CN111310884A - Optimal layout method of wind turbine generator based on data-driven evolutionary algorithm - Google Patents
Optimal layout method of wind turbine generator based on data-driven evolutionary algorithm Download PDFInfo
- Publication number
- CN111310884A CN111310884A CN202010111340.7A CN202010111340A CN111310884A CN 111310884 A CN111310884 A CN 111310884A CN 202010111340 A CN202010111340 A CN 202010111340A CN 111310884 A CN111310884 A CN 111310884A
- Authority
- CN
- China
- Prior art keywords
- wind turbine
- individual
- wind
- data
- individuals
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 43
- 238000005457 optimization Methods 0.000 claims abstract description 41
- 238000009826 distribution Methods 0.000 claims abstract description 16
- 238000013528 artificial neural network Methods 0.000 claims abstract description 12
- 238000012360 testing method Methods 0.000 claims description 47
- 230000003044 adaptive effect Effects 0.000 claims description 31
- 239000013598 vector Substances 0.000 claims description 12
- 230000006978 adaptation Effects 0.000 claims description 8
- 238000012549 training Methods 0.000 claims description 7
- 238000009499 grossing Methods 0.000 claims description 3
- 238000012935 Averaging Methods 0.000 claims 2
- 238000002790 cross-validation Methods 0.000 claims 1
- 238000007781 pre-processing Methods 0.000 claims 1
- 230000006870 function Effects 0.000 abstract description 20
- 230000000694 effects Effects 0.000 abstract description 11
- 238000010248 power generation Methods 0.000 abstract description 5
- 238000010801 machine learning Methods 0.000 abstract description 4
- 230000009466 transformation Effects 0.000 abstract description 2
- 230000007246 mechanism Effects 0.000 description 15
- 238000004364 calculation method Methods 0.000 description 10
- 238000012545 processing Methods 0.000 description 9
- 230000035772 mutation Effects 0.000 description 8
- 230000008901 benefit Effects 0.000 description 4
- 230000010354 integration Effects 0.000 description 4
- 238000012795 verification Methods 0.000 description 4
- 230000009286 beneficial effect Effects 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 238000011144 upstream manufacturing Methods 0.000 description 3
- 238000010276 construction Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000011156 evaluation Methods 0.000 description 2
- 238000012804 iterative process Methods 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- 238000009827 uniform distribution Methods 0.000 description 2
- 230000008859 change Effects 0.000 description 1
- 230000000052 comparative effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 230000010429 evolutionary process Effects 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000000717 retained effect Effects 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 230000003746 surface roughness Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- 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
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/06—Energy or water supply
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
- Y02E40/00—Technologies for an efficient electrical power generation, transmission or distribution
- Y02E40/70—Smart grids as climate change mitigation technology in the energy generation sector
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Economics (AREA)
- General Physics & Mathematics (AREA)
- Strategic Management (AREA)
- General Health & Medical Sciences (AREA)
- Human Resources & Organizations (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- General Business, Economics & Management (AREA)
- Marketing (AREA)
- Tourism & Hospitality (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Public Health (AREA)
- Water Supply & Treatment (AREA)
- Primary Health Care (AREA)
- Development Economics (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Operations Research (AREA)
- Quality & Reliability (AREA)
- Wind Motors (AREA)
Abstract
The invention discloses a data-driven evolutionary algorithm-based optimal layout method for a wind turbine generator, and belongs to the technical field of power generation, power transformation or power distribution. According to the method, a constraint target optimization model for optimizing the layout of the wind power plant units is established by taking the output power of a maximized wind power plant as a target function and taking the safety distance between wind turbines and the regional limitation of the wind power plant as constraint conditions; the method comprises the steps of solving a wind turbine generator layout optimization model based on an improved parameter self-adaptive differential evolution algorithm, adopting a machine learning model generalized regression neural network as a proxy model of an optimization objective function, and adopting a data driving method to improve the iteration efficiency of the evolution algorithm. The method has high optimization efficiency and obvious effect on a complex layout optimization model.
Description
Technical Field
The invention relates to a wind turbine micro-site selection technology, in particular to a wind turbine optimal layout method based on a data-driven evolutionary algorithm, and belongs to the technical field of power generation, power transformation or power distribution.
Background
Wind power generation is a renewable energy power generation form which is technically mature and environmentally friendly and has the greatest prospect of scale development and commercialization development, and wind power, one of the main renewable energy sources, is expected to keep steady growth for a long time.
In the planning of the wind power plant, the position of each wind turbine is required to be reasonably arranged in the wind power plant area. In a wind power plant, a wind turbine obtains energy from wind and forms a wake zone with a reduced wind speed at the downstream of the wind turbine, and the wake zone develops downstream along the wind direction. If the downstream wind turbine is located in the wake area of the upstream wind turbine, the input wind speed of the downstream wind turbine is lower than that of the upstream wind turbine, and the phenomenon called wake effect influences the efficiency of the downwind wind turbine, so that the annual energy generation of the wind power plant is influenced, and the economic benefit of the wind power plant is reduced. Therefore, the goal of researching a wind turbine layout optimization method of a wind power plant to reduce wake loss and further improve the productivity of the wind power plant becomes a hot spot of research of many scholars in recent years.
At present, most of wind power plant unit layout optimization methods are to improve and apply certain heuristic algorithm, with the expansion of a wind power plant limited area and the increase of the number of fans, a wind power plant layout optimization model develops towards the direction of refinement and accuracy, the optimization model needs to consider more constraint limiting conditions and accurately calculate a complex objective function, and the problems of overhigh calculation cost and poor optimization capability occur by adopting the traditional heuristic algorithm. In view of the above, the invention provides a wind power plant unit layout optimization method based on a data-driven evolutionary algorithm.
Disclosure of Invention
The invention aims to provide a wind turbine generator optimal layout method based on a data-driven evolutionary algorithm aiming at the defects of the background technology, and aims at a wind power plant layout optimization model taking a wake effect and taking the output power of a wind power plant as a target function, the solving time of the wind power plant layout optimization model is reduced through a parameter self-adaptive differential evolutionary algorithm, and the technical problem of high calculation cost of the existing wind power plant layout optimization model is solved.
The method is characterized in that a machine learning model Generalized Regression Neural Network (GRNN) is built by using historical data in the evolutionary process of an evolutionary algorithm, and the GRNN plays a role of an objective function proxy model. The method has the advantages that the advantages are shown under the condition that the number of wind turbines is large, the total capacity of the wind power plant can be greatly improved, and the evolution rate of the evolution algorithm is effectively improved.
The invention adopts the following technical scheme for realizing the aim of the invention:
a wind turbine generator optimal layout method based on a data-driven evolutionary algorithm comprises the following 3 steps.
(1) Establishing a wind power plant wind turbine generator layout optimization model:
the optimal layout of the wind turbines in the wind farm takes the maximization of the output power of N wind turbines in the wind farm as an optimization target, and the microscopic address of each wind turbine is expressed by coordinates (x)k,yk) Indicating that the output power of each wind turbine is PkShown. Under a model based on coordinates, the first two constraint conditions enable the wind turbine to be located in the limited area range of the wind power plant, and the wind turbine is prevented from being too close to the edge of the wind power plant and influencing the construction and operation of projects such as nearby roads and the like; the third constraint ensures that the distance between the fan k and any other fan is not less than five times of the radius R of the fan, so that adjacent fans are not too close to each other, and safety accidents are avoided.
(2) Constructing a parameter Adaptive Differential Evolution (ADE): on the basis of a basic differential evolution algorithm, all main links of a parameter self-adaptive differential evolution algorithm are sequentially constructed, and the method comprises the following steps: (21) population initialization, (22) mutation operation, (23) crossover operation, (24) selection operation, and (25) parameter adaptation mechanism.
(21) Population initialization:
the initial population is recorded as:
{xi,0＝(x1,i,0,x2,i,0,...,xD,i,0)|i＝1,2,...,NP} (2)，
in the formula (2), D is the number of decision variables in an individual; i is the serial number mark of the individual in the population; NP is one in the populationThe number of the bodies represents the size of the population; 0 characterizes the current starting population. Arbitrary element x in an individualt,i,0(t ═ 1,2, …, D) in the range
The values are randomly taken in uniform distribution. Under the wind power plant layout optimization model, each wind turbine k uses a two-dimensional coordinate (x)k,yk) Representing the position, each individual, i.e., each solution vector, is represented by the formula { (x)1,y1),(x2,y2),…,(xN,yN) The coding mode of the method represents a layout method, and the dimension of each individual is 2 × N, wherein N is the number of wind turbines. The size of the population is NP and the information contained by the population is recorded using a matrix NP × 2N.
(22) Mutation operation:
furthermore, a new variation strategy is provided for the layout optimization problem of wind turbine generators in the wind power plant, and each target individual x in the g-th generation of population is subjected to sequential variationi,g(i-1, 2, …, NP) to generate variant individual v according to formula (3)i,g(i＝1,2,…,NP)：
Wherein x isbest,gThe best individual in the g-th generation population characterizes the candidate layout with the largest wind farm output power, and the variance is a difference vector which is different from the target individual. F1i，F2iIs a scaling factor with an adaptive mechanism, x for each target individuali,gAll independently by mean value μF1，μF2The normal distribution of the standard deviation sigma takes a random value. F1i，F2iThe value range of (a) is determined by engineering problems. x is the number ofp1,gIs an individual randomly selected from the g-th generation population, xp2,gThe wind turbines are disorganized and rearranged in sequence by xp1,gThe generated new individual, randint (0,1), is randomly valued between integers 0 and 1, and the purpose of setting (-1) ^ randint (0,1) is to increase the randomness of the search direction.
(23) And (3) cross operation:
furthermore, a new variation strategy is provided for the layout optimization problem of the wind turbine, and each target individual x in the g generation is subjected to sequential variationi,gAnd variant individuals v corresponding theretoi,gPerforming crossover operation to generate test individual ui,g(i＝1,2,…,NP)：
The solution vector is expressed by the formula { (x) in reference to the above1,y1),(x2,y2),…,(xN,yN) The coding mode of the individual i represents a layout method, in the formula (4), j is a serial number mark of a j-th typhoon machine in the individual i, and the value range of j is 1 to N.
The traditional minimum crossing unit of the crossing operation is a single decision variable, and considering that the individual code has special practical significance under a wind turbine layout optimization model of a wind power plant, when the decision variable in an individual is replaced, the coordinate of a certain wind turbine is actually changed, and the position of the wind turbine in the wind field is moved, so that the overall coordinate of a single wind turbine is set to be the minimum crossing unit. For each target individual xi,gIndependent random generation of array jrandIncluding the serial numbers of all the randomly selected wind turbines to be moved, the length ujrand of the arrayiRepresents the number of wind turbines to be moved, ujrandiIs a variable parameter with an adaptive mechanism, x for each target individuali,gIndependently, the value [ u ] is randomly taken by the positive distribution of the mean value mu jrand and the standard deviation sigma2*j-1,i,g,u2*j,i,g]For storing the array of the coordinates of the jth wind turbine in the ith individual of the ith generation group after the cross operation, [ v ]2*j-1,i,g,v2*j,i,g]For storing the array of j wind turbine coordinates in the variation individual of the ith individual in the g generation population, [ x ]2*j-1,i,g,x2*j,i,g]For storing the array of the jth wind turbine coordinate in the ith individual of the ith generation group, the cross operation is performed when j belongs to a random array jrandIs derived from [ v ]2*j-1,i,g,v2*j,i,g]Obtaining the cross variation characteristics of the tested individuals, and j does not belong to the random array jrandThe previous characteristics of the test subject were maintained.
(24) Selecting operation:
further, with respect to the test subject u obtained in (23)i,gChecking, if the layout does not meet the limitation condition of the wind field layout, returning to the mutation operation in (22), reducing the mutant by one time, keeping the other parameters unchanged, and generating a new test individual ui,gRepeating the above steps until ui,gIs a feasible solution.
Respectively evaluating test individuals u by using objective functions in wind turbine generator layout optimization modeli,gAnd target individual xi,gIf f (u) is an adaptation value ofi,g)>f(xi,g) Then, using the test subject ui,gSubstitution of target individual xi,gTarget individual x to become the next generationi,g+1On the contrary, the current target individual xi,gRetention to the next generation, xi,g+1＝xi,g。
(25) A parameter updating mechanism:
in each generation, x is for each individuali,gRespectively, mean value of μF1，μF2Normal distribution of standard deviation sigma randomly generates F1i，F2i。
F1i＝randn(μF1,1) (5)，
F2i＝randn(μF2,1) (6)，
Array SF1，SF2For successful storage (so that f (u)i,g)>f(xi,g) F of (b)1i，F2i. After each generation of updating is finished, if the array SF1，SF2If the length of (d) exceeds a set threshold, the top entry, μ, of the array is sequentially fetchedF1，μF2Updated according to the following formula:
μF1＝meanA(SF1) (7)，
μF2＝meanA(SF2) (8)，
further, with F1i，F2iThe adaptation mechanism is similar, in each generation, for each individual xi,gRandomly generating ujrand by taking the mean value as mu jrand and the positive distribution of standard deviation sigmai，ujrandiHas a value range of [1, N]Any integer within.
ujrandi＝round(randn(μjrand,1)) (9)。
In the formula (9), round () is a rounding function.
The array Sjrand is used for storing successful ujrandi. After each generation of updating is finished, if the length of the array Sjrand exceeds a set threshold value, the items at the front of the array are taken out in sequence, and the mu jrand is updated according to the following formula:
μjrand＝meanA(Sjrand) (10)。
(26) and recording the maximum iteration times as maxGEs, when the iteration times are smaller than the maxGEs, respectively performing mutation operators, crossover operators and selection operators, correcting parameters based on a parameter self-adaptive mechanism, and stopping evolution when the iteration times reach the maxGEs.
(3) Constructing a data-driven evolution algorithm (ADE-GRNN) fusing the generalized regression neural network: using data generated during the iterative process of the algorithm (each generation of test individuals u)i,gAnd its corresponding adaptive value f (u) calculated from the objective functioni,g) Establishing and updating a generalized regression neural network, taking the generalized regression neural network as a proxy model of an objective function of a wind power plant layout optimization model, and only carrying out actual objective function calculation on the wind turbine layout with a higher proxy model evaluation value;
(31) data processing:
further, consider the equation in { (x)1,y1),(x2,y2),…,(xN,yN) The coding mode of the method represents a wind turbine layout, each pair of coordinates does not have any relation, and each pair of coordinates represents that the wind turbine can be located at any position of a wind field. Such individuals are directly used for GRNN (Generalized Regression Neural Network) training, and obviously do not well incorporate the layout information of the wind farm.Therefore, before the GRNN is driven by data, all test individuals in each generation are subjected to data processing, x coordinates of all wind turbine coordinates are sorted firstly, wind turbines with small x coordinates are sorted in front, then wind turbines with the same x coordinates are sorted in y coordinates, and the wind turbines with small y coordinates are sorted in front.
(32) Determining GRNN parameters:
the smoothing factor ω of GRNN needs to be set manually, for which all data-processed test subjects and their fitness values, referred to as data set S, generated in the ADE algorithm are first recorded in their entirety once. Dividing the data set into ten parts, training 9 parts in the ten parts in turn for 1 part for verification, taking the average value of results obtained after 10 times as the estimation of the precision of the proxy model, comparing the precision of the proxy model under different omega values, and finally selecting the omega value by using the data set through cross verification.
(33) ADE-GRNN algorithm process
(331) Further, accumulating samples by using an original ADE algorithm, and storing all the data-processed test individuals u in each iteration when the number of the samples does not reach a set value SNi,gAnd corresponding adaptation value f (u)i,g)。
(332) Further, when the number of samples is greater than or equal to the SN, the next SN samples in the data set are retained, and the GRNN with the number of samples being the SN is established.
(333) Further, NP test individuals generated in the iteration of the ADE algorithm are all put into GRNN after data processing to predict adaptive values, test individuals which are fifty percent of the predicted adaptive values are screened out, real adaptive values of the test individuals are calculated, a selection link is carried out on the screened test individuals and target individuals left in the last iteration, and the screened test individuals and the corresponding real adaptive values are put into a training set at the end of each iteration.
The ADE-GRNN algorithm completely reserves the evolution mechanism in the ADE algorithm, and only integrates the GRNN prediction adaptive value into the selection operation.
By adopting the technical scheme, the invention has the following beneficial effects:
(1) compared with the traditional evolutionary algorithm objective function which is high in calculation cost and low in calculation efficiency during repeated iteration, the ADE-GRNN disclosed by the application adopts a data driving idea, introduces a machine learning model GRNN, establishes a proxy model of the objective function by using data generated in the iteration process, preliminarily evaluates the wind turbine layout adaptive value through the proxy model for screening, reduces the times of calculating the adaptive value by using an actual objective function, improves the calculation efficiency, and effectively increases the output power of a wind power plant for the layout planning design of the wind power plant.
(2) The method includes the steps that differential vectors representing differences among individuals are introduced into a differential evolution algorithm to obtain variant individuals, test individuals are obtained by taking position coordinates of wind turbines as minimum cross units in cross operation, when the test individuals do not meet wind field layout limiting conditions, the variant operation is carried out again after the differential vectors are reduced, the cross operation is carried out until the obtained test individuals meet the wind field layout limiting conditions, ADE-GRNN optimization tends to be fast and accurate through parameter self-adaptive adjustment after iteration is finished every time, when the wind turbine layout problem of the wind power plant is solved, a good effect is shown, and especially the output power of the wind power plant can be greatly improved under the condition that the number of the wind turbines is increased.
Drawings
Fig. 1 is a general flow chart of the present invention.
Fig. 2 is a schematic diagram of the wake generated by fans i to j in the wind direction θ.
Fig. 3 is a flow chart of the ADE algorithm.
Fig. 4 shows layout information included in an individual in the ADE algorithm.
Fig. 5 shows the population structure in the ADE algorithm.
FIG. 6 illustrates wind turbine layout information included in an individual before and after data processing.
FIG. 7 is a flow chart of the ADE-GRNN algorithm.
FIG. 8 illustrates the optimization effect of the wind turbine layout ADE-GRNN when the number N of wind turbines is 25.
FIG. 9 illustrates the optimization effect of the wind turbine layout ADE-GRNN when the number N of wind turbines is 60, respectively.
FIG. 10 illustrates the optimization effect of the wind turbine layout ADE-GRNN when the number N of wind turbines is 100, respectively.
FIG. 11 is a comparison of the run times of the ADE-GRNN algorithm and the ADE algorithm at the same number of iterations.
FIG. 12 shows the variation trend of the adaptive parameter μ jump in the crossover operation of the ADE-GRNN algorithm.
Detailed Description
The technical scheme of the invention is explained in detail in the following with reference to the attached drawings.
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
The method for optimizing the layout of the wind turbine generator based on the data-driven evolutionary algorithm takes the output power of a maximized wind power plant as an optimization target, considers the wake effect among wind turbines, calculates the output power of the wind turbines through numerical integration, performs optimization solution based on a parameter-adaptive differential evolutionary algorithm, and improves the efficiency of the evolutionary algorithm by using a data-driven model. Referring to fig. 1, a flow chart of a method for optimal layout of wind turbines based on a data-driven evolutionary algorithm is shown, where the method includes the following steps:
(1) establishing a wind power plant wind turbine generator layout optimization model:
(11) in this embodiment, for any wind turbine, at a given wind direction θ, the wind speed v follows a Weibull distribution, expressed as:
in the formula (1), p (-) is a Weibull probability density function, v is wind speed, and c (theta) and k (theta) are a proportion parameter and a shape parameter under the wind direction theta respectivelyhThe wind direction being in the interval [ theta ]h,θh+1) The probability of wind direction from west to east is defined as 0 deg., and a counter-clockwise rotation, for example a wind direction from south to north will be defined as 90 deg.. As can be seen from Table 1, in this wind regime, the wind direction is mainly distributed over 120-In the 225 ° interval, it is therefore the objective direction of wind field layout optimization in this embodiment to emphasize the attenuation of wake losses in this part of the wind direction.
TABLE 1 wind conditions
(12) Establishing a wake loss model: and quantifying the wake effect among fans by adopting a Jensen wake model. The wake behind the turbine may be approximately conical. FIG. 2 shows a site at (x)k,yk) The cross section of the conical wake generated by fan k, a is the conical apex, κ is the entrainment constant, the angle γ is equal to arctan (κ), and the distance from point a to the turbine center is R/κ. The change in speed of fan m caused by the wake of fan k is denoted VDk,mAnd is calculated by:
dk,m＝|(xk-xm)cosθ+(yk-ym)sinθ| (3)，
VDk,m＝1-vdn/vup＝2a/(1+κdk,m/R)2(4)，
in the formulae (2) to (4), a is the axial inductance, CTIs constant thrust coefficient, z is wind turbine tower height, z0Is the surface roughness, dk,mIs the distance of the projection of the upwind fan and the downwind fan in the wind direction, vdnIs the input wind speed, v, of the downstream wind turbine mupIs the input wind speed of the upstream wind generator k and R is the fan radius. On the basis, the fan m is influenced by all other fans, and the total speed loss is as follows:
the proportional parameter of the Weibull distribution is affected by the wake effect, updated c (θ) (denoted as c)m(θ)) is calculated from the following formula:
cm(θ)＝c(θ)×(1-VDm),m＝1,2,...,N (6)。
(13) establishing a wind turbine power curve model: fitting a Logistic function according to actual power output data, wherein a fan power model is represented as follows:
in equation (7), P is the output power of the fan, α is a constant, and when the input wind speed v is greater than the cut-out wind speed v, as shown in equation (7)coIn time, the fan is shut down for protection; when the input wind speed v is less than the cut-in wind speed vciIn time, the fan will not work. When the wind speed is at the rated wind speed vrTo cut-out wind speed vcoIn between, the control system of the fan will maintain the rated output power Pr。
(14) Calculating the output power of the wind turbine based on numerical integration: for a single fan k at an arbitrary position, the output power at a constant wind direction θ calculates the expected output power of the fan i by integrating the product of the equation (1) and the equation (7) with respect to the wind speed v in the range [0, + ∞ ] and with respect to the wind direction θ in the range [0 °,360 ° ]:
in equation (8), p (θ) is a probability density function of θ. Since f (v) is a piecewise function, the integral to v in equation (8) can be divided into four intervals: [0, v ]ci)，[vci,vr)，[vr,vco)，[vco, + ∞). In the interval [0, vci) And [ v ]co, + ∞) where v is constantly equal to 0, so equation (8) is equal to 0; in the interval [ vr,vco) In (iii), formula (8) can be simplified to formula (9):
in the interval [ vci,vr) In (2), it is very difficult to directly perform integration, and thus a numerical integration technique, i.e., riemann's sum, is employed. The wind speed v is equally divided into S intervals of the same interval length: [ v ] of0,v1)，[v1,v2)，…， [vS-1,vS) Wherein v is0＝vci，vS＝vr. Similarly, the wind direction θ is equally divided into H intervals having the same interval length: [ theta ] of0,θ1)，[θ1,θ2)，…，[θH-1,θH) Wherein theta0＝0°，θH360 deg.. After processing, the expected output power of fan k can be obtained in the following discrete form:
(15) establishing a wind power plant fan layout optimization model: and establishing a target function and a limiting condition of a wind turbine layout optimization model of the wind power plant by taking the output power of N wind turbines of the wind power plant as an optimization target. The first two constraints enable the wind turbine to be located in the region range of the wind power plant, and the wind turbine is guaranteed not to be too close to the edge of the wind power plant and influence the construction and operation of projects such as nearby roads. The third constraint ensures that the distance between the wind turbine k and any other wind turbine is not less than five times of the radius of the wind turbine, so that adjacent wind turbines are not too close to each other, and safety accidents occur.
(2) Constructing a parameter adaptive differential evolution Algorithm (ADE):
on the basis of a basic differential evolution algorithm, main links of a parameter self-adaptive differential evolution algorithm are designed in sequence: population initialization, mutation operation, crossover operation, selection operation and parameter adaptive mechanism, and fig. 3 shows the basic flow of ADE algorithm;
(21) population initialization: the initial population is recorded as:
{xi,0＝(x1,i,0,x2,i,0,...,xD,i,0)|i＝1,2,...,NP} (12)，
in the formula (12), D is the number of decision variables in an individual; i is the serial number mark of the individual in the population; NP is the number of individuals in the population and represents the size of the population; 0 characterizes the current starting population. Arbitrary element x in an individualt,i,0(t ═ 1,2, …, D) in the range
The values are randomly taken in uniform distribution. Under the wind power plant layout optimization model, each wind turbine k uses a two-dimensional coordinate (x)k,yk) Representing the position, each individual, i.e., each solution vector, is represented by the formula { (x)1,y1),(x2,y2),…,(xN,yN) The coding mode of the method represents a layout method, and the dimension of each individual is 2 × N, wherein N is the number of wind turbines. The size of the population is NP and the information contained by the population is recorded using a matrix NP × 2N. Fig. 4 shows layout information contained by an individual in the ADE algorithm. Fig. 5 shows the population structure in the ADE algorithm.
(22) Mutation operation:
aiming at the layout optimization problem of the wind turbine generator, a new variation strategy is provided for each target individual x in the g generation population in sequencei,g(i-1, 2, …, NP) producing variant individuals v by the formula (13)i,g(i＝1,2,…,NP)：
xbest,gThe best individual in the g-th generation population characterizes the candidate layout with the largest wind farm output power, and the variance is a difference vector which is different from the target individual. F1i，F2iIs a scaling factor with an adaptive mechanism, x for each target individuali,gAre all independentlyMean value of μF1，μF2(initial values are all 1), and the normal distribution with the standard deviation sigma of 1 takes random values. F1i，F2iAll the value ranges of (1), (0.1) and (2)]。 xp1,gIs an individual randomly selected from the g-th generation population, xp2,gIs formed by xp1,gThe wind turbine sequence is disturbed, new individuals are rearranged, randint (0,1) refers to random values between integers of 0 and 1, and the purpose of setting (-1) ^ randint (0,1) is to increase the randomness of the search direction.
(23) And (3) cross operation:
a new variation strategy is provided for the layout optimization problem of the wind turbine generator, and each target individual x in the g generation is subjected to sequential variationi,gAnd variant individuals v corresponding theretoi,gPerforming crossover operation to generate test individuals ui,g(i＝1,2,…,NP)：
Consider the solution vector above in the equation { (x)1,y1),(x2,y2),…,(xN,yN) And j is a serial number mark of a jth fan in the individual, and the value range of j is 1 to N.
The minimum crossing unit of the traditional crossing operation is a single decision variable, and considering that the individual code has special practical significance under a wind turbine layout optimization model, when the decision variable in the individual is changed, the coordinate of a certain wind turbine is actually changed, and the position of the wind turbine in a wind field is moved. Therefore, the overall coordinate of a single wind turbine is set to be the minimum crossing unit. For each target individual xi,gIndependent random generation of array jrandIncluding the serial number of all the randomly selected wind turbines to be moved, the length ujrand of the arrayiRepresents the number of wind turbines to be moved, ujrandiIs a variable parameter with an adaptive mechanism, x for each target individuali,gBoth are independently randomly valued by a positive distribution with mean μ jrand (initial value of 5) and standard deviation σ of 1.
(24) Selecting operation:
for test subject u obtained in (23)i,gChecking, if the layout does not meet the limitation condition of the wind field layout, returning to the mutation operation in (22), reducing the mutant by one time, keeping the other parameters unchanged, and generating a new test individual ui,gRepeating the above steps until ui,gIs a feasible solution.
Comparative test subjects ui,gAnd target individual xi,gIf f (u) is an adaptation value ofi,g)>f(xi,g) Then, using the test subject ui,gSubstitution of target individual xi,gTarget individual x to become the next generationi,g+1On the contrary, the target individual x of the current generationi,gRetention to the next generation, xi,g+1＝xi,g。
(25) A parameter updating mechanism:
in each generation, x is for each individuali,gRespectively, mean value of μF1，μF2(initial values are all 1), and F is randomly generated by normal distribution with standard deviation σ of 11i，F2i，F1i，F2iHas a value range of [0.1,2 ]]。
F1i＝randn(μF1,1) (15)，
F2i＝randn(μF2,1) (16)，
Array SF1，SF2For successful storage (so that f (u)i,g)>f(xi,g) F of (b)1i，F2i. After each generation of updating is finished, if the array SF1，SF2If the length of (d) exceeds a set threshold, the top entry, μ, of the array is sequentially fetchedF1，μF2Updated according to the following formula:
μF1＝meanA(SF1) (17)，
μF2＝meanA(SF2) (18)，
and F1i，F2iThe adaptation mechanism is similar, in each generation, for each individual xi,gTo all areThe value μ jrand (initial value 5) is generated randomly by a positive-distribution of standard deviation σ ═ 1i，ujrandiHas a value range of [1, N]Any integer within.
ujrandi＝round(randn(μjrand,1)) (19)。
In equation (19), round () is a rounding function.
The array Sjrand is used for storing successful ujrandi. After each generation of updating is finished, if the length of the array Sjrand exceeds a set threshold value, the items at the front of the array are taken out in sequence, and the mu jrand is updated according to the following formula:
μjrand＝meanA(Sjrand) (20)。
(26) and recording the maximum iteration times as maxGEs, when the iteration times are smaller than the maxGEs, respectively performing mutation operators, crossover operators and selection operators, correcting parameters based on a parameter self-adaptive mechanism, and stopping evolution when the iteration times reach the maxGEs.
(3) Constructing a data-driven evolution algorithm (ADE-GRNN) fusing the generalized regression neural network: using data generated during the iterative process of the algorithm (each generation of test individuals u)i,gAnd its corresponding adaptive value f (u) calculated from the objective functioni,g) Establishing and updating a generalized regression neural network, taking the generalized regression neural network as a proxy model of an objective function of a wind power plant layout optimization model, and only carrying out actual objective function calculation on the wind turbine layout with a higher proxy model evaluation value;
(31) data processing:
further, consider the equation in { (x)1,y1),(x2,y2),…,(xN,yN) The coding mode of the method represents a wind turbine layout, each pair of coordinates does not have any relation, and each pair of coordinates represents that the wind turbine can be located at any position of a wind field. Such individuals are directly used for GRNN training, and obviously, the layout information of the wind power plant is not well integrated. Before the GRNN is driven by data, all test individuals in each generation are subjected to data processing, x coordinates of all wind turbine coordinates are sorted firstly, wind turbines with small x coordinates are sorted in front, and then wind power with the same x coordinates is subjected to wind powerThe machine carries out sorting on y coordinates, and sorting with smaller y coordinates is closer to the front. FIG. 6 illustrates wind turbine layout information variations contained by data processing individuals.
(32) Determining GRNN parameters:
the smoothing factor ω of GRNN needs to be set manually, for which all data-processed test subjects and their fitness values, referred to as data set S, generated in the ADE algorithm are first recorded in their entirety once. Dividing the data set into ten parts, training 9 parts in the ten parts in turn for 1 part for verification, taking the average value of results obtained after 10 times as the estimation of the precision of the proxy model, comparing the precision of the proxy model under different omega values, and finally selecting the omega value by using the data set through cross verification.
(33) ADE-GRNN algorithm process
(331) Further, accumulating samples by using an original ADE algorithm, and storing all the data-processed test individuals u in each iteration when the number of the samples does not reach 5000i,gAnd corresponding adaptation value f (u)i,g)。
(332) Further, when the number of samples is larger than or equal to 5000, 5000 samples behind the data set are reserved, GRNN with the number of samples being 5000 is established, NP test individuals generated in the iteration of the ADE algorithm are all placed into GRNN after data processing to predict adaptive values, fifty percent of test individuals before the predicted adaptive values are screened out, real adaptive values of the test individuals are calculated, a selection link is carried out on the screened test individuals and target individuals selected in the previous iteration, and finally the screened test individuals and the corresponding real adaptive values are placed into a training set in each iteration.
The ADE-GRNN algorithm completely reserves the evolution mechanism in the ADE algorithm, and only integrates the GRNN prediction adaptive value into the selection operation. FIG. 7 shows a flow chart of the ADE-GRNN algorithm.
FIGS. 8, 9 and 10 are diagrams illustrating layout effects of wind turbines when the number N of the wind turbines is 25, 60 and 100 respectively. It can be seen that under the condition of the wind resource distribution (the wind direction is mainly concentrated at 120-225 degrees), the wind power field wind turbine layout obtained by the ADE-GRNN solution can greatly increase the distance between the wind turbines in the wind direction.
Fig. 11 shows the ADE algorithm of the ADE-GRNN algorithm and the ADE algorithm of the non-fusion machine learning model GRNN, and the comparison of the solving time under the same iteration number shows that when N of the ADE-GRNN algorithm based on the data driving is larger, the solving speed can be improved by about 40% compared with the ADE algorithm, which shows that the idea of improving the iteration efficiency of the heuristic algorithm based on the data driving idea is effective, and the idea and the solving way of reducing the calculation cost are provided for the heuristic algorithm with larger calculation amount. FIG. 12 shows the variation trend of the parameter μ jump in the crossover operation under the parameter adaptive mechanism of the ADE-GRNN algorithm, and it can be seen that the parameter values adaptively converge from the initial value to 1, which shows that the algorithm moves one wind turbine each time the crossover operator is performed, which is most beneficial to the generation of good quality test individuals.
The method is suitable for optimizing the layout of the wind power generation sets of the wind power plant based on the continuous coordinates, can improve the total output power of the wind power plant from the technical aspect, effectively reduces the calculation loss of solving a complex optimization model, improves the evolution rate, and provides beneficial reference for related optimization problems.
Claims (9)
1. A wind turbine generator optimal layout method based on a data-driven evolutionary algorithm is characterized in that a parameter self-adaptive differential evolutionary algorithm is adopted to optimize a population representing a wind turbine generator layout scheme, when an optimized population obtained by current iterative operation of the parameter self-adaptive differential evolutionary algorithm meets the minimum sample number requirement, all test individuals generated by cross operation in the current iterative operation are preprocessed, a target function of the parameter self-adaptive differential evolutionary algorithm is pre-evaluated by a generalized regression neural network to obtain predicted adaptive values of all test individuals generated by the current iterative operation, test individuals with good represented candidate layout positions are pre-screened according to the predicted adaptive values, the pre-screened test individuals and the optimized population generated by the last iterative operation are subjected to selection operation to generate a next generation optimized population, and the optimized population generated after each iterative operation and the real adaptive values of all individuals in the optimized population are used as parameters of the optimized generalized regression neural network A data set.
2. The optimal layout method of the wind turbine generator based on the data-driven evolutionary algorithm is characterized in that differential vectors representing differences among individuals are introduced into the parameter self-adaptive differential evolutionary algorithm to obtain variant individuals, test individuals are obtained by taking position coordinates of a wind turbine as a minimum cross unit in cross operation, when the test individuals do not meet wind field layout limiting conditions, the variant operation and the cross operation are carried out again after the differential vectors are reduced by one time until the obtained test individuals meet the wind field layout limiting conditions, and for each target individual in the population before optimization, the target individual of which the adaptive value does not exceed the test individuals is replaced by the test individual.
3. The optimal layout method of the wind turbine generator based on the data-driven evolutionary algorithm as claimed in claim 1, characterized in that the expression for representing individual-specific difference vectors to obtain variant individuals is introduced as follows:
wherein v isi,gThe variation individual of the ith individual in the population of the g generation, i is 1,2, …, NP, NP is the population scale, xbest,gFor the best individual in the population of the g generation, mutant is a difference vector, F1i、F2iAs scaling factor for the i-th individual variation, F1iIs given as μ on the basis of the mean valueF1And the normal distribution with standard deviation of sigma is randomly selected, F2iIs given as μ on the basis of the mean valueF2And the normal distribution with standard deviation of sigma is randomly selected, xp1,gFor randomly selected individuals from the population of the g generation, xp2,gRearranging x for disordering wind turbine sequencep1,gThe new individual generated, randint (0,1), is a random number between 0 and 1.
4. The optimal layout method for the wind turbine generator based on the data-driven evolutionary algorithm as claimed in claim 2, wherein the expression of the test individual is obtained by taking the position coordinates of the wind turbine as the minimum intersection unit in the intersection operation as follows:
j is serial number mark of wind turbine contained by individual, jrandIs an array randomly generated according to the serial number of each wind turbine in an individual, jrandIs randomly selected according to a positive distribution with a mean value of μ jrand and a standard deviation of σ, [ u [2*j-1,i,g,u2*j,i,g]For storing the array of the coordinates of the jth wind turbine in the ith individual of the ith generation group after the cross operation, [ v ]2*j-1,i,g,v2*j,i,g]For storing the array of j wind turbine coordinates in the variation individual of the ith individual in the g generation population, [ x ]2*j-1,i,g,x2*j,i,g]The coordinate array of the jth wind turbine in the ith individual of the ith generation group is stored.
5. The optimal layout method for the wind turbine generator based on the data-driven evolutionary algorithm is characterized in that the adaptive value is determined according to an objective function established by a parameter adaptive differential evolution algorithm, and the objective function is as follows:
wherein P is the output power of the wind turbine, PkIs the output power of the kth wind turbine, (x)k,yk) Is the position coordinate of the kth wind turbine, (x)m,ym) Is the position coordinate, x, of the mth wind turbinemax、xminAs maximum and minimum values, y, of the x-axis coordinate of the wind turbinemax、yminThe maximum value and the minimum value of the y-axis coordinate of the wind turbine are shown, N is the total number of the wind turbine, and R is the radius of the wind turbine.
6. The optimal layout method for wind turbines based on the data-driven evolutionary algorithm as claimed in claim 3, wherein after each iteration operation is finished, μF1According to muF1＝meanA(SF1) Update, muF2According to muF2＝meanA(SF2) Update meanA(. is an averaging operation, SF1、SF2Test individuals for storageF with adaptive value exceeding target individual adaptive value1i、F2iAn array of (2).
7. The optimal layout method for wind turbine generators based on data-driven evolutionary algorithm as claimed in claim 4, wherein the length ujrand of the array randomly generated according to the serial number of each wind turbine in the ith individualiAccording to ujrandi＝round(randn(μ jrand,1)) update, μ jrand being in terms of μ jrand meanA(Sjrand) update, randn(. cndot.) is a random number operation, round (. cndot.) is a rounding operation, meanA(. cndot.) is an averaging operation, and Sjrand is ujrand stored so that the adaptation value of the test individual exceeds that of the target individualiAn array of (2).
8. The optimal layout method of the wind turbine generator based on the data-driven evolutionary algorithm as claimed in claim 1, wherein the method for preprocessing all the test individuals generated by the crossover operation in the current iterative operation comprises the following steps: all the wind turbine position coordinates contained in each test individual are sorted from small to large according to the x-axis coordinate size, and the wind turbine position coordinates with the same x-coordinate are sorted from small to large according to the y-axis coordinate size.
9. The optimal layout method of the wind turbine generator based on the data-driven evolutionary algorithm is characterized in that the method for optimizing the parameters of the generalized regression neural network is as follows: and performing cross validation and selecting a smoothing factor value by using the precision estimation value calculated by the data set training result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010111340.7A CN111310884B (en) | 2020-02-24 | 2020-02-24 | Optimal layout method of wind turbine generator based on data-driven evolutionary algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010111340.7A CN111310884B (en) | 2020-02-24 | 2020-02-24 | Optimal layout method of wind turbine generator based on data-driven evolutionary algorithm |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111310884A true CN111310884A (en) | 2020-06-19 |
| CN111310884B CN111310884B (en) | 2023-05-16 |
Family
ID=71162193
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010111340.7A Active CN111310884B (en) | 2020-02-24 | 2020-02-24 | Optimal layout method of wind turbine generator based on data-driven evolutionary algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111310884B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113223633A (en) * | 2021-03-13 | 2021-08-06 | 宁波大学科学技术学院 | Width GRNN model-based papermaking process drain water quality prediction method |
| CN114065916A (en) * | 2021-11-11 | 2022-02-18 | 西安工业大学 | DQN-based agent training method |
| WO2023087521A1 (en) * | 2021-11-19 | 2023-05-25 | 中国华能集团清洁能源技术研究院有限公司 | Wind power plant layout optimization method based on mathematical programming |
| TWI833188B (en) * | 2022-03-30 | 2024-02-21 | 嘉凱能源科技有限公司 | Gas turbine unit evaluation and verification system, built-in program and its device |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104077496A (en) * | 2014-07-17 | 2014-10-01 | 中国科学院自动化研究所 | Intelligent pipeline arrangement optimization method and system based on differential evolution algorithm |
| CN104200097A (en) * | 2014-08-29 | 2014-12-10 | 重庆大学 | Wind power plant wind generation set layout site selection method |
| CN108258724A (en) * | 2018-01-22 | 2018-07-06 | 佛山科学技术学院 | A kind of wind power plant unit is laid out Multipurpose Optimal Method |
-
2020
- 2020-02-24 CN CN202010111340.7A patent/CN111310884B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104077496A (en) * | 2014-07-17 | 2014-10-01 | 中国科学院自动化研究所 | Intelligent pipeline arrangement optimization method and system based on differential evolution algorithm |
| CN104200097A (en) * | 2014-08-29 | 2014-12-10 | 重庆大学 | Wind power plant wind generation set layout site selection method |
| CN108258724A (en) * | 2018-01-22 | 2018-07-06 | 佛山科学技术学院 | A kind of wind power plant unit is laid out Multipurpose Optimal Method |
Non-Patent Citations (1)
| Title |
|---|
| 朱述宝: "含分布式风力发电的电力机组组合问题研究" * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113223633A (en) * | 2021-03-13 | 2021-08-06 | 宁波大学科学技术学院 | Width GRNN model-based papermaking process drain water quality prediction method |
| CN113223633B (en) * | 2021-03-13 | 2024-04-05 | 宁波大学科学技术学院 | Width GRNN model-based water quality prediction method for sewage discharge outlet in papermaking process |
| CN114065916A (en) * | 2021-11-11 | 2022-02-18 | 西安工业大学 | DQN-based agent training method |
| WO2023087521A1 (en) * | 2021-11-19 | 2023-05-25 | 中国华能集团清洁能源技术研究院有限公司 | Wind power plant layout optimization method based on mathematical programming |
| TWI833188B (en) * | 2022-03-30 | 2024-02-21 | 嘉凱能源科技有限公司 | Gas turbine unit evaluation and verification system, built-in program and its device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111310884B (en) | 2023-05-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111310884A (en) | Optimal layout method of wind turbine generator based on data-driven evolutionary algorithm | |
| CN105046374B (en) | A kind of power interval prediction technique based on core extreme learning machine model | |
| CN106286130B (en) | Wind turbines based on SCADA data yaw Optimization about control parameter method | |
| CN105631483A (en) | Method and device for predicting short-term power load | |
| CN110942205B (en) | Short-term photovoltaic power generation power prediction method based on HIMVO-SVM | |
| Li et al. | A hybrid assembly sequence planning approach based on discrete particle swarm optimization and evolutionary direction operation | |
| CN105787592A (en) | Wind turbine generator set ultra-short period wind power prediction method based on improved RBF network | |
| CN110264012A (en) | Renewable energy power combination prediction technique and system based on empirical mode decomposition | |
| CN111339713A (en) | Optimal design method and system for wind power plant, electronic device and storage medium | |
| Askarzadeh et al. | Wind power modeling using harmony search with a novel parameter setting approach | |
| CN109858665A (en) | Photovoltaic short term power prediction technique based on Feature Selection and ANFIS-PSO | |
| CN114021483A (en) | Ultra-short-term wind power prediction method based on time domain characteristics and XGboost | |
| CN113294297B (en) | Variable weight adjusting method for wind turbine generator nonlinear model prediction torque control | |
| CN108985323A (en) | A kind of short term prediction method of photovoltaic power | |
| CN111401659A (en) | Ultra-short-term or short-term photovoltaic power generation power prediction method based on case reasoning | |
| CN113962113A (en) | Optimized arrangement method and system for offshore wind farm fans | |
| CN113033012A (en) | Hierarchical data-driven wind power plant generated power optimization scheme | |
| CN112085335A (en) | Improved random forest algorithm for power distribution network fault prediction | |
| CN116896093A (en) | Online analysis and optimization method for grid-connected oscillation stability of wind farm | |
| CN115713029A (en) | Wind power plant stochastic model prediction optimization control method considering delay | |
| CN111191815B (en) | Ultra-short-term output prediction method and system for wind power cluster | |
| Wang et al. | Short term load forecasting: A dynamic neural network based genetic algorithm optimization | |
| Yin et al. | Forecasting the intrinsic viscosity of polyester based on improved extreme learning machine | |
| CN111859780A (en) | Micro-grid operation optimization method and system | |
| CN111626465A (en) | New energy power short-term interval prediction method and system |
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 |