WO2022151890A1 - 一种风电场发电量评估及微观选址模型建立方法 - Google Patents

一种风电场发电量评估及微观选址模型建立方法 Download PDF

Info

Publication number
WO2022151890A1
WO2022151890A1 PCT/CN2021/137870 CN2021137870W WO2022151890A1 WO 2022151890 A1 WO2022151890 A1 WO 2022151890A1 CN 2021137870 W CN2021137870 W CN 2021137870W WO 2022151890 A1 WO2022151890 A1 WO 2022151890A1
Authority
WO
WIPO (PCT)
Prior art keywords
wind
model
wind farm
power generation
micro
Prior art date
Application number
PCT/CN2021/137870
Other languages
English (en)
French (fr)
Inventor
易侃
张皓
张子良
王浩
Original Assignee
中国长江三峡集团有限公司
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by 中国长江三峡集团有限公司 filed Critical 中国长江三峡集团有限公司
Publication of WO2022151890A1 publication Critical patent/WO2022151890A1/zh

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F30/00—Computer-aided design [CAD]
    • G06F30/20—Design optimisation, verification or simulation
    • G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • 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—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
    • G06Q50/06—Electricity, gas or water supply
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2111/00—Details relating to CAD techniques
    • G06F2111/06—Multi-objective optimisation, e.g. Pareto optimisation using simulated annealing [SA], ant colony algorithms or genetic algorithms [GA]
    • 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

Definitions

  • the invention relates to the field of new energy power generation, in particular to a method for evaluating the power generation capacity of a wind farm and establishing a microscopic site selection model, which can be applied to preliminary planning, power generation calculation, economic benefit evaluation, and unit discharge of offshore wind farms or flat terrain wind farms. Cloth optimization and so on.
  • micro-site selection is a very important link, which affects the power generation efficiency of the entire wind farm.
  • a reasonable micro-site selection scheme needs to accurately evaluate the expected power generation of wind turbines at different locations in the planning area of the wind farm, and on this basis, optimize the arrangement of the entire wind turbine to achieve the optimal power generation benefit of the entire farm.
  • the free flow in the planned area of the wind farm is relatively uniform in spatial distribution. Obtain the wind energy resources in the field. Therefore, the accurate assessment of the wake interaction between wind turbines has become the most important factor in determining the rationality of wind farm micro-site selection and power generation assessment calculations.
  • the wake assessment models of wind farms mainly include computational fluid dynamics (CFD) models and analytical models.
  • CFD computational fluid dynamics
  • analytical models the CFD model can obtain accurate flow details in the flow field by numerically solving the three-dimensional Navier-Stokes equations that control the wind field flow field, so the simulation of the wind field wake can obtain relatively accurate results.
  • the patent document with the application publication number CN109086534A discloses a CFD-based wake correction method and system for wind farms.
  • the wind direction rotation model is used to calculate the wake flow field of the wind farm, and the wind frequency of each wind direction is considered for weighting, so that the wake The region is more in line with the real flow field.
  • the direct CFD simulation of a wind farm with multiple wind turbines requires high computer hardware, and the huge time cost of its calculation is often a problem in engineering applications. unacceptable.
  • Analytical model is currently the most widely used wake evaluation method in the micro-sitting stage of offshore wind farms or wind farms with flat terrain.
  • the velocity of the wake region satisfies self-similarity in its derivation, and the velocity distribution on the cross-section of the wake region is generally in the form of self-similarity.
  • Uniform distribution or Gaussian distribution the maximum speed loss is determined by the wind turbine thrust coefficient.
  • Jensen model assumes that the wake region expands linearly, the speed of the wake region is only related to the downstream distance, and the wind speed has a constant distribution in the radial direction.
  • the expansion velocity of the wake region is related to the atmospheric turbulence, the additional turbulence generated by the shear layer in the wake region, and the mechanical turbulence generated by the unit itself.
  • the patent document with the application publication number CN109376389A discloses a three-dimensional wake numerical simulation method based on the 2D_k Jensen model, which comprehensively considers the distribution characteristics of wind speed and turbulence intensity at the vertical height and its impact on the wake on the basis of the Jensen model.
  • a new three-dimensional wake model is proposed.
  • the engineering wake model has a simple structure and low computational cost, but the computational accuracy of the model depends heavily on the adjustment of empirical parameters in the model under different working conditions.
  • a large number of assumptions and simplifications are used in the model derivation process, which makes the model unable to accurately reflect
  • the complex characteristics of the wake effect of wind farms lead to large uncertainty in the calculation results of power generation.
  • the purpose of the present invention is to solve the problem that the existing computational fluid dynamics model has a large amount of calculation, takes a long time to calculate, and is difficult to meet the requirements of engineering applications, and the accuracy of the analytical model depends heavily on the selection of empirical parameters, and the derivation process needs to use a large number of assumptions. There is a large uncertainty in the results, which will affect the technical problems of wind farm power generation evaluation and micro-sitting efficiency and accuracy.
  • the technical scheme adopted in the present invention is:
  • a method for evaluating the power generation capacity of a wind farm and establishing a microscopic site selection model comprising the following steps:
  • Step 1 Collect wind farm operation cases and related data information
  • Step 2 Divide a spatial grid that includes a certain range, so that the collected relative position information of wind turbines in each wind farm can correspond to different grid points;
  • Step 3 Use a single wind turbine in the data set as a sample, establish a corresponding wind farm arrangement matrix according to the divided spatial grid, and use different digital labels in the matrix for the relative positions of the sample wind turbines and other wind turbines in the wind farm. distinguish;
  • Step 4 Determine the free flow wind speed and direction data, wind turbine parameters, and wind farm arrangement matrix as the input sample set of the model, and determine the wind speed data or power generation data y at the wind turbine to be estimated as the output sample set of the model, and according to a certain proportion Divide into training data set X and test data set X';
  • Step 5 Input the preprocessed training data set X into the machine learning model, and optimize the parameters in the machine learning model;
  • Step 6 Use the test data set X' to test the trained network, if the model error reaches the preset condition, the network training is completed, otherwise the model is retrained;
  • Step 7 Use the trained machine learning model as the wind farm power generation evaluation and micro-site selection model.
  • the micro-site selection stage of offshore wind farms or onshore flat terrain wind farms input the wind speed and direction data observed by the wind tower, and the wind turbines.
  • the parameters and the proposed fan arrangement matrix can be used to calculate the corresponding power generation at the fan to be estimated.
  • the collected relevant data information includes one or more of free-flowing wind measurement data, fan parameters, fan arrangement position, fan cabin wind speed, fan operating state and power generation.
  • step 1 data cleaning is performed on the collected wind farm data, and the overall data of the wind farm in which a single or multiple wind turbines cannot operate normally due to factors such as failures and shutdowns, and when the data monitoring occurs abnormally, is eliminated, so as to ensure the wind farm data used. Centralized all fans are running normally and data monitoring is normal.
  • step 2 the spatial extent is greater than the planned extent of all individual wind farms.
  • step 2 the spatial grid has a high resolution.
  • step 2 the spatial grid has meter resolution.
  • the machine learning model is a machine learning model constructed by using a neural network.
  • the machine learning model constructed by the neural network consists of an input layer, a hidden layer and an output layer.
  • the gradient descent algorithm is used to iteratively optimize the model parameters.
  • the function selects the mean square error function. When the iteration reaches a certain number of times or the error reaches a certain value, the model stops training; and the test data set X' is used to test the trained network. If the model error reaches the preset condition, the model training is completed.
  • step 5 the mean square error function is shown in formula (1):
  • y is the actual value of the ith output variable
  • y * is the model calculated value of the ith output variable
  • ess is the mean square error cost function
  • step 7 the wind farm power generation evaluation and micro-site selection model can be combined with a heuristic algorithm to optimize the arrangement of each wind turbine to obtain the optimal value of the entire power generation.
  • the above-mentioned meta-heuristic algorithm is a genetic algorithm or a particle swarm optimization algorithm or a simulated annealing algorithm.
  • the present invention has the following technical effects:
  • the present invention can effectively improve the calculation efficiency of the existing fluid dynamics model, and reduce the calculation time, so that it can meet the harsh engineering application requirements;
  • the present invention adopts the machine learning method to establish the wind farm power generation evaluation and micro-site selection model.
  • the model has a strong self-learning ability, and can efficiently use the existing actual operation data of the wind farm for modeling and analysis, which effectively solves the problem of the current situation.
  • Fig. 1 is the flow chart of the present invention
  • Fig. 2 is a spatial grid diagram of wind farm fan arrangement in the present invention
  • FIG. 3 is a structural diagram of a machine learning model in an embodiment.
  • a method for evaluating the power generation of a wind farm and establishing a micro-site selection model includes the following steps:
  • Step 1 Collect as many wind farm operation cases and related data information as possible, including but not limited to free-flow wind speed and direction data (such as wind tower observation data), fan parameters, fan arrangement position, fan cabin wind speed, fan operation status and power generation, etc.;
  • free-flow wind speed and direction data such as wind tower observation data
  • fan parameters such as wind tower observation data
  • fan arrangement position such as fan cabin wind speed, fan operation status and power generation, etc.
  • Data cleaning is performed on the collected wind farm data, and the overall data of the wind farm in which a single or multiple wind turbines cannot operate normally due to factors such as faults and shutdowns and the abnormal period of data monitoring are excluded, so as to ensure the normal operation of all wind turbines in the used wind farm data set. And there is no abnormality in data monitoring;
  • Step 2 Referring to the computational fluid dynamics modeling method, divide a spatial grid with a certain range (the spatial range must be larger than the planned range of all individual wind farms) and with a higher resolution (meter-level resolution is recommended), so that the collection
  • the relative position information of the arrangement of wind turbines in each wind farm can correspond to different grid points, as shown in Figure 2;
  • Step 3 Take a single wind turbine in the data set as a sample, establish a corresponding wind farm arrangement matrix according to the divided spatial grid, and compare the relative positions of the sample wind turbine and other wind turbines in the wind farm through different digital labels in the matrix. distinguish;
  • Step 4 Determine the free flow wind speed and direction data, wind turbine parameters, and wind farm arrangement matrix as the input sample set of the model, and determine the wind speed data or power generation data y at the wind turbine to be estimated as the output sample set of the model, and according to a certain proportion Divide into training data set X and test data set X';
  • Step 5 Input the preprocessed training data set X into the machine learning model shown in FIG. 3 constructed by using the neural network, and the model consists of an input layer, a hidden layer and an output layer. And the gradient descent algorithm is used to iteratively optimize the model parameters.
  • the cost function is selected as the mean square error function shown in formula (1).
  • y is the actual value of the ith output variable
  • y * is the model calculated value of the ith output variable
  • ess is the mean square error cost function.
  • Step 6 Use the test data set X' to test the trained network, if the model error reaches the preset condition, the network training is completed, otherwise the model is retrained;
  • Step 7 Use the trained machine learning model as the wind farm power generation evaluation and micro-site selection model.
  • the micro-site selection stage of offshore wind farms or onshore flat terrain wind farms input the wind speed and direction data observed by the wind tower, and the wind turbines.
  • the corresponding power generation can be calculated according to the parameters and the fan arrangement matrix to be used;
  • Step 8 The model can be combined with meta-heuristic algorithms (genetic algorithm, particle swarm optimization algorithm, simulated annealing algorithm, etc.) to optimize the arrangement of each fan to obtain the optimal value of the entire power generation.
  • meta-heuristic algorithms genetic algorithm, particle swarm optimization algorithm, simulated annealing algorithm, etc.

Abstract

一种风电场发电量评估及微观选址模型建立方法，它包括：步骤1：收集风电场运行案例及相关数据信息,建立包含一定范围的空间网格，步骤2：将单个风力发电机作为样本并根据划分的空间网格建立对应的风机排布矩阵，将各风机的位置在矩阵中通过不同的数字标签进行区分，步骤3：确定输入输出样本集，采用训练样本集对机器学习模型进行迭代训练，训练完成后采用测试数据集进行模型测试，步骤4：将训练好的机器学习模型作为风电场发电量评估及微观选址模型，对待估计风机处相应的发电量进行计算等步骤；本发明的目的是为了解决现有的计算流体动力学模型计算量大、计算耗费时间长，难以满足工程应用要求的技术问题。

Description

一种风电场发电量评估及微观选址模型建立方法 技术领域
本发明涉及新能源发电领域，具体涉及一种风电场发电量评估及微观选址模型建立方法，可应用于海上风电场或平坦地形风电场的前期规划、发电量计算、经济效益评价、机组排布优化等工作中。
背景技术
在风电场开发的规划设计中，微观选址是其中十分重要的环节，影响着整个风电场的发电效益。合理的微观选址方案需要对风电场规划区域内对不同位置的风力发电机预期发电量进行准确的评估，在此基础上对整个风力发电机组进行优化排布从而达到整场发电效益的最优。对于海上风电场和平坦地形风电场，由于下垫面地势平坦，风电场规划区域内的自由来流在空间分布上较为均一，可通过在风电场内建立测风塔或放置其它观测设备进行观测获得场区内的风能资源情况。因此，对风机间的尾流相互作用进行准确评估成为了决定风电场微观选址和发电量评估计算合理性的最重要因素。
目前而言，风电场尾流评估模型主要包含计算流体动力学(CFD)模型及解析模型。其中，CFD模型通过对控制风场流场的三维纳维斯托克斯方程进行数值求解，能够获得流场内精确地流动细节，因此对风场尾流的模拟能够获得较为准确的结果。
申请公布号为CN109086534A的专利文献公开了一种基于CFD的风电场尾流订正方法和系统，采用风向旋转模型计算风场尾流流场，同时考虑了各个风向的风频进行加权，使得尾流区域更加符合真实流场。然而，由于采用了精细的网格来解决边界层问题，对一个包含多台风机的风电场直接进行CFD模拟对计算机的硬件要求较高，而且其计算耗费的巨大时间成本往往是工程应用上所不能接受的。解析模型是目前海上风电场或平坦地形风电场微观选址阶段使用最多的尾流评估方法，其在推导时通常假定尾流区的速度满足自相似性，尾流区截面上的速度分布一般呈均一分布或是高斯分布，速度损失最大值由风电机组推力系数决定。例如最为广泛应用的Jensen模型，其假设尾流区呈线性扩张，尾流区速度仅与下游距离相关，风速在径向呈常数分布。尾流区的膨胀速度则与大气湍流、尾流区内剪切层产生的附加湍流以及机组自身产生的机械湍流有关。
另外，申请公布号为CN109376389A的专利文献公开了一种基于2D_k Jensen模型的三维尾流数值模拟方法，它在Jensen模型的基础上，综合考虑风速和湍流强度在垂直高度的分布特性及其对尾流的影响，提出了一种新型的三维尾流模型。工程尾流模型结构简单、计算代价较小，但是模型的计算精度严重依赖模型中经验参数在不同工况下的调整，此外模型推导过程中采用了大量假设和简化，这使得模型无法准确地反映风电场尾流影响的复杂特性，导致发电量计算结果存在较大的不确定性。
随着风电的不断发展，越来越多的风电场建成并投入运行，有效地利用已建成的风电场实际运行数据进行建模分析，能够有效地解决现有建模方法中存在的诸多不足，对后续的风电场开发具有重要的指导和应用价值。
发明内容
本发明的目的是为了解决现有的计算流体动力学模型计算量大、计算耗费时间长，难以满足工程应用的要求，且解析模型精确度严重依赖经验参数选取，推导过程需要运用大量假设，计算结果存在较大不确定性，这样会影响风电场发电评估及微观选址效率及准确度的技术问题。
为解决上述技术问题，本发明所采用的技术方案是：
一种风电场发电量评估及微观选址模型建立方法，它包括以下步骤：
步骤1：收集风电场运行案例及相关数据信息；
步骤2：划分包含一定范围的空间网格，使收集到的每个风电场中风机排布的相对位置信息能够对应到不同的网格点上；
步骤3：使数据集中的单个风力发电机作为样本，根据划分的空间网格建立相应的风电场排布矩阵，将样本风机和风电场内其他风机的相对位置在矩阵中通过不同的数字标签进行区分；
步骤4：确定自由来流风速风向数据、风机参数、风电场排布矩阵作为模型的输入样本集，确定待估计风机处的风速数据或发电量数据y作为模型的输出样本集，并按照一定比例划分为训练数据集X与测试数据集X’；
步骤5：将预处理后的训练数据集X输入机器学习模型，并对该机器学习模型中的参数进行优化；
步骤6：采用测试数据集X’对已训练网络进行测试，若模型误差达到预先设定条件，网络训练完成，否则重新对模型进行训练；
步骤7：将训练好的机器学习模型作为风电场发电量评估及微观选址模型，在海上风电场或陆上平坦地形风电场微观选址阶段，通过输入测风塔观测的风速风向数据、风机参数和拟采用的风机排布矩阵可计算得待估计风机处相应的发电量。
在步骤1中，收集的相关数据信息包括自由来流测风数据、风机参数、风机排布位置、风机机舱风速、风机运行状态及发电量其中的一种或多种。
在步骤1中，针对收集到的风电场数据进行数据清洗，剔除单个或多个风机由于故障、停机等因素不能正常运行以及数据监测出现异常时段的风电场整体数据，保证所使用的风电场数据集中所有风机正常运行且数据监测无异常。
在步骤2中，所述空间范围大于所有单个风电场的规划范围。
在步骤2中，所述空间网格具有高分辨率。
在步骤2中，空间网格具有米级分辨率。
在步骤5中，机器学习模型为采用神经网络构建的机器学习模型，该神经网络构建的机器学习模型由输入层、隐含层及输出层组成，采用梯度下降算法对模型参数进行迭代优化，代价函数选取均方误差函数，当迭代达到一定次数或者误差达到一定取值后模型停止训练；并采用测试数据集X’对已训练网络进行测试，模型误差达到预先设定条件，模型训练完成。
在步骤5中，所述均方误差函数如式(1)所示：
e ss＝1/2(y *-y) 2                                   (1)
上式中：y为第i个输出变量的实际值，y *为第i个输出变量的模型计算值；e ss为均方误差代价函数；
在步骤7中，风电场发电量评估及微观选址模型可结合启发式算法，对各风机排布方案进行寻优，以获得整场发电量的最优值。
上述元启发式算法为遗传算法或粒子群优化算法或模拟退火算法。
与现有技术相比，本发明具有如下技术效果：
1)本发明能有效提高现有流体动力学模型的计算效率、并降低计算耗时，使之满足苛刻的工程应用要求；
2)采用机器学习方法建立风电场发电量评估及微观选址模型，可用于海上风电场或平坦地形风电场的前期规划、发电量计算、经济效益评价、机组排布优化等工作中；
3)本发明采用机器学习方法建立风电场发电量评估及微观选址模型，该模型具有强大的自学习能力，能够高效地利用现有的风电场实际运行数据进行建模分析，切实解决了现有物理模型计算复杂或经验模型准确性不足的问题，使风电场发电量计算及微观选址设计更加高效准确。
附图说明
下面结合附图和实施例对本发明作进一步说明：
图1为本发明的流程图；
图2为本发明中风电场风机排布空间网格图；
图3为实施例中机器学习模型的结构图。
具体实施方式
如图1所示，一种风电场发电量评估及微观选址模型建立方法，它包括以下步骤：
步骤1：收集尽可能多的风电场运行案例及相关数据信息，包括但不限于自由来流风速风向数据(如测风塔观测数据)、风机参数、风机排布位置、风机机舱风速、风机运行状态及发电量等；
针对收集到的风电场数据进行数据清洗，剔除单个或多个风机由于故障、停机等因素不能正常运行以及数据监测出现异常时段的风电场整体数据，保证所使用的风电场数据集中所有风机正常运行且数据监测无异常；
步骤2：参照计算流体力学建模方法，划分一个包含一定范围(空间范围需大于所有单个风电场的规划范围)且具有较高分辨率(建议使用米级分辨率)的空间网格，使收集到的每个风电场中风机排布的相对位置信息能够对应到不同的网格点上，如图2中所示；
步骤3：将数据集中的单个风力发电机作为样本，根据划分的空间网格建立相应的风电场排布矩阵，将样本风机和风电场内其他风机的相对位置在矩阵中通过不同的数字标签进行区分；
步骤4：确定自由来流风速风向数据、风机参数、风电场排布矩阵作为模型的输入样本集，确定待估计风机处的风速数据或发电量数据y作为模型的输出样本集，并按照一定比例划分为训练数据集X与测试数据集X’；
步骤5：将预处理后的训练数据集X输入图3中所示的采用神经网络构建的机器学习模型，该模型由输入层、隐含层及输出层组成。并采用梯度下降算法对模型参数 进行迭代优化，代价函数选取为公式(1)中所示的均方误差函数，当迭代达到一定次数或者e ss达到一定取值后网络停止训练：
e ss＝1/2(y *-y) 2      (1)
上式中：y为第i个输出变量的实际值；y *为第i个输出变量的模型计算值；e ss为均方误差代价函数。
步骤6：采用测试数据集X’对已训练网络进行测试，若模型误差达到预先设定条件，网络训练完成，否则重新对模型进行训练；
步骤7：将训练好的机器学习模型作为风电场发电量评估及微观选址模型，在海上风电场或陆上平坦地形风电场微观选址阶段，通过输入测风塔观测的风速风向数据、风机参数和拟采用的风机排布矩阵可计算得相应的发电量；
步骤8：本模型可结合元启发式算法(遗传算法、粒子群优化算法、模拟退火算法等)对各风机排布方案进行寻优，以获得整场发电量的最优值。

Claims (10)

  1. 一种风电场发电量评估及微观选址模型建立方法，其特征在于，它包括以下步骤:
    步骤1：收集风电场运行案例及相关数据信息；
    步骤2：划分包含一定范围的空间网格，使收集到的每个风电场中风机排布的相对位置信息能够对应到不同的网格点上；
    步骤3：使数据集中的单个风力发电机作为样本，根据划分的空间网格建立相应的风电场排布矩阵，将样本风机和风电场内其他风机的相对位置在矩阵中通过不同的数字标签进行区分；
    步骤4：确定自由来流风速风向数据、风机参数、风电场排布矩阵作为模型的输入样本集，确定待估计风机处的风速数据或发电量数据y作为模型的输出样本集，并按照一定比例划分为训练数据集X与测试数据集X’；
    步骤5：将预处理后的训练数据集X输入机器学习模型，并对该机器学习模型中的参数进行迭代优化；
    步骤6：采用测试数据集X’对已训练网络进行测试，若模型误差达到预先设定条件，网络训练完成，否则重新对模型进行训练；
    步骤7：将训练好的机器学习模型作为风电场发电量评估及微观选址模型，在海上风电场或陆上平坦地形风电场微观选址阶段，通过输入测风塔观测的风速风向数据、风机参数和拟采用的风机排布矩阵可计算得待估计风机处相应的发电量。
  2. 根据权利要求1所述的风电场发电量评估及微观选址模型建立方法，其特征在于：在步骤1中，收集的相关数据信息包括自由来流测风数据、风机参数、风机排布位置、风机机舱风速、风机运行状态及发电量其中的一种或多种。
  3. 根据权利要求2所述的风电场发电量评估及微观选址模型建立方法，其特征在于：在步骤1中，针对收集到的风电场数据进行数据清洗，剔除单个或多个风机由于故障、停机等因素不能正常运行以及数据监测出现异常时段的风电场整体数据，保证所使用的风电场数据集中所有风机正常运行且数据监测无异常。
  4. 根据权利要求1所述的风电场发电量评估及微观选址模型建立方法，其特征在于，在步骤2中，所述空间范围大于所有单个风电场的规划范围。
  5. 根据权利要求1或4所述的风电场发电量评估及微观选址模型建立方法，其 特征在于，在步骤2中，所述空间网格具有高分辨率。
  6. 根据权利要求5所述的风电场发电量评估及微观选址模型建立方法，其特征在于，在步骤2中，空间网格具有米级分辨率。
  7. 根据权利要求1所述的风电场发电量评估及微观选址模型建立方法，其特征在于，在步骤5中，机器学习模型为采用神经网络构建的机器学习模型，该神经网络构建的机器学习模型由输入层、隐含层及输出层组成，采用梯度下降算法对模型参数进行迭代优化，代价函数选取均方误差函数，当迭代达到一定次数或者误差达到一定取值后模型停止训练；并采用测试数据集X’对已训练网络进行测试，模型误差达到预先设定条件，模型训练完成。
  8. 根据权利要求7所述的风电场发电量评估及微观选址模型建立方法，其特征在于：在步骤5中，所述均方误差函数如式(1)所示：
    e ss＝1/2(y *-y) 2    (1)
    上式中：y为第i个输出变量的实际值。
  9. 根据权利要求1所述的风电场发电量评估及微观选址模型建立方法，其特征在于，在步骤7中，风电场发电量评估及微观选址模型可结合启发式算法，对各风机排布方案进行寻优，以获得整场发电量的最优值。
  10. 根据权利要求9所述的风电场发电量评估及微观选址模型建立方法，其特征在于，所述元启发式算法为遗传算法或粒子群优化算法或模拟退火算法。
PCT/CN2021/137870 2021-01-13 2021-12-14 一种风电场发电量评估及微观选址模型建立方法 WO2022151890A1 (zh)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202110045045.0 2021-01-13
CN202110045045.0A CN112818590A (zh) 2021-01-13 2021-01-13 一种风电场发电量评估及微观选址模型建立方法

Publications (1)

Publication Number Publication Date
WO2022151890A1 true WO2022151890A1 (zh) 2022-07-21

Family

ID=75869174

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/137870 WO2022151890A1 (zh) 2021-01-13 2021-12-14 一种风电场发电量评估及微观选址模型建立方法

Country Status (2)

Country Link
CN (1) CN112818590A (zh)
WO (1) WO2022151890A1 (zh)

Cited By (15)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115013261A (zh) * 2022-08-08 2022-09-06 国网浙江省电力有限公司舟山供电公司 一种用于海上风电场的状态监测方法及系统
CN115422209A (zh) * 2022-11-07 2022-12-02 东方电气风电股份有限公司 一种风电案例数据处理系统及方法
CN116169727A (zh) * 2023-02-14 2023-05-26 中节能风力发电股份有限公司 无测风数据的老旧风电场重建项目发电量评估方法及系统
CN116205079A (zh) * 2023-03-23 2023-06-02 盛东如东海上风力发电有限责任公司 风电场重复设计方案筛选方法及系统
CN116596165A (zh) * 2023-07-17 2023-08-15 国网山东省电力公司汶上县供电公司 一种风力发电功率预测方法及系统
CN116771596A (zh) * 2023-06-30 2023-09-19 渤海石油航务建筑工程有限责任公司 海上风电场尾流转向控制方法及相关设备
CN117078116A (zh) * 2023-10-17 2023-11-17 华能(浙江)能源开发有限公司清洁能源分公司 风电场选址对海洋生物群落影响的鲁棒性分析方法及系统
CN117151352A (zh) * 2023-11-01 2023-12-01 北京大学长沙计算与数字经济研究院 风资源评估方法、系统及计算机存储介质和终端设备
CN117152372A (zh) * 2023-10-30 2023-12-01 四川电力设计咨询有限责任公司 一种风电工程三维数字化地理信息服务平台
CN117217099A (zh) * 2023-11-08 2023-12-12 云南滇能智慧能源有限公司 一种续建风电机的机位确定方法、装置、设备及存储介质
CN117371299A (zh) * 2023-12-08 2024-01-09 安徽大学 一种托卡马克新经典环向粘滞力矩的机器学习方法
CN117454805A (zh) * 2023-12-22 2024-01-26 浙江远算科技有限公司 基于流体降阶仿真的风机尾流影响计算方法和系统
CN117556713A (zh) * 2024-01-11 2024-02-13 中国空气动力研究与发展中心计算空气动力研究所 一种cfd多可信度高维相关流场的不确定度量化方法
CN117852448A (zh) * 2024-03-05 2024-04-09 南京航空航天大学 一种基于区域分解的大规模风电场流场计算方法和装置
CN117852448B (zh) * 2024-03-05 2024-05-14 南京航空航天大学 一种基于区域分解的大规模风电场流场计算方法和装置

Families Citing this family (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN112818590A (zh) * 2021-01-13 2021-05-18 中国长江三峡集团有限公司 一种风电场发电量评估及微观选址模型建立方法
CN113705093B (zh) * 2021-08-25 2024-03-01 广东电网有限责任公司广州供电局 一种杆塔力学响应预测方法、装置、设备和介质
CN113919606B (zh) * 2021-12-14 2022-03-15 山东建筑大学 一种分布式光伏电站智能选址方法及系统

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150278405A1 (en) * 2014-03-31 2015-10-01 Vestas Wind Systems A/S Method for evaluating a performance prediction for a wind farm
CN108520319A (zh) * 2018-04-02 2018-09-11 太原理工大学 基于大数据的风电场微观选址研究方法
CN109992889A (zh) * 2019-04-02 2019-07-09 上海电气风电集团有限公司 风电场模型的建立方法及系统、尾流值计算方法及系统
CN111967205A (zh) * 2020-08-14 2020-11-20 中国华能集团清洁能源技术研究院有限公司 一种基于风加速因子的测风塔微观选址方法
CN112818590A (zh) * 2021-01-13 2021-05-18 中国长江三峡集团有限公司 一种风电场发电量评估及微观选址模型建立方法

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20150278405A1 (en) * 2014-03-31 2015-10-01 Vestas Wind Systems A/S Method for evaluating a performance prediction for a wind farm
CN108520319A (zh) * 2018-04-02 2018-09-11 太原理工大学 基于大数据的风电场微观选址研究方法
CN109992889A (zh) * 2019-04-02 2019-07-09 上海电气风电集团有限公司 风电场模型的建立方法及系统、尾流值计算方法及系统
CN111967205A (zh) * 2020-08-14 2020-11-20 中国华能集团清洁能源技术研究院有限公司 一种基于风加速因子的测风塔微观选址方法
CN112818590A (zh) * 2021-01-13 2021-05-18 中国长江三峡集团有限公司 一种风电场发电量评估及微观选址模型建立方法

Cited By (25)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115013261A (zh) * 2022-08-08 2022-09-06 国网浙江省电力有限公司舟山供电公司 一种用于海上风电场的状态监测方法及系统
CN115422209A (zh) * 2022-11-07 2022-12-02 东方电气风电股份有限公司 一种风电案例数据处理系统及方法
CN115422209B (zh) * 2022-11-07 2023-02-03 东方电气风电股份有限公司 一种风电案例数据处理系统及方法
CN116169727B (zh) * 2023-02-14 2023-11-17 中节能风力发电股份有限公司 无测风数据的老旧风电场重建项目发电量评估方法及系统
CN116169727A (zh) * 2023-02-14 2023-05-26 中节能风力发电股份有限公司 无测风数据的老旧风电场重建项目发电量评估方法及系统
CN116205079A (zh) * 2023-03-23 2023-06-02 盛东如东海上风力发电有限责任公司 风电场重复设计方案筛选方法及系统
CN116771596A (zh) * 2023-06-30 2023-09-19 渤海石油航务建筑工程有限责任公司 海上风电场尾流转向控制方法及相关设备
CN116596165A (zh) * 2023-07-17 2023-08-15 国网山东省电力公司汶上县供电公司 一种风力发电功率预测方法及系统
CN116596165B (zh) * 2023-07-17 2023-10-13 国网山东省电力公司汶上县供电公司 一种风力发电功率预测方法及系统
CN117078116A (zh) * 2023-10-17 2023-11-17 华能(浙江)能源开发有限公司清洁能源分公司 风电场选址对海洋生物群落影响的鲁棒性分析方法及系统
CN117078116B (zh) * 2023-10-17 2024-02-27 华能(浙江)能源开发有限公司清洁能源分公司 风电场选址对海洋生物群落影响的鲁棒性分析方法及系统
CN117152372A (zh) * 2023-10-30 2023-12-01 四川电力设计咨询有限责任公司 一种风电工程三维数字化地理信息服务平台
CN117152372B (zh) * 2023-10-30 2024-01-30 四川电力设计咨询有限责任公司 一种风电工程三维数字化地理信息服务平台
CN117151352B (zh) * 2023-11-01 2024-01-30 北京大学长沙计算与数字经济研究院 风资源评估方法、系统及计算机存储介质和终端设备
CN117151352A (zh) * 2023-11-01 2023-12-01 北京大学长沙计算与数字经济研究院 风资源评估方法、系统及计算机存储介质和终端设备
CN117217099A (zh) * 2023-11-08 2023-12-12 云南滇能智慧能源有限公司 一种续建风电机的机位确定方法、装置、设备及存储介质
CN117217099B (zh) * 2023-11-08 2024-03-26 云南滇能智慧能源有限公司 一种续建风电机的机位确定方法、装置、设备及存储介质
CN117371299B (zh) * 2023-12-08 2024-02-27 安徽大学 一种托卡马克新经典环向粘滞力矩的机器学习方法
CN117371299A (zh) * 2023-12-08 2024-01-09 安徽大学 一种托卡马克新经典环向粘滞力矩的机器学习方法
CN117454805A (zh) * 2023-12-22 2024-01-26 浙江远算科技有限公司 基于流体降阶仿真的风机尾流影响计算方法和系统
CN117454805B (zh) * 2023-12-22 2024-03-19 浙江远算科技有限公司 基于流体降阶仿真的风机尾流影响计算方法和系统
CN117556713A (zh) * 2024-01-11 2024-02-13 中国空气动力研究与发展中心计算空气动力研究所 一种cfd多可信度高维相关流场的不确定度量化方法
CN117556713B (zh) * 2024-01-11 2024-04-02 中国空气动力研究与发展中心计算空气动力研究所 一种cfd多可信度高维相关流场的不确定度量化方法
CN117852448A (zh) * 2024-03-05 2024-04-09 南京航空航天大学 一种基于区域分解的大规模风电场流场计算方法和装置
CN117852448B (zh) * 2024-03-05 2024-05-14 南京航空航天大学 一种基于区域分解的大规模风电场流场计算方法和装置

Also Published As

Publication number Publication date
CN112818590A (zh) 2021-05-18

Similar Documents

Publication Publication Date Title
WO2022151890A1 (zh) 一种风电场发电量评估及微观选址模型建立方法
Wang et al. Approaches to wind power curve modeling: A review and discussion
Liu et al. Wind power plant prediction by using neural networks
CN102663251B (zh) 基于计算流体力学模型的风电场功率物理预测方法
CN104699936A (zh) 基于cfd短期风速预测风电场的扇区管理方法
CN109492777A (zh) 一种基于机器学习算法平台的风电机组健康管理方法
CN105719002A (zh) 一种基于组合预测的风电机组状态参数异常辨识方法
CN104952000A (zh) 基于马尔科夫链的风电机组运行状态模糊综合评价方法
Famoso et al. A novel hybrid model for the estimation of energy conversion in a wind farm combining wake effects and stochastic dependability
EP3771822A1 (en) A method for computer-implemented determination of a vertical speed wind profile of a wind field
CN104218571B (zh) 一种风力发电设备的运行状态评估方法
CN111664823A (zh) 基于介质热传导系数差异的均压电极垢层厚度检测方法
CN105279384A (zh) 一种基于风力机机舱风速的来流风速计算方法及装置
CN111340307B (zh) 预测风机风力发电功率的方法以及相关装置
CN108319131A (zh) 基于数据挖掘的机组调峰能力评估方法
Zhou et al. Structural health monitoring of offshore wind power structures based on genetic algorithm optimization and uncertain analytic hierarchy process
Barber et al. Development of a wireless, non-intrusive, MEMS-based pressure and acoustic measurement system for large-scale operating wind turbine blades
CN108052963A (zh) 风电功率预测建模的数据筛选方法、装置及风力发电机组
Mirsane et al. An innovative method of investigating the wind turbine’s inflow speed in a wind farm due to the multiple wake effect issue
CN113187674A (zh) 一种风电机组变桨系统的故障确定方法及系统
Ashuri et al. Uncertainty quantification of the levelized cost of energy for a 20 MW research wind turbine model
Chen et al. A Compressor Off-Line Washing Schedule Optimization Method With a LSTM Deep Learning Model Predicting the Fouling Trend
CN115713029A (zh) 一种考虑迟延的风电场随机模型预测优化控制方法
CN112231979B (zh) 基于计算流体力学和机器学习的山区瞬时风况预报方法
Liu et al. Real-time comprehensive health status assessment of hydropower units based on multi-source heterogeneous uncertainty information

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 21919079

Country of ref document: EP

Kind code of ref document: A1

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 21919079

Country of ref document: EP

Kind code of ref document: A1