WO2023015460A1 - 一种基于风过程识别的风电功率预测集成优化方法及装置 - Google Patents
一种基于风过程识别的风电功率预测集成优化方法及装置 Download PDFInfo
- Publication number
- WO2023015460A1 WO2023015460A1 PCT/CN2021/111898 CN2021111898W WO2023015460A1 WO 2023015460 A1 WO2023015460 A1 WO 2023015460A1 CN 2021111898 W CN2021111898 W CN 2021111898W WO 2023015460 A1 WO2023015460 A1 WO 2023015460A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- wind power
- prediction
- wind
- data
- vector
- Prior art date
Links
- 238000000034 method Methods 0.000 title claims abstract description 173
- 238000005457 optimization Methods 0.000 title claims abstract description 17
- 238000012417 linear regression Methods 0.000 claims abstract description 50
- 239000013598 vector Substances 0.000 claims description 199
- 238000004590 computer program Methods 0.000 claims description 13
- 238000012952 Resampling Methods 0.000 claims description 12
- 238000003066 decision tree Methods 0.000 claims description 6
- 238000003062 neural network model Methods 0.000 claims description 6
- 238000003860 storage Methods 0.000 claims description 6
- 238000010586 diagram Methods 0.000 description 11
- 238000004364 calculation method Methods 0.000 description 4
- 230000006870 function Effects 0.000 description 3
- 230000019771 cognition Effects 0.000 description 1
- 230000000295 complement effect Effects 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000007619 statistical method Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
-
- 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"
Definitions
- the present application relates to the field of wind power prediction, in particular to an integrated optimization method and device for wind power prediction based on wind process identification.
- wind power forecasting is an important technical means to improve the level of wind power consumption, ensure the safety and stability of the power system, and improve the economic efficiency of power grid operation.
- the wind power prediction methods for engineering applications are mainly statistical methods, which establish explanatory variables (such as numerical weather prediction, historical power, meteorological observation data, etc.) and explained variables (
- explanatory variables such as numerical weather prediction, historical power, meteorological observation data, etc.
- explained variables due to the limitations of human cognition and technical level, as well as the complexity of the wind power prediction problem itself, there are large errors in wind power prediction.
- the purpose of the embodiments of the present application is to provide an integrated optimization method and device for wind power prediction based on wind process identification, which can improve the accuracy of wind power prediction.
- An embodiment of the present application provides an integrated optimization method for wind power forecasting based on wind process identification.
- the method includes: processing the numerical weather forecast data at the forecast time through one or more pre-trained wind power forecasting models, and obtaining each Power data at the prediction time corresponding to the pre-trained wind power prediction model; using the linear regression coefficient corresponding to the one or more pre-trained wind power prediction models at the prediction time and the power at the prediction time corresponding to each pre-trained wind power prediction model The data determines the optimal power data at the prediction time; wherein, the linear regression coefficient corresponding to the one or more pre-trained wind power prediction models at the prediction time is obtained based on the training data of each pre-trained wind power prediction model.
- the wind power prediction model includes: a backpropagation (BP, Back Propagation) neural network model, a support vector regression model, a decision tree regression model, and a k-nearest neighbor regression model.
- BP Back Propagation
- the linear regression coefficient corresponding to the one or more pre-trained wind power prediction models at the prediction time and the power data at the prediction time corresponding to each pre-trained wind power prediction model Determine the optimal power data at the time of prediction, including:
- the acquisition process of the linear regression coefficient corresponding to the prediction time of the one or more pre-trained wind power prediction models includes: selecting the same value from the wind process vector at each time in the historical period K similar vectors corresponding to the wind process vector at the forecast moment; k is a positive integer; the numerical weather forecast data of the historical moments corresponding to the k similar vectors are processed by m pre-trained wind power prediction models respectively, and each The similar power vectors at the prediction time corresponding to the pre-trained wind power prediction models; m is a positive integer; use the similar power vectors at the prediction time corresponding to each pre-trained wind power prediction model to determine The linear regression coefficient of .
- the selection of k similar vectors corresponding to the wind process vector at the forecast moment from the wind process vectors at each moment in the historical period includes: obtaining the wind process at each moment in the historical period vector, and determine the Euclidean distance between the wind process vector at the forecast moment and the wind process vector at each moment in the historical period; the Euclidean distance between the wind process vector at the forecast moment and the wind process vector at each moment in the historical period is performed in descending order Arrange, and select the wind process vectors corresponding to the first k Euclidean distances in the descending sequence as k similar vectors of the wind process vector at the prediction time.
- the acquiring the wind process vector at each moment in the historical period, and determining the Euclidean distance between the wind process vector at the forecast moment and the wind process vector at each moment in the historical period includes: Determine the Euclidean distance d i,j between the wind process vector at the forecast moment i and the wind process vector at the jth moment in the historical period according to the following formula:
- v i [v ib ,v i-b+1 ,...,v i ,...,v i+b-1 ,v i+b ] T
- v i is the wind process vector at the forecast moment i
- v i+b is the wind speed at the i+bth moment
- b is the time bandwidth
- v j [v jb ,v j-b+1 ,...,v j ,...,v j+b-1 ,v j+b ] T
- v j is the wind process vector at the jth moment in the historical period.
- the numerical weather forecast data at historical moments corresponding to the k similar vectors are processed through m pre-trained wind power prediction models to obtain each pre-trained wind power
- the similar power vector at the prediction time corresponding to the prediction model including: determine the similar power vector at the prediction time corresponding to the cth wind power prediction model according to the following formula
- P cr is the wind power data obtained after the c-th wind power prediction model processes the numerical weather prediction data corresponding to the r-th similar vector among the k similar vectors.
- determining the linear regression coefficients of each pre-trained wind power prediction model at the prediction time by using the similar power vectors corresponding to each pre-trained wind power prediction model at the prediction time includes: The following formula determines the linear regression coefficient [ ⁇ 1 ... ⁇ c ... ⁇ m ] of the 1st to mth pre-trained wind power prediction models at the prediction time:
- the embodiment of the present application also provides an integrated optimization device for wind power prediction based on wind process identification, the device includes: an acquisition module configured to use one or more pre-trained wind power prediction models for the numerical weather forecast at the prediction time The data is processed to obtain the power data at the prediction time corresponding to each pre-trained wind power prediction model; the determination module is configured to use the linear regression coefficient corresponding to the one or more pre-trained wind power prediction models at the prediction time and each pre-trained The power data at the prediction time corresponding to the trained wind power prediction model determines the optimal power data at the prediction time; wherein, the linear regression coefficients corresponding to the one or more pre-trained wind power prediction models at the prediction time are based on each pre-trained wind power Training data acquisition for the power prediction model.
- the device further includes a training module configured to perform y groups of bootstrap random resampling on the data set to obtain y groups of training data; the data set includes multiple moments in the historical period The numerical weather forecast data and the corresponding wind power data; y is a positive integer; it is also configured to use the numerical weather forecast data in the y group of training data as the input layer data of each wind power prediction model, and the y group of training data
- the wind power data corresponding to the numerical weather forecast data in the data are respectively used as the output layer data of each wind power prediction model for training, and m wind power prediction models are obtained;
- the wind power prediction model includes: a BP neural network model, a support vector regression model, a decision tree regression model, and a k-nearest neighbor regression model.
- the determination module is configured to: determine the optimal predicted power data according to the following formula
- the acquisition module is further configured to: select k similar vectors corresponding to the wind process vector at the forecast moment from the wind process vectors at each moment in the historical period; k is a positive integer ; Process the numerical weather prediction data of the historical moments corresponding to the k similarity vectors through m pre-trained wind power prediction models respectively, and obtain similar power vectors at the prediction times corresponding to each pre-trained wind power prediction model; m is a positive integer; the linear regression coefficient of each pre-trained wind power prediction model at the prediction time is determined by using the similar power vector at the prediction time corresponding to each pre-trained wind power prediction model.
- the acquisition module is configured to: acquire the wind process vector at each moment in the historical period, and determine the relationship between the wind process vector at the forecast moment and the wind process vector at each moment in the historical period the Euclidean distance; arrange the Euclidean distance between the wind process vector at the forecast time and the wind process vector at each time in the historical period in descending order, and select the wind process vector corresponding to the first k Euclidean distances in the descending sequence as the forecast time k similar vectors of the wind process vector of .
- the acquisition module is configured to: determine the Euclidean distance d i,j between the wind process vector at the forecast moment i and the wind process vector at the jth moment in the historical period according to the following formula:
- v i [v ib ,v i-b+1 ,...,v i ,...,v i+b-1 ,v i+b ] T
- v i is the wind process vector at the forecast moment i
- v i+b is the wind speed at the i+bth moment
- b is the time bandwidth
- v j [v jb ,v j-b+1 ,...,v j ,...,v j+b-1 ,v j+b ] T
- v j is the wind process vector at the jth moment in the historical period.
- the acquisition module is configured to: determine the similar power vector at the prediction moment output by the cth wind power prediction model according to the following formula
- P cr is the wind power data obtained after the c-th wind power prediction model processes the numerical weather prediction data corresponding to the r-th similar vector among the k similar vectors.
- the acquisition module is configured to: determine the linear regression coefficient [ ⁇ 1 ... ⁇ of the first to m pre-trained wind power prediction models at the prediction time according to the following formula c ... ⁇ m ]:
- the embodiment of the present application also provides a computer-readable storage medium, on which a computer program is stored.
- the computer program is executed by a processor, the steps of the wind power forecasting integrated optimization method based on wind process identification described in the embodiment of the present application are implemented. .
- each pre-trained wind power prediction model is used to process the numerical weather forecast data at the prediction time, and the power data at the prediction time output by each pre-trained wind power prediction model is obtained;
- the linear regression coefficient corresponding to the wind power prediction model at the prediction time and the power data at the prediction time output by each pre-trained wind power prediction model determine the optimal power data at the prediction time.
- This method can quickly and effectively match and identify the wind process at the prediction time
- the similar vectors of wind power are integrated and optimized at the same time, thereby improving the accuracy of wind power prediction.
- the prediction method is simple and fast, and can be widely promoted and applied.
- Fig. 1 is a schematic flow diagram of an integrated optimization method for wind power prediction based on wind process identification provided by an embodiment of the present application
- Fig. 2 (a) to Fig. 2 (d) are the identification and matching calculation example result figure of the similar vector of four wind processes provided by the embodiment of the present application;
- Fig. 3 is the time sequence diagram of the predicted power and actual power of the wind farm provided by the embodiment of the present application;
- Fig. 4 is a schematic structural diagram of an integrated optimization device for wind power prediction based on wind process identification provided by an embodiment of the present application.
- the embodiment of the present application provides an integrated optimization method for wind power prediction based on wind process identification, as shown in Figure 1, the method includes:
- Step 101 Process the numerical weather forecast data at the prediction time through one or more pre-trained wind power prediction models, and obtain the power data at the prediction time corresponding to each pre-trained wind power prediction model;
- Step 102 Using the linear regression coefficients corresponding to the one or more pre-trained wind power prediction models at the prediction time and the power data at the prediction time corresponding to each pre-trained wind power prediction model to determine the optimal power data at the prediction time;
- the linear regression coefficient corresponding to the one or more pre-trained wind power prediction models at the prediction time is obtained based on the training data of each pre-trained wind power prediction model.
- the wind power prediction model includes: BP neural network model, support vector regression model, decision tree regression model and k-nearest neighbor regression model.
- the linear regression coefficient corresponding to the one or more pre-trained wind power prediction models at the prediction time and the power data corresponding to each pre-trained wind power prediction model at the prediction time are used to determine the optimal Optimal power data at the time of prediction, including:
- the acquisition process of the linear regression coefficient corresponding to the one or more pre-trained wind power prediction models at the prediction time includes:
- the training data used to obtain the linear regression coefficient corresponding to each pre-trained wind power prediction model at the prediction time is the numerical weather prediction data of more than one year and its corresponding wind power data.
- the result diagrams of the identification and matching calculation examples of the similar vectors of the four wind processes are respectively given, Among them, the thick solid line is the target wind process, and the other thin curves are matching similar wind processes.
- the time bandwidth b is 4 hours, and the time resolution is 15 minutes. From the calculation example, it can be found that the matching degree of the wind process is good.
- the selection of k similar vectors corresponding to the wind process vector at the forecast moment from the wind process vectors at each moment in the historical period includes: obtaining the wind process at each moment in the historical period vector, and determine the Euclidean distance between the wind process vector at the forecast moment and the wind process vector at each moment in the historical period; the Euclidean distance between the wind process vector at the forecast moment and the wind process vector at each moment in the historical period is performed in descending order Arrange, and select the wind process vectors corresponding to the first k Euclidean distances in the descending sequence as k similar vectors of the wind process vector at the prediction time.
- the acquiring the wind process vector at each moment in the historical period, and determining the Euclidean distance between the wind process vector at the forecast moment and the wind process vector at each moment in the historical period includes: Determine the Euclidean distance d i,j between the wind process vector at the forecast moment i and the wind process vector at the jth moment in the historical period according to the following formula:
- v i [v ib ,v i-b+1 ,...,v i ,...,v i+b-1 ,v i+b ] T
- v i is the wind process vector at the forecast moment i
- v i+b is the wind speed at the i+bth moment
- b is the time bandwidth
- v j [v jb ,v j-b+1 ,...,v j ,...,v j+b-1 ,v j+b ] T
- v j is the wind process vector at the jth moment in the historical period.
- the numerical weather prediction data at historical moments corresponding to k similar vectors are respectively substituted into m pre-trained wind power prediction models, and the output of each pre-trained wind power prediction model is obtained Similar power vectors at the prediction time of , including:
- P cr is the wind power data obtained after the c-th wind power prediction model processes the numerical weather prediction data corresponding to the r-th similar vector among the k similar vectors.
- determining the linear regression coefficients of each pre-trained wind power prediction model at the prediction time by using the similar power vectors corresponding to each pre-trained wind power prediction model at the prediction time includes: The following formula determines the linear regression coefficient [ ⁇ 1 ... ⁇ c ... ⁇ m ] of the 1st to mth pre-trained wind power prediction models at the prediction time:
- Table 1-Table 3 provides the prediction results of three wind farms in a certain province, and the wind power prediction model provides a total of 96 point (with 15-minute resolution) prediction results.
- each station provides three differentiated models, namely Model 1-Model 3, and the "proposed optimization model” is the linear regression coefficient corresponding to each pre-trained wind power prediction model at the prediction time and each pre-trained wind power prediction model.
- the optimal forecasting time power data determined by the forecasting time power data output by the wind power forecasting model.
- the proposed optimization model can significantly improve the prediction accuracy.
- the time series diagram corresponding to wind farm 3 is given. It can be found that the optimization model can use the complementary effect of the difference between the prediction models to optimize the prediction results.
- the embodiment of the present application also provides an integrated optimization device for wind power prediction based on wind process identification, as shown in Figure 4, the device includes:
- the acquisition module is configured to process the numerical weather forecast data at the prediction time through one or more pre-trained wind power prediction models, and obtain the power data at the prediction time corresponding to each pre-trained wind power prediction model;
- the determination module is configured to use the linear regression coefficient corresponding to the one or more pre-trained wind power prediction models at the prediction time and the power data at the prediction time corresponding to each pre-trained wind power prediction model to determine the optimal power data at the prediction time ;
- the linear regression coefficient corresponding to the one or more pre-trained wind power prediction models at the prediction time is obtained based on the training data of each pre-trained wind power prediction model.
- the device further includes a training module configured to perform y groups of bootstrap random resampling on the data set to obtain y groups of training data;
- the data set includes Numerical weather forecast data and corresponding wind power data;
- y is a positive integer;
- the numerical weather forecast data in the y group of training data are respectively used as the input layer data of each wind power prediction model, and the numerical weather forecast data in the y group of training data
- the wind power data corresponding to the forecast data are respectively used as the output layer data of each wind power prediction model for training to obtain m wind power prediction models;
- m is a positive integer;
- the wind power prediction model includes: a BP neural network model, a support vector regression model, a decision tree regression model, and a k-nearest neighbor regression model.
- the determining module is configured to:
- the acquisition module is further configured to: select k similar vectors corresponding to the wind process vector at the forecast moment from the wind process vectors at each moment in the historical period; k is a positive integer; Process the numerical weather prediction data at historical moments corresponding to the k similarity vectors through m pre-trained wind power prediction models respectively, and obtain similar power vectors at the prediction times corresponding to each pre-trained wind power prediction model; m is Positive integer; use the similar power vectors corresponding to each pre-trained wind power prediction model at the prediction time to determine the linear regression coefficient of each pre-trained wind power prediction model at the prediction time.
- the acquisition module is configured to: acquire the wind process vector at each moment in the historical period, and determine the relationship between the wind process vector at the forecast moment and the wind process vector at each moment in the historical period Euclidean distance: Arrange the Euclidean distances between the wind process vector at the forecast moment and the wind process vector at each moment in the historical period in descending order, and select the wind process vectors corresponding to the first k Euclidean distances in the descending sequence as the Euclidean distance at the forecast moment k similarity vectors of the wind process vector.
- the acquisition module is configured to: determine the Euclidean distance d i,j between the wind process vector at the forecast moment i and the wind process vector at the jth moment in the historical period according to the following formula :
- v i [v ib ,v i-b+1 ,...,v i ,...,v i+b-1 ,v i+b ] T
- v i is the wind process vector at the forecast moment i
- v i+b is the wind speed at the i+bth moment
- b is the time bandwidth
- v j [v jb ,v j-b+1 ,...,v j ,...,v j+b-1 ,v j+b ] T
- v j is the wind process vector at the jth moment in the historical period.
- the acquisition module is configured to: determine the similar power vector at the prediction moment output by the cth wind power prediction model according to the following formula
- P cr is the wind power data obtained after the c-th wind power prediction model processes the numerical weather prediction data corresponding to the r-th similar vector among the k similar vectors.
- the acquisition module is configured to: determine the linear regression coefficient [ ⁇ 1 ... ⁇ c of the first to m pre-trained wind power prediction models at the prediction time according to the following formula ... ⁇ m ]:
- the device can be applied to electronic equipment.
- Acquisition module, determination module and training module in the described device all can be by central processing unit (CPU, Central Processing Unit), digital signal processor (DSP, Digital Signal Processor), micro control unit (MCU, Microcontroller) in practical application. Unit) or programmable gate array (FPGA, Field-Programmable Gate Array) implementation.
- CPU Central Processing Unit
- DSP Digital Signal Processor
- MCU Microcontroller
- FPGA Field-Programmable Gate Array
- the device may be implemented by one or more Application Specific Integrated Circuit (ASIC, Application Specific Integrated Circuit), DSP, Programmable Logic Device (PLD, Programmable Logic Device), Complex Programmable Logic Device (CPLD , Complex Programmable Logic Device), FPGA, general-purpose processor, controller, MCU, microprocessor (Microprocessor), or other electronic components are implemented for performing the aforementioned method.
- ASIC Application Specific Integrated Circuit
- DSP Digital Signal Processing Unit
- PLD Programmable Logic Device
- CPLD Complex Programmable Logic Device
- FPGA general-purpose processor
- controller MCU
- microprocessor Microprocessor
- the embodiment of the present application also provides a computer-readable storage medium, on which a computer program is stored.
- the computer program is executed by a processor, the steps of the wind power forecasting integrated optimization method based on wind process identification described in the embodiment of the present application are implemented. .
- the embodiments of the present application may be provided as methods, apparatuses, or computer program products. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein.
- computer-usable storage media including but not limited to disk storage, CD-ROM, optical storage, etc.
- These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing apparatus to operate in a specific manner, such that the instructions stored in the computer-readable memory produce an article of manufacture comprising instruction means, the instructions
- the device realizes the function specified in one or more procedures of the flowchart and/or one or more blocks of the block diagram.
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Human Resources & Organizations (AREA)
- Strategic Management (AREA)
- Computational Mathematics (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- Operations Research (AREA)
- Economics (AREA)
- Pure & Applied Mathematics (AREA)
- Mathematical Physics (AREA)
- Game Theory and Decision Science (AREA)
- Bioinformatics & Computational Biology (AREA)
- Algebra (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Development Economics (AREA)
- Evolutionary Biology (AREA)
- Software Systems (AREA)
- Databases & Information Systems (AREA)
- Probability & Statistics with Applications (AREA)
- Entrepreneurship & Innovation (AREA)
- Marketing (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
本申请实施例公开了一种基于风过程识别的风电功率预测集成优化方法及装置,所述方法包括:通过一个或多个预先训练的风电功率预测模型对预测时刻的数值天气预报数据进行处理,获取各预先训练的风电功率预测模型对应的预测时刻功率数据;利用所述一个或多个预先训练的风电功率预测模型在预测时刻对应的线性回归系数和各预先训练的风电功率预测模型对应的预测时刻功率数据确定最优的预测时刻功率数据;其中,所述一个或多个预先训练的风电功率预测模型在预测时刻对应的线性回归系数基于各预先训练的风电功率预测模型的训练数据获取。
Description
本申请涉及风电功率预测领域,具体涉及一种基于风过程识别的风电功率预测集成优化方法及装置。
在风电大规模并网的背景下,风电出力的随机性、波动性为电力系统的安全经济运行带来了巨大冲击,而风电功率预测可以提前给出未来的风电出力,从而作为电力系统决策优化的重要依据,因此,风电功率预测是提高风电消纳水平、保障电力系统安全稳定、提高电网运行经济效益的重要技术手段。
目前工程应用的风电功率预测方法以统计方法为主,此类方法通过统计模型或机器学习(人工智能)模型建立解释变量(如数值天气预报、历史功率、气象观测数据等)和被解释变量(风电功率)之间的映射关系,但是由于人类认知和技术水平的局限,以及风电功率预测问题本身的复杂性,风电功率预测存在大的误差。
发明内容
本申请实施例的目的是提供一种基于风过程识别的风电功率预测集成优化方法及装置,能够提升风电功率预测的精准度。
本申请实施例是采用下述技术方案实现的:
本申请实施例提供了一种基于风过程识别的风电功率预测集成优化方法,所述方法包括:通过一个或多个预先训练的风电功率预测模型对预测时刻的数值天气预报数据进行处理,获取各预先训练的风电功率预测模型 对应的预测时刻功率数据;利用所述一个或多个预先训练的风电功率预测模型在预测时刻对应的线性回归系数和各预先训练的风电功率预测模型对应的预测时刻功率数据确定最优的预测时刻功率数据;其中,所述一个或多个预先训练的风电功率预测模型在预测时刻对应的线性回归系数基于各预先训练的风电功率预测模型的训练数据获取。
在本申请的一些可选实施例中,一个或多个的风电功率预测模型的训练过程包括:对数据集进行y组自助法随机重采样,获取y组训练数据;所述数据集包括历史时段内多个时刻的数值天气预报数据及其对应的风电功率数据;y为正整数;将所述y组训练数据中数值天气预报数据分别作为各风电功率预测模型的输入层数据,所述y组训练数据中数值天气预报数据对应的风电功率数据分别作为各风电功率预测模型的输出层数据进行训练,获得m个预先训练的风电功率预测模型;m为正整数;其中,每组训练数据包括q次有放回随机重采样构成的q个样本数据,q为正整数;则m=y×g,g表示风电功率预测模型的数量。
在本申请的一些可选实施例中,所述风电功率预测模型包括:反向传播(BP,Back Propagation)神经网络模型、支持向量回归模型、决策树回归模型和k近邻回归模型。
在本申请的一些可选实施例中,所述利用所述一个或多个预先训练的风电功率预测模型在预测时刻对应的线性回归系数和各预先训练的风电功率预测模型对应的预测时刻功率数据确定最优的预测时刻功率数据,包括:
在本申请的一些可选实施例中,所述一个或多个预先训练的风电功率 预测模型在预测时刻对应的线性回归系数的获取过程包括:在历史时段内各时刻的风过程向量中选取与预测时刻的风过程向量对应的k个相似向量;k为正整数;分别通过m个预先训练的风电功率预测模型对所述k个相似向量对应的历史时刻的数值天气预报数据进行处理,获取各预先训练的风电功率预测模型对应的预测时刻的相似功率向量;m为正整数;利用各预先训练的风电功率预测模型对应的预测时刻的相似功率向量确定各预先训练的风电功率预测模型在预测时刻的线性回归系数。
在本申请的一些可选实施例中,所述在历史时段内各时刻的风过程向量中选取与预测时刻的风过程向量对应的k个相似向量,包括:获取历史时段内各时刻的风过程向量,并确定预测时刻的风过程向量与历史时段内各时刻的风过程向量之间的欧式距离;将预测时刻的风过程向量与历史时段内各时刻的风过程向量之间的欧式距离进行降序排列,并选择所述降序序列中前k个欧式距离对应的风过程向量作为预测时刻的风过程向量的k个相似向量。
在本申请的一些可选实施例中,所述获取历史时段内各时刻的风过程向量,并确定预测时刻的风过程向量与历史时段内各时刻的风过程向量之间的欧式距离,包括:按下式确定预测时刻i的风过程向量与历史时段内第j个时刻的风过程向量之间的欧式距离d
i,j:
上式中,j∈(1~N),v
i=[v
i-b,v
i-b+1,...,v
i,...,v
i+b-1,v
i+b]
T,v
i为预测时刻i的风过程向量,v
i+b为第i+b个时刻的风速,b为时间带宽,v
j=[v
j-b,v
j-b+1,...,v
j,...,v
j+b-1,v
j+b]
T,v
j为历史时段内第j个时刻的风过程向量。
在本申请的一些可选实施例中,所述分别通过m个预先训练的风电功率预测模型对所述k个相似向量对应的历史时刻的数值天气预报数据进行处理,获取各预先训练的风电功率预测模型对应的预测时刻的相似功率向 量,包括:按下式确定第c个风电功率预测模型对应的预测时刻的相似功率向量
上式中,P
cr为第c个风电功率预测模型对k个相似向量中第r个相似向量对应的数值天气预报数据进行处理后得到的风电功率数据。
在本申请的一些可选实施例中,所述利用各预先训练的风电功率预测模型对应的预测时刻的相似功率向量确定各预先训练的风电功率预测模型在预测时刻的线性回归系数,包括:按下式确定第1至第m个预先训练的风电功率预测模型在预测时刻的线性回归系数[β
1...β
c...β
m]:
上式中,
为第c个风电功率预测模型对k个相似向量对应的历史时刻的数值天气预报数据进行处理后得到的预测时刻的相似功率向量,P
k=[P
1...P
r...P
k],P
k为k个相似向量对应的实际风电功率向量,P
r为k个相似向量中第r个相似向量对应的实际风电功率,β
c为第c个风电功率预测模型在预测时刻对应的线性回归系数。
本申请实施例还提供了一种基于风过程识别的风电功率预测集成优化装置,所述装置包括:获取模块,配置为通过一个或多个预先训练的风电功率预测模型对预测时刻的数值天气预报数据进行处理,获取各预先训练的风电功率预测模型对应的预测时刻功率数据;确定模块,配置为利用所述一个或多个预先训练的风电功率预测模型在预测时刻对应的线性回归系数和各预先训练的风电功率预测模型对应的预测时刻功率数据确定最优的预测时刻功率数据;其中,所述一个或多个预先训练的风电功率预测模型在预测时刻对应的线性回归系数基于各预先训练的风电功率预测模型的训练数据获取。
在本申请的一些可选实施例中,所述装置还包括训练模块,配置为对 数据集进行y组自助法随机重采样,获取y组训练数据;所述数据集包括历史时段内多个时刻的数值天气预报数据及其对应的风电功率数据;y为正整数;还配置为将所述y组训练数据中数值天气预报数据分别作为各风电功率预测模型的输入层数据,所述y组训练数据中数值天气预报数据对应的风电功率数据分别作为各风电功率预测模型的输出层数据进行训练,获得m个风电功率预测模型;m为正整数;其中,每组训练数据包括q次有放回随机重采样构成的q个样本数据,q为正整数;则m=y×g,g表示风电功率预测模型的数量。
在本申请的一些可选实施例中,所述风电功率预测模型包括:BP神经网络模型、支持向量回归模型、决策树回归模型和k近邻回归模型。
在本申请的一些可选实施例中,所述获取模块,还配置为:在历史时段内各时刻的风过程向量中选取与预测时刻的风过程向量对应的k个相似向量;k为正整数;分别通过m个预先训练的风电功率预测模型对所述k个相似向量对应的历史时刻的数值天气预报数据进行处理,获取各预先训练的风电功率预测模型对应的预测时刻的相似功率向量;m为正整数;利用各预先训练的风电功率预测模型对应的预测时刻的相似功率向量确定各预先训练的风电功率预测模型在预测时刻的线性回归系数。
在本申请的一些可选实施例中,所述获取模块,配置为:获取历史时段内各时刻的风过程向量,并确定预测时刻的风过程向量与历史时段内各时刻的风过程向量之间的欧式距离;将预测时刻的风过程向量与历史时段内各时刻的风过程向量之间的欧式距离进行降序排列,并选择所述降序序 列中前k个欧式距离对应的风过程向量作为预测时刻的风过程向量的k个相似向量。
上式中,j∈(1~N),v
i=[v
i-b,v
i-b+1,...,v
i,...,v
i+b-1,v
i+b]
T,v
i为预测时刻i的风过程向量,v
i+b为第i+b个时刻的风速,b为时间带宽,v
j=[v
j-b,v
j-b+1,...,v
j,...,v
j+b-1,v
j+b]
T,v
j为历史时段内第j个时刻的风过程向量。
上式中,P
cr为第c个风电功率预测模型对k个相似向量中第r个相似向量对应的数值天气预报数据进行处理后得到的风电功率数据。
上式中,
为第c个风电功率预测模型对k个相似向量对应的历史时刻的数值天气预报数据进行处理后得到的预测时刻的相似功率向量,P
k=[P
1...P
r...P
k],P
k为k个相似向量对应的实际风电功率向量,P
r为k个相似向量中第r个相似向量对应的实际风电功率,β
c为第c个风电功率预测模型在预测时刻对应的线性回归系数。
本申请实施例还提供了一种计算机可读存储介质,其上存储有计算机程序,该计算机程序被处理器执行时实现本申请实施例所述基于风过程识 别的风电功率预测集成优化方法的步骤。
本申请实施例提供的技术方案,通过各预先训练的风电功率预测模型对预测时刻的数值天气预报数据进行处理,获取各预先训练的风电功率预测模型输出的预测时刻功率数据;利用各预先训练的风电功率预测模型在预测时刻对应的线性回归系数和各预先训练的风电功率预测模型输出的预测时刻功率数据确定最优的预测时刻功率数据,通过该方法可以快速有效的匹配识别预测时刻的风过程的相似的向量同时对风电功率进行集成优化,进而提高风电功率预测的精准度,预测方法简单快捷,可以广泛的推广应用。
图1是本申请实施例提供的一种基于风过程识别的风电功率预测集成优化方法的流程示意图;
图2(a)至图2(d)是本申请实施例提供的四个风过程的相似向量的识别和匹配算例结果图;
图3是本申请实施例提供的风电场的预测功率和实际功率时间序列图;
图4是本申请实施例提供的一种基于风过程识别的风电功率预测集成优化装置的结构示意图。
下面结合附图对本申请的具体实施方式作进一步的详细说明。
为使本申请实施例的目的、技术方案和优点更加清楚,下面将结合本申请实施例中的附图,对本申请实施例中的技术方案进行清楚、完整地描述,显然,所描述的实施例是本申请一部分实施例,而不是全部的实施例。基于本申请中的实施例,本领域普通技术人员在没有做出创造性劳动前提下所获得的所有其它实施例,都属于本申请实施例保护的范围。
本申请实施例提供了一种基于风过程识别的风电功率预测集成优化方法,如图1所示,所述方法包括:
步骤101:通过一个或多个预先训练的风电功率预测模型对预测时刻的数值天气预报数据进行处理,获取各预先训练的风电功率预测模型对应的预测时刻功率数据;
步骤102:利用所述一个或多个预先训练的风电功率预测模型在预测时刻对应的线性回归系数和各预先训练的风电功率预测模型对应的预测时刻功率数据确定最优的预测时刻功率数据;
其中,所述一个或多个预先训练的风电功率预测模型在预测时刻对应的线性回归系数基于各预先训练的风电功率预测模型的训练数据获取。
本申请的一些可选实施例中,一个或多个风电功率预测模型的训练过程包括:对数据集进行y组自助法随机重采样,获取y组训练数据;所述数据集包括历史时段内多个时刻的数值天气预报数据及其对应的风电功率数据;y为正整数;将所述y组训练数据中数值天气预报数据分别作为各风电功率预测模型的输入层数据,所述y组训练数据中数值天气预报数据对应的风电功率数据分别作为各风电功率预测模型的输出层数据进行训练,获得m个风电功率预测模型;m为正整数;其中,每组训练数据包括q次有放回随机重采样构成的q个样本数据,q为正整数;则m=y×g,g表示风电功率预测模型的数量。
其中,所述风电功率预测模型包括:BP神经网络模型、支持向量回归模型、决策树回归模型和k近邻回归模型。
本申请的最优实施例中,所述利用所述一个或多个预先训练的风电功率预测模型在预测时刻对应的线性回归系数和各预先训练的风电功率预测模型对应的预测时刻功率数据确定最优的预测时刻功率数据,包括:
在一些可选实施例中,所述一个或多个预先训练的风电功率预测模型在预测时刻对应的线性回归系数的获取过程包括:
在历史时段内各时刻的风过程向量中选取与预测时刻的风过程向量对应的k个相似向量;k为正整数;分别通过m个预先训练的风电功率预测模型对所述k个相似向量对应的历史时刻的数值天气预报数据进行处理,获取各预先训练的风电功率预测模型输出的预测时刻的相似功率向量;m为正整数;利用各预先训练的风电功率预测模型对应的预测时刻的相似功率向量确定各预先训练的风电功率预测模型在预测时刻的线性回归系数;
其中,用于获取各预先训练的风电功率预测模型在预测时刻对应的线性回归系数的训练数据为1年以上的数值天气预报数据和其对应的风电功率数据。
示例性的,如图2(a)、图2(b)、图2(c)和图2(d)所示,分别给出了四个风过程的相似向量的识别匹配算例结果图,其中粗实线为目标风过程,其他细曲线为匹配的相似风过程,算例中时间带宽b为4小时,时间分辨率为15分钟,从算例中可以发现风过程的匹配程度较好。
在本申请的一些可选实施例中,所述在历史时段内各时刻的风过程向量中选取与预测时刻的风过程向量对应的k个相似向量,包括:获取历史时段内各时刻的风过程向量,并确定预测时刻的风过程向量与历史时段内各时刻的风过程向量之间的欧式距离;将预测时刻的风过程向量与历史时段内各时刻的风过程向量之间的欧式距离进行降序排列,并选择所述降序序列中前k个欧式距离对应的风过程向量作为预测时刻的风过程向量的k个相似向量。
在本申请的一些可选实施例中,所述获取历史时段内各时刻的风过程向量,并确定预测时刻的风过程向量与历史时段内各时刻的风过程向量之间的欧式距离,包括:按下式确定预测时刻i的风过程向量与历史时段内第j个时刻的风过程向量之间的欧式距离d
i,j:
上式中,j∈(1~N),v
i=[v
i-b,v
i-b+1,...,v
i,...,v
i+b-1,v
i+b]
T,v
i为预测时刻i的风过程向量,v
i+b为第i+b个时刻的风速,b为时间带宽,v
j=[v
j-b,v
j-b+1,...,v
j,...,v
j+b-1,v
j+b]
T,v
j为历史时段内第j个时刻的风过程向量。
在本申请的一些可选实施例中,所述将k个相似向量对应的历史时刻的数值天气预报数据分别代入m个预先训练的风电功率预测模型中,获取各预先训练的风电功率预测模型输出的预测时刻的相似功率向量,包括:
上式中,P
cr为第c个风电功率预测模型对k个相似向量中第r个相似向量对应的数值天气预报数据进行处理后得到的风电功率数据。
在本申请的一些可选实施例中,所述利用各预先训练的风电功率预测模型对应的预测时刻的相似功率向量确定各预先训练的风电功率预测模型在预测时刻的线性回归系数,包括:按下式确定第1至第m个预先训练的风电功率预测模型在预测时刻的线性回归系数[β
1...β
c...β
m]:
上式中,
为第c个风电功率预测模型对k个相似向量对应的历史时刻的数值天气预报数据进行处理后得到的预测时刻的相似功率向量,P
k=[P
1...P
r...P
k],P
k为k个相似向量对应的实际风电功率向 量,P
r为k个相似向量中第r个相似向量对应的实际风电功率,β
c为第c个风电功率预测模型在预测时刻对应的线性回归系数。
在本申请的可选实施例中,表1-表3给出了某省三个风电场的预测结果,风电功率预测模型在每天8:00给出第二天00:00 23:45共96个点(以15分钟为分辨率)的预测结果。其中,每个场站给出了3个差异化模型,即模型1-模型3,“所提优化模型”为利用各预先训练的风电功率预测模型在预测时刻对应的线性回归系数和各预先训练的风电功率预测模型输出的预测时刻功率数据确定的最优的预测时刻功率数据。根据表格中的结果显示,所提优化模型可以明显提高预测精度。为直观说明效果,如图3所示,给出了风电场3对应的时间序列图,可以发现,优化模型可以利用预测模型间差异化的互补效应,优化预测结果。
表1 风电场1预测指标
表2 风电场2预测指标
表3 风电场3预测指标
本申请实施例还提供了一种基于风过程识别的风电功率预测集成优化装置,如图4所示,所述装置包括:
获取模块,配置为通过一个或多个预先训练的风电功率预测模型对预测时刻的数值天气预报数据进行处理,获取各预先训练的风电功率预测模型对应的预测时刻功率数据;
确定模块,配置为利用所述一个或多个预先训练的风电功率预测模型在预测时刻对应的线性回归系数和各预先训练的风电功率预测模型对应的预测时刻功率数据确定最优的预测时刻功率数据;
其中,所述一个或多个预先训练的风电功率预测模型在预测时刻对应的线性回归系数基于各预先训练的风电功率预测模型的训练数据获取。
本申请的一些可选实施例中,所述装置还包括训练模块,配置为对数据集进行y组自助法随机重采样,获取y组训练数据;所述数据集包括历史时段内多个时刻的数值天气预报数据及其对应的风电功率数据;y为正整数;将所述y组训练数据中数值天气预报数据分别作为各风电功率预测模型的输入层数据,所述y组训练数据中数值天气预报数据对应的风电功率数据分别作为各风电功率预测模型的输出层数据进行训练,获得m个风电 功率预测模型;m为正整数;
其中,每组训练数据包括q次有放回随机重采样构成的q个样本数据,q为正整数;则m=y×g,g表示风电功率预测模型的数量。
本申请的一些可选实施例中,所述风电功率预测模型包括:BP神经网络模型、支持向量回归模型、决策树回归模型和k近邻回归模型。
本申请的一些可选实施例中,所述确定模块,配置为:
本申请的一些可选实施例中,所述获取模块,还配置为:在历史时段内各时刻的风过程向量中选取与预测时刻的风过程向量对应的k个相似向量;k为正整数;分别通过m个预先训练的风电功率预测模型对所述k个相似向量对应的历史时刻的数值天气预报数据进行处理,获取各预先训练的风电功率预测模型对应的预测时刻的相似功率向量;m为正整数;利用各预先训练的风电功率预测模型对应的预测时刻的相似功率向量确定各预先训练的风电功率预测模型在预测时刻的线性回归系数。
本申请的一些可选实施例中,所述获取模块,配置为:获取历史时段内各时刻的风过程向量,并确定预测时刻的风过程向量与历史时段内各时刻的风过程向量之间的欧式距离;将预测时刻的风过程向量与历史时段内各时刻的风过程向量之间的欧式距离进行降序排列,并选择所述降序序列中前k个欧式距离对应的风过程向量作为预测时刻的风过程向量的k个相似向量。
本申请的一些可选实施例中,所述获取模块,配置为:按下式确定预测时刻i的风过程向量与历史时段内第j个时刻的风过程向量之间的欧式距 离d
i,j:
上式中,j∈(1~N),v
i=[v
i-b,v
i-b+1,...,v
i,...,v
i+b-1,v
i+b]
T,v
i为预测时刻i的风过程向量,v
i+b为第i+b个时刻的风速,b为时间带宽,v
j=[v
j-b,v
j-b+1,...,v
j,...,v
j+b-1,v
j+b]
T,v
j为历史时段内第j个时刻的风过程向量。
上式中,P
cr为第c个风电功率预测模型对k个相似向量中第r个相似向量对应的数值天气预报数据进行处理后得到的风电功率数据。
本申请的一些可选实施例中,所述获取模块,配置为:按下式确定第1至第m个预先训练的风电功率预测模型在预测时刻的线性回归系数[β
1...β
c...β
m]:
上式中,
为第c个风电功率预测模型对k个相似向量对应的历史时刻的数值天气预报数据进行处理后得到的预测时刻的相似功率向量,P
k=[P
1...P
r...P
k],P
k为k个相似向量对应的实际风电功率向量,P
r为k个相似向量中第r个相似向量对应的实际风电功率,β
c为第c个风电功率预测模型在预测时刻对应的线性回归系数。
本发明实施例中,所述装置可应用于电子设备中。所述装置中的获取模块、确定模块和训练模块,在实际应用中均可由中央处理器(CPU,Central Processing Unit)、数字信号处理器(DSP,Digital Signal Processor)、微控制单元(MCU,Microcontroller Unit)或可编程门阵列(FPGA,Field- Programmable Gate Array)实现。
在示例性实施例中,所述装置可以被一个或多个应用专用集成电路(ASIC,Application Specific Integrated Circuit)、DSP、可编程逻辑器件(PLD,Programmable Logic Device)、复杂可编程逻辑器件(CPLD,Complex Programmable Logic Device)、FPGA、通用处理器、控制器、MCU、微处理器(Microprocessor)、或其他电子元件实现,用于执行前述方法。
本申请实施例还提供了一种计算机可读存储介质,其上存储有计算机程序,该计算机程序被处理器执行时实现本申请实施例所述基于风过程识别的风电功率预测集成优化方法的步骤。
本申请所提供的几个方法或设备实施例中所揭露的特征,在不冲突的情况下可以任意组合,得到新的方法实施例或设备实施例。
本领域内的技术人员应明白,本申请的实施例可提供为方法、装置、或计算机程序产品。因此,本申请可采用完全硬件实施例、完全软件实施例、或结合软件和硬件方面的实施例的形式。而且,本申请可采用在一个或多个其中包含有计算机可用程序代码的计算机可用存储介质(包括但不限于磁盘存储器、CD-ROM、光学存储器等)上实施的计算机程序产品的形式。
本申请是参照根据本申请实施例的方法、装置、和计算机程序产品的流程图和/或方框图来描述的。应理解可由计算机程序指令实现流程图和/或方框图中的每一流程和/或方框、以及流程图和/或方框图中的流程和/或方杠的结合。可提供这些计算机程序指令到通用计算机、专用计算机、嵌入式处理机或其他可编程数据处理设备的处理器以产生一个机器,使得通过计算机或其他可编程数据处理设备的处理器执行的指令产生用于实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能的装置。
这些计算机程序指令也可存储在能引导计算机或其他可编程数据处理设备以特定方式工作的计算机可读存储器中,使得存储在该计算机可读存储器中的指令产生包括指令装置的制造品,该指令装置实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能。
这些计算机程序指令也可装载到计算机或其他可编程数据处理设备上,使得在计算机或其他可编程设备上执行一系列操作步骤以产生计算机实现的处理,从而在计算机或其他可编程设备上执行的指令提供用于实现在流程图一个流程或多个流程和/或方框图一个方框或多个方框中指定的功能的步骤。
最后应当说明的是:以上实施例仅用以说明本申请的技术方案而非对其限制,尽管参照上述实施例对本申请进行了详细的说明,所属领域的普通技术人员应当理解:依然可以对本申请的具体实施方式进行修改或者等同替换,而未脱离本申请精神和范围的任何修改或者等同替换,其均应涵盖在本申请的权利要求保护范围之内。
Claims (19)
- 一种基于风过程识别的风电功率预测集成优化方法,所述方法包括:通过一个或多个预先训练的风电功率预测模型对预测时刻的数值天气预报数据进行处理,获取各预先训练的风电功率预测模型对应的预测时刻功率数据;利用所述一个或多个预先训练的风电功率预测模型在预测时刻对应的线性回归系数和各预先训练的风电功率预测模型对应的预测时刻功率数据确定最优的预测时刻功率数据;其中,所述一个或多个预先训练的风电功率预测模型在预测时刻对应的线性回归系数基于各预先训练的风电功率预测模型的训练数据获取。
- 如权利要求1所述的方法,其中,一个或多个风电功率预测模型的训练过程包括:对数据集进行y组自助法随机重采样,获取y组训练数据;所述数据集包括历史时段内多个时刻的数值天气预报数据及其对应的风电功率数据;y为正整数;将所述y组训练数据中数值天气预报数据分别作为各风电功率预测模型的输入层数据,所述y组训练数据中数值天气预报数据对应的风电功率数据分别作为各风电功率预测模型的输出层数据进行训练,获得m个风电功率预测模型;m为正整数;其中,每组训练数据包括q次有放回随机重采样构成的q个样本数据,q为正整数;则m=y×g,g表示风电功率预测模型的数量。
- 如权利要求2所述的方法,其中,所述风电功率预测模型包括:反向传播BP神经网络模型、支持向量回归模型、决策树回归模型和k近邻回归模型。
- 如权利要求1所述的方法,其中,所述一个或多个预先训练的风电功率预测模型在预测时刻对应的线性回归系数的获取过程包括:在历史时段内各时刻的风过程向量中选取与预测时刻的风过程向量对应的k个相似向量;k为正整数;分别通过m个预先训练的风电功率预测模型对所述k个相似向量对应的历史时刻的数值天气预报数据进行处理,获取各预先训练的风电功率预测模型对应的预测时刻的相似功率向量;m为正整数;利用各预先训练的风电功率预测模型对应的预测时刻的相似功率向量确定各预先训练的风电功率预测模型在预测时刻的线性回归系数。
- 如权利要求5所述的方法,其中,所述在历史时段内各时刻的风过程向量中选取与预测时刻的风过程向量对应的k个相似向量,包括:获取历史时段内各时刻的风过程向量,并确定预测时刻的风过程向量与历史时段内各时刻的风过程向量之间的欧式距离;将预测时刻的风过程向量与历史时段内各时刻的风过程向量之间的欧式距离进行降序排列,并选择所述降序序列中前k个欧式距离对应的风过程向量作为预测时刻的风过程向量的k个相似向量。
- 一种基于风过程识别的风电功率预测集成优化装置,所述装置包括:获取模块,配置为通过一个或多个预先训练的风电功率预测模型对预测时刻的数值天气预报数据进行处理,获取各预先训练的风电功率预测模型对应的预测时刻功率数据;确定模块,配置为利用所述一个或多个预先训练的风电功率预测模型在预测时刻对应的线性回归系数和各预先训练的风电功率预测模型对应的预测时刻功率数据确定最优的预测时刻功率数据;其中,所述一个或多个预先训练的风电功率预测模型在预测时刻对应的线性回归系数基于各预先训练的风电功率预测模型的训练数据获取。
- 如权利要求10所述的装置,其中,所述装置还包括训练模块,配置为对数据集进行y组自助法随机重采样,获取y组训练数据;所述数据集包括历史时段内多个时刻的数值天气预报数据及其对应的风电功率数据;y为正整数;将所述y组训练数据中数值天气预报数据分别作为各风电功率预测模型的输入层数据,所述y组训练数据中数值天气预报数据对应的风电功率数据分别作为各风电功率预测模型的输出层数据进行训练,获得m个预先训练的风电功率预测模型;m为正整数;其中,每组训练数据包括q次有放回随机重采样构成的q个样本数据,q为正整数;则m=y×g,g表示风电功率预测模型的数量。
- 如权利要求11所述的装置,其中,所述风电功率预测模型包括: BP神经网络模型、支持向量回归模型、决策树回归模型和k近邻回归模型。
- 如权利要求10所述的装置,其中,所述获取模块,还配置为:在历史时段内各时刻的风过程向量中选取与预测时刻的风过程向量对应的k个相似向量;k为正整数;分别通过m个预先训练的风电功率预测模型对所述k个相似向量对应的历史时刻的数值天气预报数据进行处理,获取各预先训练的风电功率预测模型对应的预测时刻的相似功率向量;m为正整数;利用各预先训练的风电功率预测模型对应的预测时刻的相似功率向量确定各预先训练的风电功率预测模型在预测时刻的线性回归系数。
- 如权利要求14所述的装置,其中,所述获取模块,配置为:获取历史时段内各时刻的风过程向量,并确定预测时刻的风过程向量与历史时段内各时刻的风过程向量之间的欧式距离;将预测时刻的风过程向量与历史时段内各时刻的风过程向量之间的欧式距离进行降序排列,并选择所述降序序列中前k个欧式距离对应的风过程向量作为预测时刻的风过程向量的k个相似向量。
- 一种计算机可读存储介质,其上存储有计算机程序,该计算机程序被处理器执行时实现权利要求1至9任一项所述方法的步骤。
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
PCT/CN2021/111898 WO2023015460A1 (zh) | 2021-08-10 | 2021-08-10 | 一种基于风过程识别的风电功率预测集成优化方法及装置 |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
PCT/CN2021/111898 WO2023015460A1 (zh) | 2021-08-10 | 2021-08-10 | 一种基于风过程识别的风电功率预测集成优化方法及装置 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2023015460A1 true WO2023015460A1 (zh) | 2023-02-16 |
Family
ID=85200395
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/CN2021/111898 WO2023015460A1 (zh) | 2021-08-10 | 2021-08-10 | 一种基于风过程识别的风电功率预测集成优化方法及装置 |
Country Status (1)
Country | Link |
---|---|
WO (1) | WO2023015460A1 (zh) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116911467A (zh) * | 2023-09-12 | 2023-10-20 | 浙江华云电力工程设计咨询有限公司 | 一种可再生能源出力的预测方法、装置及存储介质 |
Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2018007312A (ja) * | 2016-06-27 | 2018-01-11 | 藤崎電機株式会社 | 発電電力予測装置、サーバ、コンピュータプログラム及び発電電力予測方法 |
CN112651576A (zh) * | 2021-01-07 | 2021-04-13 | 云南电力技术有限责任公司 | 长期风电功率预测方法及装置 |
CN113011625A (zh) * | 2019-12-19 | 2021-06-22 | 中国电力科学研究院有限公司 | 一种基于风过程识别的风电功率预测集成优化方法及装置 |
-
2021
- 2021-08-10 WO PCT/CN2021/111898 patent/WO2023015460A1/zh unknown
Patent Citations (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JP2018007312A (ja) * | 2016-06-27 | 2018-01-11 | 藤崎電機株式会社 | 発電電力予測装置、サーバ、コンピュータプログラム及び発電電力予測方法 |
CN113011625A (zh) * | 2019-12-19 | 2021-06-22 | 中国电力科学研究院有限公司 | 一种基于风过程识别的风电功率预测集成优化方法及装置 |
CN112651576A (zh) * | 2021-01-07 | 2021-04-13 | 云南电力技术有限责任公司 | 长期风电功率预测方法及装置 |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN116911467A (zh) * | 2023-09-12 | 2023-10-20 | 浙江华云电力工程设计咨询有限公司 | 一种可再生能源出力的预测方法、装置及存储介质 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN104573879A (zh) | 基于最优相似日集的光伏电站出力预测方法 | |
CN110210660B (zh) | 一种超短期风速预测方法 | |
CN106529731A (zh) | 一种区域电网光伏电站集群划分方法 | |
CN107609774B (zh) | 一种基于思维进化算法优化小波神经网络的光伏功率预测方法 | |
CN114757427B (zh) | 自回归修正的lstm智能风电场超短期功率预测方法 | |
Li et al. | Deep spatio-temporal wind power forecasting | |
CN105243461A (zh) | 一种基于人工神经网络改进训练策略的短期负荷预测方法 | |
CN113822418A (zh) | 一种风电场功率预测方法、系统、设备和存储介质 | |
CN113344288B (zh) | 梯级水电站群水位预测方法、装置及计算机可读存储介质 | |
CN109002928A (zh) | 一种基于贝叶斯网络模型的电力负荷峰值预测方法和装置 | |
WO2023015460A1 (zh) | 一种基于风过程识别的风电功率预测集成优化方法及装置 | |
CN112270439B (zh) | 超短期风电功率预测方法、装置、电子设备及存储介质 | |
CN116307291B (zh) | 一种基于小波分解的分布式光伏发电预测方法及预测终端 | |
Chen et al. | Research on wind power prediction method based on convolutional neural network and genetic algorithm | |
Cheng et al. | A novel fuzzy time series forecasting method based on fuzzy logical relationships and similarity measures | |
CN116777039A (zh) | 基于训练集分段和误差修正的双层神经网络风速预测方法 | |
CN116799796A (zh) | 一种光伏发电功率预测方法、装置、设备及介质 | |
CN115330096A (zh) | 基于时序序列的能量数据中长期预测方法、装置及介质 | |
CN114418180B (zh) | 一种风电功率的超短期预测方法、装置及存储介质 | |
CN110598947A (zh) | 一种基于改进布谷鸟-神经网络算法的负荷预测方法 | |
CN117332896A (zh) | 多层集成学习的新能源小时间尺度功率预测方法及系统 | |
CN117290673A (zh) | 一种基于多模型融合的船舶能耗高精度预测系统 | |
CN111476402A (zh) | 耦合气象信息与emd技术的风电发电能力预测方法 | |
Lijuan et al. | A novel model for wind power forecasting based on Markov residual correction | |
CN117421646A (zh) | 风电功率异常值的预测方法、装置、存储介质及计算机设备 |
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: 21953091 Country of ref document: EP Kind code of ref document: A1 |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
32PN | Ep: public notification in the ep bulletin as address of the adressee cannot be established |
Free format text: NOTING OF LOSS OF RIGHTS PURSUANT TO RULE 112(1) EPC (EPO FORM 1205A DATED 11.06.2024) |