CN111667098B - Wind power station output power prediction method based on multi-model combination optimization - Google Patents
Wind power station output power prediction method based on multi-model combination optimization Download PDFInfo
- Publication number
- CN111667098B CN111667098B CN202010405975.8A CN202010405975A CN111667098B CN 111667098 B CN111667098 B CN 111667098B CN 202010405975 A CN202010405975 A CN 202010405975A CN 111667098 B CN111667098 B CN 111667098B
- Authority
- CN
- China
- Prior art keywords
- wind power
- data
- wolf
- meteorological
- model
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 50
- 238000005457 optimization Methods 0.000 title claims abstract description 49
- 241000282461 Canis lupus Species 0.000 claims abstract description 101
- 238000012549 training Methods 0.000 claims abstract description 57
- 238000012706 support-vector machine Methods 0.000 claims abstract description 52
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 22
- 230000010354 integration Effects 0.000 claims abstract description 10
- 238000004458 analytical method Methods 0.000 claims abstract description 5
- 238000010606 normalization Methods 0.000 claims abstract description 4
- 230000006870 function Effects 0.000 claims description 70
- 238000012360 testing method Methods 0.000 claims description 26
- 239000013598 vector Substances 0.000 claims description 18
- 125000004122 cyclic group Chemical group 0.000 claims description 10
- 239000011159 matrix material Substances 0.000 claims description 9
- 210000002569 neuron Anatomy 0.000 claims description 9
- 108010074864 Factor XI Proteins 0.000 claims description 6
- 238000013528 artificial neural network Methods 0.000 claims description 6
- 238000010219 correlation analysis Methods 0.000 claims description 4
- 238000007781 pre-processing Methods 0.000 claims description 4
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 3
- 108010046685 Rho Factor Proteins 0.000 claims description 3
- 230000004913 activation Effects 0.000 claims description 3
- 238000012937 correction Methods 0.000 claims description 3
- 238000006073 displacement reaction Methods 0.000 claims description 3
- 230000002401 inhibitory effect Effects 0.000 claims description 3
- 229940050561 matrix product Drugs 0.000 claims description 3
- 230000008569 process Effects 0.000 claims description 3
- 230000000717 retained effect Effects 0.000 claims description 3
- 230000001502 supplementing effect Effects 0.000 claims description 3
- 238000004364 calculation method Methods 0.000 claims description 2
- 230000003247 decreasing effect Effects 0.000 claims description 2
- 230000009286 beneficial effect Effects 0.000 abstract description 2
- BTCSSZJGUNDROE-UHFFFAOYSA-N gamma-aminobutyric acid Chemical compound NCCCC(O)=O BTCSSZJGUNDROE-UHFFFAOYSA-N 0.000 description 19
- 238000010248 power generation Methods 0.000 description 5
- 238000011161 development Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 238000003912 environmental pollution Methods 0.000 description 1
- 230000000670 limiting effect Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000002829 reductive effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 238000004088 simulation 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
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
- Y02E40/00—Technologies for an efficient electrical power generation, transmission or distribution
- Y02E40/70—Smart grids as climate change mitigation technology in the energy generation sector
-
- 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)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Economics (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Strategic Management (AREA)
- Human Resources & Organizations (AREA)
- Software Systems (AREA)
- Computational Linguistics (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- Marketing (AREA)
- Biophysics (AREA)
- General Business, Economics & Management (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Tourism & Hospitality (AREA)
- Game Theory and Decision Science (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Entrepreneurship & Innovation (AREA)
- Development Economics (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- Primary Health Care (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention provides a wind power station output power prediction method based on multi-model combination optimization. Acquiring meteorological influence factor data and wind power data of a target wind power station, sequentially performing data missing value completion and data normalization, and analyzing the association degree of the meteorological influence factor data and the wind power data through a grey association degree analysis method; further establishing a support vector machine prediction model, and optimizing parameters through a grey wolf group optimization algorithm; taking the optimized support vector machine model as a weak regressor, constructing a strong regressor by establishing an AdaBoost integration model, and predicting by using the strong regressor to obtain a training set wind power predicted value sequence; further constructing a gated cycle unit network to predict the error sequence E; and superposing the corrected prediction error value and the prediction result of the AdaBoost integrated prediction model to obtain the final prediction result. The method improves the prediction efficiency and the prediction precision of the wind power prediction model, and is beneficial to scientific scheduling and grid-connected operation safety and stability of the power system.
Description
Technical Field
The invention belongs to the technical field of wind power generation, and relates to a wind power station output power prediction method based on multi-model combination optimization.
Background
Along with the rapid development of economy, the demand of each country on energy sources is greater and greater, and the requirement on the environment is higher and higher. The wind energy is used as a clean energy, so that the outstanding problems of energy shortage, environmental pollution, greenhouse effect and the like at present can be effectively relieved; meanwhile, as a renewable energy source, the wind energy has a wide development prospect under the background that fossil energy is gradually exhausted. Data shows that in 2018, the newly added grid-connected wind power installed capacity in China is 2059 thousands of kilowatts, the accumulated grid-connected installed capacity reaches 1.84 hundred million kilowatts, and the accumulated grid-connected installed capacity accounts for 9.7 percent of the total installed capacity of power generation. In 2019, in 1-6 months, 909 million kilowatts of wind power installed capacity are newly added in China, wherein 40 million kilowatts of offshore wind power are added, and the accumulated grid-connected installed capacity reaches 1.93 hundred million kilowatts.
Due to uncertainty, volatility and intermittency of wind power, when wind power generation is carried out, severe examination can be brought to safe and stable operation of a power grid and safety and reliability of grid connection of a power system. The method has great significance for reducing the influence of wind power generation on a power system and realizing accurate prediction of wind power of a wind power plant.
Disclosure of Invention
The invention aims to provide a wind power station output power prediction method based on multi-model combination optimization, so as to improve the prediction precision of wind power output power and realize scientific scheduling and optimized operation of a power system.
In order to solve the above technical problem, an embodiment of the present invention provides the following technical solutions:
a wind power station output power prediction method based on multi-model combination optimization predicts wind power generation power in a future time period according to historical wind power characteristic data.
The wind power station output power prediction method based on the combined model comprises the following steps:
step 6, overlapping the corrected prediction error value with a prediction result of the AdaBoost integrated prediction model to obtain a final prediction result;
preferably, the preprocessed meteorological influence factor data in step 1 is:
X i (k)，i∈[1，n]，k∈[1，m]
wherein, X i (k) Representing kth point data in the ith preprocessed meteorological influence factor, wherein n is the type number of the meteorological influence factors, and m is the number of data points in each meteorological factor;
X o (k)，k∈[1，m]
wherein, X o (k) Representing preprocessed wind power data when the kth point data is selected from the preprocessed meteorological influence factor data, wherein m is the number of data points in each meteorological factor;
selecting kth point data from the preprocessed meteorological influence factor data as follows:
{X 1 (k)，X 2 (k)，...，X n (k)}
calculating a gray correlation coefficient:
i∈[1，n]，k∈[1，m]
wherein, X i (k) Represents the kth point data theta in the ith preprocessed meteorological influence factor i (k) Representing the gray correlation coefficient, X, of the kth point in the weather influence factor after the ith preprocessing o (k) Representing the preprocessed wind power data when the kth point data is selected from the preprocessed meteorological influence factor data, wherein n is the type number of the meteorological influence factors, m is the number of data points in each meteorological factor, rho is a resolution coefficient,
represents the two-stage minimum difference between the wind power and the ith meteorological influence factor, and is used for judging whether the wind power is greater than or equal to the preset value>
Representing the two-stage maximum difference between the wind power and the ith meteorological influence factor;
γ i ∈[0，1]
wherein, gamma is i Correlation degree of ith meteorological influence factor and wind power, gamma i The smaller the error between the power value and 1 is, the higher the correlation degree between the ith meteorological influence factor and the wind power is;
preferably, the meteorological influence factor data after the feature selection in step 2 is:
X j (k)，j∈[1，c]，k∈[1，m]，c＜n
wherein, X j (k) Representing kth point data in the meteorological influence factors after the jth characteristic selection, wherein c is the type number of the meteorological influence factors after the characteristic selection, and m is the number of data points in each meteorological factor;
simultaneously, a relaxation factor xi is introduced to avoid the relaxation factor xi i If the size is too large, adding a penalty factor C, and supporting the target function expression of a vector machine prediction model as follows:
wherein, C represents the punishment parameter of the support vector machine as the parameter to be optimized, xi i Represents the relaxation factor, x, of the ith sample data i Represents input sample data, y i Representing a label, b representing a displacement term, determining the distance between the hyperplane and the origin, ω representing a normal vector, determining the direction of the hyperplane,
kernel function K (x) of support vector machine p ，x q ) The method is characterized by comprising the following steps of (1) selecting a Gaussian radial basis kernel function (RBF) as follows:
x p ＝[X 1 (p) X 2 (p) ... X j (p) ... X c (p)]
x q ＝[X 1 (q) X 2 (q) ... X j (q) ... X c (q)]
wherein x is p Denotes the p-th sample data, x q It is indicated that the q-th sample data,
representing the Euclidean distance between two points of the square of the distance between the two points, σ being the width of the function, g =1/2 σ 2 Representing the kernel function parameter of the support vector machine as a parameter to be optimized;
preferably, in step 3, the penalty parameter of the support vector machine and the kernel function parameter of the support vector machine are optimized by a grey wolf group optimization algorithm, respectively, as follows:
step 3.1, initializing algorithm parameters of the gray wolf colony optimization algorithm: the population scale N, the maximum iteration times maxgen, the penalty parameter C, the value range of the kernel function parameter g and the position of the wolf individual;
step 3.2, randomly generating a population of N wolf individuals according to the population scale N, wherein the positions of the N wolf individuals D are both composed of a penalty parameter C and a kernel function parameter g, and random initialization is carried out according to the value ranges of the penalty parameter C and the kernel function parameter g;
the positions of the N individual wolfs are as follows in sequence: (C) b ，g b )，b∈[1，N]；
Step 3.3, the mean square error MSE is used as a fitness function, the population of N wolf individuals is divided into four different levels of alpha, beta, delta and omega, and the positions of the wolfs are respectively updated according to the fitness function value and the fitness function value relations of the alpha wolf, the beta wolf, the delta wolf and the omega wolf;
predicting the meteorological influence factors after the feature selection through the SVM model established in the step 2 to obtain a wind power predicted value
The fitness function is the sum of the mean square errors of the predicted value of the wind power and the actual value of the wind power, and specifically comprises the following steps:
k∈[1，S]S＜m
wherein, X o (k) Selecting the actual value of the wind power when the k point data is selected for the meteorological influence factor data after each characteristic is selected,
when the kth point data is selected for the meteorological influence factor data after the features are selected, the wind power predicted value predicted by a support vector machine prediction model is obtained, S is the number of wind power data points in a training set, and m is the number of the wind power data points;
and 3.4, updating the direction and the position of the omega wolf searching individual in the grey wolf group according to a formula, wherein the specific formula is as follows:
D s (t)＝|C u X s (t)-X u (t)|u＝1，2，3，s＝α，β，δ
X u (t+1)＝X s (t)-A u |C u X s (t)-X u (t)|u＝1，2，3，s＝α，β，δ
where t represents the current number of iterations, D s (t) watchShowing the distances between the alpha, beta and delta level wolfs and the omega level wolf, X s (t) represents the current position of the alpha-level wolf, the beta-level wolf, the delta-level wolf, X u (t) represents the current position of the omega level wolf, X u (t + 1) represents the alpha-level wolf, the beta-level wolf, the delta-level wolf determines the position of the next movement of the omega-level wolf, and X (t + 1) represents the position of the next moment of the omega-level wolf;
and 3.5, updating parameters alpha, A and C according to a formula, and determining a new position of an alpha-level wolf, a new position of a beta-level wolf and a new position of a delta-level wolf according to the updated optimal fitness function value, wherein the specific formula is as follows:
A＝α*(2r 1 -1)
C＝2r 2
wherein, denoted as matrix product, A, C is the first cooperative coefficient and the second cooperative coefficient respectively, a is linearly decreased from 2 to 0 in the whole iteration process, r1 and r2 are the first random vector and the second random variable between [0,1 ];
and 3.6, if the maximum iteration times maxgen are exceeded, terminating the algorithm, outputting the global optimal positions in the N wolf group individuals, namely the penalty parameter C of the optimized support vector machine * And kernel function parameter g of optimized support vector machine * Otherwise, skipping to the step 3.4 to continue optimization;
wherein i is E [1,c]，w zi Weight of the kth sample representing the z-th weak regressor, e zi Expressed as the relative error of the kth sample of the z-th weak regressor, X o (k) After selecting for each featureWind power actual value F when the k point data is selected from the meteorological influence factor data z (X i (k) Is the predicted value of the wind power of the z-th weak regressor at the k-th data point, E z The maximum error of the z-th weak regressor on the training set;
and 4, calculating error rates of a plurality of weak regressors by combining the relative errors, wherein the error rates are as follows:
according to the relative error, the error rate e of the z-th weak regressor can be obtained kz The method comprises the following steps:
determining the weight coefficient alpha of a weak regressor z The method comprises the following steps:
the sample weights are updated as follows:
wherein, ω is z+1，i Sample weight coefficient, Z, representing the Z +1 th weak regressor a In order to normalize the factors, the method comprises the steps of,
and 4, constructing a strong regressor by establishing an AdaBoost integration model: superposing the product of the weight of each weak regression and the predicted value to obtain the final strong regression;
and 4, predicting each sample in the training set by using a strong regressor to obtain a training set wind power predicted value sequence as follows:
the training set is the meteorological influence factor after the feature selection in the step 2;
training a wind power predicted value sequence, which is specifically as follows:
the sequence of the predicted values of the wind power is tested as follows:
wherein S represents the number of wind power data points in the training set, m represents the number of wind power data points,
represents the sequence of k wind power predicted values in the training set, and>
representing k wind power predicted value sequences, X, in a test set f (k) Represents the k wind power predicted value, X, in the training set g (k) Representing the k wind power predicted value in the test set,
calculating a training set predicted power value
Historical wind power value X of wind power station o (k) The difference in (c) is denoted as error sequence E, as follows:
preferably, the step 5 of constructing the gated cyclic unit network is as follows:
constructing a gated circulation unit network through an input layer, a plurality of gated circulation unit layers and an output layer;
an updating gate and a resetting gate corresponding to each gating circulation unit layer exist on each gating circulation unit layer;
the update gate determines information that was retained to the current time at a previous time, and the reset gate determines the extent to which information was discarded;
the updating gate and the placing gate are sequentially defined as follows:
Z t ＝σ(W Z X t +U Z h t - 1 +b Z )
r t ＝σ(W r X t +U r h t-1 +b r )
wherein z is t To refresh the door, r t To reset the gate, X t Input representing the current time, W Z And U Z Respectively, updating the weight matrix of the gate, W r And U r Weight matrices, b, of reset gates, respectively Z And b r Respectively are offset vectors, and sigma is a sigmoid activation function;
the hyper-parameters of the gated loop unit include: the number of layers of the neural network, the number of neurons in each layer and the dropout rate; wherein the dropout rate is a proportion of the number of discarded neurons and is used for inhibiting an overfitting phenomenon of the model, and the dropout belongs to (0,1);
by setting the value range of the parameters to be optimized simultaneously, finding out the corresponding parameters, namely the optimal number of neural network layers, the optimal number of neurons in each layer and the optimal dropout rate of the GRU model according to the minimum principle of a loss function (MAE) on a training set;
k∈[1，S] S＜m
wherein, X o (k) Selecting the actual value of the wind power when the k point data is selected for the meteorological influence factor data after each characteristic is selected,
selecting a wind power predicted value predicted by the gate control cycle unit network when the kth point data is selected for the meteorological influence factor data after each feature is selected;
and 5, optimizing the parameters of the gated cyclic unit network as follows: optimizing and solving the gating circulation unit network through an Adam algorithm to obtain a weight matrix Q of the updated gate after optimization Z 、U* Z Weight matrix W of reset gate after optimization r 、U* r Updating the offset vector b of the gate after optimization *z Bias vector b of reset gate after optimization r ；
After parameter optimization is carried out on the training set, prediction is carried out on the test set to obtain a predicted value sequence of the test set errors, and the predicted value sequence is recorded as
The following were used:
preferably, the final prediction result in step 6 is as follows:
wherein X' (k) is the final prediction result of k wind power in the test set,
for a sequence of predictors in a test set error>
And m-S is the number of samples of the test set.
Compared with the prior art, the invention has the following advantages:
the innovative combined model prediction scheme is as follows: because the single model has low prediction accuracy, a large error exists. The invention innovatively applies a technical scheme of predicting five model combinations of gray correlation analysis GRA, support vector machine SVM, gray wolf group optimization algorithm GWOO, adaBoost integration and gating circulation unit GRU, and improves the prediction precision of wind power to a certain extent.
And performing feature selection by utilizing a Gray Relevance Analysis (GRA), and selecting the first few influence factors with larger influence from a plurality of features as the input of the model, thereby improving the prediction efficiency of the model.
The parameters of the SVM prediction model are solved by constructing a complete optimization problem model, and a gray wolf group optimization algorithm GWO is adopted in the solving method.
And (3) constructing a strong regression by using an AdaBoost integration model, integrating on the basis of the shallow model, and improving the prediction precision of the model.
And finally, predicting the error by using a gated circulation unit GRU, and superposing the corrected predicted error and the predicted value of the wind power so as to further improve the prediction precision of the model.
The invention has the beneficial technical effects that:
according to the invention, the accurate prediction of the wind power is realized through the GRA-SVM-GWO-AdaBoost-GRU combined model, and the safety and stability of scientific scheduling and grid-connected operation of the power system are facilitated.
Drawings
FIG. 1: is a flow chart of the method of the present invention.
FIG. 2 is a schematic diagram: the invention provides a flow chart of a method for optimizing penalty parameters and kernel function parameters in a support vector machine by a wolf group optimization algorithm.
FIG. 3: the invention provides a comparison graph of regression prediction data and original data obtained by predicting wind power by a combined model.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention. In addition, the technical features involved in the embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.
In order to verify the effectiveness of the model and the method provided by the invention, the method is applied to a Germany wind power plant for short-term wind power prediction, in order to prove the effectiveness and generalization capability of the model, the data of two time periods are respectively predicted, wind power experimental data of 2016 year 1, month 1 and day 10 and 2016 year 11, month 11 and day 18 are respectively selected for simulation experiments, and the sampling interval between experimental data points is 5 minutes. In the first section of data, the data (2021 data points) of the first 7 days are used as a training set, the data (671 data points) of the last 3 days are used as a test set, and the wind power in the next 3 days is predicted; and the second stage data also adopts the data of the first 7 days as a training set to predict the wind power within 1 day in the future.
FIG. 1 shows a flow chart of the method of the present invention.
The following describes a specific embodiment of the present invention with reference to fig. 1 to 3, and a method for predicting output power of a wind power plant based on multi-model combination optimization specifically includes the following steps:
X i (k)，i∈[1，n]，k∈[1，m]
wherein, X i (k) Representing kth point data in the ith preprocessed meteorological influence factor, wherein n is the type number of the meteorological influence factors, and m is the number of data points in each meteorological factor;
X o (k)，k∈[1，m]
wherein X o (k) Representing the preprocessed wind power data when the kth point data is selected from the preprocessed meteorological influence factor data, wherein m is the number of data points in each meteorological factor;
selecting the kth point data from the preprocessed meteorological influence factor data as follows:
{X 1 (k)，X 2 (k)，...，X n (k)}
calculating a gray correlation coefficient:
i∈[1，n]，k∈[1，m]
wherein, X i (k) Represents the kth point data in the meteorological influence factor after the ith preprocessing, theta i (k) Representing the gray correlation coefficient, X, of the kth point in the weather influence factor after the ith preprocessing o (k) Representing the preprocessed wind power data when the kth point data is selected from the preprocessed meteorological influence factor data, wherein n is the type number of the meteorological influence factors, m is the number of data points in each meteorological factor, rho is a resolution coefficient,
represents the two-stage minimum difference between the wind power and the i-th weather influence factor>
Representing the two-stage maximum difference between the wind power and the ith meteorological influence factor;
γ i ∈[0，1]
wherein, gamma is i Correlation degree of ith meteorological influence factor and wind power, gamma i The smaller the error between the power value and 1 is, the higher the correlation degree between the ith meteorological influence factor and the wind power is;
X j (k)，j∈[1，c]，k∈[1，m]，c＜n
wherein X j (k) Representing kth point data in the meteorological influence factors after the jth characteristic selection, wherein c is the type number of the meteorological influence factors after the characteristic selection, and m is the number of data points in each meteorological factor;
simultaneously, a relaxation factor xi is introduced to avoid the relaxation factor xi i If the size is too large, adding a penalty factor C, and supporting the target function expression of a vector machine prediction model as follows:
wherein, C represents the punishment parameter of the support vector machine as the parameter to be optimized, xi i Represents the relaxation factor, x, of the ith sample data i Representing input sample data, y i Representing a label, b representing a displacement term, determining the distance between the hyperplane and the origin, ω representing a normal vector, determining the direction of the hyperplane,
kernel function K (x) of support vector machine p ，x q ) The method is characterized by comprising the following steps of (1) selecting a Gaussian Radial Basis Function (RBF) as follows:
x p ＝[X 1 (p) X 2 (p) ... X j (p) ... X c (p)]
x q ＝[X 1 (q) X 2 (q) ... X j (q) ... X c (q)]
wherein x is p Denotes the p-th sample data, x q It means that the q-th sample data,
representing the Euclidean distance between two points of the square of the distance between the two points, σ being the width of the function, g =1/2 σ 2 Representing the kernel function parameter of the support vector machine as a parameter to be optimized;
preferably, the preset threshold is 0.9;
and 3, respectively optimizing the punishment parameters of the support vector machine and the kernel function parameters of the support vector machine through a grey wolf colony optimization algorithm to obtain the punishment parameters C of the optimized support vector machine * And kernel function parameter g of optimized support vector machine * ；
As shown in fig. 2, in step 3, the penalty parameter of the support vector machine and the kernel function parameter of the support vector machine are respectively optimized by the gray wolf group optimization algorithm as follows:
step 3.1, initializing algorithm parameters of a graying wolf colony optimization algorithm: the population scale N, the maximum iteration times maxgen, the penalty parameter C, the value range of the kernel function parameter g and the position of the wolf individual;
step 3.2, randomly generating a population of N wolf individuals according to the population scale N, wherein the positions of the N wolf individuals D are both composed of a penalty parameter C and a kernel function parameter g, and random initialization is carried out according to the value ranges of the penalty parameter C and the kernel function parameter g;
the positions of the N grey wolf individuals are as follows in sequence: (C) b ，g b )，b∈[1，N]；
Step 3.3, taking the mean square error MSE as a fitness function, dividing the population of N wolf individuals into four different levels of alpha, beta, delta and omega, and respectively updating the positions of the wolfs according to the fitness function value and the fitness function value relations of the alpha wolf, the beta wolf, the delta wolf and the omega wolf;
predicting the meteorological influence factors after the feature selection through the SVM model established in the step 2 to obtain a wind power predicted value
The fitness function is the sum of the mean square errors of the predicted value of the wind power and the actual value of the wind power, and specifically comprises the following steps:
k∈[1，S] S＜m
wherein, X o (k) Selecting the actual value of the wind power when the k point data is selected for the meteorological influence factor data after each characteristic is selected,
when the kth point data is selected for the meteorological influence factor data after the features are selected, the wind power predicted value predicted by a support vector machine prediction model is obtained, S is the number of wind power data points in a training set, and m is the number of the wind power data points;
and 3.4, updating the direction and the position of the omega wolf searching individual in the grey wolf group according to a formula, wherein the specific formula is as follows:
D s (t)＝|C u X s (t)-X u (t)|u＝1，2，3，s＝α，β，δ
X u (t+1)＝X s (t)-A u |C u X s (t)-X u (t)|u＝1，2，3，s＝α，β，δ
where t represents the current number of iterations, D s (t) represents the distance between the alpha-level wolf, the beta-level wolf and the delta-level wolf and the omega-level wolf respectively in the t iteration, X s (t) represents the current position of the alpha-level wolf, the beta-level wolf, the delta-level wolf, X u (t) represents the current position of the omega level wolf, X u (t + 1) represents the alpha-level wolf, the beta-level wolf, the delta-level wolf determines the position of the next movement of the omega-level wolf, and X (t + 1) represents the position of the next moment of the omega-level wolf;
step 3.5, parameters alpha, A and C are updated according to a formula, a new position of the alpha-level wolf, a new position of the beta-level wolf and a new position of the delta-level wolf are determined according to the updated optimal fitness function value, and the specific formula is as follows:
A＝α*(2r 1 -1)
C＝2r 2
wherein, denotes as matrix product, A, C is the first and second cooperative coefficient respectively, a is reduced from 2 to 0 in whole iteration process, r1 and r2 are the first and second random vector between [0,1 ];
and 3.6, if the maximum iteration times maxgen are exceeded, terminating the algorithm, outputting the global optimal positions in the N wolf group individuals, namely the penalty parameter C of the optimized support vector machine * And kernel function parameter g of optimized support vector machine * Otherwise, skipping to the step 3.4 to continue optimization;
wherein i is E [1,c]，w zi Weight of the kth sample representing the z-th weak regressor, e zi Expressed as the relative error of the kth sample of the z-th weak regressor, X o (k) Selecting the actual value of wind power F when the k point data is selected for the meteorological influence factor data after each characteristic is selected z (X i (k) Is the predicted value of the wind power of the z-th weak regressor at the k-th data point, E z The maximum error of the z-th weak regressor on the training set;
and 4, calculating error rates of a plurality of weak regressors by combining the relative errors, wherein the error rates are as follows:
according to the relative error, the error rate e of the z-th weak regressor can be obtained kz The method comprises the following steps:
determining weight coefficients alpha of a weak regressor z The method comprises the following steps:
the sample weights are updated as follows:
wherein, ω is z+1，i Sample weight coefficient, Z, representing the Z +1 th weak regressor a In order to normalize the factors, the method comprises the steps of,
and 4, predicting each sample in the training set by using a strong regressor to obtain a training set wind power predicted value sequence as follows:
the training set is the meteorological influence factor after the feature selection in the step 2;
training a wind power predicted value sequence, which is specifically as follows:
the sequence of the predicted values of the wind power is tested as follows:
wherein S represents the number of wind power data points in the training set, m represents the number of wind power data points,
represents a sequence of k wind power prediction values in the training set, in combination>
Representing k wind power predicted value sequences, X, in a test set f (k) Represents the k wind power predicted value, X, in the training set g (k) Representing the k wind power predicted value in the test set，/>
calculating a training set predicted power value
Historical wind power value X of wind power station o (k) The difference in (c) is denoted as error sequence E, as follows:
And 5, constructing a gating cycle unit network:
constructing a gated circulation unit network through an input layer, a plurality of gated circulation unit layers and an output layer;
an updating gate and a resetting gate corresponding to each gating circulation unit layer exist on each gating circulation unit layer;
the update gate determines information that was retained to the current time at a previous time, and the reset gate determines the extent to which information was discarded;
the updating gate and the placing gate are sequentially defined as follows:
Z t ＝σ(W Z X t +U Z h t-1 +b Z )
r t ＝σ(W r X t +U r h t-1 +b r )
wherein z is t To refresh the door, r t To reset the gate, X t Input representing the current time, W Z And U Z Respectively, updating the weight matrix of the gate, W r And U r Weight matrices, b, of reset gates, respectively Z And b r Respectively are offset vectors, and sigma is a sigmoid activation function;
the hyper-parameters of the gated loop unit include: the number of neural network layers, the number of neurons in each layer and the dropout rate; wherein the dropout rate is a proportion of the number of discarded neurons and is used for inhibiting an overfitting phenomenon of the model, and the dropout belongs to (0,1);
by setting the value range of the parameters to be optimized simultaneously, finding out the corresponding parameters, namely the optimal number of neural network layers, the optimal number of neurons in each layer and the optimal dropout rate of the GRU model according to the minimum principle of a loss function (MAE) on a training set;
k∈[1，S] S＜m
wherein, X o (k) Selecting the actual value of the wind power when the k point data is selected for the meteorological influence factor data after each characteristic is selected,
selecting a wind power predicted value predicted by the gate control cycle unit network when the kth point data is selected for the meteorological influence factor data after each feature is selected;
and 5, optimizing the parameters of the gated loop unit network as follows: optimizing and solving the gate control cycle unit network through an Adam algorithm to obtain a weight matrix W of the updated gate after optimization Z 、U* Z Weight matrix W of reset gate after optimization r 、U* r Updating bias vector b of gate after optimization Z Bias vector b of reset gate after optimization r ；
After parameter optimization is carried out on the training set, prediction is carried out on the test set to obtain a predicted value sequence of the test set errors, and the predicted value sequence is recorded as
The following were used:
and 6, superposing the corrected prediction error value and the prediction result of the AdaBoost integrated prediction model to obtain a final prediction result, wherein the method specifically comprises the following steps:
wherein X' (k) is the final prediction result of k wind power in the test set,
for the predictor sequence of the test set error, ->
And m-S is the number of samples of the test set.
Fig. 3 is a comparison diagram of regression prediction data obtained by predicting wind power by applying the combined model of the invention according to the third embodiment of the invention and original data. The black curve in the graph is a prediction curve graph of the method, the blue curve is a real value of the wind power, and the other color curves are prediction curves of the comparison model.
The feasibility and the accuracy of the method are verified by carrying out comparison research by totally adopting 5 models, wherein the model 1 is a support vector machine (SVR), the model 2 is a gray relevance analysis-support vector machine (GRA-SVR), the model 3 is a gray relevance analysis-Grey wolf group optimization algorithm-support vector machine (GRA-GWO-SVR), the model 4 is a gray relevance analysis-Grey wolf group optimization algorithm-support vector machine integration model (GRA-GWO-SVR-AdaBoost), and the model 5 is a gray relevance analysis-Grey wolf group optimization algorithm-support vector machine integration-gating cycle unit (GRA-GWO-SVR-AdaBoost-GRU).
Specific evaluation indexes of the 5 models are shown in table 1.
TABLE 1 evaluation indexes of 5 different prediction models in two different time periods
As can be seen from Table 1, the prediction accuracy of the model and method provided by the invention is very high, the root mean Square error RMSE is about 0.1, the average absolute error MAE is about 0.07, and the R-Square exceeds 0.995. This demonstrates the effectiveness and superiority of the proposed model and method of the present invention.
In conclusion, the invention provides the wind power station output power prediction method based on the combined model of GRA-SVM-GWO-AdaBoost-GRU, so that the accurate prediction of the short-term output power of the wind power station is realized, and the scientific scheduling of a power system and the safety and stability of grid-connected operation are facilitated.
The above-mentioned embodiments only express the embodiments of the present invention, and the description thereof is more specific and detailed, but not construed as limiting the scope of the present invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the inventive concept, which falls within the scope of the present invention. Therefore, the protection scope of the present patent should be subject to the appended claims.
Claims (7)
1. A wind power station output power prediction method based on multi-model combination optimization is characterized by comprising the following steps:
step 1, acquiring meteorological influence factor data and wind power data of a target wind power station, sequentially supplementing the meteorological influence factor data and the wind power data through data missing values and a data normalization method to obtain preprocessed meteorological influence factor data and preprocessed wind power data, taking the preprocessed meteorological influence factor data as a comparison array, taking the preprocessed wind power data as a reference array, and obtaining the association degree of each meteorological influence factor and the wind power through a grey association degree analysis method;
step 2, sorting the relevance degrees of the meteorological influence factors from large to small, taking the meteorological influence factors with the relevance degrees larger than a preset threshold value as weather influence factors after feature selection, taking the meteorological influence factors after feature selection as a training set, and further establishing a support vector machine prediction model;
step 3, respectively optimizing the punishment parameters of the support vector machine and the kernel function parameters of the support vector machine through a grey wolf colony optimization algorithm to obtain the punishment parameters of the optimized support vector machine and the kernel function parameters of the optimized support vector machine;
step 4, dividing the meteorological influence factors after feature selection into a training set and a test set according to a certain proportion, taking an optimized support vector machine model as a weak regressor, initializing sample weights, obtaining a plurality of weak regressors, calculating the relative error of each sample in the training set, calculating the error rates of the plurality of weak regressors by combining the relative errors, further calculating the weight coefficients of the weak regressors to update the sample weights, establishing an AdaBoost integration model to construct a strong regressor, optimizing the number, the learning rate and the loss function model of the weak regressors by a grid search method, predicting each sample in the training set by the strong regressor, obtaining a predicted value sequence of the wind power in the training set, and further calculating an error sequence;
step 5, constructing a gated circulation unit network, predicting an error sequence E through the gated circulation unit network, taking the meteorological factor with the selected characteristics as input data of the gated circulation unit network during error correction, and optimizing parameters of the gated circulation unit network according to a loss function model of the gated circulation unit network to obtain the optimized gated circulation unit network;
and 6, overlapping the corrected prediction error value with the prediction result of the AdaBoost integrated prediction model to obtain a final prediction result.
2. The multi-model combinatorial optimization-based wind power plant output power prediction method of claim 1, characterized in that:
step 1, the preprocessed meteorological influence factor data are as follows:
X i (k)，i∈[1，n]，k∈[1，m]
wherein X i (k) Representing kth point data in the ith preprocessed meteorological influence factor, wherein n is the type number of the meteorological influence factors, and m is the number of data points in each meteorological factor;
step 1, the preprocessed wind power data are as follows:
X o (k)，k∈[1，m]
wherein, X o (k) Representing the preprocessed wind power data when the kth point data is selected from the preprocessed meteorological influence factor data, wherein m is the number of data points in each meteorological factor;
selecting the kth point data from the preprocessed meteorological influence factor data as follows:
{X 1 (k)，X 2 (k)，...，X n (k)}
step 1, obtaining the correlation of each weather influence factor by a grey correlation analysis method, specifically:
calculating a grey correlation coefficient:
i∈[1，n]，k∈[1，m]
wherein, X i (k) Represents the kth point data in the meteorological influence factor after the ith preprocessing, theta i (k) Representing the k point gray correlation coefficient, X, in the meteorological influence factor after the i type of pretreatment o (k) Representing the preprocessed wind power data when the kth point data is selected from the preprocessed meteorological influence factor data, wherein n is the type number of the meteorological influence factors, m is the number of data points in each meteorological factor, rho is a resolution coefficient,
representing the two-stage minimum difference between the wind power and the ith meteorological influence factor,
representing the two-stage maximum difference between the wind power and the ith meteorological influence factor;
step 1, the correlation degree of each meteorological influence factor is as follows:
γ i ∈[0，1]
wherein, γ i Correlation degree of the ith meteorological influence factor and wind power, gamma i The smaller the error between 1, the higher the correlation degree of the ith meteorological influence factor and the wind power.
3. The wind power plant output power prediction method based on multi-model combinatorial optimization according to claim 1, characterized in that:
step 2, the meteorological influence factor data after the feature selection is as follows:
X j (k)，j∈[1，c]，k∈[1，m]，c＜n
wherein, X j (k) Representing kth point data in the meteorological influence factors after the jth characteristic selection, wherein c is the type number of the meteorological influence factors after the characteristic selection, and m is the number of data points in each meteorological factor;
at the same time, a relaxation factor xi is introduced to avoid the relaxation factor xi i If the size is too large, adding a penalty factor C, and supporting the target function expression of a vector machine prediction model as follows:
wherein, C represents the punishment parameter of the support vector machine as the to-be-searchedOptimization parameter xi i Represents the relaxation factor, x, of the ith sample data i Representing input sample data, y i Representing a label, b representing a displacement term, determining the distance between the hyperplane and the origin, ω representing a normal vector, determining the direction of the hyperplane,
kernel function K (x) of support vector machine p ，x q ) The method is characterized by comprising the following steps of (1) selecting a Gaussian radial basis kernel function (RBF) as follows:
x p ＝[X 1 (p)X 2 (p)...X j (p)...X c (p)]
x q ＝[X 1 (q)X 2 (q)...X j (q)...X c (q)]
wherein x is p Denotes the p-th sample data, x q It is indicated that the q-th sample data,
representing the Euclidean distance between two points of the square of the distance between the two points, σ being the width of the function, g =1/2 σ 2 And representing the kernel function parameters of the support vector machine as the parameters to be optimized.
4. The multi-model combinatorial optimization-based wind power plant output power prediction method of claim 1, characterized in that:
in the step 3, the penalty parameter of the support vector machine and the kernel function parameter of the support vector machine are respectively optimized by a gray wolf group optimization algorithm as follows:
step 3.1, initializing algorithm parameters of the gray wolf colony optimization algorithm: the population scale N, the maximum iteration times maxgen, the penalty parameter C, the value range of the kernel function parameter g and the position of the wolf individual;
step 3.2, randomly generating a population of N wolf individuals according to the population scale N, wherein the positions of the N wolf individuals D are both composed of a penalty parameter C and a kernel function parameter g, and random initialization is carried out according to the value ranges of the penalty parameter C and the kernel function parameter g;
the positions of the N grey wolf individuals are as follows in sequence: (C) b ，g b )，b∈[1，N]；
Step 3.3, taking the mean square error MSE as a fitness function, dividing the population of N wolf individuals into four different levels of alpha, beta, delta and omega, and respectively updating the positions of the wolfs according to the fitness function value and the fitness function value relations of the alpha wolf, the beta wolf, the delta wolf and the omega wolf;
predicting the meteorological influence factors after the feature selection through the SVM model established in the step 2 to obtain a wind power predicted value
The fitness function is the sum of the mean square errors of the predicted value of the wind power and the actual value of the wind power, and specifically comprises the following steps:
k∈[1，S]S＜m
wherein, X o (k) Selecting the actual value of the wind power when the kth point data is selected for the meteorological influence factor data after each characteristic is selected,
when the kth point data is selected for the meteorological influence factor data after the features are selected, the wind power predicted value predicted by a support vector machine prediction model is obtained, S is the number of wind power data points in a training set, and m is the number of the wind power data points;
and 3.4, updating the direction and the position of the omega wolf searching individual in the grey wolf group according to a formula, wherein the specific formula is as follows:
D s (t)＝|C u X s (t)-X u (t)|u＝1，2，3，s＝α，β，δ
X u (t+1)＝X s (t)-A u |C u X s (t)-X u (t)|u＝1，2，3，s＝α，β，δ
where t represents the current number of iterations, D s (t) represents the distance between the alpha-level wolf, the beta-level wolf and the delta-level wolf and the omega-level wolf respectively in the t-th iteration, X s (t) represents the current position of the alpha-level wolf, the beta-level wolf, the delta-level wolf, X u (t) represents the current position of the omega level wolf, X u (t + 1) represents the alpha-level wolf, the beta-level wolf, the delta-level wolf determines the position of the next movement of the omega-level wolf, and X (t + 1) represents the position of the next moment of the omega-level wolf;
step 3.5, parameters alpha, A and C are updated according to a formula, a new position of the alpha-level wolf, a new position of the beta-level wolf and a new position of the delta-level wolf are determined according to the updated optimal fitness function value, and the specific formula is as follows:
A＝α*(2r 1 -1)
C＝2r 2
wherein, denoted as matrix product, A, C is the first cooperative coefficient and the second cooperative coefficient respectively, a is linearly decreased from 2 to 0 in the whole iteration process, r1 and r2 are the first random vector and the second random variable between [0,1 ];
and 3.6, if the maximum iteration times maxgen are exceeded, terminating the algorithm, outputting the global optimal positions in the N wolf group individuals, namely the penalty parameter C of the optimized support vector machine * And kernel function parameter g of optimized support vector machine * And otherwise, skipping to step 3.4 to continue optimization.
5. The multi-model combinatorial optimization-based wind power plant output power prediction method of claim 1, characterized in that:
step 4, the weight of the initialization sample is w zi ；
Step 4, the plurality of weak regressors are z weak regressors F z (x)；
Step 4, calculating the relative error of each sample in the training set, specifically as follows:
wherein i is E [1,c]，w zi Weight of the kth sample representing the z-th weak regressor, e zi Expressed as the relative error of the kth sample of the z-th weak regressor, X o (k) Selecting the actual value of wind power F when the k point data is selected for the meteorological influence factor data after each characteristic is selected z (X i (k) Is the predicted value of the wind power at the kth data point for the z-th weak regressor, E z The maximum error of the z-th weak regressor on the training set;
and 4, calculating error rates of a plurality of weak regressors by combining the relative errors, wherein the error rates are as follows:
according to the relative error, the error rate e of the z-th weak regressor can be obtained kz The method comprises the following steps:
step 4, further calculating the weight coefficient of the weak regressor to update the sample weight as follows:
determining weight coefficients alpha of a weak regressor z The method comprises the following steps:
the sample weights are updated as follows:
wherein, ω is z+1，i Sample weight coefficient, Z, representing the Z +1 th weak regressor a In order to normalize the factors, the method comprises the steps of,
step 4, establishing an AdaBoost integration model to construct a strong regressor: superposing the product of the weight of each weak regression and the predicted value to obtain the final strong regression;
and 4, predicting each sample in the training set by using a strong regressor to obtain a training set wind power predicted value sequence as follows:
the training set is the meteorological influence factor after the feature selection in the step 2;
training a wind power predicted value sequence, which is specifically as follows:
the sequence of the predicted values of the wind power in the test set is as follows:
wherein S represents the number of wind power data points in the training set, m represents the number of wind power data points,
represents a sequence of k wind power prediction values in the training set, in combination>
Representing k wind power predicted value sequences, X, in a test set f (k) Represents the k-th wind power predicted value, X, in the training set g (k) Representing the k wind power predicted value in the test set,
step 4, the calculation error sequence is as follows:
calculating a training set predicted power value
Historical wind power value X of wind power station o (k) The difference of (d), denoted as error sequence E, is as follows:
6. the multi-model combinatorial optimization-based wind power plant output power prediction method of claim 1, characterized in that:
step 5, constructing the gating cycle unit network comprises the following steps:
constructing a gated cyclic unit network through an input layer, a plurality of gated cyclic unit layers and an output layer;
an updating gate and a resetting gate corresponding to each gating circulation unit layer exist on each gating circulation unit layer;
the update gate determines information that was retained to the current time from a previous time, and the reset gate determines the extent to which the information is discarded;
the updating gate and the placing gate are sequentially defined as follows:
z t ＝σ(W Z X t +U Z h t-1 +b Z )
r t ＝σ(W r X t +U r h t-1 +b r )
wherein z is t To refresh the door, r t To reset the gate, X t Input representing the current time, W Z And U Z Respectively, updating the weight matrix of the gate, W r And U r Weight matrices, b, of reset gates, respectively Z And b r Respectively are offset vectors, and sigma is a sigmoid activation function;
the hyper-parameters of the gated loop unit include: the number of neural network layers, the number of neurons in each layer and the dropout rate; wherein, the dropout rate is a proportion of the number of discarded neurons and is used for inhibiting the overfitting phenomenon of the model, and dropout belongs to (0,1);
by setting the value range of the parameters to be optimized simultaneously, finding out the corresponding parameters, namely the optimal number of neural network layers, the optimal number of neurons in each layer and the optimal dropout rate of the GRU model according to the minimum principle of a loss function (MAE) on a training set;
step 5, the loss function model of the gated cyclic unit network is an average absolute error, and the loss function model of the gated cyclic unit network specifically includes the following steps:
k∈[1，S]S＜m
wherein, X o (k) Selecting the actual value of the wind power when the kth point data is selected for the meteorological influence factor data after each characteristic is selected,
selecting a wind power predicted value predicted by the gate control cycle unit network when the kth point data is selected for the meteorological influence factor data after each feature is selected;
and 5, optimizing the parameters of the gated loop unit network as follows: optimizing and solving the gate control cycle unit network through an Adam algorithm to obtain a weight matrix W of the updated gate after optimization Z 、U* Z Weight matrix W of reset gate after optimization r 、U* r Updating bias vector b of gate after optimization Z Bias vector b of reset gate after optimization r ；
After parameter optimization is carried out on the training set, prediction is carried out on the test set to obtain a predicted value sequence of the test set errors, and the predicted value sequence is recorded as
The following were used:
7. the multi-model combinatorial optimization-based wind power plant output power prediction method of claim 1, characterized in that:
and 6, the final prediction result is as follows:
wherein X' (k) is the final prediction result of k wind power in the test set,
to test the sequence of predicted values of the set error,
and m-S is the number of samples of the test set. />
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010405975.8A CN111667098B (en) | 2020-05-14 | 2020-05-14 | Wind power station output power prediction method based on multi-model combination optimization |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010405975.8A CN111667098B (en) | 2020-05-14 | 2020-05-14 | Wind power station output power prediction method based on multi-model combination optimization |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111667098A CN111667098A (en) | 2020-09-15 |
| CN111667098B true CN111667098B (en) | 2023-04-18 |
Family
ID=72383603
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010405975.8A Active CN111667098B (en) | 2020-05-14 | 2020-05-14 | Wind power station output power prediction method based on multi-model combination optimization |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111667098B (en) |
Families Citing this family (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112671043B (en) * | 2021-01-14 | 2023-12-12 | 国网山东省电力公司枣庄供电公司 | Hyperopt-GRU model-based distribution transformer operation optimization method |
| CN113051769B (en) * | 2021-04-09 | 2022-09-30 | 中南大学 | Power curve modeling method based on asymmetric loss and hybrid intelligent optimization algorithm |
| CN113033917B (en) * | 2021-04-19 | 2022-04-12 | 重庆工商大学 | Sewage treatment plant prediction planning operation management method based on peripheral data |
| CN115204532A (en) * | 2022-09-16 | 2022-10-18 | 中国地质大学（北京） | Oil-gas yield prediction method and system based on multivariable error correction combined model |
| CN115204533A (en) * | 2022-09-16 | 2022-10-18 | 中国地质大学（北京） | Oil-gas yield prediction method and system based on multivariable weighted combination model |
| CN116231646B (en) * | 2023-05-09 | 2023-07-07 | 湖北工业大学 | PMU optimal configuration method and system based on electric power system weakness and economy |
| CN116415739A (en) * | 2023-05-29 | 2023-07-11 | 广东电网有限责任公司湛江供电局 | Photovoltaic power short-term prediction method and related device |
| CN116777243B (en) * | 2023-06-21 | 2024-02-09 | 中国联合网络通信有限公司深圳市分公司 | Resident trip index evaluation method and device and computer readable storage medium |
| CN117933298A (en) * | 2024-03-20 | 2024-04-26 | 湖南科技大学 | Wind speed reconstruction method, wind speed reconstruction device, electronic equipment and storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106374534A (en) * | 2016-11-17 | 2017-02-01 | 云南电网有限责任公司玉溪供电局 | Multi-target grey wolf optimization algorithm-based large scale household energy management method |
| CN109214449A (en) * | 2018-08-28 | 2019-01-15 | 华北电力大学 | A kind of electric grid investment needing forecasting method |
| CN109993270A (en) * | 2019-03-27 | 2019-07-09 | 东北大学 | Lithium ion battery residual life prediction technique based on grey wolf pack optimization LSTM network |
| CN110414734A (en) * | 2019-07-29 | 2019-11-05 | 河北工业大学 | A method of meter and the assessment of wind-resources usage forecast |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7991512B2 (en) * | 2007-08-28 | 2011-08-02 | General Electric Company | Hybrid robust predictive optimization method of power system dispatch |
-
2020
- 2020-05-14 CN CN202010405975.8A patent/CN111667098B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106374534A (en) * | 2016-11-17 | 2017-02-01 | 云南电网有限责任公司玉溪供电局 | Multi-target grey wolf optimization algorithm-based large scale household energy management method |
| CN109214449A (en) * | 2018-08-28 | 2019-01-15 | 华北电力大学 | A kind of electric grid investment needing forecasting method |
| CN109993270A (en) * | 2019-03-27 | 2019-07-09 | 东北大学 | Lithium ion battery residual life prediction technique based on grey wolf pack optimization LSTM network |
| CN110414734A (en) * | 2019-07-29 | 2019-11-05 | 河北工业大学 | A method of meter and the assessment of wind-resources usage forecast |
Non-Patent Citations (1)
| Title |
|---|
| 石峰 ; 楼文高 ; 张博 ; .基于灰狼群智能最优化的神经网络PM_(2.5)浓度预测.计算机应用.2017,(第10期), * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111667098A (en) | 2020-09-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111667098B (en) | Wind power station output power prediction method based on multi-model combination optimization | |
| CN109063911B (en) | Load aggregation grouping prediction method based on gated cycle unit network | |
| CN108549929B (en) | A kind of photovoltaic power prediction technique based on deep layer convolutional neural networks | |
| CN115293415A (en) | Multi-wind-farm short-term power prediction method considering time evolution and space correlation | |
| CN111860982A (en) | Wind power plant short-term wind power prediction method based on VMD-FCM-GRU | |
| CN108596327B (en) | Seismic velocity spectrum artificial intelligence picking method based on deep learning | |
| Liu et al. | Heating load forecasting for combined heat and power plants via strand-based LSTM | |
| CN110059867A (en) | A kind of wind speed forecasting method of SWLSTM combination GPR | |
| CN114462718A (en) | CNN-GRU wind power prediction method based on time sliding window | |
| CN111898825A (en) | Photovoltaic power generation power short-term prediction method and device | |
| CN116628510A (en) | Self-training iterative artificial intelligent model training method | |
| CN115759415A (en) | Power consumption demand prediction method based on LSTM-SVR | |
| Eze et al. | Enhanced recurrent neural network for short-term wind farm power output prediction | |
| CN111222689A (en) | LSTM load prediction method, medium, and electronic device based on multi-scale temporal features | |
| CN116014722A (en) | Sub-solar photovoltaic power generation prediction method and system based on seasonal decomposition and convolution network | |
| Schreiber et al. | Model selection, adaptation, and combination for transfer learning in wind and photovoltaic power forecasts | |
| CN113393119B (en) | Stepped hydropower short-term scheduling decision method based on scene reduction-deep learning | |
| Aribowo | Optimizing feed forward backpropagation neural network based on teaching-learning-based optimization algorithm for long-term electricity forecasting | |
| CN114219126A (en) | Method for predicting power supply load of small hydropower station permeable regional network based on residual correction | |
| CN115392387B (en) | Low-voltage distributed photovoltaic power generation output prediction method | |
| Lops et al. | A deep learning approach for climate parameter estimations and renewable energy sources | |
| CN116227677A (en) | Power prediction correction method considering photovoltaic power climbing characteristics | |
| CN115759343A (en) | E-LSTM-based user electric quantity prediction method and device | |
| Wilms et al. | On the necessity of exogenous variables for load, pv and wind day-ahead forecasts using recurrent neural networks | |
| CN115293406A (en) | Photovoltaic power generation power prediction method based on Catboost and Radam-LSTM |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |