CN112819238A - Short-term wind power prediction method based on chaotic chicken flock optimization algorithm - Google Patents
Short-term wind power prediction method based on chaotic chicken flock optimization algorithm Download PDFInfo
- Publication number
- CN112819238A CN112819238A CN202110189031.6A CN202110189031A CN112819238A CN 112819238 A CN112819238 A CN 112819238A CN 202110189031 A CN202110189031 A CN 202110189031A CN 112819238 A CN112819238 A CN 112819238A
- Authority
- CN
- China
- Prior art keywords
- prediction
- data
- wind power
- elm
- chicken flock
- 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.)
- Withdrawn
Links
- 238000004422 calculation algorithm Methods 0.000 title claims abstract description 100
- 241000287828 Gallus gallus Species 0.000 title claims abstract description 67
- 238000000034 method Methods 0.000 title claims abstract description 64
- 244000144992 flock Species 0.000 title claims abstract description 63
- 238000005457 optimization Methods 0.000 title claims abstract description 51
- 230000000739 chaotic effect Effects 0.000 title claims abstract description 36
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 13
- 238000013528 artificial neural network Methods 0.000 claims abstract description 11
- 238000012549 training Methods 0.000 claims description 37
- 230000006870 function Effects 0.000 claims description 27
- 230000008569 process Effects 0.000 claims description 14
- 210000002569 neuron Anatomy 0.000 claims description 11
- 238000011156 evaluation Methods 0.000 claims description 10
- 238000007781 pre-processing Methods 0.000 claims description 8
- 238000010835 comparative analysis Methods 0.000 claims description 7
- 238000010606 normalization Methods 0.000 claims description 7
- 238000012360 testing method Methods 0.000 claims description 7
- 108090000672 Annexin A5 Proteins 0.000 claims description 6
- 238000004458 analytical method Methods 0.000 claims description 6
- 238000007621 cluster analysis Methods 0.000 claims description 5
- 230000004913 activation Effects 0.000 claims description 4
- 238000002156 mixing Methods 0.000 claims description 4
- 238000012544 monitoring process Methods 0.000 claims description 4
- 101001095088 Homo sapiens Melanoma antigen preferentially expressed in tumors Proteins 0.000 claims 2
- 102100037020 Melanoma antigen preferentially expressed in tumors Human genes 0.000 claims 2
- 230000002159 abnormal effect Effects 0.000 description 10
- 238000004364 calculation method Methods 0.000 description 8
- 238000010586 diagram Methods 0.000 description 8
- 230000008859 change Effects 0.000 description 5
- 230000000694 effects Effects 0.000 description 5
- 238000005070 sampling Methods 0.000 description 3
- 239000011159 matrix material Substances 0.000 description 2
- 239000002245 particle Substances 0.000 description 2
- 238000000053 physical method Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000007619 statistical method Methods 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 1
- 230000002411 adverse Effects 0.000 description 1
- 238000013398 bayesian method Methods 0.000 description 1
- 238000005291 chaos (dynamical) Methods 0.000 description 1
- 150000001875 compounds Chemical class 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000001186 cumulative effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000002028 premature Effects 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 238000012706 support-vector machine 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
-
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N7/00—Computing arrangements based on specific mathematical models
- G06N7/08—Computing arrangements based on specific mathematical models using chaos models or non-linear system models
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Business, Economics & Management (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Economics (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Human Resources & Organizations (AREA)
- Computing Systems (AREA)
- Strategic Management (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Biomedical Technology (AREA)
- General Business, Economics & Management (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computational Linguistics (AREA)
- Tourism & Hospitality (AREA)
- Bioinformatics & Computational Biology (AREA)
- Biophysics (AREA)
- Marketing (AREA)
- Water Supply & Treatment (AREA)
- Pure & Applied Mathematics (AREA)
- Public Health (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Nonlinear Science (AREA)
- Algebra (AREA)
- Computational Mathematics (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Primary Health Care (AREA)
- Development Economics (AREA)
- Game Theory and Decision Science (AREA)
Abstract
The invention discloses a short-term wind power prediction method based on a chaotic chicken flock optimization algorithm, belongs to the technical field of wind power prediction, combines a chicken flock algorithm and a chaotic theory, provides the chaotic chicken flock optimization algorithm, improves the population diversity and the local search capability of the algorithm, and improves the optimization performance of the algorithm. Aiming at the problem of randomness of the wind speed sequence, the wind speed sequence is decomposed into a plurality of subsequences with more regularity by adopting a clustering empirical mode decomposition method, and prediction is respectively carried out. Aiming at the problem that point prediction cannot provide more quantitative information, probabilistic interval prediction and an interval prediction model based on lower upper and lower limit estimation (LUBE) of a neural network are adopted, a dual-output neural network structure is adopted to directly output a prediction interval, the model is simple and efficient, the prediction precision is effectively improved, and the method has important practical application value.
Description
Technical Field
The invention belongs to the technical field of wind power prediction, and particularly relates to a short-term wind power prediction method based on a chaotic chicken flock optimization algorithm.
Background
Wind energy is used as a clean renewable energy source, has intermittency and randomness, and causes non-negligible influence on a power grid, and accurate wind power prediction is a necessary means for solving the problem. The current research situation of wind power at home and abroad is mainly divided into a physical method based on NWP data and a statistical method based on historical data. The physical method needs to analyze and model the internal structure and the surrounding environment of the wind turbine in detail, and the model is complex in structure. The statistical method needs a large amount of historical data, and the nonlinear relation between the wind power and the influence factors of the wind power is analyzed by using methods such as an artificial neural network and a support vector machine. However, the existing research on the optimization problem of the intelligent algorithm model does not well solve many problems existing in the model optimization algorithm, such as the problems of insufficient population diversity, easy falling into local optimal solution caused by weak local search capability, and the like. In addition, the prediction methods mainly focus on the aspect of deterministic point prediction, and because the wind power time sequence has the characteristics of non-stationarity and randomness, point prediction errors always exist and cannot be eliminated, and from the decision point, the use of the point prediction has certain influence on the stable and reliable operation of the power system.
Compared with a common point prediction method, the probabilistic interval prediction can provide more quantitative information for wind power uncertainty. The reliability and the definition of the prediction interval can be evaluated respectively through the confidence level and the average bandwidth, and the decision of the power system is facilitated. The method has the disadvantages that the traditional interval construction method is usually carried out after a deterministic prediction model with a specific priori assumption, the interval is generally calculated by using quantile regression, Bootstrap and Bayes methods, and the methods have large calculation amount and complex model.
Disclosure of Invention
The purpose of the invention is as follows: the invention provides a short-term wind power prediction method based on a chaotic chicken flock optimization algorithm, which aims to overcome the defects of large calculation amount and inaccurate prediction caused by complex models in the prior art.
The technical scheme is as follows: in order to achieve the purpose, the invention is realized by the following technical scheme:
the short-term wind power prediction method based on the chaotic chicken flock optimization algorithm comprises the following steps:
1) taking the acquired fan data as sample data;
2) performing clustering empirical mode decomposition on sample data;
3) establishing an ELM prediction model according to data obtained by clustering analysis, and optimizing the weight of an output layer by adopting a CCSO algorithm, which is called a CCSO-ELM prediction model;
4) optimizing the ELM training process by adopting PSO, CSO, CPSO and CCSO algorithms to verify the effectiveness of the prediction model;
5) detailed comparative analysis of prediction error using Mean Absolute Percent Error (MAPE) and Root Mean Square Error (RMSE);
6) and analyzing the uncertainty of the wind power, taking the proposed CCSO-ELM as an interval prediction model of the wind power, and proposing a new evaluation index considering relative deviation as a fitness function to reduce the deviation degree of points outside the interval.
Further, the step 1) is specifically as follows: the fan data acquisition method is based on the following steps: the historical data of the wind power plant comprises NWP data and data provided by a data acquisition and monitoring control system, and wind speed, wind direction and temperature are selected as input variables of the model;
the method for carrying out cluster analysis on the sample data comprises the following steps:
for signal x (t), the decomposition steps for EMD are:
1.1) finding and recording all maximum value points in the signal, and fitting by adopting cubic spline interpolation to obtain an upper envelope line e of the signalup(t)；
1.2) finding and recording all minimum value points in the signal, and fitting by adopting cubic spline interpolation to obtainLower envelope e of the signallow(t) taking the average of the two envelopes as
1.3)h1(t)＝x(t)-m1(t) mixing h1(t) as a new signal, repeating step 1.1) and step 1.2), and calculating for k times until a new signal h1(t) satisfying the IMF condition, let c1(t)＝h1(t), then c1(t) is the first IMF component;
1.4) subtracting c from x (t)1(t) obtaining a new signal r with the high frequency removed1(t) then r1(t) performing the above steps to finally obtain N IMF components and a non-resolvable margin;
the original signal x (t) is decomposed into
3. The short-term wind power prediction method based on the chaotic chicken flock optimization algorithm according to claim 1, wherein the step 3) is specifically as follows: the method for establishing an ELM prediction model according to data obtained by clustering analysis and optimizing the weight of an output layer by adopting a CCSO algorithm comprises the following steps:
3.1) selection of input and output variables
Constructing a training sample set; determining an input vector X ═ X1,…x4,x5,x6,x7]Wherein x is1-x4Wind power data, x, for the first 4 time points5、x6And x7Wind speed, wind direction sine and temperature of the predicted point are respectively; the output data is the wind power value of the prediction point; selecting a sigmid function as the activation function of the ELM, implying layer neuronsThe number l is determined by an empirical formula and a heuristic method and is used as a network structure for single-step prediction of the wind power; the empirical formula is as follows:
wherein alpha is a constant between 1 and 10, the number n of input layer neurons of the ELM is 7, and the number m of output layer neurons is 1; selecting a value with a smaller prediction error as a value of l according to the up-and-down floating trial of an empirical formula;
3.2) data preprocessing
In order to avoid errors caused by calculation of data with different dimensions, normalization needs to be carried out before data is input, and the normalization range selected in the text is [ -1,1 ]; the specific formula is as follows:
wherein x is the original input data, x' is the normalized value, xmaxIs the maximum value, x, in the raw dataminThe minimum in the raw data;
3.3) parameter initialization
The output layer weight of the ELM is used as the individual beta of the chicken flock of the CCSO, the prediction error is used as the individual fitness according to a fitness formula, and the fitness formula is as follows:
in the formula, yiIs the actual output of the ith node of the neural network, oiIs the predicted output of the ith node. The fitness is represented by fitness, and the number of nodes of the output layer is represented by n; initializing all parameters of the CCSO according to a chaotic chicken flock optimization algorithm, and randomly generating an initial population within an upper limit and a lower limit;
3.4) iterative optimization and termination conditions
Updating and optimizing the chicken flock according to the contents of sections (c) - (f) 3.1.3, and outputting an optimal chicken flock individual beta' and the corresponding fitness thereof when the maximum iteration times is reached or the global optimal individual fitness is not reduced any more; and (5) directly calculating and predicting the wind power value by taking beta as the weight of the output layer of the ELM.
Further, the step 4) is specifically as follows: the PSO, CSO, CPSO and CCSO algorithms are used for optimizing the ELM training process to verify the effectiveness of the prediction model, and the method comprises the following steps:
the iteration times of the algorithm are all set to be 300 times, the population size is 100, and other parameters are shown in the following table:
the step 5) is specifically as follows: the detailed comparative analysis of the Mean Absolute Percent Error (MAPE) and the Root Mean Square Error (RMSE) on the prediction error includes the following formula:
wherein N is the number of samples, PiIs the measured value of the wind power at the ith moment,
and the predicted value is the wind power predicted value at the ith moment.
Further, the step 6) is specifically: the method for analyzing the uncertainty of the wind power and taking the proposed CCSO-ELM as the interval prediction model of the wind power and the new evaluation index considering the relative deviation as the fitness function to reduce the deviation degree of the points outside the interval can be obtained by the following steps:
6.1) preprocessing data; normalizing the training data and the test data to [ -1,1], so as to avoid errors caused by data of different dimensions;
6.2) constructing a training sample set; determining an input vector X ═ X1,…x4,x5,x6,x7]Wherein x is1-x4Wind power data, x, for the first 4 time points5、x6And x7Wind speed, wind direction sine and temperature of the predicted point are respectively; for original training sample pair (X)i,ti) To float to a small extent, i.e.
Wherein Rand represents [0, 1]]Random numbers uniformly distributed therebetween to obtain UiAnd LiTwo outputs as training data;
6.3) initializing parameters; randomly generating an input weight and bias of ELM, initializing various parameters of CCSO by a chaotic chicken flock algorithm, and randomly generating an initial population;
6.4) constructing an interval prediction model; constructing a dual-output extreme learning machine prediction model according to the LUBE method, taking an output weight value as an individual beta of the chicken flock, taking a cost function F as the fitness of the individual, wherein the lower the fitness is, the better the individual is;
6.5) iterative optimization; updating and optimizing the chicken flock according to a chaotic chicken flock algorithm, and outputting an optimal chicken flock individual beta' and a corresponding fitness F when the maximum iteration times is reached or the global optimal individual fitness is not reduced any more;
6.6) calculating intervals; and 6.5) taking the optimal chicken group individual beta' obtained in the step 6.5) as an output layer weight of the ELM, directly calculating a prediction interval, and evaluating PICP, PINAW and PIRD indexes of the interval.
The invention principle is as follows: based on the analysis, aiming at the problem of insufficient optimization capacity, the invention combines the chicken flock algorithm and the chaos theory, provides the chaotic chicken flock optimization algorithm, improves the population diversity and the local search capacity of the algorithm, and improves the optimization performance of the algorithm. Aiming at the problem of randomness of the wind speed sequence, the wind speed sequence is decomposed into a plurality of subsequences with more regularity by adopting a clustering empirical mode decomposition method, and prediction is respectively carried out. Aiming at the problem that point prediction cannot provide more quantitative information, probabilistic interval prediction and an interval prediction model based on Lower Upper Bound Estimation (LUBE) of a neural network are adopted, a dual-output neural network structure is adopted to directly output a prediction interval, and the model is simple and efficient.
Has the advantages that: compared with the prior art, the method has the advantages that the problem of over-high nonlinearity of the wind power sequence is considered, the original wind power sequence is decomposed by using a clustering empirical mode decomposition method, and IMF components with lower nonlinearity are obtained; the internal rules and trends of the data can be better analyzed by using ELM, and the output weight of the extreme learning machine is optimized by using a chaotic chicken flock optimization algorithm CCSO; and establishing a CCSO-ELM prediction model for each IMF component, and finally superposing the prediction results of the components to obtain the final wind power point prediction value. Through comparative analysis, the CCSO-ELM model provided by the invention shows better prediction accuracy; in order to better provide more quantitative information for wind power uncertainty, the invention further provides a wind power interval prediction model, an upper interval and a lower interval are directly generated by adopting a LUBE method, a PIRD index is innovatively provided, the influence of a farther point falling outside the interval on a prediction result is considered, the relative deviation of the points outside the interval is effectively reduced, and the overall quality of interval prediction is improved.
Drawings
FIG. 1 is a sequence diagram of the original wind power in the present invention;
FIG. 2 is a wind power decomposition subsequence chart in the invention;
FIG. 3 is a graph of the fitness function convergence curve of the present invention;
FIG. 4 is a diagram of the wind power prediction results of algorithms in the present invention;
FIG. 5 is a graph of the PSO algorithm convergence in the present invention;
FIG. 6 is a CSO algorithm convergence graph in accordance with the present invention;
FIG. 7 is a graph of the convergence of the CPSO algorithm of the present invention;
FIG. 8 is a CCSO algorithm convergence graph in accordance with the present invention;
FIG. 9 is a PSO wind power prediction interval diagram in the present invention;
FIG. 10 is a CSO predicted wind power interval diagram in the present invention;
FIG. 11 is a CPSO predicted wind power interval diagram in the present invention;
FIG. 12 is a diagram of a wind power interval predicted by CCSO;
FIG. 13 is a flow chart of EEMD calculation in the present invention;
FIG. 14 is a diagram of an ELM network according to the present invention;
FIG. 15 is a flow chart of CCSO-ELM prediction in the present invention;
FIG. 16 is a short-term wind power prediction flow chart based on a chaotic chicken flock optimization algorithm.
Detailed description of the preferred embodiments
The present invention will be described in detail below with reference to the accompanying drawings and specific embodiments.
As shown in fig. 1-6 and fig. 16, the short-term wind power prediction method based on the chaotic chicken flock optimization algorithm includes the following steps:
taking the acquired fan data as sample data;
performing clustering empirical mode decomposition on sample data;
establishing an ELM prediction model according to data obtained by clustering analysis, and optimizing the weight of an output layer by adopting a CCSO algorithm;
optimizing the ELM training process by adopting PSO, CSO, CPSO and CCSO algorithms to verify the effectiveness of the prediction model;
detailed comparative analysis of prediction error using Mean Absolute Percent Error (MAPE) and Root Mean Square Error (RMSE);
and analyzing the uncertainty of the wind power, taking the proposed CCSO-ELM as an interval prediction model of the wind power, and proposing a new evaluation index considering relative deviation as a fitness function to reduce the deviation degree of points outside the interval.
In this embodiment, the fan data obtaining method has the following basis:
the wind power plant is provided with 25 fans, and the total installed capacity is 20 MW. The historical data of the wind power plant mainly comprises NWP data and data provided by a data acquisition and monitoring control System (SCADA), the input variable dimension is too large, so that a prediction model is too complicated, and in order to simplify the model, wind speed, wind direction and temperature which have large correlation with wind power are selected as input variables of the model;
in this embodiment, the method for performing cluster analysis on sample data includes the following steps:
for signal x (t), the decomposition steps for EMD are:
(1) finding and recording all maximum value points in the signal, and fitting by adopting cubic spline interpolation to obtain an upper envelope line e of the signalup(t)。
(2) Finding and recording all minimum value points in the signal, and fitting by adopting cubic spline interpolation to obtain a lower envelope line e of the signallow(t) taking the average of the two envelopes as
(3)h1(t)＝x(t)-m1(t) mixing h1(t) as a new signal, repeating the steps (1) and (2) for k times until a new signal h1(t) satisfying the IMF condition, let c1(t)＝h1(t), then c1(t) is the first IMF component.
(4) Subtracting c from x (t)1(t) obtaining a new signal r with the high frequency removed1(t) then r1(t) performing the above steps to finally obtain N IMF components and a margin which cannot be decomposed.
The original signal x (t) is decomposed into
In this embodiment, the step of establishing an ELM prediction model according to data obtained by cluster analysis and optimizing the output layer weight by using a CCSO algorithm includes the following steps:
(1) selection of input and output variables
A training sample set is constructed. Determining an input vector X ═ X1,…x4,x5,x6,x7]Wherein x is1-x4Wind power data, x, for the first 4 time points5、x6And x7Wind speed, wind direction sine and temperature at the predicted point, respectively. And the output data is the wind power value of the predicted point. The number n of neurons of an input layer of the ELM is 7, the number m of neurons of an output layer of the ELM is 1, a sigmid function is selected as an activation function of the ELM, and the number l of neurons of an implicit layer is determined by an empirical formula and a heuristic method and serves as a network structure for single-step prediction of wind power. The empirical formula is as follows:
where alpha is a constant between 1 and 10. And selecting a value with a smaller prediction error as a value of l according to the up-and-down floating test of an empirical formula.
(2) Data pre-processing
In order to avoid errors caused by calculation of data with different dimensions, normalization needs to be carried out before inputting the data, and the normalization range is selected to be [ -1,1 ]. The specific formula is as follows:
wherein x is the original input data, x' is the normalized value, xmaxIs the maximum value, x, in the raw dataminThe minimum in the raw data.
(3) Parameter initialization
The output layer weight of the ELM is used as the individual beta of the chicken flock of the CCSO, the prediction error is used as the individual fitness according to a fitness formula, and the fitness formula is as follows:
in the formula, yiIs the actual output of the ith node of the neural network, oiIs the predicted output of the ith node. The fitness is represented by fitness, and the number of nodes of the output layer is represented by n; initializing all parameters of the CCSO according to a chaotic chicken flock optimization algorithm, and randomly generating an initial population within an upper limit and a lower limit;
(4) iterative optimization and termination conditions
And updating and optimizing the chicken flocks according to a chaos chicken flock optimization algorithm, and outputting the optimal individual beta and the fitness corresponding to the optimal individual beta when the maximum iteration number is reached or the global optimal individual fitness is not reduced any more. And (5) directly calculating and predicting the wind power value by taking beta as the weight of the output layer of the ELM.
In this embodiment, the optimization of the PSO, CSO, CPSO and CCSO algorithms on the ELM training process to verify the validity of the prediction model includes the following steps:
the iteration times of the algorithm are all set to be 300 times, the population size is 100, and other parameters are shown in the following table:
in this embodiment, the detailed comparative analysis of the prediction error by Mean Absolute Percent Error (MAPE) and Root Mean Square Error (RMSE) includes the following equations:
wherein N is the number of samples, PiFor the wind power at the ith momentThe measured value of the rate is measured,
and the predicted value is the wind power predicted value at the ith moment.
In this embodiment, the uncertainty of the wind power is analyzed, the proposed CCSO-ELM is used as an interval prediction model of the wind power, and a new evaluation index considering relative deviation is proposed as a fitness function to reduce the deviation degree of the point outside the interval, and the method can be obtained by the following steps:
(1) and (4) preprocessing data. And normalizing the training data and the test data to [ -1,1], so as to avoid errors caused by data with different dimensions.
(2) A training sample set is constructed. Determining an input vector X ═ X1,…x4,x5,x6,x7]Wherein x is1-x4Wind power data, x, for the first 4 time points5、x6And x7Wind speed, wind direction sine and temperature at the predicted point, respectively. For original training sample pair (X)i,ti) To float to a small extent, i.e.
Wherein Rand represents [0, 1]]Random numbers uniformly distributed therebetween to obtain UiAnd LiAs two outputs of the training data.
(3) And initializing parameters. The output layer weight of the ELM is used as the individual beta of the chicken flock of the CCSO, the prediction error is used as the individual fitness according to a fitness formula, and the fitness formula is as follows:
in the formula, yiIs the actual output of the ith node of the neural network, oiIs the predicted output of the ith node. The fitness is represented by fitness, and the number of nodes of the output layer is represented by n; initializing each parameter of CCSO according to chaotic chicken flock optimization algorithmRandomly generating an initial population within the upper limit and the lower limit;
(4) and constructing an interval prediction model. A dual-output extreme learning machine prediction model is constructed according to the LUBE method, an output weight value beta is used as an individual of the chicken flock, a cost function F is used as the fitness of the individual, and the lower the fitness is, the better the individual is.
(5) And (6) iterative optimization. And updating and optimizing the chicken flock according to the chaotic chicken flock algorithm, and outputting the optimal chicken flock individual beta' and the corresponding fitness F when the maximum iteration times is reached or the global optimal individual fitness is not reduced any more.
(6) And calculating the interval. And (5) taking the individual beta obtained in the step (5) as an output layer weight of the ELM, directly calculating a prediction interval, and evaluating PICP, PINAW and PIRD indexes of the interval.
Based on the above, the method comprises the following specific implementation steps:
s1: the data assembly contains 360 sampling points in total, the time interval is 1h, the first 360 sampling points are used as a training sample set, and the last 60 sampling points are used as a test sample set. The prediction model is excessively complicated due to the fact that the input variable dimension is too large, and in order to simplify the model, wind speed, wind direction and temperature which are relatively high in wind power correlation are selected as input variables of the model. And for abnormal values or missing values which do not accord with the sequence change rule, the average value of adjacent data is adopted to replace the abnormal values or missing values.
S2: decomposing the original sequence by adopting an EEMD tool box in Matlab, setting the standard deviation of Gaussian white noise to be 0.4, and setting the number of times of adding noise to be 100;
s3: respectively establishing an ELM prediction model for each component and each margin, and optimizing the weight of an output layer by adopting a CCSO algorithm;
s4: in order to verify the effectiveness of the EEMD-CCSO-ELM-based wind power prediction model, the ELM training process is optimized by adopting PSO, CSO, CPSO and CCSO algorithms respectively, the model is trained by adopting a training sample set, and the prediction performances of the 4 algorithms are analyzed and compared. The iteration times of the algorithm are set to be 300 times, and the population size is 100.
S5: and analyzing the uncertainty of the wind power, taking the proposed CCSO-ELM as an interval prediction model of the wind power, and proposing a new evaluation index considering relative deviation as a fitness function to reduce the deviation degree of points outside the interval.
The LUBE method adopts a double-output neural network to directly construct a prediction interval, and two outputs respectively correspond to the upper bound and the lower bound of the interval without considering distribution hypothesis. The traditional Della method and the Bayesian method respectively need to calculate a Jacobian matrix and a blackplug matrix, are complex in calculation and possibly cause singularity problems, and influence the reliability and robustness of interval prediction.
In step S1, 25 fans are arranged in the wind power plant, and the total installed capacity is 20 MW. The historical data of the wind power plant mainly comprises NWP data and data provided by a data acquisition and monitoring control System (SCADA), the input variable dimension is too large, a prediction model is too complicated, and in order to simplify the model, wind speed, wind direction and temperature which have large correlation with wind power are selected as input variables of the model.
In step S2, decomposing the original sequence by using the EEMD toolbox in Matlab, setting the standard deviation of white gaussian noise to 0.4, setting the number of times of adding noise to 100, and for the signal x (t), the decomposition step of EMD is:
(1) finding and recording all maximum value points in the signal, and fitting by adopting cubic spline interpolation to obtain an upper envelope line e of the signalup(t)。
(2) Finding and recording all minimum value points in the signal, and fitting by adopting cubic spline interpolation to obtain a lower envelope line e of the signallow(t) taking the average of the two envelopes as
(3)h1(t)＝x(t)-m1(t) mixing h1(t) as a new signal, repeating the steps (1) and (2) for k times until a new signal h1(t) satisfying the IMF condition, let c1(t)＝h1(t), then c1(t) is the first IMF scoreAmount of the compound (A).
(4) Subtracting c from x (t)1(t) obtaining a new signal r with the high frequency removed1(t) then r1(t) performing the above steps to finally obtain N IMF components and a margin which cannot be decomposed.
The original signal x (t) is decomposed into
FIG. 13 is a flow chart of EEMD calculation.
In step S3, respectively establishing an ELM prediction model for each component and the margin, optimizing the output layer weight using a CCSO algorithm, establishing an ELM prediction model according to the data obtained by the cluster analysis, and optimizing the output layer weight using the CCSO algorithm include the steps of:
(1) selection of input and output variables
A training sample set is constructed. Determining an input vector X ═ X1,…x4,x5,x6,x7]Wherein x is1-x4Wind power data, x, for the first 4 time points5、x6And x7Wind speed, wind direction sine and temperature at the predicted point, respectively. And the output data is the wind power value of the predicted point. The number n of neurons of an input layer of the ELM is 7, the number m of neurons of an output layer of the ELM is 1, a sigmid function is selected as an activation function of the ELM, and the number l of neurons of an implicit layer is determined by an empirical formula and a heuristic method and serves as a network structure for single-step prediction of wind power. The empirical formula is as follows:
where alpha is a constant between 1 and 10. And selecting a value with a smaller prediction error as a value of l according to the up-and-down floating test of an empirical formula.
(2) Data pre-processing
In order to avoid errors caused by calculation of data with different dimensions, normalization needs to be carried out before inputting the data, and the normalization range is selected to be [ -1,1 ]. The specific formula is as follows:
wherein x is the original input data, x' is the normalized value, xmaxIs the maximum value, x, in the raw dataminThe minimum in the raw data.
(3) Parameter initialization
The output layer weight of the ELM is used as the individual beta of the chicken flock of the CCSO, the prediction error is used as the individual fitness according to a fitness formula, and the fitness formula is as follows:
in the formula, yiIs the actual output of the ith node of the neural network, oiIs the predicted output of the ith node. The fitness is represented by fitness, and the number of nodes of the output layer is represented by n; initializing all parameters of the CCSO according to a chaotic chicken flock optimization algorithm, and randomly generating an initial population within an upper limit and a lower limit;
(4) iterative optimization and termination conditions
And updating and optimizing the chicken flock according to the chaotic chicken flock algorithm, and outputting the optimal chicken flock individual beta 'and the fitness corresponding to the optimal chicken flock individual beta' when the maximum iteration times is reached or the global optimal individual fitness is not reduced any more. And (5) directly calculating and predicting the wind power value by taking beta as the weight of the output layer of the ELM.
FIG. 14 is a diagram of an ELM network structure, and FIG. 15 is a prediction flow chart of the above CCSO-ELM step.
In step S4, the wind power prediction model based on EEMD-CCSO-ELM provided by the present invention is verified for validity, the ELM training process is optimized by using PSO, CSO, CPSO and CCSO algorithms, respectively, the model is trained by using the training sample set, and the prediction performance of the 4 algorithms is analyzed and compared. The iteration times of the algorithm are set to be 300 times, and the population size is 100. The PSO, CSO, CPSO and CCSO algorithms for optimizing the ELM training process to verify the effectiveness of the prediction model comprise the following steps:
the iteration times of the algorithm are all set to be 300 times, the population size is 100, and other parameters are shown in the following table:
in the step S5, the uncertainty of the wind power is analyzed, the proposed CCSO-ELM is used as an interval prediction model of the wind power, and a new evaluation index considering relative deviation is proposed as a fitness function to reduce the deviation degree of points outside the interval. The method comprises the following steps:
(1) and (4) preprocessing data. And normalizing the training data and the test data to [ -1,1], so as to avoid errors caused by data with different dimensions.
(2) A training sample set is constructed. Determining an input vector X ═ X1,…x4,x5,x6,x7]Wherein x is1-x4Wind power data, x, for the first 4 time points5、x6And x7Wind speed, wind direction sine and temperature at the predicted point, respectively. For original training sample pair (X)i,ti) To float to a small extent, i.e.
Wherein Rand represents [0, 1]]Random numbers uniformly distributed therebetween to obtain UiAnd LiAs two outputs of the training data.
(3) And initializing parameters. And (3) randomly generating an input weight and bias of the ELM, initializing various parameters of the CCSO according to the chaotic chicken swarm algorithm, and randomly generating an initial population.
(4) And constructing an interval prediction model. A dual-output extreme learning machine prediction model is constructed according to the LUBE method, an output weight value beta is used as an individual of the chicken flock, a cost function F is used as the fitness of the individual, and the lower the fitness is, the better the individual is.
(5) And (6) iterative optimization. And updating and optimizing the chicken flock according to the chaotic chicken flock algorithm, and outputting the optimal chicken flock individual beta' and the corresponding fitness F when the maximum iteration times is reached or the global optimal individual fitness is not reduced any more.
(6) And calculating the interval. And (5) taking the optimal chicken flock individual beta' obtained in the step 5 as an output layer weight of the ELM, directly calculating a prediction interval, and evaluating PICP, PINAW and PIRD indexes of the interval.
The solution according to the invention is further illustrated below in a specific example:
step 1: the method is based on the fan data of a certain domestic wind power plant: the wind power plant is provided with 25 fans, and the total installed capacity is 20 MW. The historical data of the wind power plant mainly comprises NWP data and data provided by a supervisory control and data acquisition (SCADA) system. NWP provides the input data needed for prediction: wind speed, direction, temperature, humidity and atmospheric pressure; the SCADA provides actual wind power data corresponding to the time nodes. The time resolution of the data set is 1 h.
The prediction model is excessively complicated due to the fact that the input variable dimension is too large, and in order to simplify the model, wind speed, wind direction and temperature which are relatively high in wind power correlation are selected as input variables of the model. However, due to external factors such as environment and personnel, situations such as data loss and abnormal values of the collected data set inevitably occur, and when the abnormal factors are excessive, the prediction result is negatively affected to a certain extent, so that the prediction accuracy is affected. It is therefore necessary to deal with data loss and anomalies accordingly: reasonable data are filled in the missing values as far as possible, and other data are replaced for abnormal values, so that the abnormal values are more in line with the change rule of the sequence. Regarding the wind power value exceeding the total installed capacity as an abnormal value, and modifying the abnormal value into the installed total capacity; and for abnormal values or missing values which do not accord with the sequence change rule, the average value of adjacent data is adopted to replace the abnormal values or missing values.
Step 2: the original wind power time sequence is shown in figure 1, which has high non-stationarity, the original sequence is decomposed by adopting an EEMD tool box in Matlab, the standard deviation of Gaussian white noise is set to be 0.4, the number of times of adding noise is set to be 100, and the corresponding IMF component is obtained as shown in figure 2. As can be seen from the figure, the original wind power sequence is decomposed into 7 IMF components and 1 residual component RES, the non-stationarity of the decomposed components and residual is obviously reduced, and even a certain change rule is presented.
And step 3: after clustering empirical mode decomposition, an ELM prediction model is respectively established for each component and the margin, and the CCSO algorithm is adopted to optimize the weight of the output layer. In order to verify the effectiveness of the wind power prediction model based on EEMD-CCSO-ELM, PSO, CSO, CPSO and CCSO algorithms are respectively adopted to optimize the ELM training process, a training sample set is adopted to train the model, and the prediction performances of the 4 algorithms are analyzed and compared. The iteration times of the algorithm are set to be 300 times, the population size is 100, and other parameters are shown in table 1.
TABLE 1 cumulative Effect coefficients
The convergence process of the fitness function fitness in the training process of each algorithm is shown in fig. 3, and the finally predicted wind power curve is shown in fig. 4. As can be seen from the figure, the optimization performance of the PSO algorithm is general, and the PSO algorithm falls into a local extreme value and cannot jump out in about the 100 th generation, and the downward trend of the fitness function of the globally optimal individual tends to be gentle, so that an ideal optimization effect is not achieved; the learning mode of dividing subgroups by the chicken swarm algorithm of CSO can greatly enhance the diversity of the population, but the descending speed of the fitness function is obviously slowed down until the fitness function is not reduced any more due to the lack of good local searching capability in the later period; the CPSO algorithm is an improved method combining chaotic search with the particle swarm algorithm, the local search capability of the particle swarm algorithm is enhanced, after the 150 th generation, the fitness function tends to be smooth, and the optimization effect is good; the CCSO algorithm combines the chicken flock algorithm and the chaotic search, the population diversity and the local search capability are improved, in addition, the adaptive adjustment of the variation probability and the search space is adopted, the optimization efficiency of the algorithm is enhanced, the possibility of being trapped into the local optimum is greatly reduced, the optimization result is more ideal, and compared with the CPSO, the model has stronger optimization capability.
And 4, step 4: as can be seen from FIG. 4, after EEMD decomposition, the fitting effect of PSO-ELM is not very good, especially the error is large at the inflection point with large fluctuation, the change direction of power cannot be tracked well, the RMSE is 105.24kW, and the MAPE is 17.43%; CSO has certain advantages in population diversity relative to PSO, the prediction error of the optimized model CSO-ELM is lower than that of PSO-ELM, MAPE reaches 14.85%, and the average absolute percentage error is reduced by about 3%; the CPSO adds chaotic search in the PSO, so that the local search capability of the algorithm is improved, and the prediction error of the CPSO-ELM is reduced by about 5 percent compared with that of the PSO-ELM; the CCSO-ELM model provided by the invention has a better fitting effect, can be tightly attached to a real wind power value at an inflection point, combines the population diversity of a chicken flock algorithm and the local search capability of chaotic optimization, greatly reduces the possibility of falling into local optimum, and can better find out the internal trend and the rule of data. The MAPE of the model is reduced by 6.79%, 4.21% and 2.17% relative to PSO-ELM, CSO-ELM and CPSO-ELM respectively, the RMSE of the model is reduced by 66.25kW, 22.85kW and 17.2kW respectively, and the prediction errors of different models are listed in Table 2.
Simulation results show that the prediction model provided by the invention has higher prediction precision and has certain reference value for the dispatching and operation of the power grid.
TABLE 2 Point prediction evaluation index for each model
And 5: optimizing the ELM training process by adopting PSO, CSO, CPSO and CCSO algorithms, training the model by adopting a training sample set, and analyzing and comparing the interval prediction performance of the 4 algorithms. The iteration times of the algorithm are set to 300 times, the population size is 100, the confidence coefficient mu is 90%, and eta in the objective function F (beta)1And η2Set to 50 and 0.3, respectively, and the other parameters are as shown in table 1. The PICP, PINAW, and PIRD convergence processes in the four algorithm training processes are shown in FIGS. 5-8.
As can be seen from fig. 5 to 8, the PICP values of the 4 algorithms are always over 90%, because due to the existence of the penalty term, the fitness function F will rise to a larger value once the constraint condition is not satisfied. The primary PINAW value of the PSO algorithm is reduced rapidly, but the PINAW value is not changed much after 40 generations, which is premature convergence caused by the lack of population diversity in the later period of the algorithm.
The PINAW is reduced by 30% by the CSO algorithm in about 10 th generation, and the fact that the learning mode of dividing subgroups by the chicken swarm algorithm can greatly enhance the diversity of the populations is proved. The descending trend of PINAW index in CPSO algorithm is obviously improved, and PINAW is reduced to about 28.5% after the 150 th generation.
The PINAW in the CCSO algorithm is reduced to about 27.7% in the 50 th generation, the optimization performance is outstanding, and the optimization capability of the model is better than that of the other three optimization algorithms.
TABLE 3 prediction evaluation index for each model interval
As can be seen from fig. 9 to 12, under the condition of 90% confidence, compared with the other 3 algorithms, the prediction interval width of the CCSO is narrower, and only a few points fall on the periphery of the interval, and due to the presence of the PIRD index, the deviation of the points on the periphery of the interval is greatly reduced, so that when the points fall outside the interval, the adverse effect is minimized.
Table 3 shows the evaluation indexes of the four prediction models for predicting the wind power interval, with the confidence levels of 90% and 80%, respectively. It can be seen that different optimization algorithms have a large impact on the prediction results at the same confidence level. Due to the existence of penalty terms, the coverage rate of each algorithm is higher than the confidence coefficient mu, and the comparative analysis of PINAW and PIRD indexes under 90% confidence level shows that the coverage rate of CCSO is slightly lower than that of PSO algorithm, the average bandwidth and the relative deviation are greatly reduced, the bandwidth indexes are respectively reduced by 5.33%, 3.52% and 1.81% relative to PSO, CSO and CPSO, and the relative deviations are respectively reduced by 1.9%, 0.83% and 0.46%. In general, the method provided by the invention effectively improves the prediction definition of the wind power interval on the premise of ensuring the reliability, and has a certain reference value for the operation of a decision maker.
Claims (5)
1. The short-term wind power prediction method based on the chaotic chicken flock optimization algorithm is characterized by comprising the following steps of:
1) taking the acquired fan data as sample data;
2) performing clustering empirical mode decomposition on sample data;
3) establishing an ELM prediction model according to data obtained by clustering analysis, and optimizing the weight of an output layer by adopting a CCSO algorithm, which is called a CCSO-ELM prediction model;
4) optimizing the ELM training process by adopting PSO, CSO, CPSO and CCSO algorithms to verify the effectiveness of the prediction model;
5) comparing and analyzing the prediction error by adopting the average absolute percentage error MAPE and the root mean square error RMSE;
6) and taking the CCSO-ELM prediction model as an interval prediction model of the wind power, and reducing the deviation degree of the points outside the interval by taking the evaluation index considering the relative deviation as a fitness function.
2. The short-term wind power prediction method based on the chaotic chicken flock optimization algorithm according to claim 1, wherein the step 1) is specifically as follows: the fan data acquisition method is based on the following steps: the historical data of the wind power plant comprises NWP data and data provided by a data acquisition and monitoring control system, and wind speed, wind direction and temperature are selected as input variables of the model;
the method for carrying out cluster analysis on the sample data comprises the following steps:
for signal x (t), the decomposition steps for EMD are:
1.1) finding and recording all maximum value points in the signal, and fitting by adopting cubic spline interpolation to obtain an upper envelope line e of the signalup(t)；
1.2) finding and recording all minimum value points in the signal, and fitting by adopting cubic spline interpolation to obtain a lower envelope line e of the signallow(t) taking the average of the two envelopes as
1.3)h1(t)＝x(t)-m1(t) mixing h1(t) as a new signal, repeating step 1.1) and step 1.2), and calculating for k times until a new signal h1(t) satisfying the IMF condition, let c1(t)＝h1(t), then c1(t) is the first IMF component;
1.4) subtracting c from x (t)1(t) obtaining a new signal r with the high frequency removed1(t) then r1(t) performing the above steps to finally obtain N IMF components and a non-resolvable margin;
the original signal x (t) is decomposed into
3. The short-term wind power prediction method based on the chaotic chicken flock optimization algorithm according to claim 1, wherein the step 3) is specifically as follows: the method for establishing an ELM prediction model according to data obtained by clustering analysis and optimizing the weight of an output layer by adopting a CCSO algorithm comprises the following steps:
3.1) selection of input and output variables
Constructing a training sample set; determining an input vector; the output data is the wind power value of the prediction point; the sigmid function is selected as the activation function of ELM, and the number l of hidden layer neurons is expressed as follows:
wherein alpha is a constant between 1 and 10, the number n of input layer neurons of the ELM, and the number m of output layer neurons;
3.2) data preprocessing
Normalization is performed, and the formula is as follows:
wherein x is the original input data, x' is the normalized value, xmaxIs the maximum value, x, in the raw dataminThe minimum in the raw data;
3.3) parameter initialization
The output layer weight of the ELM is used as the individual beta of the chicken flock of the CCSO, the prediction error is used as the individual fitness according to a fitness formula, and the fitness formula is as follows:
in the formula, yiIs the actual output of the ith node of the neural network, oiIs the predicted output of the ith node; the fitness is represented by fitness, and the number of nodes of the output layer is represented by n; initializing all parameters of the CCSO according to a chaotic chicken flock optimization algorithm, and randomly generating an initial population within an upper limit and a lower limit;
3.4) iterative optimization and termination conditions
Updating and optimizing the chicken flock according to a chaotic chicken flock optimization algorithm, and outputting an optimal chicken flock individual beta 'and a fitness corresponding to the optimal chicken flock individual beta' when the maximum iteration times is reached or the global optimal individual fitness is not reduced any more; and (5) directly calculating and predicting the wind power value by taking beta as the weight of the output layer of the ELM.
4. The short-term wind power prediction method based on the chaotic chicken flock optimization algorithm according to claim 1, wherein the step 5) is specifically as follows:
the detailed comparative analysis of the prediction error by the mean absolute percent error MAPE and the root mean square error RMSE includes the following equations:
wherein N is the number of samples, PiIs the measured value of the wind power at the ith moment,
and the predicted value is the wind power predicted value at the ith moment.
5. The short-term wind power prediction method based on the chaotic chicken flock optimization algorithm according to claim 1, characterized in that the step 6) is obtained by the following steps:
6.1) preprocessing data; normalizing both the training data and the test data to [ -1,1 ];
6.2) constructing a training sample set, i.e.
Wherein Rand represents [0, 1]]Random numbers uniformly distributed therebetween to obtain UiAnd LiTwo outputs as training data;
6.3) initializing parameters; randomly generating an input weight and bias of ELM, initializing various parameters of CCSO according to a chaotic chicken flock optimization algorithm, and randomly generating an initial population;
6.4) constructing an interval prediction model; constructing a dual-output extreme learning machine prediction model according to the LUBE method, taking an output weight value as an individual beta of the chicken flock, and taking a cost function F as the fitness of the individual;
6.5) iterative optimization; updating and optimizing the chicken flock according to a chaotic chicken flock optimization algorithm, and outputting an optimal chicken flock individual beta' and a corresponding fitness F when the maximum iteration times is reached or the global optimal individual fitness is not reduced any more;
6.6) calculating intervals; and 6.5) taking the optimal chicken flock individual beta' obtained in the step 6.5) as an output layer weight of the ELM, directly calculating a prediction interval, and evaluating PICP, PINAW and PIRD indexes of the interval.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110189031.6A CN112819238A (en) | 2021-02-19 | 2021-02-19 | Short-term wind power prediction method based on chaotic chicken flock optimization algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110189031.6A CN112819238A (en) | 2021-02-19 | 2021-02-19 | Short-term wind power prediction method based on chaotic chicken flock optimization algorithm |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112819238A true CN112819238A (en) | 2021-05-18 |
Family
ID=75865615
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110189031.6A Withdrawn CN112819238A (en) | 2021-02-19 | 2021-02-19 | Short-term wind power prediction method based on chaotic chicken flock optimization algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112819238A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113361769A (en) * | 2021-06-04 | 2021-09-07 | 南方电网科学研究有限责任公司 | Stability margin value prediction method and device based on PRMSE and CRMSE evaluation indexes |
| CN113468817A (en) * | 2021-07-13 | 2021-10-01 | 淮阴工学院 | Ultra-short-term wind power prediction method based on IGOA (optimized El-electric field model) |
| CN114169252A (en) * | 2021-12-27 | 2022-03-11 | 广东工业大学 | Short-term region wind power prediction method for dynamically selecting representative wind power plant |
| CN115130743A (en) * | 2022-06-22 | 2022-09-30 | 国网山东省电力公司电力科学研究院 | Wind turbine generator regulation rate prediction method and system based on variation inference |
-
2021
- 2021-02-19 CN CN202110189031.6A patent/CN112819238A/en not_active Withdrawn
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113361769A (en) * | 2021-06-04 | 2021-09-07 | 南方电网科学研究有限责任公司 | Stability margin value prediction method and device based on PRMSE and CRMSE evaluation indexes |
| CN113468817A (en) * | 2021-07-13 | 2021-10-01 | 淮阴工学院 | Ultra-short-term wind power prediction method based on IGOA (optimized El-electric field model) |
| CN113468817B (en) * | 2021-07-13 | 2023-08-22 | 淮阴工学院 | Ultra-short-term wind power prediction method based on IGOA (insulated gate bipolar transistor) optimized ELM (ELM) |
| CN114169252A (en) * | 2021-12-27 | 2022-03-11 | 广东工业大学 | Short-term region wind power prediction method for dynamically selecting representative wind power plant |
| CN115130743A (en) * | 2022-06-22 | 2022-09-30 | 国网山东省电力公司电力科学研究院 | Wind turbine generator regulation rate prediction method and system based on variation inference |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112819238A (en) | Short-term wind power prediction method based on chaotic chicken flock optimization algorithm | |
| CN110619360A (en) | Ultra-short-term wind power prediction method considering historical sample similarity | |
| CN111401755B (en) | Multi-new-energy output scene generation method, device and system based on Markov chain | |
| CN110717610B (en) | Wind power prediction method based on data mining | |
| CN110766200A (en) | Method for predicting generating power of wind turbine generator based on K-means mean clustering | |
| CN111027775A (en) | Step hydropower station generating capacity prediction method based on long-term and short-term memory network | |
| CN111709454B (en) | Multi-wind-field output clustering evaluation method based on optimal copula model | |
| CN114861533A (en) | Wind power ultra-short-term prediction method based on time convolution network | |
| CN114362175B (en) | Wind power prediction method and system based on depth certainty strategy gradient algorithm | |
| CN114511132A (en) | Photovoltaic output short-term prediction method and prediction system | |
| CN115239029B (en) | Wind power prediction method and system considering power time sequence and meteorological dependent characteristics | |
| CN115511657A (en) | Wind power output and photovoltaic output evaluation method based on combined prediction model | |
| CN110991701A (en) | Wind power plant fan wind speed prediction method and system based on data fusion | |
| Chen et al. | Research on wind power prediction method based on convolutional neural network and genetic algorithm | |
| CN116702937A (en) | Photovoltaic output day-ahead prediction method based on K-means mean value clustering and BP neural network optimization | |
| Fang et al. | Wind turbine rotor speed design optimization considering rain erosion based on deep reinforcement learning | |
| CN114298132A (en) | Wind power prediction method and device and electronic equipment | |
| CN116821610B (en) | Method for optimizing wind power generation efficiency by utilizing big data | |
| CN111582567B (en) | Wind power probability prediction method based on hierarchical integration | |
| CN112633565A (en) | Photovoltaic power aggregation interval prediction method | |
| CN115526376A (en) | Multi-feature fusion generation countermeasure network ultra-short-term wind power prediction method | |
| CN114897260A (en) | Short-term wind speed prediction model modeling method and prediction method based on LSTM neural network | |
| CN114139783A (en) | Wind power short-term power prediction method and device based on nonlinear weighted combination | |
| CN113807019A (en) | MCMC wind power simulation method based on improved scene classification and coarse grain removal | |
| CN113112085A (en) | New energy station power generation load prediction method based on BP neural network |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20210518 |
|
| WW01 | Invention patent application withdrawn after publication |