CN115018156A - Short-term wind power prediction method - Google Patents
Short-term wind power prediction method Download PDFInfo
- Publication number
- CN115018156A CN115018156A CN202210633967.8A CN202210633967A CN115018156A CN 115018156 A CN115018156 A CN 115018156A CN 202210633967 A CN202210633967 A CN 202210633967A CN 115018156 A CN115018156 A CN 115018156A
- Authority
- CN
- China
- Prior art keywords
- imf component
- wind power
- firework
- wind speed
- imf
- 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 30
- 241000039077 Copula Species 0.000 claims abstract description 23
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 23
- 238000004880 explosion Methods 0.000 claims description 29
- 238000013507 mapping Methods 0.000 claims description 17
- 230000000739 chaotic effect Effects 0.000 claims description 14
- 230000006870 function Effects 0.000 claims description 14
- 238000012549 training Methods 0.000 claims description 6
- 238000004364 calculation method Methods 0.000 claims description 4
- 230000001174 ascending effect Effects 0.000 claims description 3
- 238000005315 distribution function Methods 0.000 claims description 3
- 238000002922 simulated annealing Methods 0.000 claims description 3
- 238000010248 power generation Methods 0.000 description 6
- 238000000354 decomposition reaction Methods 0.000 description 5
- 238000013528 artificial neural network Methods 0.000 description 3
- 230000007774 longterm Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 238000007637 random forest analysis Methods 0.000 description 2
- 230000000717 retained effect Effects 0.000 description 2
- 238000012706 support-vector machine Methods 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 230000006978 adaptation Effects 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000000053 physical method Methods 0.000 description 1
- 230000008569 process Effects 0.000 description 1
- 238000004445 quantitative analysis Methods 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
- 239000000126 substance Substances 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J3/00—Circuit arrangements for ac mains or ac distribution networks
- H02J3/004—Generation forecast, e.g. methods or systems for forecasting future energy generation
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J3/00—Circuit arrangements for ac mains or ac distribution networks
- H02J3/38—Arrangements for parallely feeding a single network by two or more generators, converters or transformers
- H02J3/381—Dispersed generators
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J2203/00—Indexing scheme relating to details of circuit arrangements for AC mains or AC distribution networks
- H02J2203/20—Simulating, e g planning, reliability check, modelling or computer assisted design [CAD]
-
- H—ELECTRICITY
- H02—GENERATION; CONVERSION OR DISTRIBUTION OF ELECTRIC POWER
- H02J—CIRCUIT ARRANGEMENTS OR SYSTEMS FOR SUPPLYING OR DISTRIBUTING ELECTRIC POWER; SYSTEMS FOR STORING ELECTRIC ENERGY
- H02J2300/00—Systems for supplying or distributing electric power characterised by decentralized, dispersed, or local generation
- H02J2300/20—The dispersed energy generation being of renewable origin
- H02J2300/28—The renewable source being wind energy
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Economics (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Strategic Management (AREA)
- Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Human Resources & Organizations (AREA)
- Power Engineering (AREA)
- General Health & Medical Sciences (AREA)
- Tourism & Hospitality (AREA)
- Marketing (AREA)
- General Business, Economics & Management (AREA)
- Data Mining & Analysis (AREA)
- Water Supply & Treatment (AREA)
- Public Health (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Primary Health Care (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Development Economics (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The invention discloses a short-term wind power prediction method, which comprises the following steps: decomposing historical wind power data and wind speed data by using a VMD algorithm to obtain a wind power IMF component and a wind speed IMF component; selecting wind speed IMF components which are closely corresponding to each wind power IMF component by utilizing a Copula entropy method and a Hampel rule; establishing an LSTM prediction model, taking the wind speed IMF component with close corresponding relation of each wind power IMF component as input, and taking the wind power IMF component as output; and taking the IMF component of the predicted daily wind speed closely corresponding to each IMF component of the wind power as input, outputting a predicted value of the corresponding IMF component of the predicted daily wind power, and finally adding the predicted values of the IMF components of the wind power to obtain a predicted value of the wind power data of the final predicted day. The method can reduce the error influence of noise on wind power prediction, so that the prediction precision is higher.
Description
Technical Field
The invention belongs to the technical field of wind power prediction, and particularly relates to a short-term wind power prediction method.
Background
In recent years, the wind power generation amount continuously rises, but the wind power generation is influenced by weather and has great intermittence and fluctuation, which improves great difficulty for large-scale grid-connected operation of the wind power generation, and if the output power of the wind power generation can be effectively predicted, the problems caused by grid-connected wind power generation can be greatly improved. At present, two methods are mainly used for predicting the short-term wind power generation power, the first method is a physical method, the predicted value of the output power is calculated through a wind power output curve based on hydrodynamics modeling and mainly based on detailed geographic information and numerical weather forecast around a wind power station, meteorological data and physical data such as solar irradiation time, irradiation intensity, wind speed, wind direction and temperature are used as input parameters, and the prediction performance is greatly influenced when the climate changes. The other is a statistical prediction method, which is to find an internal statistical rule from the mathematical perspective according to a large amount of long-term historical output power data and predict the future output power by using the past output power by using a learning model. Common methods in the Artificial intelligence technology include Artificial Neural Networks (ANN), Support Vector Machines (SVM), Random Forest (RF), and the like. However, because the wind power has great fluctuation, the single artificial neural network has unstable prediction performance, and further improvement is needed to improve the prediction accuracy.
Disclosure of Invention
The invention aims to: aiming at the problems in the prior art, the invention discloses a short-term wind power prediction method, which realizes short-term wind power prediction and achieves the aim of reducing wind power prediction errors.
The technical scheme is as follows: in order to realize the purpose of the invention, the invention adopts the following technical scheme:
a short-term wind power prediction method comprises the following steps:
s1, decomposing historical wind power data and wind speed data by utilizing a VMD algorithm to respectively obtain K 1 Wind power IMF component sum K of different central frequencies 2 An individual wind speed IMF component;
s2, establishing a prediction input and output model by utilizing a Copula entropy method and a Hampel criterion, and selecting a wind speed IMF component which is closely corresponding to each wind power IMF component;
s3, respectively establishing an LSTM prediction model for each wind power IMF component and training, wherein the wind speed IMF component with close corresponding relation to each wind power IMF component in the step S2 is used as the input of the prediction model, and the wind power IMF component is used as the output of the prediction model, wherein the number of hidden layer nodes and the learning efficiency of the LSTM prediction model are optimized by adopting an improved firework algorithm;
and S4, taking the predicted daily wind speed IMF component which is closely corresponding to each wind power IMF component in the step S2 as input, outputting the corresponding predicted daily wind power IMF component predicted values, and finally adding the predicted values of each wind power IMF component to obtain the wind power data predicted value of the final predicted day.
Preferably, step S2 includes the steps of:
s21, estimating an empirical Copula density function f (x) of the decomposed wind power IMF component and wind speed IMF component:
f(x)=rank(x)/N
wherein x is a wind power IMF component or a wind speed IMF component, N is the number of data contained in the IMF component x, and rank (x) represents that the data in the IMF component x are ranked according to the numerical value in a reverse order;
the k1 th wind power IMF component IMF1 k1 Empirical Copula density function thereofComprises the following steps:
the k2 th wind speed IMF component IMF2 k2 Empirical Copula density functionComprises the following steps:
s22, estimating Copula entropy values between each wind power IMF component and all wind speed IMF components through a K-nearest neighbor algorithm, and calculating the K1 th wind power IMF component IMF1 k1 And the k2 th wind speed IMF component IMF2 k2 For example, compute the Copula entropy value CE between them k1,k2 : note the book
Where ψ (·) is a Digamma function, ε (n) is the Euler distance of x (n) to its 3 rd neighbor, specifically: distance ifThe Euler distance of (A) is 3 rdThen:
by analogy, the k1 th wind power IMF component IMF1 is solved k1 And Copula entropy values of other wind speed IMF components and form a set CE k1 Then CE k1 Has a total of K 2 Value, CE k1 ={CE k1,1 ,CE k1,2 ,…,CE k1,K2 };
S23, calculating the IMF1 component of the k1 th wind power k1 And the k2 th wind speed IMF component IMF2 k2 Hampel distance between:
d k1 (k2)=|CE k1,k2 -CE k1 (0.5) |
d k1 ={d k1 (1),d k1 (2),…,d k1 (K2)}
wherein H k1 (k2) For the k1 th wind power IMF component IMF1 k1 And the k2 th wind speed IMF component IMF2 k2 Hampel distance, CE of k1 (0.5) As a set CE k1 The median of (a) is determined,is a set d k1 A median of (d); if H is k1 (k2) If the wind speed IMF component is greater than 3, selecting the k2 th wind speed IMF component as a wind speed IMF component which is closely corresponding to the k1 th wind power IMF component;
by analogy, according to the step S22 and the step S23, the wind speed IMF component having close correspondence with each wind power IMF component can be obtained.
Preferably, the step S3 of optimizing the number of hidden layer nodes and learning efficiency of the LSTM prediction model by using the improved firework algorithm includes: the objective function is:
wherein N is the number of prediction time nodes on the prediction day,predicting a wind power IMF component prediction value x at the nth predicted time node of a day n Predicting the actual value of the IMF component of the wind power at the nth predicted time node of the day;
the specific steps of the improved firework algorithm for optimizing the number of hidden layer nodes and the learning efficiency of the LSTM prediction model are as follows:
s31, initializing parameters including firework number M, maximum iteration number T, variant spark number M, explosion number S, explosion radius initial value r, variable upper limit UB and variable lower limit LB;
s32, generating an initial firework population, and introducing a Tent chaotic mapping and reverse population learning strategy, wherein the Tent chaotic mapping expression is as follows:
Z i,j =(UB-LB)×rand+LB
wherein Z is i,j Representing Tent chaotic mapping initial firework individual, i is firework serial number, J is 1,2, …, J-1, J is variable number, rand is represented in interval [0,1]Randomly draw a number, u ∈ [0, 1]]And after a Tent chaotic mapping sequence is obtained, obtaining an initial population according to the following formula, wherein the initial population comprises M fireworks:
X=(UB-LB)×Z+LB
reverse population, including M fireworks:
X′=UB+LB-X
calculating the firework fitness in the initial population and the reverse population, selecting the first M fireworks according to the fitness value, and performing iterative operation;
and S33, iteration is carried out, the initial iteration number is 0, and the explosion radius and the explosion number of each firework are respectively calculated according to the following formulas:
wherein fr (i) is the explosion radius of the ith firework, f (i) is the individual fitness of the ith firework, f min As fireworksMinimum fitness in the population, f max The maximum fitness of the firework population is fn (i), the number of sparks generated by the ith firework is fn, theta is a machine minimum quantity and is used for avoiding zero operation, and the minimum quantity value generated in each time is different;
s34, selecting the ith firework individual x (i) to generate an explosion spark ex i :
ex i,j =x i,j +fr(i)×(2×rand(1,J)-1)
Therein, ex i,j For explosion spark ex i The j (th) variable, x i,j For J variable in the firework individual x (i), rand (1, J) is a number randomly extracted in the interval (1, J), and J is the number of the variable;
combining the firework population and the generated explosion spark population into a new population, and randomly selecting m individuals from the new population to generate Cauchy variant sparks mx i And whether the Cauchy variant spark is reserved is considered according to the simulated annealing principle, and the variant formula is as follows:
mx i,j =x i,j +x i,j ×Cauchy(0,1)
wherein mex i,j Is a Coxiv spark mx i The jth variable in the middle, Cauchy (0,1), is a standard Cauchy distribution function;
wherein, f (x) i ) Is the individual fitness of the fireworks before Cauchy variation, f (mx) i ) The individual fitness of Cauchi's variation spark, W is the current temperature, W 0 The initial temperature is t, and the current iteration times is t; if P is 1, keeping the Cauchy variant spark, otherwise not keeping;
checking whether the newly generated sparks exceed the limit, mapping the sparks exceeding the limit, taking an upper limit value if the sparks exceed the upper limit, and taking a lower limit value if the sparks exceed the lower limit;
s35, calculating the fitness values of the explosion sparks and the Cauchy variation sparks, comparing the fitness values with the fitness values of the fireworks, arranging all individuals in an ascending order according to the fitness values, selecting the first M individuals to enter next iteration, and adding 1 to the iteration times;
and S36, judging whether the current iteration number is smaller than the maximum iteration number T, if so, returning to the step S33 to continue the iteration, otherwise, terminating the calculation, and outputting the firework individual with the minimum fitness and the fitness value, wherein the firework individual with the minimum fitness is the optimal hidden layer node number and the optimal learning efficiency of the LSTM prediction model.
Has the advantages that: compared with the prior art, the invention has the following remarkable beneficial effects:
1. the method uses the VMD algorithm to decompose the original wind speed data and the wind power data, can reduce the error influence of noise on wind power prediction, and enables the prediction precision to be higher;
2. the method selects the relationship between the wind speed IMF component and the wind power IMF component after quantitative analysis and decomposition by using a Copula entropy method and a Hampel rule, thereby establishing an accurate prediction input and output model;
3. the method improves the traditional firework algorithm, enables the firework algorithm to have the capability of global optimization and escape from the local optimal solution, and is used for optimizing the LSTM network, so that the final wind power prediction is more accurate.
Drawings
FIG. 1 is a flow chart of a wind power prediction method according to the present invention;
FIG. 2 is a schematic diagram of wind power data and IMF components of wind power decomposed by VMD;
FIG. 3 is a diagram illustrating wind speed data and wind speed IMF components after VMD decomposition;
FIG. 4 is a flow chart of the improved fireworks algorithm optimizing LSTM predictive model;
FIG. 5 is a comparison graph of wind power of a wind farm predicted by the wind power prediction method of the present invention and actual wind power.
Detailed Description
The present invention will be further described with reference to the accompanying drawings.
The invention discloses a short-term wind power prediction method, which comprises the following steps as shown in figure 1:
step S1: decomposing original wind power data and wind speed data by using a Variational Modal Decomposition (VMD) algorithm to respectively obtain K 1 Wind power Intrinsic Mode Function (IMF) components and K of different central frequencies 2 Individual wind speed IMF component.
The VMD algorithm is to decompose a sequence f (n) into K IMF components u with specific center frequencies k And with all IMF components u k Is the minimum of the sum of the estimated bandwidths of the two components as an objective function, and all IMF components u k And the sum equals f (n) as a constraint, the problem translates into:
wherein, { u k }={u 1 ,u 2 ,…,u K The IMF component set obtained by decomposition, K is the total number of IMF components obtained by decomposition, { omega } k }={ω 1 ,ω 2 ,…,ω K Is the set of center frequencies, ω, of the decomposed IMF components k Is the IMF component u k Central frequency of (d) n In order to have a dirac distribution, the distribution,to calculate the partial derivative for n, n represents a time node, | · | calculation 2 Representing a 2 norm.
Converting the constrained questions into unconstrained questions through Lagrange transformation, wherein the objective function of the transformed variational constraint model is as follows:
wherein, λ is Lagrange multiplier, α is quadratic penalty factor.
IMF component u k And its center frequency omega k Update equations such asThe following:
wherein the content of the first and second substances,respectively representf (n), λ (n), l being the current number of iterations,respectively the k-th IMF component u at the next iteration k And the ith IMF component u at the current iteration i 。
The original wind power data and wind speed data are substituted into the formula to obtain K 1 Individual wind power IMF component and K 2 Individual wind speed IMF component.
In an embodiment of the invention, as shown in fig. 2, the VMD algorithm is used to decompose the original wind power data to obtain K 1 6 wind power IMF components; as shown in FIG. 3, the original wind speed data is decomposed by VMD algorithm to obtain K 2 6 wind speed IMF components.
Step S2: and establishing an accurate prediction input and output model by using a Copula entropy method and a Hampel criterion, and selecting a wind speed IMF component which is closely corresponding to each wind power IMF component. The method comprises the following specific steps:
step S21: estimating an empirical Copula density function f (x) of the decomposed wind power IMF component and wind speed IMF component by an order statistic (rank):
f(x)=rank(x)/N
wherein x is a wind power IMF component or a wind speed IMF component, N is the number of data contained in the IMF component x, and rank (x) represents that the data in the IMF component x are ranked according to the numerical value in a reverse order.
According to the formula, the k1 th wind power IMF component IMF1 k1 Empirical Copula density function thereofComprises the following steps:
the k2 th wind speed IMF component IMF2 k2 Empirical Copula density functionComprises the following steps:
step S22: estimating Copula entropy value between each wind power IMF component and all wind speed IMF components by K-nearest neighbor (KNN) algorithm, and calculating the K1 th wind power IMF component IMF1 k1 And the k2 th wind speed IMF component IMF2 k2 For example, compute the Copula entropy CE between them k1,k2 : note the book
Where ψ (·) is a Digamma function, ε (n) is the Euler distance of x (n) to its 3 rd neighbor, specifically: distance ifIs near the No. 3 Euler (except the nearest and second nearest, i.e., near the No. 3 Euler), e.g., {1,2,3,4} has a distance of 4 nearest to 5, 3 next nearest, 3 near to 3Is 2) isThen:
by analogy, the k1 th wind power IMF component IMF1 is solved k1 And Copula entropy values of other wind speed IMF components, and form a set CE k1 Then CE k1 Has a total of K 2 Value, CE k1 ={CE k1,1 ,CE k1,2 ,…,CE k1,K2 }。
Step S23: calculating the k1 th wind power IMF component IMF1 according to the following formula k1 And the k2 th wind speed IMF component IMF2 k2 Hampel distance between:
d k1 (k2)=|CE k1,k2 -CE k1 (0.5) |
d k1 ={d k1 (1),d k1 (2),…,d k1 (K2)} (7)
wherein H k1 (k2) For the k1 th wind power IMF component IMF1 k1 And the k2 th wind speed IMF component IMF2 k2 Hampel distance, CE of k1 (0.5) As a set CE k1 The median of (a) is determined,is a set d k1 The median of (3). If H is present k1 (k2) And if the wind speed IMF component is larger than 3, selecting the k2 th wind speed IMF component as the wind speed IMF component which is closely corresponding to the k1 th wind power IMF component.
By analogy, according to the step S22 and the step S23, the wind speed IMF component having close correspondence with each wind power IMF component can be obtained.
Step S3: and taking the wind power IMF component and the wind speed IMF component before the prediction day as a training set, and taking the wind power IMF component and the wind speed IMF component on the prediction day as a test set. And (4) respectively establishing an LSTM prediction model for each wind power IMF component for the training set, optimizing the LSTM prediction model by adopting an improved firework algorithm, taking the wind speed IMF component with close corresponding relation to each wind power IMF component in the step S2 as the input of the prediction model, and taking the wind power IMF component as the output of the prediction model.
The optimized variables of the improved firework algorithm optimized LSTM prediction model are the number of hidden layer nodes and the learning efficiency of the LSTM prediction model, and the objective function is as follows:
wherein N is the number of prediction time nodes of the prediction day,predicting a wind power IMF component prediction value x at the nth predicted time node of a day n And predicting the actual value of the IMF component of the wind power at the nth predicted time node of the day.
As shown in fig. 4, the specific steps of the improved firework algorithm for optimizing the number of hidden layer nodes and learning efficiency of the LSTM prediction model are as follows:
step S31: initializing parameters including firework number M, maximum iteration number T, variation spark number M, explosion number s, an initial explosion radius value r, a variable upper limit UB and a variable lower limit LB;
in one embodiment of the invention, the number of fireworks M is 10, the maximum iteration number T is 20, the number of variant sparks M is 10, the number of explosions s is 10, the explosion radius r is 10, the number of hidden layer nodes and the upper limit vector UB of learning efficiency are [200,0.15], and the number of hidden layer nodes and the lower limit vector LB of learning efficiency are [10,0.01 ];
step S32: generating an initial firework population, and introducing a Tent chaotic mapping and reverse population learning strategy, wherein a Tent chaotic mapping expression is as follows:
Z i,j =(UB-LB)×rand+LB (9)
wherein Z is i,j Representing Tent chaotic mapping initial firework individual, i is the firework serial number, J is 1,2, …, J-1, J is the variable number, J takes the value of 2 in the embodiment, and rand represents in the interval [0,1]Randomly draw a number, u ∈ [0, 1]]Obtaining a Tent chaotic mapping sequence through the formula (9), and obtaining an initial population according to the following formula:
X=(UB-LB)×Z+LB (10)
the reverse population is:
X′=UB+LB-X (11)
according to the formula (11), the number of the reverse population fireworks is the same as that of the population fireworks subjected to Tent chaotic mapping. In summary, the improved AO initial population generation process is:
(1) generating an initial population subjected to Tent chaotic mapping according to an equation (9) and an equation (10), wherein the initial population comprises M fireworks;
(2) generating a reverse population according to equation (11), comprising M fireworks;
(3) and calculating the fitness of the 2M fireworks (including the fireworks in the initial population and the fireworks in the reverse population), selecting the first M fireworks according to the fitness value, and performing iterative operation.
Step S33: and (4) entering iteration, wherein the initial iteration number is 0, and the explosion radius and the explosion number of each firework are respectively calculated according to the following formulas:
wherein fr (i) is the explosion radius of the ith firework, f (i) is the individual fitness of the ith firework, f min Is the minimum fitness, f, in the firework population max The maximum fitness of the firework population is fn (i), the number of sparks generated by the ith firework is fn, theta is a machine minimum quantity and is used for avoiding zero operation, and the minimum quantity value generated in each time is different;
step S34: selecting the ith firework individual x (i), and generating an explosion spark ex according to a formula (14) i :
ex i,j =x i,j +fr(i)×(2×rand(1,J)-1) (14)
Therein, ex i,j For explosion spark ex i The j (th) variable, x i,j For the jth variable in the firework individual x (i), rand (1, J) is a number randomly extracted in the interval (1, J), and J is the number of the variable.
Combining the firework population and the generated explosion spark population into a new population, randomly selecting m individuals from the new population, and generating Cauchy variant sparks mx according to a formula (15) i And whether the Cauchy variant spark is reserved is considered according to the simulated annealing principle, and the variant formula is as follows:
mx i,j =x i,j +x i,j ×Cauchy(0,1) (15)
wherein mex i,j Is a Coxiv spark mx i The j-th variable, Cauchy (0,1), is a standard Cauchy distribution function.
Wherein, f (x) i ) Is the individual fitness of the fireworks before Cauchy variation, f (mx) i ) Is the individual fitness of Cauchy's variant spark, W is the currentTemperature, W 0 The initial temperature is t, and the current iteration times is t; if P is 1, the cauchy variant spark is retained, otherwise it is not retained.
Checking whether the newly generated spark exceeds the limit, mapping the spark exceeding the boundary, and taking the upper limit value if the spark exceeds the upper limit and taking the lower limit value if the spark exceeds the lower limit.
Step S35: calculating the fitness values of the explosion sparks and the Cauchy variation sparks, comparing the fitness values with the fitness values of the fireworks, arranging all individuals in an ascending order according to the fitness values, selecting the first M individuals to enter next iteration, and adding 1 to the iteration times;
step S36: and judging whether the current iteration number is less than the maximum iteration number T, if so, returning to the step S33 to continue the iteration, otherwise, terminating the calculation and outputting the individual with the minimum fitness and the fitness value. The individual with the minimum fitness is the optimal hidden layer node number and the learning efficiency of the LSTM prediction model.
Step S4: respectively establishing LSTM prediction models trained in the step S3 for each wind power IMF component of the prediction day, taking the IMF component of the prediction day wind speed closely corresponding to each wind power IMF component in the step S2 as input, outputting the predicted value of the corresponding IMF component of the prediction day wind power, and finally adding the predicted values of the IMF components of each wind power to obtain the predicted value of the wind power data of the final prediction day.
Taking data of a certain wind power plant as an example, the sampling period of the data is 15 minutes, wind speed data of the previous 364 days and wind power data of the previous 364 days are used as training sets, wind speed data of 365 days are used as test sets to predict the wind power data of 365 days, a prediction model is established by the wind power prediction method, and the percentage of the average absolute error of the final prediction result and the actual ratio is 2.31% as shown in fig. 5.
The method can be used for short-term wind power prediction and also can be used for medium-term and long-term wind power prediction: the short-term wind power prediction is generally to predict the wind power within one day to one week, and the needed historical data is less; the wind power prediction of the medium-long term generally needs to predict the wind power in the next years, a large amount of historical data is needed for training, and if the method is used for predicting the long-term wind power, the required time is greatly increased.
The above description is only of the preferred embodiments of the present invention, and it should be noted that: it will be apparent to those skilled in the art that various modifications and adaptations can be made without departing from the principles of the invention and these are intended to be within the scope of the invention.
Claims (3)
1. A short-term wind power prediction method is characterized by comprising the following steps:
s1, decomposing historical wind power data and wind speed data by utilizing a VMD algorithm to respectively obtain K 1 Wind power IMF component sum K of different central frequencies 2 An individual wind speed IMF component;
s2, establishing a prediction input and output model by utilizing a Copula entropy method and a Hampel criterion, and selecting a wind speed IMF component which is closely corresponding to each wind power IMF component;
s3, respectively establishing an LSTM prediction model for each wind power IMF component and training, wherein the wind speed IMF component with close corresponding relation to each wind power IMF component in the step S2 is used as the input of the prediction model, and the wind power IMF component is used as the output of the prediction model, wherein the number of hidden layer nodes and the learning efficiency of the LSTM prediction model are optimized by adopting an improved firework algorithm;
and S4, taking the predicted daily wind speed IMF component which is closely corresponding to each wind power IMF component in the step S2 as input, outputting the corresponding predicted daily wind power IMF component predicted values, and finally adding the predicted values of each wind power IMF component to obtain the wind power data predicted value of the final predicted day.
2. The short-term wind power prediction method according to claim 1, wherein the step S2 includes the following steps:
s21, estimating an empirical Copula density function f (x) of the decomposed wind power IMF component and wind speed IMF component:
f(x)=rank(x)/N
wherein x is a wind power IMF component or a wind speed IMF component, N is the number of data contained in the IMF component x, and rank (x) represents that the data in the IMF component x are ranked according to the numerical value in a reverse order;
the k1 th wind power IMF component IMF1 k1 Empirical Copula density functionComprises the following steps:
the k2 th wind speed IMF component IMF2 k2 Empirical Copula density functionComprises the following steps:
s22, estimating Copula entropy values between each wind power IMF component and all wind speed IMF components through a K-nearest neighbor algorithm, and calculating the K1 th wind power IMF component IMF1 k1 And the k2 th wind speed IMF component IMF2 k2 For example, compute the Copula entropy CE between them k1,k2 : note the book
Where ψ (·) is a Digamma function, ε (n) is the Euler distance of x (n) to its 3 rd neighbor, specifically: distance ifThe Euler distance of (A) is 3 rdThen:
by analogy, the k1 th wind power IMF component IMF1 is solved k1 And Copula entropy values of other wind speed IMF components, and form a set CE k1 Then CE k1 Has a total of K 2 Value, CE k1 ={CE k1,1 ,CE k1,2 ,…,CE k1,K2 };
S23, calculating IMF1 component of k1 wind power k1 And the k2 th wind speed IMF component IMF2 k2 Hampel distance between:
d k1 (k2)=|CE k1,k2 -CE k1 (0.5) |
d k1 ={d k1 (1),d k1 (2),…,d k1 (K2)}
wherein H k1 (k2) For the k1 th wind power IMF component IMF1 k1 And the k2 th wind speed IMF component IMF2 k2 Hampel distance, CE of k1 (0.5) As a set CE k1 The median of (a) is determined,is a set d k1 A median of (d); if H is k1 (k2) If the wind speed IMF component is greater than 3, selecting the k2 th wind speed IMF component as a wind speed IMF component which is closely corresponding to the k1 th wind power IMF component;
by analogy, according to the step S22 and the step S23, the wind speed IMF component having close correspondence with each wind power IMF component can be obtained.
3. The short-term wind power prediction method according to claim 1, wherein optimizing the number of hidden layer nodes and learning efficiency of the LSTM prediction model by using the improved firework algorithm in step S3 comprises: the objective function is:
wherein N is the number of prediction time nodes on the prediction day,predicting a wind power IMF component prediction value x at the nth predicted time node of a day n Predicting the actual value of the IMF component of the wind power at the nth predicted time node of the day;
the specific steps of the improved firework algorithm for optimizing the number of hidden layer nodes and the learning efficiency of the LSTM prediction model are as follows:
s31, initializing parameters including firework number M, maximum iteration number T, varied spark number M, explosion number S, explosion radius initial value r, variable upper limit UB and variable lower limit LB;
s32, generating an initial firework population, and introducing a Tent chaotic mapping and reverse population learning strategy, wherein the Tent chaotic mapping expression is as follows:
Z i,j =(UB-LB)×rand+LB
wherein Z is i,j Representing Tent chaotic mapping initial firework individual, i is firework serial number, J is 1,2, …, J-1, J is variable number, rand is represented in interval [0,1]Randomly draw a number, u ∈ [0, 1]]And after a Tent chaotic mapping sequence is obtained, obtaining an initial population according to the following formula, wherein the initial population comprises M fireworks:
X=(UB-LB)×Z+LB
reverse population, including M fireworks:
X′=UB+LB-X
calculating the firework fitness in the initial population and the reverse population, selecting the first M fireworks according to the fitness value, and performing iterative operation;
and S33, iteration is carried out, the initial iteration number is 0, and the explosion radius and the explosion number of each firework are respectively calculated according to the following formulas:
wherein fr (i) is the explosion radius of the ith firework, f (i) is the individual fitness of the ith firework, f min Is the minimum fitness, f, in the firework population max The maximum fitness of the firework population is fn (i), the number of sparks generated by the ith firework is fn, theta is a machine minimum quantity and is used for avoiding zero operation, and the minimum quantity value generated in each time is different;
s34, selecting the ith firework individual x (i) to generate an explosion spark ex i :
ex i,j =x i,j +fr(i)×(2×rand(1,J)-1)
Therein, ex i,j For explosion spark ex i The j (th) variable, x i,j For the jth variable in the firework individual x (i), rand (1, J) is a number randomly extracted in an interval (1, J), and J is the number of the variable;
combining the firework population and the generated explosion spark population into a new population, and randomly selecting m individuals from the new population to generate Cauchy variant sparks mx i And whether the Cauchy variant spark is reserved is considered according to the simulated annealing principle, and the variant formula is as follows:
mx i,j =x i,j +x i,j ×Cauchy(0,1)
wherein mex i,j Is a Cauchi variant sparkmx i The jth variable in the middle, Cauchy (0,1), is a standard Cauchy distribution function;
wherein, f (x) i ) Is the individual fitness of the fireworks before Cauchy variation, f (mx) i ) The individual fitness of Cauchi's variation spark, W is the current temperature, W 0 The initial temperature is t, and the current iteration times is t; if P is 1, keeping the Cauchy variant spark, otherwise not keeping;
checking whether the newly generated sparks exceed the limit, mapping the sparks exceeding the limit, taking an upper limit value if the sparks exceed the upper limit, and taking a lower limit value if the sparks exceed the lower limit;
s35, calculating the fitness values of the explosion sparks and the Cauchy variation sparks, comparing the fitness values with the fitness values of the fireworks, arranging all individuals in an ascending order according to the fitness values, selecting the first M individuals to enter next iteration, and adding 1 to the iteration times;
and S36, judging whether the current iteration number is smaller than the maximum iteration number T, if so, returning to the step S33 to continue the iteration, otherwise, terminating the calculation, and outputting the firework individual with the minimum fitness and the fitness value, wherein the firework individual with the minimum fitness is the optimal hidden layer node number and the optimal learning efficiency of the LSTM prediction model.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210633967.8A CN115018156A (en) | 2022-06-06 | 2022-06-06 | Short-term wind power prediction method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202210633967.8A CN115018156A (en) | 2022-06-06 | 2022-06-06 | Short-term wind power prediction method |
Publications (1)
Publication Number | Publication Date |
---|---|
CN115018156A true CN115018156A (en) | 2022-09-06 |
Family
ID=83073227
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202210633967.8A Pending CN115018156A (en) | 2022-06-06 | 2022-06-06 | Short-term wind power prediction method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN115018156A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117313043A (en) * | 2023-10-25 | 2023-12-29 | 四川大学 | Wind power generation power prediction method |
-
2022
- 2022-06-06 CN CN202210633967.8A patent/CN115018156A/en active Pending
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN117313043A (en) * | 2023-10-25 | 2023-12-29 | 四川大学 | Wind power generation power prediction method |
CN117313043B (en) * | 2023-10-25 | 2024-04-30 | 四川大学 | Wind power generation power prediction method |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Shamshirband et al. | A survey of deep learning techniques: application in wind and solar energy resources | |
Wang et al. | A review of deep learning for renewable energy forecasting | |
Lu et al. | Review of meta-heuristic algorithms for wind power prediction: Methodologies, applications and challenges | |
Pousinho et al. | A hybrid PSO–ANFIS approach for short-term wind power prediction in Portugal | |
CN112733462B (en) | Ultra-short-term wind power plant power prediction method combining meteorological factors | |
JP5888640B2 (en) | Photovoltaic power generation prediction apparatus, solar power generation prediction method, and solar power generation prediction program | |
CN111401755B (en) | Multi-new-energy output scene generation method, device and system based on Markov chain | |
Wang et al. | A regional pretraining-classification-selection forecasting system for wind power point forecasting and interval forecasting | |
CN114298377A (en) | Photovoltaic power generation prediction method based on improved extreme learning machine | |
CN115310669A (en) | Ultra-short-term wind power prediction method based on secondary decomposition and IWOA-LSSVM | |
CN115018156A (en) | Short-term wind power prediction method | |
CN116702937A (en) | Photovoltaic output day-ahead prediction method based on K-means mean value clustering and BP neural network optimization | |
CN116960978A (en) | Offshore wind power prediction method based on wind speed-power combination decomposition reconstruction | |
CN114429248A (en) | Transformer apparent power prediction method | |
CN113435595A (en) | Two-stage optimization method for extreme learning machine network parameters based on natural evolution strategy | |
CN113128666A (en) | Mo-S-LSTMs model-based time series multi-step prediction method | |
CN117132132A (en) | Photovoltaic power generation power prediction method based on meteorological data | |
CN116681154A (en) | Photovoltaic power calculation method based on EMD-AO-DELM | |
CN116205359A (en) | Method, device, terminal and storage medium for predicting daily power load | |
CN115796327A (en) | Wind power interval prediction method based on VMD (vertical vector decomposition) and IWOA-F-GRU (empirical mode decomposition) -based models | |
CN115456286A (en) | Short-term photovoltaic power prediction method | |
CN115759458A (en) | Load prediction method based on comprehensive energy data processing and multi-task deep learning | |
Zhang et al. | A Novel Combined Model Based on Hybrid Data Decomposition, MSWOA and ENN for Short-term Wind Speed Forecasting | |
Zhang et al. | Temperature prediction and analysis based on improved GA-BP neural network | |
CN113962454A (en) | LSTM energy consumption prediction method based on dual feature selection and particle swarm optimization |
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 |