CN113837499A - Ultra-short-term wind power prediction method and system - Google Patents
Ultra-short-term wind power prediction method and system Download PDFInfo
- Publication number
- CN113837499A CN113837499A CN202111398570.7A CN202111398570A CN113837499A CN 113837499 A CN113837499 A CN 113837499A CN 202111398570 A CN202111398570 A CN 202111398570A CN 113837499 A CN113837499 A CN 113837499A
- Authority
- CN
- China
- Prior art keywords
- wind power
- data
- data set
- gru
- cnn
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 49
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 49
- 238000012545 processing Methods 0.000 claims abstract description 41
- 238000010606 normalization Methods 0.000 claims abstract description 37
- 238000013528 artificial neural network Methods 0.000 claims abstract description 24
- 238000012549 training Methods 0.000 claims abstract description 18
- 238000012216 screening Methods 0.000 claims abstract description 17
- 239000011159 matrix material Substances 0.000 claims description 54
- 239000013598 vector Substances 0.000 claims description 16
- 239000000284 extract Substances 0.000 claims description 10
- 230000004913 activation Effects 0.000 claims description 9
- 238000000605 extraction Methods 0.000 claims description 8
- 230000008569 process Effects 0.000 claims description 7
- 238000004458 analytical method Methods 0.000 claims description 6
- 230000002596 correlated effect Effects 0.000 claims description 6
- NAWXUBYGYWOOIX-SFHVURJKSA-N (2s)-2-[[4-[2-(2,4-diaminoquinazolin-6-yl)ethyl]benzoyl]amino]-4-methylidenepentanedioic acid Chemical compound C1=CC2=NC(N)=NC(N)=C2C=C1CCC1=CC=C(C(=O)N[C@@H](CC(=C)C(O)=O)C(O)=O)C=C1 NAWXUBYGYWOOIX-SFHVURJKSA-N 0.000 claims description 4
- 238000004364 calculation method Methods 0.000 claims description 4
- 230000003213 activating effect Effects 0.000 claims description 3
- 238000006243 chemical reaction Methods 0.000 claims description 3
- 238000011156 evaluation Methods 0.000 claims description 3
- 230000014759 maintenance of location Effects 0.000 claims description 3
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Chemical class O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 3
- 230000001276 controlling effect Effects 0.000 claims description 2
- 238000010248 power generation Methods 0.000 abstract description 3
- 230000008901 benefit Effects 0.000 description 4
- 238000000053 physical method Methods 0.000 description 4
- 238000007781 pre-processing Methods 0.000 description 4
- 238000007619 statistical method Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- VNWKTOKETHGBQD-UHFFFAOYSA-N methane Chemical compound C VNWKTOKETHGBQD-UHFFFAOYSA-N 0.000 description 2
- 238000011176 pooling Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 230000002411 adverse Effects 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000003245 coal Substances 0.000 description 1
- 238000002485 combustion reaction Methods 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 230000000875 corresponding effect Effects 0.000 description 1
- 125000004122 cyclic group Chemical group 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 239000006185 dispersion Substances 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 239000003345 natural gas Substances 0.000 description 1
- 239000003208 petroleum Substances 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- 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
- G06N3/045—Combinations of networks
-
- 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
- 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)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Economics (AREA)
- Software Systems (AREA)
- Biomedical Technology (AREA)
- Data Mining & Analysis (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Biophysics (AREA)
- Strategic Management (AREA)
- Artificial Intelligence (AREA)
- Human Resources & Organizations (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- General Business, Economics & Management (AREA)
- Marketing (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 relates to the technical field of wind power generation power prediction, and discloses an ultra-short-term wind power prediction method and system, which comprise the following steps: analyzing the type of the required data, and acquiring a wind power data set; filling missing values of the acquired wind power data set; screening the optimal meteorological factors and meteorological factor and historical power combination; selecting a time window length meeting a preset requirement according to the wind power data set; carrying out normalization processing on the wind power data set, and slicing the data set according to the length of the selected time window; building a parallel model of CNN and GRU neural networks, setting parameters, and training the model by using a data set after slice processing; performing inverse normalization processing on a prediction result obtained by the model; and adjusting the anti-normalization processing result according to preset upper and lower thresholds of the wind power. The method improves the wind power prediction precision, and has application value for stable operation of the power system and wind power scheduling.
Description
Technical Field
The invention relates to the technical field of wind power generation power prediction, in particular to an ultra-short-term wind power prediction method and system.
Background
The total demand of world energy is increasing continuously, the consumption rate of fossil energy such as coal, natural gas and petroleum is increasing day by day, and the combustion products of the fossil energy also harm the global ecological environment. Therefore, the great development and use of green and clean renewable energy sources are of great significance. Among them, wind energy is widely concerned by countries in the world due to its wide distribution and low cost, and is one of the most potential renewable energy sources. However, wind power generation has strong randomness, volatility and intermittency, so that large-scale wind power integration brings severe challenges to safe, stable and economic operation and scheduling of a power system, and simultaneously, problems can be brought to operation, maintenance and the like of a wind power plant. The wind power prediction technology is one of effective means for relieving the adverse effects, and provides technical support for safe, economic and stable operation of a power system and a wind power plant.
At present, wind power prediction methods are mainly divided into physical methods and statistical methods. The physical method mainly utilizes information such as physical factors, numerical weather forecast and the like to predict the wind power; the statistical method mainly excavates the relation among historical data of the wind power plant to predict the wind speed and the wind power. The data required in the physical method is difficult to obtain, higher meteorological bases are required, and the calculation is complex and complicated, so that the method is not suitable for ultra-short-term prediction. Compared with a physical method, the statistical method has the advantages of easily obtained data and relatively simple calculation, and is suitable for ultra-short-term wind power prediction. However, the existing statistical method for ultra-short-term wind power prediction based on parallel CNN-GRUs cannot effectively extract effective characteristic information of a complex data set and cannot capture time sequence characteristics of the wind power data set.
In view of this, we propose an ultra-short-term wind power prediction method.
Disclosure of Invention
The ultra-short-term wind power prediction method comprises the steps that meteorological factors with high correlation with wind power and corresponding historical wind power are screened through Pearson correlation coefficients to serve as input, data dimensionality is reduced, and the method is beneficial to efficient and accurate prediction of a model; meanwhile, effective time sequence characteristics in the original data set are extracted through a CNN network, time sequence dependency relations in the data set are captured through a GRU network, the characteristics obtained through the CNN network and the GRU network are fused, and a wind power predicted value is obtained through full-connection layer analysis.
In order to solve the technical problems, the technical scheme of the invention is as follows:
an ultra-short-term wind power prediction method comprises the following steps:
s1, analyzing the type of the required data, and acquiring a wind power data set;
s2, filling missing values of the acquired wind power data set;
s3, screening the optimal meteorological factor and historical power combination from the wind power data set by using a Pearson correlation coefficient;
s4, selecting a time window length meeting the preset requirement according to the wind power data set;
s5, carrying out normalization processing on the wind power data set, and slicing the data set according to the length of the selected time window;
s6, building a parallel model of the CNN and GRU neural networks, setting parameters, and training the model by using the data set after slice processing;
s7, performing inverse normalization processing on the prediction result obtained by the model;
and S8, adjusting the result of the inverse normalization processing according to the preset wind power upper and lower thresholds in the wind power plant.
Preferably, the type of the required data comprises wind speed, atmospheric density, wind direction, temperature, air pressure and wind power; wherein, the wind power refers to the active power output by the fan,
the active power output by the fan is calculated according to the following formula (1) and formula (2):
wherein,Pactive power output by a fan is kW;C p the wind energy conversion coefficient;Aswept area of the wind wheel in m2；ρIn terms of atmospheric density, in kg/m3；vIs the wind speed, with the unit of m/s:
wherein,ρis at atmospheric density;P a is atmospheric pressure in kPa;Tthermodynamic temperature, in K;P b is the saturated water vapor pressure in kPa;
is the relative humidity of air in%.
The above formula shows that the wind power has direct relation with the wind speed, the atmospheric density and the swept area of the wind wheel; wherein the swept area of the wind wheel is influenced by the wind direction; the atmospheric density is affected by temperature and air pressure, so the selected data set needs to contain 6-dimensional data of wind speed, atmospheric density, wind direction, temperature, air pressure and wind power.
Preferably, the filling of the missing data in the acquired wind power data set includes:
filling by using an undeleted antecedent value;
or padding with an unneeded later value;
or based on the wind power plant data and a preset time scale, filling by taking a missing value as a center and taking an average value of the un-missing values in the preset time scale;
missing data is padded using equation (17):
wherein,x 2 representing padding data;x 0 representing the original data without missing value padding;fillna(. to) represents the padding function used;methodrepresenting a filling method;padindicates padding with an un-missing antecedent;bfillindicating padding with an unneeded late value;
Preferably, the meteorological factors are screened, including: by obtaining of formula (3)xVariable sumyThe correlation coefficient of the variable is screened based on the magnitude of the correlation coefficient;
wherein,xvariable sumyThe variables are two variables in meteorological factors;
are respectively asxVariables of、yThe average value of the variables;r xy is a correlation coefficient;inumbering the predicted points;x i 、y i are respectively a predicted pointiCorresponding toxThe variables,yA variable;npredicting the number of data;
correlation coefficientr xy The larger the size, the correspondingxVariable sumyThe more strongly correlated these 2 factors are the variables, the more strongly correlated the factors are screened using equation (18):
wherein,if (. cndot.) represents a judgment function.
The correlation between each meteorological factor and power in different wind power plants is different, and the optimal combination of the meteorological factors and the wind power in the wind power plant can be obtained by screening the meteorological factors according to the specific wind power plant condition.
When the time window length is too long, the prediction accuracy of the model may be reduced due to capturing misleading time sequence characteristics (when wind power data suddenly changes), or when the time window length is too short, the prediction accuracy is not high due to the lack of necessary time sequence characteristics of the model, so that when the time window length meeting the preset requirement is selected, the appropriate time window length needs to be selected according to specific wind power plant data. Preferably, the normalization processing and slicing of the wind power data set include:
firstly, carrying out normalization processing on data by using a formula (4);
wherein,x′is the value after normalization;xis an actual value;x max 、x min respectively a maximum value and a minimum value in each dimension data;
then, slicing the result data obtained by the normalization processing by using an equation (14), and converting the result data into a time sequence data matrix to obtain input data presenting time sequence as input data of the parallel CNN-GRU model;
wherein,in t to representtA time series data matrix of the time;Wrepresents a time window length;y t to representtHistorical power at the time and meteorological factor matrix.
Preferably, the establishment of the parallel model of the CNN and GRU neural networks comprises the following steps: s6.1, carrying out time sequence feature capture on the wind power data by the CNN neural network through different convolution modes shown as a formula (5) and a formula (6) to obtain a feature value of each layer of neural network:
wherein, y l j is as followslLayer onejA characteristic of the convolution kernel output;f(. h) is an activation function;M i is as followsl-feature set of layer 1 output;y l-1 i is as followsl-1 layer output of the secondiA feature;K l j is as alLayer onejA convolution;b l j is an offset; relu (x 1) Activating a function for Relu;x 1 is an input vector;
s6.2, after the eigenvalue of each layer of neural network is obtained according to the formula (5) and the formula (6), because the matrix obtained by the CNN is a three-dimensional matrix, the parallel model uses the full-connection layer to process the characteristics, and needs to be transmitted into a two-dimensional matrix, the characteristics need to be extended; the invention uses formula (15) to extend the characteristic matrix finally obtained by the CNN network and stores the characteristic matrix inout CNN So as to build a parallel CNN-GRU model;
wherein,out CNN representing a feature matrix of the CNN extraction;ConvId(. h) represents the way the convolutional neural network extracts features;Flatten (. to) represents a feature extension;in t to representtA time series data matrix of the time;
s6.3, controlling the retention degree of information at different moments by the GRU neural network through the reset gate and the update gate to extract the feature matrixout GRU The formulas used for extraction are shown as formula (7), formula (8), formula (9) and formula (10);
wherein,x t the input vector of the current moment is taken as the input vector of the current moment;r t 、z t respectively a reset gate and an update gate;h t-1 、h t 、
respectively representing a hidden layer state at the previous moment, a hidden layer state at the current moment and a candidate state at the current moment;W rx 、W zx 、W hx is prepared by reacting withx t A related weight matrix; W rh 、W zh 、W hh is prepared by reacting withh t-1 A related weight matrix;b r 、b z 、b h is a bias vector; sigma is an activation function sigmoid; an indication of a dot product;
s6.4, using a formula (16) to store the extracted feature matrix of the GRUout GRU So as to build a parallel CNN-GRU model;
wherein,out GRU representing a feature matrix extracted by GRU;GRU (. to) shows the way the gated loop unit extracts features;in t to representtA time series data matrix of the time;
s6.5, splicing the results after the CNN and GRU neural network processing by using a full connection layer to obtain a CNN-GRU parallel model, wherein the formula (11) is as follows:
wherein,y 2representing the output result of the parallel model; Φan activation function representing a fully connected layer;w 1 、w 2representing a weight matrix;brepresenting a bias vector;out CNN representing a feature matrix of the CNN extraction;out GRU representing a feature matrix extracted by GRU;
s6.6, inputting the data set after the slicing processing into the constructed CNN-GRU parallel model for training.
Preferably, the denormalizing process is performed on the prediction result obtained by the model, and includes:
and (3) performing inverse normalization processing on the prediction result obtained by the model by adopting an equation (12):
wherein,x′is the value after normalization;xis an actual value;x max 、x min the maximum value and the minimum value in each dimension of data are respectively.
Preferably, the capacity of each wind farm is fixed, the maximum capacity of the wind farm is set as an upper limit threshold, the exceeding part of the processed prediction result is determined as an upper limit, and the result of which the prediction part is lower than 0 is determined as a 0 value and is a lower limit; the formula used is shown in equation (19):
wherein,x 3 representing the result of the denormalization process;threpresents an upper threshold; threshold (·) represents the thresholding function.
Preferably, in the wind power plant, according to a preset wind power upper threshold and a preset wind power lower threshold, the accuracy of the wind power prediction model is verified by adopting RMSE as an evaluation standard, the RMSE is used for measuring the discrete degree of the error between the predicted value and the actual value, and the calculation formula of the RMSE is shown as the formula (13):
wherein,
the predicted value of the wind power is;p i the actual value of the wind power is obtained;npredicting the number of data;ithe predicted points are numbered.
An ultra-short term wind power prediction system, the system comprising:
the analysis module is used for analyzing the type of the required data and acquiring a wind power data set;
the filling module is used for filling missing values of the acquired wind power data set;
the screening module is used for screening the optimal meteorological factor and historical power combination from the wind power data set by using the Pearson correlation coefficient;
the selection module is used for selecting the length of a time window meeting the preset requirement according to the wind power data set;
the slicing module is used for carrying out normalization processing on the wind power data set and slicing the data set according to the length of the selected time window;
the training module is used for building a parallel model of the CNN and GRU neural networks, setting parameters and training the model by using a data set after slice processing;
and the adjusting module is used for performing inverse normalization adjustment on the prediction result obtained by the model.
According to the technical scheme, the required wind power data set type is selected and preprocessing operation is carried out; screening the optimal meteorological factor and historical power combination by using the Pearson correlation coefficient; selecting a proper time window length according to the original data; building a parallel structure model of a convolutional neural network and a gated circulation unit; and carrying out an upper threshold setting method and a lower threshold setting method on the power obtained by the model. The method combines the advantages of CNN and GRU, can effectively extract effective characteristic information of a complex data set, and can capture time sequence characteristics of the wind power data set. The invention uses single BP and LSTM methods as comparison to verify the prediction effect of the method.
Drawings
FIG. 1 is a block flow diagram of the ultra-short term wind power prediction method of the present invention;
FIG. 2 is a convolutional neural network structure;
figure 3 is a gated loop element network basic unit.
Detailed Description
The following further describes the embodiments of the present invention. It should be noted that the description of the embodiments is provided to help understanding of the present invention, but the present invention is not limited thereto. 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.
As shown in one of fig. 1 to 3, the ultra-short-term wind power prediction method of the present invention includes the following steps:
s1, analyzing the type of the required data, and acquiring a wind power data set;
s2, filling missing values of the acquired wind power data set;
s3, screening the optimal meteorological factor and historical power combination from the wind power data set by using a Pearson correlation coefficient;
s4, selecting a time window length meeting the preset requirement according to the wind power data set;
s5, carrying out normalization processing on the wind power data set, and slicing the data set according to the length of the selected time window;
s6, building a parallel model of the CNN and GRU neural networks, setting parameters, and training the model by using the data set after slice processing;
s7, performing inverse normalization processing on the prediction result obtained by the model;
and S8, adjusting the result of the inverse normalization processing according to the preset wind power upper and lower thresholds in the wind power plant.
Example 1
In the specific implementation process of the invention, the content of the steps can be summarized as follows: the method comprises the following five parts of data selection, data preprocessing, model building, prediction result obtaining and result analysis:
firstly, selecting data;
firstly, analyzing the specific type of the required wind power data,
the active power output by the fan is calculated according to the following formula (1) and formula (2):
wherein,Pactive power output by a fan is kW;C p the wind energy conversion coefficient;Aswept area of the wind wheel in m2；ρIn terms of atmospheric density, in kg/m3；vIs the wind speed, with the unit of m/s:
wherein,ρis at atmospheric density;P a is atmospheric pressure in kPa;Tthermodynamic temperature, in K;P b is the saturated water vapor pressure in kPa;
is the relative humidity of air in%.
From the equation (1), the wind power has a direct relationship with the wind speed, the atmospheric density and the swept area of the wind wheel, wherein the swept area of the wind wheel is affected by the wind direction. From the formula (2), the atmospheric density is affected by temperature and pressure. For example, data of a certain wind power plant is selected as an example, the sampling time is 5min, and the data comprises 6-dimensional data of wind speed, atmospheric density, wind direction, temperature, air pressure and wind power.
Second, data preprocessing
And filling missing values of the original data set, wherein the filling mode is selected to fill by using non-missing front values.
Missing data is padded using equation (17):
wherein,x 2 representing padding data;x 0 representing the original data without missing value padding;fillna(. to) represents the padding function used;methodrepresenting a filling method;padindicates padding with an un-missing antecedent;bfillindicating padding with an unneeded late value;
Then, the Pearson correlation coefficient is used for screening the optimal meteorological factors, and the correlation formula is shown as the formula (3):
wherein,xvariable sumyThe variables are two variables in meteorological factors;
are respectively asxVariables of、yThe average value of the variables;r xy is a correlation coefficient;inumbering the predicted points;x i 、y i are respectively a predicted pointiCorresponding toxThe variables,yA variable;npredicting the number of data;
correlation coefficientr xy The larger the size, the correspondingxVariable sumyThe more strongly correlated these 2 factors are the variables, the more strongly correlated the factors are screened using equation (18):
wherein,if (. cndot.) represents a judgment function.
The correlation between each meteorological factor and power in different wind power plants is different, and the meteorological factors need to be screened according to specific wind power plant conditions to obtain the optimal combination of the meteorological factors and the wind power in the wind power plant.
In this example, the correlation coefficients of the meteorological factors and the historical power are shown in table 1, only the correlation between the wind speed and the historical wind power is strong and is 0.955, the absolute values of the correlation coefficients of the other meteorological factors and the historical power are all below 0.3, and the degree of correlation is not high, so the dimensionality of the data input by the model is 2-dimensional: wind speed, wind power.
TABLE 1 Pearson correlation coefficients for different variables
Firstly, the data is normalized, so that the numerical range after processing is [0, 1], and the formula is shown as formula (4):
wherein,x′is the value after normalization;xis an actual value;x max 、x min respectively a maximum value and a minimum value in each dimension data;
then, slicing the result data obtained by the normalization processing by using an equation (14), and converting the result data into a time sequence data matrix to obtain input data presenting time sequence as input data of the parallel CNN-GRU model;
wherein,in t to representtA time series data matrix of the time;Wrepresents a time window length;y t to representtHistorical power at the time and meteorological factor matrix.
When the time window length is too long, the model may cause the prediction accuracy to be reduced due to capturing misleading time sequence characteristics (when the wind power data suddenly changes); when the time window length is too short, the model lacks necessary timing characteristics, resulting in low prediction accuracy. Therefore, a suitable time window length needs to be selected according to specific wind farm data, and the time window size selected in the example is 30, namely 2.5 hours. And slicing the normalized data according to the selected time window length.
Model building
1) Convolutional neural network
CNN networks can be generally classified into one-dimensional convolution, two-dimensional convolution, and three-dimensional convolution, where one-dimensional convolution has a great advantage in processing multidimensional data. Therefore, the present invention selects a one-dimensional convolution network to extract the features of the original data, and the structural diagram thereof is shown in fig. 2.
The CNN network is mainly characterized by local connection and weight sharing compared with the conventional neural network, and generally consists of convolutional layers, pooling layers and fully-connected layers. The convolutional layers are the core part of the CNN network, and each convolutional layer is provided with a plurality of convolutional cores and is responsible for carrying out feature capture on input data; the pooling layer samples the feature vectors extracted from the convolutional layer, so that the overfitting is prevented, and the complexity of data is reduced; the full connection layer is used for connecting the extracted features of each channel.
The most important part in the CNN is a convolution part, and wind power data is subjected to time sequence characteristic capture through different convolution modes to obtain a characteristic value of each layer of neural network. As shown in formulas (5) and (6):
wherein, y l j is as followslLayer onejA characteristic of the convolution kernel output;f(. h) is an activation function;M i is as followsl-feature set of layer 1 output;y l-1 i is as followsl-1 layer output of the secondiA feature;K l j is as alLayer onejA convolution;b l j is an offset; relu (x 1) Activating a function for Relu;x 1 is the input vector.
And obtaining the characteristic value of each layer of neural network according to the formula (5) and the formula (6), wherein the matrix obtained by the CNN is a three-dimensional matrixThe parallel model uses a full connection layer to process the characteristics, and needs to be transmitted into a two-dimensional matrix, so that the characteristics need to be extended; the invention uses formula (15) to extend the characteristic matrix finally obtained by the CNN network and stores the characteristic matrix inout CNN So as to build a parallel CNN-GRU model;
wherein,out CNN representing a feature matrix of the CNN extraction;ConvId(. h) represents the way the convolutional neural network extracts features;Flatten (. to) represents a feature extension;in t to representtTime series data matrix of time of day.
2) Gated cyclic unit
The LSTM network is a kind of recurrent neural network, and is widely used because it can capture time series long and short term dependencies. The GRU is an optimized structure based on LSTM, and is widely applied to the field of time series data processing. The GRU integrates a forgetting gate and an input gate in the LSTM into a single updating gate, so that the model training efficiency is higher than that of the LSTM, and meanwhile, the prediction accuracy is guaranteed. The basic structure of a GRU network is shown in fig. 3.
FIG. 3 shows a reset gate in the left dashed box and an update gate in the right dashed box; tanh is the hidden state tangent function, and the associated formula is equation (9). Solid small circles outside the rounded rectangles represent input and output variables, and dashed small circles represent state variables of the hidden layer. The GRU internal principle formula is adopted as formula (7), formula (8), formula (9) and formula (10) to extract the characteristic matrixout GRU ：
Wherein,x t the input vector of the current moment is taken as the input vector of the current moment;r t 、z t respectively a reset gate and an update gate;h t-1 、h t 、
respectively representing a hidden layer state at the previous moment, a hidden layer state at the current moment and a candidate state at the current moment;W rx 、W zx 、W hx is prepared by reacting withx t A related weight matrix; W rh 、W zh 、W hh is prepared by reacting withh t-1 A related weight matrix;b r 、b z 、b h is a bias vector; sigma is an activation function sigmoid; an as dot product.
The update gate and the reset gate are the core part of the GRU unit and can decide which information to keep. Reset doorr t The hidden layer state at the last moment can be reseth t-1 Determining the retention degree of the information at the previous time; updating doorz t The GRU unit may be controlled to remember or forget information, i.e. to update the content of the new state.
Extracting a feature matrix of the GRU according to the formula (7), the formula (8), the formula (9) and the formula (10)out GRU Then, storing by using a formula (16) so as to build a parallel CNN-GRU model;
wherein,out GRU representing a feature matrix extracted by GRU;GRU (. to) shows the way the gated loop unit extracts features;in t to representtA time series data matrix of the time; 3) building a parallel model of CNN and GRU neural networks, fusing the features extracted by the CNN and the GRU networks respectively, and splicing by using a full connection layer to obtain the CNN-GRU parallel model, wherein the formula (11) is as follows:
wherein,y 2 representing the output result of the parallel model; Φ an activation function representing a fully connected layer;w 1 、w 2representing a weight matrix;brepresenting a bias vector;out CNN representing a feature matrix of the CNN extraction;out GRU representing the GRU extracted feature matrix.
The model specific parameters are shown in table 2. And distributing a training set and a test set according to the proportion of 9:1, and inputting the data of the training set into the built parallel CNN-GRU model for training.
TABLE 2 parameter settings for parallel CNN-GRU model
Fourthly, obtaining a prediction result
Inputting the test set into a model to obtain a wind power prediction value, and performing inverse normalization processing to make the prediction data have physical significance, as shown in formula (12):
wherein,x′is the value after normalization;xis an actual value;x max 、x min the maximum value and the minimum value in each dimension of data are respectively.
For example, the upper and lower thresholds of the wind power are set according to the specific wind farm, and the upper threshold of the wind farmx max 10MW, lower thresholdx min The obtained result is subjected to threshold correction to 0.
And fifthly, setting upper and lower thresholds of the wind power according to the specific wind power plant, and adjusting the predicted result.
The capacity of each wind power plant is fixed, the maximum capacity of the wind power plant is set as an upper limit threshold, the exceeding part of the processed prediction result is determined as an upper limit, and the result of which the prediction part is lower than 0 is determined as a 0 value and is a lower limit; the formula used is shown in equation (19):
wherein,x 3 representing the result of the denormalization process;threpresents an upper threshold; threshold (·) represents the thresholding function.
In order to verify the accuracy of the wind power prediction model, RMSE is selected as an evaluation standard. The RMSE can be used to measure the degree of dispersion of the error between the predicted value and the true value, and the smaller the value, the higher the model accuracy. The formula for RMSE is as given in equation (13):
wherein,
the predicted value of the wind power is;p i the actual value of the wind power is obtained;npredicting the number of data;ithe predicted points are numbered.
The final prediction results are shown in table 3:
TABLE 3 different models predict RMSE error
To sum up, the embodiment of the invention selects the type of the required wind power data set and carries out preprocessing operation; screening the optimal meteorological factor and historical power combination by using the Pearson correlation coefficient; selecting a proper time window length according to the original data; building a parallel structure model of a convolutional neural network and a gated circulation unit; and carrying out an upper threshold setting method and a lower threshold setting method on the power obtained by the model. The method combines the advantages of CNN and GRU, can effectively extract effective characteristic information of a complex data set, and can capture time sequence characteristics of the wind power data set. The invention uses single BP and LSTM methods as comparison, and verifies that the prediction effect of the method is better.
Example 2
An ultra-short term wind power prediction system, the system comprising:
the analysis module is used for analyzing the type of the required data and acquiring a wind power data set;
the filling module is used for filling missing values of the acquired wind power data set;
the screening module is used for screening the optimal meteorological factor and historical power combination from the wind power data set by using the Pearson correlation coefficient;
the selection module is used for selecting the length of a time window meeting the preset requirement according to the wind power data set;
the slicing module is used for carrying out normalization processing on the wind power data set and slicing the data set according to the length of the selected time window;
the training module is used for building a parallel model of the CNN-GRU neural network, setting parameters and training the model by using a data set after slice processing;
and the adjusting module is used for performing inverse normalization adjustment on the prediction result obtained by the model.
In the embodiment, the wind power data set can be obtained through the analysis module; filling missing values of the acquired wind power data set through a filling module; screening out the optimal meteorological factor and historical power combination through a screening module; selecting the length of a time window which meets the preset requirement through a selection module; slicing the data set by a slicing module; training the model through a training module; and performing inverse normalization adjustment on the prediction result obtained by the model through an adjusting module.
The ultra-short-term wind power prediction system based on the parallel CNN-GRU can extract effective characteristic information of a complex data set and capture time sequence characteristics of the wind power data set, so that data dimensionality is reduced, and more efficient and accurate prediction of a model is facilitated.
The embodiments of the present invention have been described in detail, but the present invention is not limited to the described embodiments. It will be apparent to those skilled in the art that various changes, modifications, substitutions and alterations can be made in these embodiments without departing from the principles and spirit of the invention, and the scope of protection is still within the scope of the invention.
Claims (10)
1. An ultra-short-term wind power prediction method is characterized by comprising the following steps: the ultra-short-term wind power prediction method comprises the following steps:
s1, analyzing the type of the required data, and acquiring a wind power data set;
s2, filling missing values of the acquired wind power data set;
s3, screening the optimal meteorological factor and historical power combination from the wind power data set by using a Pearson correlation coefficient;
s4, selecting a time window length meeting the preset requirement according to the wind power data set;
s5, carrying out normalization processing on the wind power data set, and slicing the data set according to the length of the selected time window;
s6, building a parallel model of the CNN and GRU neural networks, setting parameters, and training the model by using the data set after slice processing;
s7, performing inverse normalization processing on the prediction result obtained by the model;
and S8, adjusting the result of the inverse normalization processing according to the preset wind power upper and lower thresholds in the wind power plant.
2. The ultra-short-term wind power prediction method of claim 1, characterized in that: the type of the required data comprises wind speed, atmospheric density, wind direction, temperature, air pressure and wind power; wherein, the wind power refers to the active power output by the fan,
the active power output by the fan is calculated according to the following formula (1) and formula (2):
wherein,Pactive power output by a fan is kW;C p the wind energy conversion coefficient;Aswept area of the wind wheel in m2；ρIn terms of atmospheric density, in kg/m3；vIs the wind speed, with the unit of m/s:
wherein,ρis at atmospheric density;P a is atmospheric pressure in kPa;Tthermodynamic temperature, in K;P b is the saturated water vapor pressure in kPa;
is the relative humidity of air in%.
3. The ultra-short-term wind power prediction method of claim 1, characterized in that: filling the obtained missing data in the wind power data set, comprising the following steps:
filling by using an undeleted antecedent value;
or padding with an unneeded later value;
or based on the wind power plant data and a preset time scale, filling by taking the missing value as the center and taking the average value of the un-missing values in the preset time scale;
missing data is padded using equation (17):
wherein,x 2 representing padding data;x 0 representing the original data without missing value padding;fillna(. to) represents the padding function used;methodrepresenting a filling method;padindicates padding with an un-missing antecedent;bfillindicating padding with an unneeded late value;
4. The ultra-short-term wind power prediction method of claim 1, characterized in that: screening meteorological factors, including: by obtaining of formula (3)xVariable sumyThe correlation coefficient of the variable is screened based on the magnitude of the correlation coefficient;
wherein,xvariable sumyThe variables are two variables in meteorological factors;
are respectively asxVariables of、yThe average value of the variables;r xy is a correlation coefficient;inumbering the predicted points;x i 、y i are respectively a predicted pointiCorresponding toxThe variables,yA variable;npredicting the number of data;
correlation coefficientr xy The larger the size, the correspondingxVariable sumyThe more strongly correlated these 2 factors are the variables, the more strongly correlated the factors are screened using equation (18):
wherein,if (. cndot.) represents a judgment function.
5. The ultra-short-term wind power prediction method of claim 1, characterized in that: the normalization processing and slicing of the wind power data set comprise:
firstly, carrying out normalization processing on data by using a formula (4);
wherein,x′is the value after normalization;xis an actual value;x max 、x min respectively a maximum value and a minimum value in each dimension data;
then, slicing the result data obtained by the normalization processing by using an equation (14), and converting the result data into a time sequence data matrix to obtain input data presenting time sequence as input data of the parallel CNN-GRU model;
wherein,in t to representtTime sequence of momentsA column data matrix;Wrepresents a time window length;y t to representtHistorical power at the time and meteorological factor matrix.
6. The ultra-short-term wind power prediction method of claim 1, characterized in that: the method for constructing the parallel model of the CNN and GRU neural networks comprises the following steps:
s6.1, carrying out time sequence feature capture on the wind power data by the CNN neural network through different convolution modes to obtain a feature value of each layer of neural network, wherein the feature values are shown in a formula (5) and a formula (6):
wherein, y l j is as followslLayer onejA characteristic of the convolution kernel output;f (. h) is an activation function;M i is as followsl-feature set of layer 1 output;y l-1 i is as followsl-1 layer output of the secondiA feature;K l j is as alLayer onejA convolution;b l j is an offset; relu (x 1) Activating a function for Relu;x 1 is an input vector;
s6.2, after the characteristic value of each layer of neural network is obtained, extending the characteristic matrix finally obtained by the CNN network by using a formula (15), and storing the characteristic matrix until the characteristic matrix is storedout CNN The method is used for building a parallel CNN-GRU model;
wherein,out CNN representing a feature matrix of the CNN extraction;ConvId(. h) represents the way the convolutional neural network extracts features;Flatten (. to) represents a feature extension;in t to representtA time series data matrix of the time;
s6.3, controlling the retention degree of information at different moments by the GRU neural network through the reset gate and the update gate to extract the feature matrixout GRU The formulas used for extraction are shown as formula (7), formula (8), formula (9) and formula (10);
wherein,x t the input vector of the current moment is taken as the input vector of the current moment;r t 、z t respectively a reset gate and an update gate;h t-1 、h t 、
respectively representing a hidden layer state at the previous moment, a hidden layer state at the current moment and a candidate state at the current moment;W rx 、W zx 、W hx is prepared by reacting withx t CorrelationThe weight matrix of (2); W rh 、W zh 、W hh is prepared by reacting withh t-1 A related weight matrix;b r 、b z 、b h is a bias vector;σis an activation function sigmoid; an indication of a dot product;
s6.4, extracting feature matrix from GRUout GRU Then, storing by using a formula (16) so as to build a parallel CNN-GRU model;
wherein,out GRU representing a feature matrix extracted by GRU;GRU (. to) shows the way the gated loop unit extracts features;in t to representtA time series data matrix of the time;
s6.5, splicing the results after the CNN and GRU neural network processing by using a full connection layer to obtain a CNN-GRU parallel model, wherein the formula (11) is as follows:
wherein,y 2representing the output result of the parallel model; Φan activation function representing a fully connected layer;w 1 、w 2representing a weight matrix;brepresenting a bias vector;out CNN representing a feature matrix of the CNN extraction;out GRU representing a feature matrix extracted by GRU;
s6.6, inputting the data set after the slicing processing into the constructed CNN-GRU parallel model for training.
7. The ultra-short-term wind power prediction method of claim 1, characterized in that: performing inverse normalization processing on a prediction result obtained by the model, wherein the inverse normalization processing comprises the following steps:
and (3) performing inverse normalization processing on the prediction result obtained by the model by adopting an equation (12):
wherein,x′is the value after normalization;xis an actual value;x max 、x min the maximum value and the minimum value in each dimension of data are respectively.
8. The ultra-short-term wind power prediction method of claim 1, characterized in that: the capacity of each wind power plant is fixed, the maximum capacity of the wind power plant is set as an upper limit threshold, the exceeding part of the processed prediction result is determined as an upper limit, and the result of which the prediction part is lower than 0 is determined as a 0 value and is a lower limit; the formula used is shown in equation (19):
wherein,x 3 representing the result of the denormalization process;threpresents an upper threshold; threshold (·) represents the thresholding function.
9. The ultra-short-term wind power prediction method of claim 1, characterized in that: in a wind power plant, according to a preset wind power upper threshold and a preset wind power lower threshold, RMSE is used as an evaluation standard to verify the precision of a wind power prediction model, the RMSE is used for measuring the discrete degree of an error between a predicted value and a true value, and the calculation formula of the RMSE is shown as a formula (13):
wherein,
the predicted value of the wind power is;p i the actual value of the wind power is obtained;npredicting the number of data;ithe predicted points are numbered.
10. An ultra-short-term wind power prediction system, characterized in that the system comprises:
the analysis module is used for analyzing the type of the required data and acquiring a wind power data set;
the filling module is used for filling missing values of the acquired wind power data set;
the screening module is used for screening the optimal meteorological factor and historical power combination from the wind power data set by using the Pearson correlation coefficient;
the selection module is used for selecting the length of a time window meeting the preset requirement according to the wind power data set;
the slicing module is used for carrying out normalization processing on the wind power data set and slicing the data set according to the length of the selected time window;
the training module is used for building a parallel model of the CNN and GRU neural networks, setting parameters and training the model by using a data set after slice processing;
and the adjusting module is used for performing inverse normalization adjustment on the prediction result obtained by the model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111398570.7A CN113837499A (en) | 2021-11-24 | 2021-11-24 | Ultra-short-term wind power prediction method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111398570.7A CN113837499A (en) | 2021-11-24 | 2021-11-24 | Ultra-short-term wind power prediction method and system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113837499A true CN113837499A (en) | 2021-12-24 |
Family
ID=78971631
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111398570.7A Pending CN113837499A (en) | 2021-11-24 | 2021-11-24 | Ultra-short-term wind power prediction method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113837499A (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113988394A (en) * | 2021-10-21 | 2022-01-28 | 中国电建集团华东勘测设计研究院有限公司 | Wind power ultra-short-term power prediction method based on gram matrix and convolutional neural network |
| CN114819377A (en) * | 2022-05-11 | 2022-07-29 | 云南电网有限责任公司电力科学研究院 | Distributed wind power prediction method, system, device and storage medium |
| CN115330040A (en) * | 2022-08-05 | 2022-11-11 | 江苏润和软件股份有限公司 | Deep learning-based comprehensive energy distributed wind power generation prediction method and system |
| CN115375002A (en) * | 2022-07-12 | 2022-11-22 | 国网江苏省电力有限公司电力科学研究院 | Short-term power load prediction method, system, storage medium and computing device |
| CN115545315A (en) * | 2022-10-12 | 2022-12-30 | 重庆移通学院 | PM2.5 prediction method based on three-dimensional convolutional neural network and gated cyclic unit |
| CN116757340A (en) * | 2023-08-22 | 2023-09-15 | 国能日新科技股份有限公司 | Short-term wind speed fusion prediction method and device for wind farm |
| CN117893030A (en) * | 2024-03-14 | 2024-04-16 | 中智(福建)科技有限公司 | Power system risk early warning method based on big data |
| CN118013300A (en) * | 2024-04-08 | 2024-05-10 | 江苏海龙风电科技股份有限公司 | Short-term wind power prediction method and system for wind turbine generator |
| CN118611059A (en) * | 2024-08-07 | 2024-09-06 | 浙江大学海南研究院 | Offshore wind power prediction method of attention mechanism double-channel neural network |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108448610A (en) * | 2018-03-12 | 2018-08-24 | 华南理工大学 | A kind of short-term wind power prediction method based on deep learning |
| CN111475909A (en) * | 2019-10-28 | 2020-07-31 | 国网江西省电力有限公司电力科学研究院 | Wind turbine generator output correlation mapping modeling method based on long-term and short-term memory network |
| CN112001486A (en) * | 2020-08-28 | 2020-11-27 | 河北工业大学 | Load decomposition method based on deep learning |
| CN112232577A (en) * | 2020-10-23 | 2021-01-15 | 浙江八达电子仪表有限公司 | Power load probability prediction system and method for multi-core intelligent meter |
| CN112733462A (en) * | 2021-01-21 | 2021-04-30 | 国网辽宁省电力有限公司阜新供电公司 | Ultra-short-term wind power plant power prediction method combining meteorological factors |
| US20210183108A1 (en) * | 2019-12-16 | 2021-06-17 | X Development Llc | Edge-based crop yield prediction |
-
2021
- 2021-11-24 CN CN202111398570.7A patent/CN113837499A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108448610A (en) * | 2018-03-12 | 2018-08-24 | 华南理工大学 | A kind of short-term wind power prediction method based on deep learning |
| CN111475909A (en) * | 2019-10-28 | 2020-07-31 | 国网江西省电力有限公司电力科学研究院 | Wind turbine generator output correlation mapping modeling method based on long-term and short-term memory network |
| US20210183108A1 (en) * | 2019-12-16 | 2021-06-17 | X Development Llc | Edge-based crop yield prediction |
| CN112001486A (en) * | 2020-08-28 | 2020-11-27 | 河北工业大学 | Load decomposition method based on deep learning |
| CN112232577A (en) * | 2020-10-23 | 2021-01-15 | 浙江八达电子仪表有限公司 | Power load probability prediction system and method for multi-core intelligent meter |
| CN112733462A (en) * | 2021-01-21 | 2021-04-30 | 国网辽宁省电力有限公司阜新供电公司 | Ultra-short-term wind power plant power prediction method combining meteorological factors |
Non-Patent Citations (1)
| Title |
|---|
| 薛阳,王琳,王舒,张亚飞,张宁: "一种结合CNN和GRU网络的超短期风电预测模型", 《可再生能源》 * |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113988394A (en) * | 2021-10-21 | 2022-01-28 | 中国电建集团华东勘测设计研究院有限公司 | Wind power ultra-short-term power prediction method based on gram matrix and convolutional neural network |
| CN114819377A (en) * | 2022-05-11 | 2022-07-29 | 云南电网有限责任公司电力科学研究院 | Distributed wind power prediction method, system, device and storage medium |
| CN115375002A (en) * | 2022-07-12 | 2022-11-22 | 国网江苏省电力有限公司电力科学研究院 | Short-term power load prediction method, system, storage medium and computing device |
| CN115330040A (en) * | 2022-08-05 | 2022-11-11 | 江苏润和软件股份有限公司 | Deep learning-based comprehensive energy distributed wind power generation prediction method and system |
| CN115545315A (en) * | 2022-10-12 | 2022-12-30 | 重庆移通学院 | PM2.5 prediction method based on three-dimensional convolutional neural network and gated cyclic unit |
| CN116757340A (en) * | 2023-08-22 | 2023-09-15 | 国能日新科技股份有限公司 | Short-term wind speed fusion prediction method and device for wind farm |
| CN116757340B (en) * | 2023-08-22 | 2023-10-13 | 国能日新科技股份有限公司 | Short-term wind speed fusion prediction method and device for wind farm |
| CN117893030A (en) * | 2024-03-14 | 2024-04-16 | 中智(福建)科技有限公司 | Power system risk early warning method based on big data |
| CN117893030B (en) * | 2024-03-14 | 2024-05-28 | 中智(福建)科技有限公司 | Power system risk early warning method based on big data |
| CN118013300A (en) * | 2024-04-08 | 2024-05-10 | 江苏海龙风电科技股份有限公司 | Short-term wind power prediction method and system for wind turbine generator |
| CN118013300B (en) * | 2024-04-08 | 2024-06-04 | 江苏海龙风电科技股份有限公司 | Short-term wind power prediction method and system for wind turbine generator |
| CN118611059A (en) * | 2024-08-07 | 2024-09-06 | 浙江大学海南研究院 | Offshore wind power prediction method of attention mechanism double-channel neural network |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113837499A (en) | Ultra-short-term wind power prediction method and system | |
| CN109492822B (en) | Air pollutant concentration time-space domain correlation prediction method | |
| CN111210082B (en) | Precipitation prediction method based on optimized BP neural network algorithm | |
| Alencar et al. | Hybrid approach combining SARIMA and neural networks for multi-step ahead wind speed forecasting in Brazil | |
| CN114092832B (en) | High-resolution remote sensing image classification method based on parallel hybrid convolutional network | |
| CN115293415A (en) | Multi-wind-farm short-term power prediction method considering time evolution and space correlation | |
| CN108428017A (en) | Wind power interval prediction method based on core extreme learning machine quantile estimate | |
| CN114462718A (en) | CNN-GRU wind power prediction method based on time sliding window | |
| CN110427654A (en) | A kind of predictive model of landslide construction method and system based on sensitiveness | |
| CN108399470B (en) | Indoor PM2.5 prediction method based on multi-example genetic neural network | |
| Liu et al. | Heating load forecasting for combined heat and power plants via strand-based LSTM | |
| CN111242351A (en) | Tropical cyclone track prediction method based on self-encoder and GRU neural network | |
| CN117290810B (en) | Short-time strong precipitation probability prediction fusion method based on cyclic convolutional neural network | |
| CN107609667A (en) | Heating load forecasting method and system based on Box_cox conversion and UFCNN | |
| CN108062720A (en) | A kind of load forecasting method based on similar day selection and random forests algorithm | |
| CN116187540B (en) | Wind power station ultra-short-term power prediction method based on space-time deviation correction | |
| CN116345555A (en) | CNN-ISCA-LSTM model-based short-term photovoltaic power generation power prediction method | |
| CN112307672A (en) | BP neural network short-term wind power prediction method based on cuckoo algorithm optimization | |
| CN112381282A (en) | Photovoltaic power generation power prediction method based on width learning system | |
| CN114357670A (en) | Power distribution network power consumption data abnormity early warning method based on BLS and self-encoder | |
| CN114498619A (en) | Wind power prediction method and device | |
| CN111488974B (en) | Ocean wind energy downscaling method based on deep learning neural network | |
| CN116611702A (en) | Integrated learning photovoltaic power generation prediction method for building integrated energy management | |
| CN117388953B (en) | Weather forecast method for improving MIM-rwkv by SADBO based on big data frame | |
| CN114692981A (en) | Medium-and-long-term runoff forecasting method and system based on Seq2Seq model |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20211224 |
|
| RJ01 | Rejection of invention patent application after publication |