CN114429078B - Short-term wind power prediction method and system - Google Patents
Short-term wind power prediction method and system Download PDFInfo
- Publication number
- CN114429078B CN114429078B CN202111585164.1A CN202111585164A CN114429078B CN 114429078 B CN114429078 B CN 114429078B CN 202111585164 A CN202111585164 A CN 202111585164A CN 114429078 B CN114429078 B CN 114429078B
- Authority
- CN
- China
- Prior art keywords
- wind power
- wind
- theta
- power plant
- matrix
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/23—Design optimisation, verification or simulation using finite element methods [FEM] or finite difference methods [FDM]
-
- 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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2113/00—Details relating to the application field
- G06F2113/06—Wind turbines or wind farms
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/06—Power analysis or power optimisation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/08—Thermal analysis or thermal optimisation
-
- 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)
- Evolutionary Computation (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- General Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Geometry (AREA)
- Computer Hardware Design (AREA)
- Medical Informatics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a short-term wind power prediction method and a short-term wind power prediction system, which are characterized in that meteorological characteristic parameters of wind power plants in different geographic positions are firstly acquired, a characteristic matrix input prediction model is formed after preliminary processing, the spatial relation between different wind power plants and a target wind power plant is mined by adopting a spatial attention mechanism in the prediction model, the implicit relation existing in the characteristic matrix is mined by a gated circulation unit network, then a hyperparameter matrix in the gated circulation unit network is optimized by adopting a criss-cross optimization algorithm to complete the training of the prediction model, and finally the prediction model is used for predicting the wind power. The method can realize the quantification of the influence degree of the adjacent wind power plants on the target wind power plant so as to improve the prediction precision of the wind power.
Description
Technical Field
The invention relates to prediction of wind power, in particular to a short-term wind power prediction method and system based on a space attention System (SA) and a cross optimization algorithm (CSO).
Background
Wind energy is a new energy source, and the large-scale grid connection of the wind energy brings challenges to the economical, safe and stable operation of a power system. Therefore, accurate wind power prediction has important significance on the power system.
The wind power is used as a time sequence with strong randomness and volatility, and the prediction precision has a great relation with the quality and the time scale of meteorological characteristic data. Wind speed has a certain correlation between different positions within a certain range due to meteorological features. The influence degree of the adjacent wind power plants on the target wind power plant is effectively quantized, and the accuracy of wind power plant power prediction can be further improved. Hitherto, existing methods for predicting wind power through spatial correlation quantify adjacent wind power plants or meteorological measurement points based on a correlation coefficient method, so that the prediction accuracy of the wind power is low.
Therefore, how to adaptively quantify the influence degree of the adjacent wind power plants on the target wind power plant becomes a technical problem which needs to be solved urgently.
Disclosure of Invention
The invention provides a short-term wind power prediction method and system, which are used for realizing the quantification of the influence degree of an adjacent wind power plant on a target wind power plant so as to improve the prediction precision of wind power.
The technical scheme of the invention is as follows:
a short-term wind power prediction method comprises the following steps:
s1: acquiring power, wind speed, wind direction and temperature data of a target wind power plant and an adjacent wind power plant, and performing primary processing;
s2: forming a characteristic matrix X by the processed power, wind speed, wind direction and temperature data of the wind power plant;
s3: constructing a prediction model comprising a space attention mechanism (SA) and a gated cyclic unit (GRU) network;
s4: inputting the characteristic matrix X into a prediction model, mining the spatial relationship between different wind power plants and a target wind power plant through a spatial attention mechanism in the prediction model, and respectively giving corresponding weights to the corresponding wind power plants according to the mined spatial relationship;
s5: transmitting the feature matrix X endowed with the weight to a gated circulation unit network, and mining an implicit relation existing in the feature matrix X by the gated circulation unit network;
s6: the prediction model adopts a longitudinal and transverse cross optimization algorithm (CSO) to optimize a hyper-parameter matrix theta in a gating cycle unit network, and the training of the prediction model is completed;
s7: and predicting the power of the target wind power plant by using the trained prediction model to obtain a power time sequence of the corresponding wind power plant.
The invention provides a short-term wind power prediction method based on a space attention mechanism (SA) and a cross optimization algorithm (CSO). Firstly, meteorological characteristic parameters of wind power plants in different geographic positions are required to be obtained, after preliminary treatment, the space attention mechanism is adopted to mine the spatial relation between different wind power plants and a target wind power plant, then a gate control circulation unit (GRU) network is adopted to mine the implicit relation existing in a characteristic matrix, finally, a cross optimization algorithm is adopted to optimize a hyper-parameter matrix in the gate control circulation unit network to finish the training of a prediction model, and the prediction model can be used for predicting the wind power. The method can effectively improve the accuracy of short-term wind power prediction.
Further, in step S1, the specific process of performing the preliminary processing on the data is as follows:
and performing min-max normalization processing on the power sequence, the wind speed sequence and the temperature sequence to obtain a processed power sequence P, a processed wind speed sequence WS and a processed temperature sequence Tem, and performing sine-cosine processing on the wind direction sequence to obtain a wind direction sine WDS and a wind direction cosine WDC.
Further, in step S2, X = [ D = 1 ,D 2 ,...,D m ]Wherein D is m Matrix of features representing the m-th wind farm at times t-1 to t-n, D m The specific expression form of (A) is as follows:
in the formula (I), the compound is shown in the specification,andand respectively representing the power, the wind speed, the wind direction sine, the wind direction cosine and the temperature of the wind power plant at the t-n moment of the mth wind power plant.
Further, the spatial attention mechanism in the step S4 includes two layers of neural networks, and the number of neurons in the input layer and the output layer is the same as that in the wind farm, and is m;
w=f(WX+b)
in the formula, W and b are weights and offsets of a full connection layer in a space attention mechanism, f () is an activation function, W is an unnormalized wind farm weight sequence, and α is a sequence of influence degrees of each wind farm on target wind farm power, and the following relations are satisfied: w = [ w = 1 ,w 2 ,...,w m ]And α = [ α = 1 ,α 2 ,...,α m ]Wherein w is i For the ith unnormalized wind farm weight, α i And the influence degree of the ith wind power plant on the target wind power plant power is shown.
Further, in step S4, after the spatial attention mechanism mines the spatial relationship, the specific process of giving the weight is as follows:
multiplying the obtained weights by the characteristics of the corresponding wind power plants respectively to obtain weighted characteristic matrixes X;
X * =X⊙α
X * =[α 1 D 1 ,α 2 D 2 ,...,α m D m ]。
further, in step S5, the gated loop unit network is constructed as follows:
building two layers of gate control circulation unit networks by taking X as input, wherein the number of neurons is 4 and 8 respectively, and the activation function is tanh;
in the formula, W r 、W z 、W h 、U r 、U z 、U h As a matrix of weight parameters, b r 、b z 、b h In the form of a matrix of offset parameters,for matrix multiplication, σ is the Sigmod function, r t To reset the gate, z t In order to update the door or doors,for the candidate state of the hidden layer at the current time, y t For the current implicit state, y t-1 Is an implicit state of the previous moment, x t The input state at the current moment.
Further, in step S6, the hyper-parameter matrix θ is formed by the weight parameter and the bias parameter of the output layer of the gated cyclic unit network, and is as follows:
θ=[W z ,W r ,W h ,U z ,U r ,U h ,b z ,b r ,b h ]。
further, in step S6, a specific process of optimizing the hyperparametric matrix θ in the gated cyclic unit network by using the criss-cross optimization algorithm is as follows:
and the crisscross optimization algorithm takes the preliminarily trained hyper-parameter matrix theta as an initial value, and obtains a better hyper-parameter matrix theta after multiple iterations.
Further, the specific steps of performing optimization iteration by using the crisscross optimization algorithm are as follows:
s61: aiming at improving the prediction precision, the vertical and horizontal cross optimization algorithm adopts the minimum root mean square error as a fitness function, which is as follows;
in the formula (f) obj As a fitness function, N is the number of samples,andrespectively representing a real value and a predicted value;
s62: randomly selecting nth dimensions of parent particles theta (i) and theta (j) in the parameters to intersect with each other;
in the formula, MS hc (i,n)、MS hc (j, n) are respectively the nth dimension filial generation generated by the transverse intersection of theta (i, n) and theta (j, n), r1, r2 and c1, c2 are respectively random numbers of (0, 1) and (-1, 1), theta (i, n) is the nth dimension of the particle theta (i), and theta (j, n) is the nth dimension of the particle theta (j);
s63: randomly selecting a v-th dimension and a k-th dimension to cross each other by parent particles theta (q) in the parameters;
MS vc (q,v)=rθ(q,v)+(1-r)θ(q,k)
in the formula, MS vc (q, v) is a descendant in the v-th dimension generated by longitudinally crossing theta (q, v) and theta (q, k), r is a random number of (0, 1), theta (q, v) is the v-th dimension of the particle theta (q), and theta (q, k) is the k-th dimension of the particle theta (q);
s64: and repeating the steps S62 and S63 according to the set iteration times, and stopping iteration when the preset iteration times are reached.
The invention also provides a short-term wind power prediction system which comprises an acquisition module and a processing module which are in communication connection;
the acquisition module is used for acquiring power, wind speed, wind direction and temperature data of a target wind power plant and an adjacent wind power plant, performing primary processing, and outputting a characteristic matrix X formed by the processed power, wind speed, wind direction and temperature data of the wind power plant to the processing module;
the processing module is constructed with a prediction model comprising a spatial attention mechanism and a gated loop unit network;
after the characteristic matrix X is input into the prediction model, a spatial attention mechanism in the prediction model excavates the spatial relationship between different wind power plants and a target wind power plant, and corresponding weights are respectively given to the corresponding wind power plants according to the excavated spatial relationship; then, the feature matrix X endowed with the weight is transmitted to a gated circulation unit network, and the gated circulation unit network excavates the implicit relation existing in the feature matrix X; then, optimizing a hyper-parameter matrix theta in a gating cycle unit network by the prediction model by adopting a criss-cross optimization algorithm to finish the training of the prediction model; and finally, predicting the power of the target wind power plant by using the trained prediction model to obtain a power time sequence of the corresponding wind power plant.
The invention has the beneficial effects that:
the method is realized based on a space attention mechanism and a criss-cross optimization algorithm, wherein the space attention mechanism can effectively mine and quantify the space correlation existing between adjacent wind power plants, and has certain help for improving the wind power prediction precision; and the cross optimization algorithm is adopted to optimize the super-parameter matrix of the GRU network, so that the local optimization problem possibly existing in the GRU network built by the neural network can be solved, and the method has certain practical significance for short-term wind power prediction. The method can effectively improve the prediction accuracy of the wind power.
Drawings
FIG. 1 is a schematic flow chart of a short-term wind power prediction method according to the present invention;
FIG. 2 is a schematic diagram of a predictive model;
FIG. 3 is a network architecture diagram of a spatial attention mechanism;
FIG. 4 is a graph of weights assigned to different wind farms by a spatial attention mechanism;
fig. 5 is a diagram of the predicted effect of wind power.
Detailed Description
The drawings are for illustrative purposes only and are not to be construed as limiting the patent; for the purpose of better illustrating the present embodiments, certain elements of the drawings may be omitted, enlarged or reduced, and do not represent the size of an actual product; it will be understood by those skilled in the art that certain well-known structures in the drawings and descriptions thereof may be omitted. The positional relationships depicted in the drawings are for illustrative purposes only and are not to be construed as limiting the present patent.
Example 1:
as shown in fig. 1 to 3, a short-term wind power prediction method includes the following steps:
s1: acquiring power, wind speed, wind direction and temperature data of a target wind power plant and an adjacent wind power plant, and performing primary processing;
s2: forming a characteristic matrix X by the processed power, wind speed, wind direction and temperature data of the wind power plant;
s3: constructing a prediction model comprising a space attention mechanism (SA) and a gated cyclic unit (GRU) network;
s4: inputting the characteristic matrix X into a prediction model, mining the spatial relationship between different wind power plants and a target wind power plant through a spatial attention mechanism in the prediction model, and respectively giving corresponding weights to the corresponding wind power plants according to the mined spatial relationship;
s5: transmitting the feature matrix X endowed with the weight to a gated circulation unit network, and mining an implicit relation existing in the feature matrix X by the gated circulation unit network;
s6: the prediction model adopts a longitudinal and transverse cross optimization algorithm (CSO) to optimize a hyper-parameter matrix theta in a gating cycle unit network, and the training of the prediction model is completed;
s7: and predicting the power of the target wind power plant by using the trained prediction model to obtain a power time sequence corresponding to the wind power plant.
In this embodiment, the specific process of performing the preliminary processing on the data in step S1 is as follows:
and performing min-max normalization processing on the power sequence, the wind speed sequence and the temperature sequence to obtain a processed power sequence P, a processed wind speed sequence WS and a processed temperature sequence Tem, and performing sine-cosine processing on the wind direction sequence to obtain a wind direction sine WDS and a wind direction cosine WDC.
In the present embodiment, the feature matrix X = [ D ] of step S2 1 ,D 2 ,...,D m ]Wherein D is m Matrix of features representing the m-th wind farm at times t-1 to t-n, D m The specific expression form of (A) is as follows:
in the formula (I), the compound is shown in the specification,andand respectively representing the power, the wind speed, the wind direction sine, the wind direction cosine and the temperature of the mth wind power plant at the t-n moment.
In this embodiment, the spatial attention mechanism in step S4 includes two layers of neural networks, and the number of neurons in the input layer and the output layer is the same as the number of neurons in the wind farm, and is m;
w=f(WX+b)
in the formula, W and b are weights and offsets of full connection layers in a space attention mechanism, f () is an activation function, W is an unnormalized wind farm weight sequence, and alpha is a sequence of influence degrees of each wind farm on target wind farm power and satisfies the following relations: w = [ w = 1 ,w 2 ,...,w m ]And α = [ α ] 1 ,α 2 ,...,α m ]Wherein w is i For the ith unnormalized wind farm weight, α i And the influence degree of the ith wind power plant on the target wind power plant power is shown.
In this embodiment, after the spatial attention mechanism mines the spatial relationship in step S4, the specific process of giving weight is as follows:
multiplying the obtained weights by the characteristics of the corresponding wind power plants respectively to obtain weighted characteristic matrixes X;
X * =X⊙α
X * =[α 1 D 1 ,α 2 D 2 ,...,α m D m ]。
further, in step S5, the gated loop unit network is constructed as follows:
building two layers of gate control circulation unit networks by taking X as input, wherein the number of neurons is 4 and 8 respectively, and the activation function is tanh;
in the formula, W r 、W z 、W h 、U r 、U z 、U h As a matrix of weight parameters, b r 、b z 、b h In the form of a matrix of offset parameters,for matrix multiplication, σ is the Sigmod function, r t To reset the gate, z t In order to update the door,for the candidate state of the hidden layer at the current time, y t For the current implicit state, y t-1 Is an implicit state of the previous moment, x t The input state at the current moment.
In this embodiment, the hyperparametric matrix θ in step S6 is formed by the weight parameters and bias parameters of the output layer of the gated cyclic unit network, and is as follows:
θ=[W z ,W r ,W h ,U z ,U r ,U h ,b z ,b r ,b h ]。
in this embodiment, the specific process of optimizing the hyperparametric matrix θ in the gated cyclic unit network by using the criss-cross optimization algorithm in step S6 is as follows:
and the crisscross optimization algorithm takes the preliminarily trained hyper-parameter matrix theta as an initial value, and obtains a better hyper-parameter matrix theta after multiple iterations.
The specific steps of carrying out optimization iteration by the crisscross optimization algorithm are as follows:
s61: aiming at improving the prediction precision, the vertical and horizontal cross optimization algorithm adopts the minimum root mean square error as a fitness function, which is as follows;
in the formula, f obj For the fitness function, N is the number of samples,andrespectively representing a real value and a predicted value;
s62: randomly selecting nth dimensions of parent particles theta (i) and theta (j) in the parameters to intersect with each other;
in the formula, MS hc (i,n)、MS hc (j, n) are respectively the nth dimension filial generation generated by the transverse intersection of theta (i, n) and theta (j, n), r1, r2 and c1, c2 are respectively random numbers of (0, 1) and (-1, 1), theta (i, n) is the nth dimension of the particle theta (i), and theta (j, n) is the nth dimension of the particle theta (j);
s63: randomly selecting a v-th dimension and a k-th dimension to intersect with each other by using a parent particle theta (q) in the parameter;
MS vc (q,v)=rθ(q,v)+(1-r)θ(q,k)
in the formula, MS vc (q, v) is a v-th dimension child generated by longitudinally crossing theta (q, v) and theta (q, k), r is a random number of (0, 1), theta (q, v) is a v-th dimension of the particle theta (q), and theta (q, k) is a k-th dimension of the particle theta (q);
s64: and repeating the steps S62 and S63 according to the set iteration times, and stopping iteration when the preset iteration times are reached.
In summary, the invention provides a short-term wind power prediction method based on a spatial attention mechanism (SA) and a cross optimization algorithm (CSO), which includes acquiring meteorological characteristic parameters of wind farms in different geographic positions, mining spatial relationships between different wind farms and a target wind farm by using the spatial attention mechanism after preliminary processing, mining implicit relationships existing in a characteristic matrix through a gated circulation unit (GRU) network, and finally optimizing a hyper-parameter matrix in the gated circulation unit network by using the cross optimization algorithm to complete training of a prediction model, i.e., performing prediction of wind power by using the prediction model. The method can effectively improve the accuracy of short-term wind power prediction.
Example 2:
in order to verify the effectiveness of the short-term wind power prediction method in embodiment 1 of the present invention, in this embodiment:
in step S1, power, wind speed, wind direction and temperature data at 70m horizontal altitude for four wind farms in inner Mongolia region 2018/12/16/0;
in step S4, the activation function of the spatial attention mechanism is a Linear rectification function (strained Linear Unit, reLU), and the number of neurons in the input layer and the output layer is 4;
in step S6, the population number of the vertical and horizontal cross optimization algorithm is set to be 25, the horizontal cross rate is 1, the vertical cross rate is 0.6, and the iteration times are 200;
the data are substituted into the short-term wind power prediction method in embodiment 1 to perform prediction.
Finally, the weights given to the four wind power sites as shown in fig. 4 and the wind power prediction effect as shown in fig. 5 are obtained.
As can be seen from the prediction effect in fig. 5, the short-term wind power prediction method of the present invention can effectively improve the accuracy of short-term wind power prediction.
Example 3:
the embodiment is similar to embodiment 1, except that the embodiment provides a system applying the short-term wind power prediction method, and the system comprises an acquisition module and a processing module which are in communication connection;
the acquisition module is used for acquiring power, wind speed, wind direction and temperature data of a target wind power plant and an adjacent wind power plant, performing primary processing, and outputting a characteristic matrix X formed by the processed power, wind speed, wind direction and temperature data of the wind power plant to the processing module;
the processing module is constructed with a prediction model comprising a space attention mechanism and a gated loop unit network;
after the characteristic matrix X is input into the prediction model, a spatial attention mechanism in the prediction model excavates the spatial relationship between different wind power plants and a target wind power plant, and corresponding weights are respectively given to the corresponding wind power plants according to the excavated spatial relationship; then, the feature matrix X endowed with the weight is transmitted to a gated circulation unit network, and the gated circulation unit network excavates the implicit relation existing in the feature matrix X; then, optimizing a hyper-parameter matrix theta in a gating cycle unit network by the prediction model by adopting a crisscross optimization algorithm to finish the training of the prediction model; and finally, predicting the power of the target wind power plant by using the trained prediction model to obtain a power time sequence of the corresponding wind power plant.
The acquisition module and the processing module in the system can be integrated in a circuit board or can be arranged separately.
It should be understood that the above-described embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. Other variations and modifications will be apparent to persons skilled in the art in light of the above description. And are neither required nor exhaustive of all embodiments. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present invention should be included in the protection scope of the claims of the present invention.
Claims (5)
1. A short-term wind power prediction method is characterized by comprising the following steps:
s1: acquiring power, wind speed, wind direction and temperature data of a target wind power plant and an adjacent wind power plant, and performing primary processing;
s2: forming a characteristic matrix X by the processed power, wind speed, wind direction and temperature data of the wind power plant;
s3: constructing a prediction model comprising a space attention mechanism and a gating cycle unit network;
s4: inputting the characteristic matrix X into a prediction model, mining the spatial relationship between different wind power plants and a target wind power plant through a spatial attention mechanism in the prediction model, and respectively giving corresponding weights to the corresponding wind power plants according to the mined spatial relationship;
s5: transmitting the feature matrix X endowed with the weight to a gated circulation unit network, and mining an implicit relation existing in the feature matrix X by the gated circulation unit network;
s6: optimizing a hyper-parameter matrix theta in a gating cycle unit network by adopting a longitudinal and transverse cross optimization algorithm to complete the training of the prediction model;
s7: predicting the power of a target wind power plant by using the trained prediction model to obtain a power time sequence of the corresponding wind power plant;
in step S1, the specific process of data preliminary processing is as follows:
performing min-max normalization processing on the power sequence, the wind speed sequence and the temperature sequence to obtain a processed power sequence P, a processed wind speed sequence WS and a processed temperature sequence Tem, and performing sine-cosine processing on the wind direction sequence to obtain a wind direction sine WDS and a wind direction cosine WDC;
in step S2, X = [ D = 1 ,D 2 ,...,D m ]In which D is m A matrix representing the formation of the characteristics of the m-th wind farm at times t-1 to t-n, D m The specific expression form of (A) is as follows:
in the formula (I), the compound is shown in the specification,andrespectively representing the power, the wind speed, the wind direction sine, the wind direction cosine and the temperature of the wind power plant at the t-n moment of the mth wind power plant;
the space attention mechanism in the step S4 comprises two layers of neural networks, wherein the number of neurons of an input layer and an output layer is the same as that of the wind power plant, and the number of the neurons is m;
w=f(WX+b)
in the formula, W and b are weights and offsets of a full connection layer in a space attention mechanism, f () is an activation function, W is an unnormalized wind farm weight sequence, and α is a sequence of influence degrees of each wind farm on target wind farm power, and the following relations are satisfied: w = [ w = 1 ,w 2 ,...,w m ]And α = [ α ] 1 ,α 2 ,...,α m ]Wherein w is i For the ith unnormalized wind farm weight, α i The influence degree of the ith wind power plant on the target wind power plant power is obtained;
in step S4, after the spatial attention mechanism mines the spatial relationship, the specific process of giving the weight is as follows:
multiplying the obtained weights by the characteristics of the corresponding wind power plants respectively to obtain weighted characteristic matrixes X;
X * =X⊙α
X * =[α 1 D 1 ,α 2 D 2 ,...,α m D m ]
in step S5, the gated loop unit network is constructed as follows:
building two layers of gated cyclic unit networks by taking X as input, wherein the number of neurons is 4 and 8 respectively, and an activation function is tanh;
in the formula, W r 、W z 、W h 、U r 、U z 、U h As a matrix of weight parameters, b r 、b z 、b h Is a matrix of the bias parameters and is,for matrix multiplication, σ is a Sigmod function, r t To reset the gate, z t In order to update the door or doors,as candidate state of the hidden layer at the current time, y t For the current implicit State, y t-1 Is an implicit state of the previous moment, x t The input state at the current moment.
2. The short-term wind power prediction method according to claim 1, wherein in step S6, the hyper-parameter matrix θ is formed by the weight parameters and the bias parameters of the output layer of the gated cyclic unit network, and is represented by the following formula:
θ=[W z ,W r ,W h ,U z ,U r ,U h ,b z ,b r ,b h ]。
3. the short-term wind power prediction method according to claim 2, wherein in step S6, the specific process of optimizing the hyper-parameter matrix θ in the gated cyclic unit network by using the crossbar cross optimization algorithm is as follows:
and the crisscross optimization algorithm takes the preliminarily trained hyper-parameter matrix theta as an initial value, and obtains a better hyper-parameter matrix theta after multiple iterations.
4. The short-term wind power prediction method according to claim 3, characterized in that the specific steps of performing optimization iteration by using a crisscross optimization algorithm are as follows:
s61: aiming at improving the prediction precision, the vertical and horizontal cross optimization algorithm adopts the minimum root mean square error as a fitness function, which is as follows;
in the formula (f) obj As a fitness function, N is the number of samples,andrespectively representing a real value and a predicted value;
s62: randomly selecting nth dimensions of parent particles theta (i) and theta (j) in the parameters to intersect with each other;
in the formula, MS hc (i,n)、MS hc (j, n) are respectively the nth dimension filial generation generated by the transverse intersection of theta (i, n) and theta (j, n), r1, r2 and c1, c2 are respectively random numbers of (0, 1) and (-1, 1), theta (i, n) is the nth dimension of the particle theta (i), and theta (j, n) is the nth dimension of the particle theta (j);
s63: randomly selecting a v-th dimension and a k-th dimension to intersect with each other by using a parent particle theta (q) in the parameter;
MS vc (q,v)=rθ(q,v)+(1-r)θ(q,k)
in the formula, MS vc (q, v) is a v-th dimension child generated by longitudinally crossing theta (q, v) and theta (q, k), r is a random number of (0, 1), theta (q, v) is a v-th dimension of the particle theta (q), and theta (q, k) is a k-th dimension of the particle theta (q);
s64: and repeating the steps S62 and S63 according to the set iteration times, and stopping iteration when the preset iteration times are reached.
5. A short-term wind power prediction system is characterized by being used for realizing the short-term wind power prediction method of claim 1, and the system comprises an acquisition module and a processing module which are in communication connection;
the acquisition module is used for acquiring power, wind speed, wind direction and temperature data of a target wind power plant and an adjacent wind power plant, performing primary processing, and outputting a characteristic matrix X formed by the processed power, wind speed, wind direction and temperature data of the wind power plant to the processing module;
the processing module is constructed with a prediction model comprising a spatial attention mechanism and a gated loop unit network;
after the characteristic matrix X is input into the prediction model, a spatial attention mechanism in the prediction model excavates the spatial relationship between different wind power plants and a target wind power plant, and corresponding weights are respectively given to the corresponding wind power plants according to the excavated spatial relationship; then, the feature matrix X endowed with the weight is transmitted to a gated circulation unit network, and the gated circulation unit network excavates the implicit relation existing in the feature matrix X; then, optimizing a hyper-parameter matrix theta in a gating cycle unit network by the prediction model by adopting a criss-cross optimization algorithm to finish the training of the prediction model; and finally, predicting the power of the target wind power plant by using the trained prediction model to obtain a power time sequence of the corresponding wind power plant.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111585164.1A CN114429078B (en) | 2021-12-22 | 2021-12-22 | Short-term wind power prediction method and system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202111585164.1A CN114429078B (en) | 2021-12-22 | 2021-12-22 | Short-term wind power prediction method and system |
Publications (2)
Publication Number | Publication Date |
---|---|
CN114429078A CN114429078A (en) | 2022-05-03 |
CN114429078B true CN114429078B (en) | 2022-10-18 |
Family
ID=81311626
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202111585164.1A Active CN114429078B (en) | 2021-12-22 | 2021-12-22 | Short-term wind power prediction method and system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN114429078B (en) |
Families Citing this family (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN115983494B (en) * | 2023-02-10 | 2023-09-12 | 广东工业大学 | Short-term wind power prediction method and system for newly-built small-sample wind power plant |
CN117892113A (en) * | 2024-03-13 | 2024-04-16 | 广东工业大学 | Wind power prediction method of self-adaptive VMD and dual dimension-reduction attention mechanism |
CN117893362A (en) * | 2024-03-18 | 2024-04-16 | 广东工业大学 | Multi-time-space-scale offshore wind power characteristic screening and enhanced power prediction method |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102682207A (en) * | 2012-04-28 | 2012-09-19 | 中国科学院电工研究所 | Ultrashort combined predicting method for wind speed of wind power plant |
CN106529741A (en) * | 2016-12-12 | 2017-03-22 | 国网电力科学研究院 | Space relevant characteristic-based ultra-short-period wind power prediction method |
CN106650982A (en) * | 2016-08-30 | 2017-05-10 | 华北电力大学 | Depth learning power prediction method based on multi-point NWP |
CN108520094A (en) * | 2018-03-07 | 2018-09-11 | 南方电网科学研究院有限责任公司 | A kind of Time-space serial analogy method of output of wind electric field |
CN112308322A (en) * | 2020-11-03 | 2021-02-02 | 华翔翔能科技股份有限公司 | Multi-wind-field space-time wind speed prediction method and device and electronic equipment |
CN112653142A (en) * | 2020-12-18 | 2021-04-13 | 武汉大学 | Wind power prediction method and system for optimizing depth transform network |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102055188B (en) * | 2011-01-07 | 2012-12-26 | 西北电网有限公司 | Ultra-short term wind power forecasting method based on time series method |
US10598157B2 (en) * | 2017-02-07 | 2020-03-24 | International Business Machines Corporation | Reducing curtailment of wind power generation |
-
2021
- 2021-12-22 CN CN202111585164.1A patent/CN114429078B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102682207A (en) * | 2012-04-28 | 2012-09-19 | 中国科学院电工研究所 | Ultrashort combined predicting method for wind speed of wind power plant |
CN106650982A (en) * | 2016-08-30 | 2017-05-10 | 华北电力大学 | Depth learning power prediction method based on multi-point NWP |
CN106529741A (en) * | 2016-12-12 | 2017-03-22 | 国网电力科学研究院 | Space relevant characteristic-based ultra-short-period wind power prediction method |
CN108520094A (en) * | 2018-03-07 | 2018-09-11 | 南方电网科学研究院有限责任公司 | A kind of Time-space serial analogy method of output of wind electric field |
CN112308322A (en) * | 2020-11-03 | 2021-02-02 | 华翔翔能科技股份有限公司 | Multi-wind-field space-time wind speed prediction method and device and electronic equipment |
CN112653142A (en) * | 2020-12-18 | 2021-04-13 | 武汉大学 | Wind power prediction method and system for optimizing depth transform network |
Non-Patent Citations (2)
Title |
---|
Very short-term spatial and temporal wind power forecasting: A deep learning approach;Hu Tianyu;《CSEE Journal of Power and Energy Systems》;20190801;第28-36页 * |
计及气象因素和风速空间相关性的风电功率预测模型;胡帅;《电力系统自动化》;20201112;第434-443页 * |
Also Published As
Publication number | Publication date |
---|---|
CN114429078A (en) | 2022-05-03 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN114429078B (en) | Short-term wind power prediction method and system | |
CN110059878B (en) | Photovoltaic power generation power prediction model based on CNN LSTM and construction method thereof | |
Tian | Modes decomposition forecasting approach for ultra-short-term wind speed | |
CN111292525B (en) | Traffic flow prediction method based on neural network | |
CN110852902A (en) | Photovoltaic power generation power prediction method based on BAS-BP | |
CN112116162B (en) | Power transmission line icing thickness prediction method based on CEEMDAN-QFAOA-LSTM | |
CN110110434B (en) | Initialization method for probability load flow deep neural network calculation | |
CN110942205A (en) | Short-term photovoltaic power generation power prediction method based on HIMVO-SVM | |
CN111784061B (en) | Training method, device and equipment for power grid engineering cost prediction model | |
Chinnathambi et al. | Deep neural networks (DNN) for day-ahead electricity price markets | |
CN115169543A (en) | Short-term photovoltaic power prediction method and system based on transfer learning | |
CN105628868B (en) | A kind of undersampled signal im-pact location processing method and system of composite structure | |
CN104732067A (en) | Industrial process modeling forecasting method oriented at flow object | |
CN117319231A (en) | Network flow prediction method and system for optimizing VMD (virtual model device) based on dung beetle algorithm | |
Tian et al. | A wind speed prediction system based on new data preprocessing strategy and improved multi-objective optimizer | |
CN112183721B (en) | Construction method of combined hydrological prediction model based on self-adaptive differential evolution | |
CN117154690A (en) | Photovoltaic power generation power prediction method and system based on neural network | |
CN115796327A (en) | Wind power interval prediction method based on VMD (vertical vector decomposition) and IWOA-F-GRU (empirical mode decomposition) -based models | |
CN114372634B (en) | Short-term wind power prediction method and system based on neural network imbalance optimization | |
CN115936062A (en) | Photovoltaic power day-ahead interval prediction method based on scene generation and deep learning | |
CN115640868A (en) | Short-term prediction method for minority data wind power of newly-built wind power plant | |
CN115034432A (en) | Wind speed prediction method for wind generating set of wind power plant | |
Zhao et al. | A dendritic neuron based very short-term prediction model for photovoltaic power | |
Wang et al. | Robust active yaw control for offshore wind farms using stochastic predictive control based on online adaptive scenario generation | |
Xie et al. | Short-term Wind Power Prediction Model Based on GSA Optimized GRU Neural Network |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |