, mixingComparing with a preset threshold value to determineThe number of the distance is less than the preset threshold value, and the sum of the number and the distance isThe ratio was calculated as:
the formula for calculating the sample entropy value is as follows:
s216, differentiating the entropy value of the sampleInitial single component ofAdding to obtain new componentsNamely:
in the formula,the component obtained by the octyl geometric modal decomposition is represented, and a represents the a-th component sequence.
In a specific embodiment, step S4 specifically includes:
s401, establishing an echo state network as follows:
wherein,in order to input the dimension number, the dimension number is input,the number of internal neurons, l is the output dimension, and u (v), x (v) and y (v) are the input vector, the state vector and the output vector of the echo state network respectively;
s402, according to the input vector, the state vector and the output vector of the echo state network, training the echo state network by the following formula to obtain:
wherein f () is the internal neuron activation function Sigmoid, f out () As a function of the output layer(s),Wa connection weight matrix for internal states to internal states,for randomly generated input layers to a reserve pool×The order of the connection weight matrix is,for feeding back randomly generated output layers to reserve tanksA connection weight matrix of order x 1,is from reserve pool to output layer++ l) order output weight matrix;
in which the echo-state network、WAndall are randomly generated and are not changed in the learning process once generated, and only need to be adjusted in the training process of the reserve pool networkThe value of (2) is sufficient.
S403, optimizing the echo state network through a dimension competition slime algorithm, which specifically comprises the following steps:
1) Setting the total number U of individual slime organisms and the maximum iteration number of the dimension competition slime organism algorithmProportional parameter z, decreasing parameterRandom numberDetermining parameters of the updating method of the slime locationDimension D of individual slime bacteria, dimension competition probability Pv, and Gaussian variation probability;
2) Randomly generating a group of solutions as initial parameters to fit a dimension competition myxomycete algorithm to optimize an echo state network:
wherein i =1,2.., U;in order to store the scale of the neurons in the pool,is the radius of the spectrum,in order to achieve the degree of sparseness,in order to input the dimensions of the cell,in order to displace the input unit, the displacement of the input unit,in order to output the dimensions of the cell,displacing the output unit;
3) Virtually exploring a target space through an initial parameter, and in t +1 iterations of the target space, updating the positions of the slime individuals in the following modes:
in the formula,andin order to search for the upper and lower boundaries of the range,is composed ofZ is a ratio parameter for determining the ratio of randomly distributed slime bacteria individuals to slime bacteria,the position of the highest food odor concentration currently found, namely the optimal solution position;the current position of the slime mold;andthe positions of two individuals randomly selected from the group are respectively;is the current iteration number;is a coefficient having a value ofAnd gradually approaches 0 with the increase of the iteration number, wherein,,is the maximum iteration number;for a decreasing parameter from 1 to 0,is composed ofA random number in between, and a random number,to determine the parameters of the updating method of the slime location,,is shown asThe fitness value of each individual slime mold,expressing the optimal fitness value of the slime under the current iteration times;
the fitness value calculation formula is as follows:
in the formula,、respectively predicting an actual value of the user side short-term power load and a predicted output value of the user side short-term power load; g is the number of training samples;
in the formula,is composed ofA random number in between, and a random number,representing fitness values in ascending orderThe index of the individual position after the column,to representThe middle half of the top ranked population,represents the optimal fitness value obtained by the current iterative process,representing the worst fitness value obtained by the current iteration process;
4) For slime mold individualAll dimensions are randomly paired without repeating pairwise pairs, and the pairs are D/2 pairs, and any pair of dimensions is paired, if rand<Pv, performing a dimension cross operator on the pair of dimensions according to the following formula;
it should be noted that, because the basic SMA algorithm is easy to have the problems of unstable optimization result, slow convergence speed, local optimization and the like when optimizing a high-dimensional complex function and a function of which the optimal solution is not at the origin, and the local optimization is often caused by the fact that one or more dimensions of the solution fall into the local optimization, a dimension competition operator is introduced into the slime mold algorithm.
The calculation amount of the dimension intersection operator can be controlled by setting different dimension intersection probability controls, so that the calculation speed of the algorithm is ensured.
5) Calculating progeny according to the formulaWith parent Myxomycetes individualsAnd updating the individual positions of the slime bacteria, and recording the current global optimal solution;
6) If it is,Then entering a Gaussian mutation operator to carry out optimization on the optimal individualsPerforming Gaussian mutation operation, further performing local search, and updatingThe position of (2):
in the formula,the particles are the optimal particles after Gaussian variation, N (0, 1) is a Gaussian distribution random quantity with the mean value of 0 and the variance of 1;
7) Judging the current iteration numberWhether or not the maximum number of iterations has been reachedIf so, the iteration is finished, the optimal solution is output, otherwise,+1, return to step 2) and continue searching until the current iteration numberTo maximum number of iterationsAfter iteration is stopped, outputting the current global optimal solution, and updating the echo by using the global optimal solutionAnd obtaining an initial power load prediction model according to the initial parameters of the state network.
The above is a detailed description of an embodiment of a short-term power load prediction method based on simmered geometry mode decomposition provided by the present invention, and the following is a detailed description of an embodiment of a short-term power load prediction system based on simmered geometry mode decomposition provided by the present invention.
For convenience of understanding, referring to fig. 2, the present invention provides a short-term power load prediction system based on simmer geometry decomposition, including:
the load acquisition module 100 is used for acquiring historical power loads of a user side and constructing a historical power load time sequence;
the decomposition module 200 is configured to decompose the historical power load time series by using a sine-shaped geometric mode to obtain a plurality of corresponding load components;
a dividing module 300 for dividing all load components into a training data set and a test data set;
the network construction module 400 is used for constructing an echo state network, and optimizing the echo state network through a dimension competition myxobacteria algorithm to obtain an initial power load prediction model;
the training module 500 is configured to input the load components in the training data set to the initial power load prediction model one by one for training, so as to obtain a power load prediction submodel corresponding to each load component;
the prediction module 600 is configured to input the test data set into the power load prediction submodel to perform prediction, so as to obtain a corresponding load prediction value;
and the superposition module 700 is configured to perform superposition processing on all the load predicted values to obtain a final user-side power load prediction result.
It can be clearly understood by those skilled in the art that, for convenience and simplicity of description, the specific working processes of the above-described systems, apparatuses and units may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again.
In the embodiments provided in the present invention, it should be understood that the disclosed apparatus and method may be implemented in other ways. For example, the above-described apparatus embodiments are merely illustrative, and for example, the division of the units is only one type of logical functional division, and other divisions may be realized in practice, for example, multiple units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, devices or units, and may be in an electrical, mechanical or other form.
The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.
In addition, functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.
The above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (4)
1. The short-term power load prediction method based on the octyl geometric modal decomposition is characterized by comprising the following steps of:
s1, obtaining historical power loads of a user side, and constructing a historical power load time sequence;
s2, decomposing the historical power load time sequence by adopting a sine geometric mode to obtain a plurality of corresponding load components;
s3, dividing all load components into a training data set and a testing data set;
s4, constructing an echo state network, and optimizing the echo state network through a dimension competition myxomycete algorithm to obtain an initial power load prediction model;
s5, inputting the load components in the training data set to the initial power load prediction model one by one for training to obtain power load prediction submodels corresponding to the load components;
s6, inputting the test data set into the power load forecasting submodel for forecasting to obtain a corresponding load forecasting value;
and S7, performing superposition processing on all the load predicted values to obtain a final user side power load prediction result.
2. The short-term power load forecasting method based on simpsoidal modal decomposition according to claim 1, wherein step S2 specifically includes:
s201, setting a historical power load time sequence asWhereinFor the data length, the time sequence of the power load is reconstructed by adopting a time sequence delay topological equivalent method to obtain a track matrixComprises the following steps:
in the formula,in order to embed the dimension number, the number of the embedded dimension,in order to delay the time of the delay,;
s202, performing autocorrelation analysis on the track matrix X to obtain a covariance symmetric matrixComprises the following steps:
in the formula, T is a transposed symbol;
s203, a covariance symmetric matrix is matchedConstructing a Hamilton matrixComprises the following steps:
s204, aiming at the Hamilton matrixSquaring to obtain another Hamilton matrixComprises the following steps:
s205, constructing a sine orthogonal matrix by the following formulaFor Hamilton matrixAnd (3) carrying out octyl orthogonal transformation:
in the formula,in the form of a real matrix,for upper triangular matrices, the matrix elements of an upper triangular matrixWherein i and j are respectively the number of rows and columns of the matrix、Andis expressed as:
s206, calculating an upper triangular matrix by QR decompositionCharacteristic value of (D) is notedFrom the upper triangular matrixEigenvalue calculation matrix ofThe characteristic values of (A) are:
for matrixThe characteristic value of the vector is vectorized to obtain a corresponding characteristic vector which is recorded as;
S207, passing matrixFeature vector ofAnd calculating the sum track matrix X to obtain a conversion coefficient matrixComprises the following steps:
using eigenvector matricesAnd a matrix of conversion coefficientsCalculating to obtain a reconstruction matrix:
The initial single-component matrix is represented,,wherein the matrix is reconstructedZIs composed ofMatrix ofAn initial single component matrixIs prepared from (A) and (B);
s208, for any initial single-component matrixDefining each element of the matrix asWhereinLet us orderAnd anWhereinrepresenting a single component matrixThe minimum of the row and the column of (c),representing a single component matrixOf rows and columns, wherein whenWhen it is takenWhen is coming into contact withTaking outWhereinrepresents the diagonal average transformation matrix calculation element values, the diagonal average transformation matrix is represented as follows:
in the formula,representing the diagonal average transformation matrix calculation element values,andrespectively representing the values of the calculated elements of the diagonal averaging transformation matrixAnd row and columnThe columns of the image data are,representing the kth element value of a diagonal averaging transformation matrix;
s209, converting the initial single component matrix by diagonal averagingConversion to a one-dimensional time series of single-component signals, denoted as,Representing the data length, i is more than or equal to 1 and less than or equal to d, and sequentially performing matrix comparison on all initial single componentsIs obtained by carrying out diagonal averagingA one-dimensional time series, pairSuperposing the one-dimensional time sequences to obtain a single-component total signal matrix as follows:
s210, initial single componentForm a group in sequenceVector of vitamin the vector is a function of the number of bits,
representing a single component from the originTo middleThe points beginning to be continuousAnA value of (d);
s211, defining vectorAnddistance between themThe element with the largest difference is the one with the largest difference between the two corresponding elements, namely:
in the formula,representing a single component from the originThe j point in the figure begins continuouslyAnThe value of (a) is,;,;
s212, mixingComparing with a preset threshold value, wherein the preset threshold value is 0.1 to 0.25Whereindetermining standard deviation for single component signalThe number of the distance is less than the preset threshold value, and the sum of the number and the distance isThe ratio was calculated as:to find out their pairsOf vector of dimension average values are:
s213. DimensionPlus 1, the sequence of the serial number arrangement forms a new groupThe + 1-dimensional vector is,
s214, defining a vectorAnddistance between themThe element with the largest difference is the one with the largest difference between the two corresponding elements, namely:
in the formula,indicating continuation from the j-th point+1 piecesThe value of (a) is set to (b),;,;
s215, mixingComparing with a preset threshold value to determineThe number of the distance is less than the preset threshold value, and the sum of the number and the distance isThe ratio was calculated as:
the formula for calculating the sample entropy value is as follows:
s216, differentiating the entropy value of the sampleInitial single component ofAdding to obtain new componentsNamely:
3. The short-term power load prediction method based on simutaneous geometric modal decomposition according to claim 1, wherein the step S4 specifically comprises:
s401, establishing an echo state network as follows:
wherein,in order to input the dimension number, the dimension number is input,the number of internal neurons, l is the output dimension, and u (v), x (v) and y (v) are the input vector, the state vector and the output vector of the echo state network respectively;
s402, according to the input vector, the state vector and the output vector of the echo state network, training the echo state network by the following formula to obtain:
wherein f () is the internal neuron activation function Sigmoid, f out () As a function of the output layer(s),Wa connection weight matrix for internal states to internal states,for randomly generated input layers to a reserve pool×The order of the connection weight matrix is such that,for feeding back randomly generated output layers to reserve tanksA connection weight matrix of order x 1,is from reserve pool to output layer × (++ l) order output weight matrix;
s403, optimizing the echo state network through a dimension competition myxomycete algorithm, which specifically comprises the following steps:
1) Setting the total number U of individual slime organisms and the maximum iteration number of the dimension competition slime organism algorithmProportional parameter z, decreasing parameterRandom numberDetermining parameters of the updating method of the slime locationDimension D of individual slime bacteria, dimension competition probability Pv, and Gaussian variation probability;
2) Randomly generating a group of solutions as initial parameters to fit a dimension competition myxomycete algorithm to optimize an echo state network:
wherein i =1,2.., U;in order to store the scale of the neurons in the pool,in order to be the radius of the spectrum,in order to be sparse in degree,in order to input the dimensions of the cell,in order to input the displacement of the unit,in order to output the dimensions of the cell,displacing the output unit;
3) Virtually exploring a target space through an initial parameter, and in t +1 iterations of the target space, updating the positions of the slime individuals in the following modes:
in the formula,andin order to search the upper and lower boundaries of the range,is composed ofZ is a ratio parameter for determining the ratio of randomly distributed slime bacteria individuals to slime bacteria,the position of the highest food odor concentration currently found, namely the optimal solution position;the current position of the slime mold;andthe positions of two individuals randomly selected from the group are respectively;the current iteration number is;is a coefficient having a value ofAnd gradually approaches 0 with the increase of the iteration number, wherein,,is the maximum iteration number;for a decreasing parameter from 1 to 0,is composed ofA random number in between, and a random number,to determine the parameters of the updating method of the slime location,,representFirst, theThe fitness value of each individual slime mold,representing the optimal fitness value of the slime mold under the current iteration times;
the fitness value calculation formula is as follows:
in the formula,、respectively predicting an actual value of the user side short-term power load and a predicted output value of the user side short-term power load; g is the number of training samples;
in the formula,is composed ofA random number in between, and a random number,indicating the individual position indexes after the fitness values are arranged in ascending order,to representThe population of the first half of the middle rank,represents the optimal fitness value obtained by the current iteration process,representing the worst fitness value obtained by the current iteration process;
4) To slime bacteria individualAll dimensions are randomly paired without repeating pairwise pairs, the total is D/2 pairs, and any pair of dimensions is paired, if rand<Pv, performing a dimension cross operator on the pair of dimensions according to the following formula;
in the formula,is a slime mold individualTo (1) aAnd the firstDimension generation by dimension crossing;
5) Calculating progeny according to the formulaWith parent Myxomycetes individualsAnd updating the individual positions of the slime bacteria, and recording the current global optimal solution;
6) If it is,Then entering a Gaussian mutation operator to carry out optimization on the optimal individualsPerforming Gaussian mutation operation, further performing local search, and updatingThe position of (c):
in the formula,the particles are the optimal particles after Gaussian variation, N (0, 1) is a Gaussian distribution random quantity with the mean value of 0 and the variance of 1;
7) Judging the current iteration numberWhether or not the maximum number of iterations has been reachedIf so, the iteration is finished, the optimal solution is output, otherwise,+1, return to step 2) and continue searching until the current iteration numberTo a maximum number of iterationsAfter iteration is stopped, outputting the current global optimal solution, and updating the initial parameters of the echo state network by using the global optimal solution to obtain the initial parametersA power load prediction model.
4. Short-term power load prediction system based on octyl geometric modal decomposition is characterized by comprising the following components:
the load acquisition module is used for acquiring historical power loads of a user side and constructing a historical power load time sequence;
the decomposition module is used for decomposing the historical power load time sequence by adopting a sine-shaped geometric mode to obtain a plurality of corresponding load components;
the dividing module is used for dividing all load components into a training data set and a testing data set;
the network construction module is used for constructing an echo state network, and optimizing the echo state network through a dimension competition myxobacteria algorithm to obtain an initial power load prediction model;
the training module is used for inputting the load components in the training data set to the initial power load prediction model one by one for training to obtain a power load prediction submodel corresponding to each load component;
the prediction module is used for inputting the test data set into the power load prediction submodel for prediction to obtain a corresponding load prediction value;
and the superposition module is used for carrying out superposition processing on all the load predicted values to obtain a final user side power load prediction result.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211395000.7A CN115456312A (en) | 2022-11-09 | 2022-11-09 | Short-term power load prediction method and system based on octyl geometric modal decomposition |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202211395000.7A CN115456312A (en) | 2022-11-09 | 2022-11-09 | Short-term power load prediction method and system based on octyl geometric modal decomposition |
Publications (1)
Publication Number | Publication Date |
---|---|
CN115456312A true CN115456312A (en) | 2022-12-09 |
Family
ID=84311095
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202211395000.7A Pending CN115456312A (en) | 2022-11-09 | 2022-11-09 | Short-term power load prediction method and system based on octyl geometric modal decomposition |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN115456312A (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115796231A (en) * | 2023-01-28 | 2023-03-14 | 湖南赛能环测科技有限公司 | Ultrashort-term wind speed prediction method based on temporal analysis |
CN115860277A (en) * | 2023-02-27 | 2023-03-28 | 西安骏硕通信技术有限公司 | Data center energy consumption prediction method and system |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110705640A (en) * | 2019-09-30 | 2020-01-17 | 温州大学 | Method for constructing prediction model based on slime mold algorithm |
US20200169085A1 (en) * | 2017-06-28 | 2020-05-28 | Silvio Becher | Method for recognizing contingencies in a power supply network |
CN113193556A (en) * | 2021-05-12 | 2021-07-30 | 燕山大学 | Short-term wind power prediction method based on probability prediction model |
-
2022
- 2022-11-09 CN CN202211395000.7A patent/CN115456312A/en active Pending
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20200169085A1 (en) * | 2017-06-28 | 2020-05-28 | Silvio Becher | Method for recognizing contingencies in a power supply network |
CN110705640A (en) * | 2019-09-30 | 2020-01-17 | 温州大学 | Method for constructing prediction model based on slime mold algorithm |
CN113193556A (en) * | 2021-05-12 | 2021-07-30 | 燕山大学 | Short-term wind power prediction method based on probability prediction model |
Non-Patent Citations (1)
Title |
---|
程正阳等: "辛几何模态分解方法及其分解能力研究", 《振动与冲击》 * |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115796231A (en) * | 2023-01-28 | 2023-03-14 | 湖南赛能环测科技有限公司 | Ultrashort-term wind speed prediction method based on temporal analysis |
CN115796231B (en) * | 2023-01-28 | 2023-12-08 | 湖南赛能环测科技有限公司 | Temporal analysis ultra-short term wind speed prediction method |
CN115860277A (en) * | 2023-02-27 | 2023-03-28 | 西安骏硕通信技术有限公司 | Data center energy consumption prediction method and system |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN112508275B (en) | Power distribution network line load prediction method and equipment based on clustering and trend indexes | |
Tian et al. | Multi-step short-term wind speed prediction based on integrated multi-model fusion | |
CN115456312A (en) | Short-term power load prediction method and system based on octyl geometric modal decomposition | |
CN105354646B (en) | Power load forecasting method for hybrid particle swarm optimization and extreme learning machine | |
CN110598854A (en) | GRU model-based transformer area line loss rate prediction method | |
CN112733417B (en) | Abnormal load data detection and correction method and system based on model optimization | |
CN113141008B (en) | Data-driven power distribution network distributed new energy consumption capability assessment method | |
CN106022614A (en) | Data mining method of neural network based on nearest neighbor clustering | |
CN113065704B (en) | Super-parameter optimization and post-processing method of non-invasive load decomposition model | |
CN115115125B (en) | Photovoltaic power interval probability prediction method based on deep learning fusion model | |
CN113780679B (en) | Load prediction method and device based on ubiquitous power Internet of things | |
CN109492748A (en) | A kind of Mid-long term load forecasting method for establishing model of the electric system based on convolutional neural networks | |
CN114219216B (en) | Operation scheduling method, system, device and medium of power system | |
CN115964932A (en) | Gas prediction method based on EMD-BilSTM-Attention mechanism transformer digital twin model | |
CN116169670A (en) | Short-term non-resident load prediction method and system based on improved neural network | |
CN115409369A (en) | Comprehensive energy system reliability evaluation method based on mechanism and data hybrid driving | |
CN115713144A (en) | Short-term wind speed multi-step prediction method based on combined CGRU model | |
Wang et al. | Causal carbon price interval prediction using lower upper bound estimation combined with asymmetric multi-objective evolutionary algorithm and long short-term memory | |
Ni et al. | An improved attention-based bidirectional LSTM model for cyanobacterial bloom prediction | |
CN109408896B (en) | Multi-element intelligent real-time monitoring method for anaerobic sewage treatment gas production | |
CN113762591B (en) | Short-term electric quantity prediction method and system based on GRU and multi-core SVM countermeasure learning | |
CN114757107A (en) | Gravity energy storage power distribution method based on load prediction model | |
Zhang et al. | Optimal echo state network parameters based on behavioural spaces | |
Cao et al. | Short-term forecasting and uncertainty analysis of photovoltaic power based on the FCM-WOA-BILSTM model | |
Wang et al. | Sustainable investment forecasting of power grids based on the deep restricted Boltzmann machine optimized by the lion algorithm |
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: 20221209 |