WO2023115425A1 - Ultra-short-time wind speed prediction method and system - Google Patents

Ultra-short-time wind speed prediction method and system Download PDF

Info

Publication number
WO2023115425A1
WO2023115425A1 PCT/CN2021/140648 CN2021140648W WO2023115425A1 WO 2023115425 A1 WO2023115425 A1 WO 2023115425A1 CN 2021140648 W CN2021140648 W CN 2021140648W WO 2023115425 A1 WO2023115425 A1 WO 2023115425A1
Authority
WO
WIPO (PCT)
Prior art keywords
wind speed
prediction
neural network
ultra
short
Prior art date
Application number
PCT/CN2021/140648
Other languages
French (fr)
Chinese (zh)
Inventor
邵宜祥
刘剑
胡丽萍
过亮
方渊
Original Assignee
南瑞集团有限公司
国网湖北省电力有限公司
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 南瑞集团有限公司, 国网湖北省电力有限公司 filed Critical 南瑞集团有限公司
Publication of WO2023115425A1 publication Critical patent/WO2023115425A1/en

Links

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
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/0464—Convolutional networks [CNN, ConvNet]
    • 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 invention relates to an ultra-short-time wind speed prediction method and system, belonging to the technical field of wind speed prediction in wind farms.
  • Wind power is an important representative of renewable energy power generation technology. With the proposal of the "double carbon" goal, the proportion of its installed capacity in the power system will gradually increase. In order to ensure that the output power of the wind turbine is maintained near the rated value and ensure the safe and stable operation of the power system, a wind turbine pitch system is usually used.
  • the feedforward control method of wind turbines based on wind speed prediction is an advanced control method for wind turbine pitch systems that has been studied more at present, and accurate wind speed prediction technology is a key factor affecting the effect of feedforward control.
  • the lidar wind measurement technology can obtain the required forecasted wind speed, but the implementation of this scheme is restricted by the cost of the lidar, so it is becoming more and more important to develop a reliable and accurate wind speed prediction model to replace the lidar remote sensing equipment to reduce the cost of wind turbines.
  • the focus of the research is restricted by the cost of the lidar, so it is becoming more and more important to develop a reliable and accurate wind speed prediction model to replace the lidar remote sensing equipment to reduce the cost of wind turbines.
  • the use of artificial neural network is a main method for wind speed prediction. For example, process historical wind speed data through wavelet transform, and use backpropagation (BP) neural network for wind speed prediction; use deep belief network (DBN) for short-term wind speed prediction; use Ensemble Empirical Mode Decomposition (EEMD) for wind speed sequence processing, and use the fused long-short-term memory neural network (LSTM) neural network to obtain ultra-short-term wind speed prediction results; build a wind speed prediction model based on wavelet neural network (WNN) for short-term wind speed prediction.
  • BP backpropagation
  • DBN deep belief network
  • EEMD Ensemble Empirical Mode Decomposition
  • LSTM fused long-short-term memory neural network
  • WNN wavelet neural network
  • the technical problem to be solved by the present invention is to overcome the defects of the prior art, provide an ultra-short-time wind speed prediction method and system, reduce the occurrence of large errors at each prediction point, improve the prediction accuracy, and replace the laser radar wind measurement equipment to achieve good wind speed. Ultra-short-term wind speed prediction performance.
  • the present invention provides a method for ultra-short-term wind speed prediction, including:
  • the wind speed values of the first several units of time are input into the pre-trained parallel neural network prediction model based on different neural networks, and the ultra-short-term wind speed prediction results from the moment to be predicted are obtained.
  • the unit time is second.
  • the parallel neural network prediction model based on different neural networks is a parallel neural network prediction model based on NARX neural network and LSTM neural network.
  • the training process of the parallel neural network prediction model based on NARX neural network and LSTM neural network includes:
  • the interpolation method is used to correct the collected historical wind speed data to obtain the wind speed time series suitable for the wind speed prediction of wind turbine feedforward control;
  • the wind speed samples include continuous n+m wind speed values per unit time.
  • the wind speed value per unit time is used as output;
  • the parallel neural network prediction model based on NARX neural network and LSTM neural network is trained, verified and tested, and the trained parallel neural network based on NARX neural network and LSTM neural network is obtained. predictive model.
  • wind speed samples obtained by dividing the wind speed time series include:
  • L is the data length of wind speed time series.
  • the wind speed values of the first several units of time are input into the pre-trained parallel neural network prediction model based on different neural networks, and the ultra-short-term wind speed prediction results from the moment to be predicted are obtained, including:
  • wind speed values of the first several unit times are input to the NARX neural network to obtain the first forecast result from the moment to be predicted;
  • the wind speed values of the first several units of time are input to the LSTM neural network to obtain the second forecast result from the moment to be predicted;
  • An average value of the first prediction result and the second prediction result is calculated to obtain an ultra-short-term wind speed prediction result from the moment to be predicted.
  • wind speed values of the first several units of time are input into the NARX neural network to obtain the first prediction result from the moment to be predicted, including:
  • wind speed values of the first several units of time are input into the LSTM neural network to obtain the second prediction result from the moment to be predicted, including:
  • the calculation of the average value of the first prediction result and the second prediction result to obtain the ultra-short-term wind speed prediction result from the moment to be predicted includes:
  • the P-step prediction using an iterative method includes:
  • the collection module is used to collect the wind speed values of several units of time before the moment when the current wind farm is to be predicted;
  • the processing module is used to input the wind speed values of the first several units of time into the pre-trained parallel neural network prediction model based on different neural networks, and obtain the ultra-short-term wind speed prediction result from the moment to be predicted.
  • the unit time is second.
  • the parallel neural network prediction model based on different neural networks is a parallel neural network prediction model based on NARX neural network and LSTM neural network.
  • processing module includes:
  • the acquisition unit is used to correct the collected historical wind speed data by interpolation method to obtain a wind speed time series suitable for wind speed prediction of fan feed-forward control;
  • the setting unit is used to set the input data dimension n and the output data dimension m;
  • the division unit is used to divide the wind speed time series according to the input data dimension n and the output data dimension m to obtain several wind speed samples, the wind speed samples include continuous n+m wind speed values per unit time, and the wind speed values of the first n unit times are used as Input, the wind speed value of the next m unit time is output;
  • a determination unit is used to determine a training set, a verification set and a test set for wind speed prediction according to the divided fan samples;
  • the training unit is used to train, verify and test the parallel neural network prediction model based on NARX neural network and LSTM neural network according to the training set, verification set and test set of wind speed prediction, and obtain the trained neural network based on NARX neural network and LSTM neural network.
  • a Parallel Neural Network Predictive Model for the Web is used to train, verify and test the parallel neural network prediction model based on NARX neural network and LSTM neural network according to the training set, verification set and test set of wind speed prediction, and obtain the trained neural network based on NARX neural network and LSTM neural network.
  • K K ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇
  • L is the data length of wind speed time series.
  • processing module includes:
  • the first prediction unit is used to input the wind speed values of the first several units of time into the NARX neural network to obtain the first prediction result from the moment to be predicted;
  • the second prediction unit is used to input the wind speed values of the first several units of time into the LSTM neural network to obtain the second prediction result from the moment to be predicted;
  • a calculation unit configured to calculate the average value of the first forecast result and the second forecast result to obtain an ultra-short-term wind speed forecast result from the moment to be predicted.
  • both the first prediction unit and the second prediction unit include an iterative processing unit
  • An electronic device comprising,
  • one or more processors memory, and one or more programs, wherein one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs include Instructions for performing any of the methods described.
  • a readable storage medium the one or more programs include instructions, and the instructions, when executed by a computing device, cause the computing device to execute any one of the methods described above.
  • the selection of a parallel neural network prediction model composed of different neural networks provides a basis for the prediction model to have good prediction performance, and solves the problem of low prediction accuracy and large deviation when a single neural network is overfitting or underfitting , which reduces the occurrence of large errors at each prediction point and effectively improves the accuracy of the prediction results.
  • the invention has good ultra-short-time wind speed prediction performance, can replace laser radar wind measuring equipment to realize wind speed prediction to meet the requirements of feedforward control, and achieve the purpose of saving the composition and manufacturing cost of wind turbines.
  • Fig. 1 is a schematic diagram of the prediction process of the present invention
  • Fig. 2 is a schematic diagram of the model training process of the present invention.
  • a method for ultra-short-time wind speed prediction its steps include:
  • the wind speed values of the first several units of time are input into the pre-trained parallel neural network prediction model based on different neural networks, and the ultra-short-term wind speed prediction results from the moment to be predicted are obtained.
  • the unit time is seconds, so as to realize that the wind speed data collection interval is at the second level.
  • the parallel neural network prediction model based on different neural networks is a parallel neural network prediction model based on NARX neural network and LSTM neural network.
  • the training process of the parallel neural network prediction model based on NARX neural network and LSTM neural network includes:
  • the interpolation method is used to correct the collected historical wind speed data to obtain the wind speed time series suitable for the wind speed prediction of wind turbine feedforward control;
  • the wind speed time series is divided to obtain the wind speed samples
  • the NARX neural network model and LSTM neural network model are trained, verified and tested respectively, and the trained parallel neural network prediction model based on NARX neural network and LSTM neural network is obtained.
  • the wind speed sample is obtained by dividing the wind speed time series according to the data dimension 10, including:
  • L is the data length of wind speed time series.
  • the wind speed values of the first several units of time are input to the pre-trained parallel neural network prediction model based on different neural networks, and the ultra-short-term wind speed prediction results from the moment to be predicted are obtained, including:
  • wind speed values of the first several unit times are input to the NARX neural network to obtain the first forecast result from the moment to be predicted;
  • the wind speed values of the first several units of time are input to the LSTM neural network to obtain the second forecast result from the moment to be predicted;
  • An average value of the first prediction result and the second prediction result is calculated to obtain an ultra-short-term wind speed prediction result from the moment to be predicted.
  • the wind speed values of the first several units of time are input to the NARX neural network to obtain the first prediction result from the moment to be predicted, including:
  • the wind speed values of the first several units of time are input to the LSTM neural network to obtain the second prediction result from the moment to be predicted, including:
  • the calculation of the average value of the first prediction result and the second prediction result to obtain the ultra-short-term wind speed prediction result from the moment to be predicted includes:
  • the P-step prediction using an iterative method includes:
  • the present invention also provides an ultra-short-term wind speed prediction system, including:
  • the collection module is used to collect the wind speed values of several units of time before the moment when the current wind farm is to be predicted;
  • the processing module is used to input the wind speed values of the first several units of time into the pre-trained parallel neural network prediction model based on different neural networks, and obtain the ultra-short-term wind speed prediction result from the moment to be predicted.
  • the unit time is second.
  • the parallel neural network prediction model based on different neural networks is a parallel neural network prediction model based on NARX neural network and LSTM neural network.
  • processing module includes:
  • the acquisition unit is used to correct the collected historical wind speed data by interpolation method to obtain a wind speed time series suitable for wind speed prediction of fan feed-forward control;
  • the setting unit is used to set the data dimension n of input and output, and is used to determine that the output of the training set is the wind speed value of 1 unit time, and the input is the wind speed value of the first n unit time corresponding to the output unit time;
  • the division unit is used to divide the wind speed time series according to the data dimension n to obtain wind speed samples
  • the determination unit is used to segment the wind speed samples, and determine the training set, verification set and test set of wind speed prediction according to the input and output data dimensions;
  • the training unit is used to train, verify and test the parallel neural network prediction model based on NARX neural network and LSTM neural network according to the training set, verification set and test set of wind speed prediction, and obtain the trained neural network based on NARX neural network and LSTM neural network.
  • a Parallel Neural Network Predictive Model for the Web is used to train, verify and test the parallel neural network prediction model based on NARX neural network and LSTM neural network according to the training set, verification set and test set of wind speed prediction, and obtain the trained neural network based on NARX neural network and LSTM neural network.
  • K K ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇ ⁇
  • L is the data length of wind speed time series.
  • processing module includes:
  • the first prediction unit is used to input the wind speed values of the first several units of time into the NARX neural network to obtain the first prediction result from the moment to be predicted;
  • the second prediction unit is used to input the wind speed values of the first several units of time into the LSTM neural network to obtain the second prediction result from the moment to be predicted;
  • a calculation unit configured to calculate the average value of the first forecast result and the second forecast result to obtain an ultra-short-term wind speed forecast result from the moment to be predicted.
  • both the first prediction unit and the second prediction unit include an iterative processing unit
  • an electronic device including:
  • one or more processors memory, and one or more programs, wherein one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs include Instructions for performing any of the methods described.
  • the present invention also provides a readable storage medium, wherein the one or more programs include instructions, and the instructions, when executed by a computing device, cause the computing device to execute any one of the above methods.
  • NARX neural network and LSTM neural network which are suitable for time series prediction and have strong nonlinear learning ability, are selected to form a parallel neural network prediction model, which provides a basis for the prediction model to have good prediction performance.
  • the prediction results of the two neural networks are considered comprehensively, and the average value is used as the final prediction result, so as to show the central tendency of the data, and solve the problem of low prediction accuracy when a single neural network is overfitting or underfitting.
  • the problem of large deviation reduces the occurrence of large errors at each prediction point and effectively improves the accuracy of the prediction results.
  • the invention has good ultra-short-time wind speed prediction performance, can replace laser radar wind measuring equipment to realize wind speed prediction to meet the requirements of feedforward control, and achieve the purpose of saving the composition and manufacturing cost of wind turbines.
  • the embodiments of the present application may be provided as methods, systems, 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)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Business, Economics & Management (AREA)
  • General Physics & Mathematics (AREA)
  • Strategic Management (AREA)
  • Human Resources & Organizations (AREA)
  • Economics (AREA)
  • Software Systems (AREA)
  • Mathematical Physics (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Computational Linguistics (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Health & Medical Sciences (AREA)
  • Development Economics (AREA)
  • Game Theory and Decision Science (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Marketing (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Tourism & Hospitality (AREA)
  • General Business, Economics & Management (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)
  • Feedback Control In General (AREA)

Abstract

An ultra-short-time wind speed prediction method and system. The method comprises: collecting wind speed values of a plurality of previous units of time of a prediction pending moment of a current wind power plant; and inputting the wind speed values of the plurality of previous units of time into a pre-trained parallel neural network prediction model based on different neural networks to obtain an ultra-short-time wind speed prediction result starting from the prediction pending moment. According to the method, a parallel neural network prediction model consisting of different neural networks is selected, a basis is provided for the prediction model to have good prediction performance, the problems of low prediction precision and large deviation when overfitting or underfitting occurs in a single neural network are solved, the occurrence of large errors of each prediction point is reduced, and the accuracy of the prediction result is effectively improved; the method has good ultra-short-time wind speed prediction performance, can replace laser radar wind measurement equipment to achieve wind speed prediction so as to meet feedforward control requirements, and achieves the purpose of saving the manufacturing costs of a wind turbine.

Description

一种超短时风速预测方法及系统A method and system for ultra-short-term wind speed prediction 技术领域technical field
本发明涉及一种超短时风速预测方法及系统，属于风电场风速预测技术领域。The invention relates to an ultra-short-time wind speed prediction method and system, belonging to the technical field of wind speed prediction in wind farms.
背景技术Background technique
风电是可再生能源发电技术的重要代表，随着“双碳”目标的提出，其装机容量在电力系统的比重将会逐渐增大。为了确保风电机组输出功率维持在额定值附近，保障电力系统的安全稳定运行，通常采用风机变桨系统。基于风速预测的风机前馈控制方法是目前研究较多的风机变桨系统先进控制方法，而精确的风速预测技术是影响前馈控制效果一个关键因素。在风速预测技术中，激光雷达测风技术可以获得所需的预测风速，但该方案的实施受激光雷达成本的制约，因此开发可靠准确的风速预测模型来替代激光雷达遥感设备以降低风机成本成为研究的重点。Wind power is an important representative of renewable energy power generation technology. With the proposal of the "double carbon" goal, the proportion of its installed capacity in the power system will gradually increase. In order to ensure that the output power of the wind turbine is maintained near the rated value and ensure the safe and stable operation of the power system, a wind turbine pitch system is usually used. The feedforward control method of wind turbines based on wind speed prediction is an advanced control method for wind turbine pitch systems that has been studied more at present, and accurate wind speed prediction technology is a key factor affecting the effect of feedforward control. In the wind speed prediction technology, the lidar wind measurement technology can obtain the required forecasted wind speed, but the implementation of this scheme is restricted by the cost of the lidar, so it is becoming more and more important to develop a reliable and accurate wind speed prediction model to replace the lidar remote sensing equipment to reduce the cost of wind turbines. The focus of the research.
目前，采用人工神经网络是进行风速预测的一种主要方法。例如，通过小波变换处理历史风速数据，并采用反向传播(BP)神经网络进行风速预测；基于深度置信网络(DBN)进行短期风速预测；采用集合经验模态分解法(EEMD)对风速序列进行处理，并利用融合长短时记忆神经网络(LSTM)神经网络得到超短时风速预测结果；基于小波神经网络(WNN)构建风速预测模型对风速进行短期预测。现有研究侧重于单一神经网络模型开发以及某种改进算法研究，但单个神经网络用于风速预测时往往由于参数调试不当而出现过拟合和欠拟合问题，使部分预测结果与实际测量结果误差过大，从而降低整体预测结果的可信度。At present, the use of artificial neural network is a main method for wind speed prediction. For example, process historical wind speed data through wavelet transform, and use backpropagation (BP) neural network for wind speed prediction; use deep belief network (DBN) for short-term wind speed prediction; use Ensemble Empirical Mode Decomposition (EEMD) for wind speed sequence processing, and use the fused long-short-term memory neural network (LSTM) neural network to obtain ultra-short-term wind speed prediction results; build a wind speed prediction model based on wavelet neural network (WNN) for short-term wind speed prediction. Existing research focuses on the development of a single neural network model and the research on an improved algorithm. However, when a single neural network is used for wind speed prediction, overfitting and underfitting problems often occur due to improper parameter adjustment, which makes some prediction results different from actual measurement results. The error is too large, which reduces the credibility of the overall forecast results.
发明内容Contents of the invention
本发明所要解决的技术问题是克服现有技术的缺陷，提供一种超短时风速预测方法及系统，减少各预测点较大误差的出现，提高预测精度，代替激光雷达测风设备实现良好的超短时风速预测性能。The technical problem to be solved by the present invention is to overcome the defects of the prior art, provide an ultra-short-time wind speed prediction method and system, reduce the occurrence of large errors at each prediction point, improve the prediction accuracy, and replace the laser radar wind measurement equipment to achieve good wind speed. Ultra-short-term wind speed prediction performance.
为解决上述技术问题，本发明提供一种超短时风速预测方法，包括：In order to solve the above technical problems, the present invention provides a method for ultra-short-term wind speed prediction, including:
采集当前风电场待预测时刻的前若干个单位时间的风速值；Collect the wind speed values of several units of time before the current wind farm to be predicted;
将所述前若干个单位时间的风速值输入到预先训练好的基于不同神经网络的并行神经网络预测模型，得到从待预测时刻起的超短时风速预测结果。The wind speed values of the first several units of time are input into the pre-trained parallel neural network prediction model based on different neural networks, and the ultra-short-term wind speed prediction results from the moment to be predicted are obtained.
进一步的，所述单位时间为秒。Further, the unit time is second.
进一步的，所述基于不同神经网络的并行神经网络预测模型为基于NARX神经网络与LSTM神经网络的并行神经网络预测模型。Further, the parallel neural network prediction model based on different neural networks is a parallel neural network prediction model based on NARX neural network and LSTM neural network.
进一步的，所述基于NARX神经网络与LSTM神经网络的并行神经网络预测模型的训练过程包括：Further, the training process of the parallel neural network prediction model based on NARX neural network and LSTM neural network includes:
采用插值法对采集的历史风速数据进行修正，得到适用于风机前馈控制风速预测的风速时间序列；The interpolation method is used to correct the collected historical wind speed data to obtain the wind speed time series suitable for the wind speed prediction of wind turbine feedforward control;
设置输入数据维度n与输出数据维度m；Set input data dimension n and output data dimension m;
根据输入数据维度n与输出数据维度m划分风速时间序列得到若干个风速样本，所述风速样本包括连续n+m个单位时间的风速值，前n个单位时间的风速值作为输入，后m个单位时间的风速值作为输出；Divide the wind speed time series according to the input data dimension n and the output data dimension m to obtain several wind speed samples. The wind speed samples include continuous n+m wind speed values per unit time. The wind speed value per unit time is used as output;
根据划分的风扇样本确定风速预测的训练集、验证集和测试集；Determine the training set, verification set and test set for wind speed prediction according to the divided fan samples;
根据风速预测的训练集、验证集和测试集对基于NARX神经网络与LSTM神经网络的并行神经网络预测模型进行训练、验证以及测试，得到训练好的基于NARX神经网络与LSTM神经网络的并行神经网络预测模型。According to the training set, verification set and test set of wind speed prediction, the parallel neural network prediction model based on NARX neural network and LSTM neural network is trained, verified and tested, and the trained parallel neural network based on NARX neural network and LSTM neural network is obtained. predictive model.
进一步的，所述划分风速时间序列得到风速样本，包括：Further, the wind speed samples obtained by dividing the wind speed time series include:
将风速时间序列划分为K个长度为N的，有一定重叠的数据段，每一个数据段看作一个样本，得到K个样本，K值通过下式确定：Divide the wind speed time series into K data segments with a length of N and a certain overlap. Each data segment is regarded as a sample, and K samples are obtained. The K value is determined by the following formula:
K＝L-nK=L-n
其中，L为风速时间序列的数据长度。Among them, L is the data length of wind speed time series.
进一步的，所述将所述前若干个单位时间的风速值输入到预先训练好的基于不同神经网络的并行神经网络预测模型，得到从待预测时刻起的超短时风速预测结果，包括：Further, the wind speed values of the first several units of time are input into the pre-trained parallel neural network prediction model based on different neural networks, and the ultra-short-term wind speed prediction results from the moment to be predicted are obtained, including:
将所述前若干个单位时间的风速值输入到所述NARX神经网络，得到从待预测时刻起的第一预测结果；The wind speed values of the first several unit times are input to the NARX neural network to obtain the first forecast result from the moment to be predicted;
将所述前若干个单位时间的风速值输入到所述LSTM神经网络，得到从待预测时刻起的第二预测结果；The wind speed values of the first several units of time are input to the LSTM neural network to obtain the second forecast result from the moment to be predicted;
计算所述第一预测结果与所述第二预测结果的平均值，得到从待预测时刻起的超短时风速预测结果。An average value of the first prediction result and the second prediction result is calculated to obtain an ultra-short-term wind speed prediction result from the moment to be predicted.
进一步的，所述将所述前若干个单位时间的风速值输入到所述NARX神经网络，得到从待预测时刻起的第一预测结果，包括：Further, the wind speed values of the first several units of time are input into the NARX neural network to obtain the first prediction result from the moment to be predicted, including:
将前若干个单位时间的风速值输入到训练好的NARX神经网络模型，并采用迭代法进行P步预测，得到基于NARX神经网络模型的超短时风速预测结果
Figure PCTCN2021140648-appb-000001
表示利用NARX神经网络模型预测的第p次预测值，p＝1,2…,P。 Input the wind speed values of the previous several units of time into the trained NARX neural network model, and use the iterative method for P-step prediction, and obtain the ultra-short-term wind speed prediction results based on the NARX neural network model
Figure PCTCN2021140648-appb-000001
Indicates the p-th predicted value predicted by the NARX neural network model, p=1,2...,P.
进一步的，所述将所述前若干个单位时间的风速值输入到所述LSTM神经网络，得到从待预测时刻起的第二预测结果，包括：Further, the wind speed values of the first several units of time are input into the LSTM neural network to obtain the second prediction result from the moment to be predicted, including:
将前若干个单位时间的风速值输入到训练好的LSTM神经网络模型，并采用迭代法进行P步预测，得到基于LSTM神经网络模型的超短时风速预测结果
Figure PCTCN2021140648-appb-000002
表示利用LSTM神经网络模型预测的第p步预测值，p＝1,2…,P。 Input the wind speed values of the previous several units of time into the trained LSTM neural network model, and use the iterative method for P-step prediction, and obtain the ultra-short-term wind speed prediction results based on the LSTM neural network model
Figure PCTCN2021140648-appb-000002
Indicates the predicted value of the pth step predicted by the LSTM neural network model, p=1,2...,P.
进一步的，所述计算所述第一预测结果与所述第二预测结果的平均值，得到从待预测时刻起的超短时风速预测结果，包括：Further, the calculation of the average value of the first prediction result and the second prediction result to obtain the ultra-short-term wind speed prediction result from the moment to be predicted includes:
取
Figure PCTCN2021140648-appb-000003
和
Figure PCTCN2021140648-appb-000004
平均值得到最终的超短时风速预测结果[y 1,y 2,…,y P]，表示为： Pick
Figure PCTCN2021140648-appb-000003
and
Figure PCTCN2021140648-appb-000004
The final ultra-short-time wind speed prediction results [y 1 ,y 2 ,…,y P ] are obtained by the average value, expressed as:
Figure PCTCN2021140648-appb-000005
Figure PCTCN2021140648-appb-000005
y p表示利用并行神经网络预测模型预测的第p步预测值，p＝1,2…,P。 y p represents the p-th step prediction value predicted by the parallel neural network prediction model, p=1,2...,P.
进一步的，所述采用迭代法进行P步预测，包括：Further, the P-step prediction using an iterative method includes:
利用输入的前n个单位时间的风速值，预测后m个单位时长的风速值，利用前n-m个单位时间的风速值和所述后m个单位时长的风速值，预测上一轮预测之后的m个单位时长的风速值，直到最后预测得到第P个单位时长的风速值。Use the input wind speed value of the first n unit time to predict the wind speed value of the next m unit time, and use the wind speed value of the first n-m unit time and the wind speed value of the last m unit time to predict the wind speed after the last round of prediction The wind speed value of the m unit time length until the wind speed value of the Pth unit time length is finally predicted.
一种超短时风速预测系统，其特征在于，包括：An ultra-short-term wind speed forecasting system is characterized in that it includes:
采集模块，用于采集当前风电场待预测时刻的前若干个单位时间的风速值；The collection module is used to collect the wind speed values of several units of time before the moment when the current wind farm is to be predicted;
处理模块，用于将所述前若干个单位时间的风速值输入到预先训练好的基于不同神经网络的并行神经网络预测模型，得到从待预测时刻起的超短时风速预测结果。The processing module is used to input the wind speed values of the first several units of time into the pre-trained parallel neural network prediction model based on different neural networks, and obtain the ultra-short-term wind speed prediction result from the moment to be predicted.
进一步的，所述单位时间为秒。Further, the unit time is second.
进一步的，所述基于不同神经网络的并行神经网络预测模型为基于NARX神经网络与LSTM神经网络的并行神经网络预测模型。Further, the parallel neural network prediction model based on different neural networks is a parallel neural network prediction model based on NARX neural network and LSTM neural network.
进一步的，所述处理模块包括：Further, the processing module includes:
采集单元，用于采用插值法对采集的历史风速数据进行修正，得到适用于风机前馈控制风速预测的风速时间序列；The acquisition unit is used to correct the collected historical wind speed data by interpolation method to obtain a wind speed time series suitable for wind speed prediction of fan feed-forward control;
设置单元，用于设置输入数据维度n与输出数据维度m；The setting unit is used to set the input data dimension n and the output data dimension m;
划分单元，用于根据输入数据维度n与输出数据维度m划分风速时间序列得到若干个风速样本，所述风速样本包括连续n+m个单位时间的风速值，前n个单位时间的风速值作为输入，后m个单位时间的风速值作为输出；The division unit is used to divide the wind speed time series according to the input data dimension n and the output data dimension m to obtain several wind speed samples, the wind speed samples include continuous n+m wind speed values per unit time, and the wind speed values of the first n unit times are used as Input, the wind speed value of the next m unit time is output;
确定单元，用于根据划分的风扇样本确定风速预测的训练集、验证集和测试集；A determination unit is used to determine a training set, a verification set and a test set for wind speed prediction according to the divided fan samples;
训练单元，用于根据风速预测的训练集、验证集和测试集对基于NARX神经网络与LSTM神经网络的并行神经网络预测模型进行训练、验证以及测试，得到训练好的基于NARX神经网络与LSTM神经网络的并行神经网络预测模型。The training unit is used to train, verify and test the parallel neural network prediction model based on NARX neural network and LSTM neural network according to the training set, verification set and test set of wind speed prediction, and obtain the trained neural network based on NARX neural network and LSTM neural network. A Parallel Neural Network Predictive Model for the Web.
进一步的，所述划分单元，Further, the division unit,
用于将风速时间序列划分为K个长度为N的，有一定重叠的数据段，每一个数据段看作一个样本，得到K个样本，K值通过下式确定：It is used to divide the wind speed time series into K data segments with a length of N and a certain overlap. Each data segment is regarded as a sample, and K samples are obtained. The K value is determined by the following formula:
K＝L-nK=L-n
其中，L为风速时间序列的数据长度。Among them, L is the data length of wind speed time series.
进一步的，所述处理模块包括：Further, the processing module includes:
第一预测单元，用于将所述前若干个单位时间的风速值输入到所述NARX神经网络，得到从待预测时刻起的第一预测结果；The first prediction unit is used to input the wind speed values of the first several units of time into the NARX neural network to obtain the first prediction result from the moment to be predicted;
第二预测单元，用于将所述前若干个单位时间的风速值输入到所述LSTM神经网络，得到从待预测时刻起的第二预测结果；The second prediction unit is used to input the wind speed values of the first several units of time into the LSTM neural network to obtain the second prediction result from the moment to be predicted;
计算单元，用于计算所述第一预测结果与所述第二预测结果的平均值，得到从待预测时刻起的超短时风速预测结果。A calculation unit, configured to calculate the average value of the first forecast result and the second forecast result to obtain an ultra-short-term wind speed forecast result from the moment to be predicted.
进一步的，所述第一预测单元，Further, the first prediction unit,
用于将前若干个单位时间的风速值输入到训练好的NARX神经网络模型，并采用迭代法进行P步预测，得到基于NARX神经网络模型的超短时风速预测结果
Figure PCTCN2021140648-appb-000006
表示利用NARX神经网络模型预测的第p次预测值，p＝1,2…,P。 It is used to input the wind speed values of the previous several units of time into the trained NARX neural network model, and use the iterative method for P-step prediction to obtain ultra-short-term wind speed prediction results based on the NARX neural network model
Figure PCTCN2021140648-appb-000006
Indicates the p-th predicted value predicted by the NARX neural network model, p=1,2...,P.
进一步的，所述第二预测单元，Further, the second prediction unit,
用于将前若干个单位时间的风速值输入到训练好的LSTM神经网络模型，并采用迭代法进行P步预测，得到基于LSTM神经网络模型的超短时风速预测结果
Figure PCTCN2021140648-appb-000007
表示利用LSTM神经网络模型预测的第p步预测值，p＝1,2…,P。 It is used to input the wind speed values of the previous several units of time into the trained LSTM neural network model, and use the iterative method for P-step prediction, and obtain the ultra-short-term wind speed prediction results based on the LSTM neural network model
Figure PCTCN2021140648-appb-000007
Indicates the predicted value of the pth step predicted by the LSTM neural network model, p=1,2...,P.
进一步的，所述计算单元，Further, the computing unit,
用于取
Figure PCTCN2021140648-appb-000008
和
Figure PCTCN2021140648-appb-000009
平均值得到最终的超短时风速预测结果[y 1,y 2,…,y P]，表示为： used to fetch
Figure PCTCN2021140648-appb-000008
and
Figure PCTCN2021140648-appb-000009
The final ultra-short-time wind speed prediction results [y 1 ,y 2 ,…,y P ] are obtained by the average value, expressed as:
Figure PCTCN2021140648-appb-000010
Figure PCTCN2021140648-appb-000010
y p表示利用并行神经网络预测模型预测的第p步预测值，p＝1,2…,P。 y p represents the p-th step prediction value predicted by the parallel neural network prediction model, p=1,2...,P.
进一步的，所述第一预测单元和第二预测单元均包括迭代处理单元，Further, both the first prediction unit and the second prediction unit include an iterative processing unit,
用于利用输入的前n个单位时间的风速值，预测后m个单位时长的风速值，利用前n-m个单位时间的风速值和所述m个单位时长的风速值，预测上一轮预测之后的m个单位时长的风速值，直到最后预测得到第P个单位时长的风速值。It is used to predict the wind speed value of the next m unit time using the wind speed value of the first n unit time input, and use the wind speed value of the first n-m unit time and the wind speed value of the m unit time to predict after the last round of prediction The wind speed value of the m unit time length until the wind speed value of the Pth unit time length is finally predicted.
一种电子设备，包括，An electronic device comprising,
一个或多个处理器、存储器以及一个或多个程序，其中一个或多个程序存储在所述存储器中并被配置为由所述一个或多个处理器执行，所述一个或多个程序包括用于执行所述的方法中的任一方法的指令。one or more processors, memory, and one or more programs, wherein one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs include Instructions for performing any of the methods described.
一种可读存储介质，所述一个或多个程序包括指令，所述指令当由计算设备执行时，使得所述计算设备执行所述的方法中的任一方法。A readable storage medium, the one or more programs include instructions, and the instructions, when executed by a computing device, cause the computing device to execute any one of the methods described above.
本发明所达到的有益效果：The beneficial effect that the present invention reaches:
选择由不同神经网络组成的并行神经网络预测模型，为预测模型具备良好的预测性能提供了基础，解决了单个神经网络在发生过拟合或欠拟合时预测精度较低、偏差较大的问题，减少了各预测点较大误差的出现，有效提高了预测结果的准确度。本发明具备了良好的超短时风速预测性能，可以代替激光雷达测风设备实现风速预测来满足前馈控制的要求，达到节约风电机组成制造本的目的。The selection of a parallel neural network prediction model composed of different neural networks provides a basis for the prediction model to have good prediction performance, and solves the problem of low prediction accuracy and large deviation when a single neural network is overfitting or underfitting , which reduces the occurrence of large errors at each prediction point and effectively improves the accuracy of the prediction results. The invention has good ultra-short-time wind speed prediction performance, can replace laser radar wind measuring equipment to realize wind speed prediction to meet the requirements of feedforward control, and achieve the purpose of saving the composition and manufacturing cost of wind turbines.
附图说明Description of drawings
图1是本发明的预测流程示意图；Fig. 1 is a schematic diagram of the prediction process of the present invention;
图2是本发明的模型训练过程示意图。Fig. 2 is a schematic diagram of the model training process of the present invention.
具体实施方式Detailed ways
下面结合附图对本发明作进一步描述。以下实施例仅用于更加清楚地说明本发明的技术方案，而不能以此来限制本发明的保护范围。The present invention will be further described below in conjunction with the accompanying drawings. The following examples are only used to illustrate the technical solution of the present invention more clearly, but not to limit the protection scope of the present invention.
如图1所示，一种超短时风速预测方法，其步骤包括：As shown in Figure 1, a method for ultra-short-time wind speed prediction, its steps include:
采集当前风电场待预测时刻的前若干个单位时间的风速值；Collect the wind speed values of several units of time before the current wind farm to be predicted;
将所述前若干个单位时间的风速值输入到预先训练好的基于不同神经网络的并行神经网络预测模型，得到从待预测时刻起的超短时风速预测结果。The wind speed values of the first several units of time are input into the pre-trained parallel neural network prediction model based on different neural networks, and the ultra-short-term wind speed prediction results from the moment to be predicted are obtained.
所述单位时间为秒，以实现风速数据采集间隔为秒级。The unit time is seconds, so as to realize that the wind speed data collection interval is at the second level.
所述基于不同神经网络的并行神经网络预测模型为基于NARX神经网络与LSTM神经网络的并行神经网络预测模型。The parallel neural network prediction model based on different neural networks is a parallel neural network prediction model based on NARX neural network and LSTM neural network.
如图2所示，所述基于NARX神经网络与LSTM神经网络的并行神经网络预测模型的训练过程包括：As shown in Figure 2, the training process of the parallel neural network prediction model based on NARX neural network and LSTM neural network includes:
采集风电场的历史风速数据；Collect historical wind speed data of wind farms;
采用插值法对采集的历史风速数据进行修正，得到适用于风机前馈控制风速预测的风速时间序列；The interpolation method is used to correct the collected historical wind speed data to obtain the wind speed time series suitable for the wind speed prediction of wind turbine feedforward control;
设置输入与输出的数据维度10，用于确定训练集的输出为1个单位时间的风速值，输入为输出对应的单位时间的前10个单位时间的风速值；Set the data dimension of input and output to 10, which is used to determine that the output of the training set is the wind speed value of 1 unit time, and the input is the wind speed value of the first 10 unit time corresponding to the output unit time;
根据数据维度10划分风速时间序列得到风速样本；According to the data dimension 10, the wind speed time series is divided to obtain the wind speed samples;
对风速样本进行切分，并依据输入与输出的数据维度，确定风速预测的训练集、验证集和测试集；Segment the wind speed samples, and determine the training set, verification set, and test set for wind speed prediction according to the input and output data dimensions;
根据风速预测的训练集、验证集和测试集分别对NARX神经网络模型和LSTM神经网络模型进行训练、验证和测试，得到训练好的基于NARX神经网络与LSTM神经网络的并行神经网络预测模型。According to the training set, verification set and test set of wind speed prediction, the NARX neural network model and LSTM neural network model are trained, verified and tested respectively, and the trained parallel neural network prediction model based on NARX neural network and LSTM neural network is obtained.
所述根据数据维度10划分风速时间序列得到风速样本，包括：The wind speed sample is obtained by dividing the wind speed time series according to the data dimension 10, including:
将风速时间序列划分为K个长度为11的，有一定重叠的数据段，每一个数据段看作一个样本，得到K个样本，K值通过下式确定：Divide the wind speed time series into K data segments with a length of 11 and a certain overlap. Each data segment is regarded as a sample, and K samples are obtained. The K value is determined by the following formula:
K＝L-nK=L-n
其中，L为风速时间序列的数据长度。Among them, L is the data length of wind speed time series.
风速样本的映射结构如表1所示：The mapping structure of wind speed samples is shown in Table 1:
表1风速样本映射结构Table 1 Wind speed sample mapping structure
Figure PCTCN2021140648-appb-000011
Figure PCTCN2021140648-appb-000011
所述将所述前若干个单位时间的风速值输入到预先训练好的基于不同神经网络的并行神经网络预测模型，得到从待预测时刻起的超短时风速预测结果， 包括：The wind speed values of the first several units of time are input to the pre-trained parallel neural network prediction model based on different neural networks, and the ultra-short-term wind speed prediction results from the moment to be predicted are obtained, including:
将所述前若干个单位时间的风速值输入到所述NARX神经网络，得到从待预测时刻起的第一预测结果；The wind speed values of the first several unit times are input to the NARX neural network to obtain the first forecast result from the moment to be predicted;
将所述前若干个单位时间的风速值输入到所述LSTM神经网络，得到从待预测时刻起的第二预测结果；The wind speed values of the first several units of time are input to the LSTM neural network to obtain the second forecast result from the moment to be predicted;
计算所述第一预测结果与所述第二预测结果的平均值，得到从待预测时刻起的超短时风速预测结果。An average value of the first prediction result and the second prediction result is calculated to obtain an ultra-short-term wind speed prediction result from the moment to be predicted.
所述将所述前若干个单位时间的风速值输入到所述NARX神经网络，得到从待预测时刻起的第一预测结果，包括：The wind speed values of the first several units of time are input to the NARX neural network to obtain the first prediction result from the moment to be predicted, including:
将前若干个单位时间的风速值输入到训练好的NARX神经网络模型，并采用迭代法进行P步预测，得到基于NARX神经网络模型的超短时风速预测结果
Figure PCTCN2021140648-appb-000012
表示利用NARX神经网络模型预测的第p次预测值，p＝1,2…,P。 Input the wind speed values of the previous several units of time into the trained NARX neural network model, and use the iterative method for P-step prediction, and obtain the ultra-short-term wind speed prediction results based on the NARX neural network model
Figure PCTCN2021140648-appb-000012
Indicates the p-th predicted value predicted by the NARX neural network model, p=1,2...,P.
所述将所述前若干个单位时间的风速值输入到所述LSTM神经网络，得到从待预测时刻起的第二预测结果，包括：The wind speed values of the first several units of time are input to the LSTM neural network to obtain the second prediction result from the moment to be predicted, including:
将前若干个单位时间的风速值输入到训练好的LSTM神经网络模型，并采用迭代法进行P步预测，得到基于LSTM神经网络模型的超短时风速预测结果
Figure PCTCN2021140648-appb-000013
表示利用LSTM神经网络模型预测的第p步预测值，p＝1,2…,P。 Input the wind speed values of the previous several units of time into the trained LSTM neural network model, and use the iterative method for P-step prediction, and obtain the ultra-short-term wind speed prediction results based on the LSTM neural network model
Figure PCTCN2021140648-appb-000013
Indicates the predicted value of the pth step predicted by the LSTM neural network model, p=1,2...,P.
所述计算所述第一预测结果与所述第二预测结果的平均值，得到从待预测时刻起的超短时风速预测结果，包括：The calculation of the average value of the first prediction result and the second prediction result to obtain the ultra-short-term wind speed prediction result from the moment to be predicted includes:
取
Figure PCTCN2021140648-appb-000014
和
Figure PCTCN2021140648-appb-000015
平均值得到最终的超短时风速预测结果[y 1,y 2,…,y P]，表示为： Pick
Figure PCTCN2021140648-appb-000014
and
Figure PCTCN2021140648-appb-000015
The final ultra-short-time wind speed prediction results [y 1 ,y 2 ,…,y P ] are obtained by the average value, expressed as:
Figure PCTCN2021140648-appb-000016
Figure PCTCN2021140648-appb-000016
y p表示利用并行神经网络预测模型预测的第p步预测值，p＝1,2…,P。 y p represents the p-th step prediction value predicted by the parallel neural network prediction model, p=1,2...,P.
所述采用迭代法进行P步预测，包括：The P-step prediction using an iterative method includes:
首先利用预测起始单位时间t+1之前的10个单位时间的风速历史数据x t-9,x t-8,…,x t-1,x t预测该起始单位时间t+1的风速值y 1＝x t+1，其次将上一步的预测输出y 1＝x t+1作为新一次的预测的输入，再利用x t-8,x t-7,…,x t,y 1预测y 2＝x t+2，最后通过P-1次迭代实现P步预测，得到超短时风速预测结果。 Firstly, use the historical wind speed data x t-9 , x t-8 ,..., x t-1 , x t of the 10 unit time before the forecast start unit time t+1 to predict the wind speed at the start unit time t+1 value y 1 ＝x t+1 , and then take the forecast output y 1 ＝x t+1 of the previous step as the input of the new forecast, and then use x t-8 , x t-7 ,…,x t ,y 1 Predict y 2 =x t+2 , and finally realize P-step prediction through P-1 iterations, and obtain ultra-short-time wind speed prediction results.
相应的本发明还提供一种超短时风速预测系统，包括：Correspondingly, the present invention also provides an ultra-short-term wind speed prediction system, including:
采集模块，用于采集当前风电场待预测时刻的前若干个单位时间的风速值；The collection module is used to collect the wind speed values of several units of time before the moment when the current wind farm is to be predicted;
处理模块，用于将所述前若干个单位时间的风速值输入到预先训练好的基于不同神经网络的并行神经网络预测模型，得到从待预测时刻起的超短时风速预测结果。The processing module is used to input the wind speed values of the first several units of time into the pre-trained parallel neural network prediction model based on different neural networks, and obtain the ultra-short-term wind speed prediction result from the moment to be predicted.
进一步的，所述单位时间为秒。Further, the unit time is second.
进一步的，所述基于不同神经网络的并行神经网络预测模型为基于NARX神经网络与LSTM神经网络的并行神经网络预测模型。Further, the parallel neural network prediction model based on different neural networks is a parallel neural network prediction model based on NARX neural network and LSTM neural network.
进一步的，所述处理模块包括：Further, the processing module includes:
采集单元，用于采用插值法对采集的历史风速数据进行修正，得到适用于风机前馈控制风速预测的风速时间序列；The acquisition unit is used to correct the collected historical wind speed data by interpolation method to obtain a wind speed time series suitable for wind speed prediction of fan feed-forward control;
设置单元，用于设置输入与输出的数据维度n，用于确定训练集的输出为1个单位时间的风速值，输入为输出对应的单位时间的前n个单位时间的风速值；The setting unit is used to set the data dimension n of input and output, and is used to determine that the output of the training set is the wind speed value of 1 unit time, and the input is the wind speed value of the first n unit time corresponding to the output unit time;
划分单元，用于根据数据维度n划分风速时间序列得到风速样本；The division unit is used to divide the wind speed time series according to the data dimension n to obtain wind speed samples;
确定单元，用于对风速样本进行切分，并依据输入与输出的数据维度，确定风速预测的训练集、验证集和测试集；The determination unit is used to segment the wind speed samples, and determine the training set, verification set and test set of wind speed prediction according to the input and output data dimensions;
训练单元，用于根据风速预测的训练集、验证集和测试集对基于NARX神经网络与LSTM神经网络的并行神经网络预测模型进行训练、验证以及测试，得到训练好的基于NARX神经网络与LSTM神经网络的并行神经网络预测模型。The training unit is used to train, verify and test the parallel neural network prediction model based on NARX neural network and LSTM neural network according to the training set, verification set and test set of wind speed prediction, and obtain the trained neural network based on NARX neural network and LSTM neural network. A Parallel Neural Network Predictive Model for the Web.
进一步的，所述划分单元，Further, the division unit,
用于将风速时间序列划分为K个长度为N的，有一定重叠的数据段，每一个数据段看作一个样本，得到K个样本，K值通过下式确定：It is used to divide the wind speed time series into K data segments with a length of N and a certain overlap. Each data segment is regarded as a sample, and K samples are obtained. The K value is determined by the following formula:
K＝L-nK=L-n
其中，L为风速时间序列的数据长度。Among them, L is the data length of wind speed time series.
进一步的，所述处理模块包括：Further, the processing module includes:
第一预测单元，用于将所述前若干个单位时间的风速值输入到所述NARX神经网络，得到从待预测时刻起的第一预测结果；The first prediction unit is used to input the wind speed values of the first several units of time into the NARX neural network to obtain the first prediction result from the moment to be predicted;
第二预测单元，用于将所述前若干个单位时间的风速值输入到所述LSTM神经网络，得到从待预测时刻起的第二预测结果；The second prediction unit is used to input the wind speed values of the first several units of time into the LSTM neural network to obtain the second prediction result from the moment to be predicted;
计算单元，用于计算所述第一预测结果与所述第二预测结果的平均值，得到从待预测时刻起的超短时风速预测结果。A calculation unit, configured to calculate the average value of the first forecast result and the second forecast result to obtain an ultra-short-term wind speed forecast result from the moment to be predicted.
进一步的，所述第一预测单元，Further, the first prediction unit,
用于将前若干个单位时间的风速值输入到训练好的NARX神经网络模型，并采用迭代法进行P步预测，得到基于NARX神经网络模型的超短时风速预测结果
Figure PCTCN2021140648-appb-000017
表示利用NARX神经网络模型预测的第p次预测值，p＝1,2…,P。 It is used to input the wind speed values of the previous several units of time into the trained NARX neural network model, and use the iterative method for P-step prediction to obtain ultra-short-term wind speed prediction results based on the NARX neural network model
Figure PCTCN2021140648-appb-000017
Indicates the p-th predicted value predicted by the NARX neural network model, p=1,2...,P.
进一步的，所述第二预测单元，Further, the second prediction unit,
用于将前若干个单位时间的风速值输入到训练好的LSTM神经网络模型，并采用迭代法进行P步预测，得到基于LSTM神经网络模型的超短时风速预测结果
Figure PCTCN2021140648-appb-000018
表示利用LSTM神经网络模型预测的第p步预测值，p＝1,2…,P。 It is used to input the wind speed values of the previous several units of time into the trained LSTM neural network model, and use the iterative method for P-step prediction to obtain the ultra-short-term wind speed prediction results based on the LSTM neural network model
Figure PCTCN2021140648-appb-000018
Indicates the predicted value of the pth step predicted by the LSTM neural network model, p=1,2...,P.
进一步的，所述计算单元，Further, the computing unit,
用于取
Figure PCTCN2021140648-appb-000019
和
Figure PCTCN2021140648-appb-000020
平均值得到最终的超短时风速预测结果[y 1,y 2,…,y P]，表示为： used to fetch
Figure PCTCN2021140648-appb-000019
and
Figure PCTCN2021140648-appb-000020
The final ultra-short-time wind speed prediction results [y 1 ,y 2 ,…,y P ] are obtained by the average value, expressed as:
Figure PCTCN2021140648-appb-000021
Figure PCTCN2021140648-appb-000021
y p表示利用并行神经网络预测模型预测的第p步预测值，p＝1,2…,P。 y p represents the p-th step prediction value predicted by the parallel neural network prediction model, p=1,2...,P.
进一步的，所述第一预测单元和第二预测单元均包括迭代处理单元，Further, both the first prediction unit and the second prediction unit include an iterative processing unit,
用于利用输入的前n个单位时间的风速值，预测待预测的第1个单位时长的风速值，利用前n-1个单位时间的风速值和所述第1个单位时长的风速值，预测待预测的第2个单位时长的风速值，直到最后预测得到第P个单位时长的风速值。It is used to predict the wind speed value of the first unit duration to be predicted by using the input wind speed value of the first n units of time, using the wind speed value of the first n-1 unit time and the wind speed value of the first unit duration, Predict the wind speed value of the second unit time to be predicted until the wind speed value of the Pth unit time is finally predicted.
相应的本发明还提供一种电子设备，包括，Correspondingly, the present invention also provides an electronic device, including:
一个或多个处理器、存储器以及一个或多个程序，其中一个或多个程序存储在所述存储器中并被配置为由所述一个或多个处理器执行，所述一个或多个程序包括用于执行所述的方法中的任一方法的指令。one or more processors, memory, and one or more programs, wherein one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs include Instructions for performing any of the methods described.
相应的本发明还提供一种可读存储介质，所述一个或多个程序包括指令，所述指令当由计算设备执行时，使得所述计算设备执行所述的方法中的任一方法。Correspondingly, the present invention also provides a readable storage medium, wherein the one or more programs include instructions, and the instructions, when executed by a computing device, cause the computing device to execute any one of the above methods.
本实施例的方法选择适用于时间序列预测且具有较强非线性学习能力的NARX神经网络和LSTM神经网络组成并行神经网络预测模型，为预测模型具备良好的预测性能提供了基础。采用统筹方法，综合考虑两种神经网络的预测结果，将平均值作为最终的预测结果，从而表现数据的集中趋势，解决了单个神经网络在发生过拟合或欠拟合时预测精度较低、偏差较大的问题，减少了各预测点较大误差的出现，有效提高了预测结果的准确度。本发明具备了良好的超短时风速预测性能，可以代替激光雷达测风设备实现风速预测来满足前馈控制的要求，达到节约风电机组成制造本的目的。In the method of this embodiment, NARX neural network and LSTM neural network, which are suitable for time series prediction and have strong nonlinear learning ability, are selected to form a parallel neural network prediction model, which provides a basis for the prediction model to have good prediction performance. Using the overall planning method, the prediction results of the two neural networks are considered comprehensively, and the average value is used as the final prediction result, so as to show the central tendency of the data, and solve the problem of low prediction accuracy when a single neural network is overfitting or underfitting. The problem of large deviation reduces the occurrence of large errors at each prediction point and effectively improves the accuracy of the prediction results. The invention has good ultra-short-time wind speed prediction performance, can replace laser radar wind measuring equipment to realize wind speed prediction to meet the requirements of feedforward control, and achieve the purpose of saving the composition and manufacturing cost of wind turbines.
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、系统、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。Those skilled in the art should understand that the embodiments of the present application may be provided as methods, systems, 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.
本申请是参照根据本申请实施例的方法、设备(系统)、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/或方框的结合。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。The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the application. It should be understood that each procedure and/or block in the flowchart and/or block diagram, and a combination of procedures and/or blocks in the flowchart and/or block diagram can be realized by computer program instructions. These computer program instructions may be provided to a general purpose computer, special purpose computer, embedded processor, or processor of other programmable data processing equipment to produce a machine such that the instructions executed by the processor of the computer or other programmable data processing equipment produce a An apparatus for realizing the functions specified in one or more procedures of the flowchart and/or one or more blocks of the block diagram.
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。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.
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。These computer program instructions can also be loaded onto a computer or other programmable data processing device, causing a series of operational steps to be performed on the computer or other programmable device to produce a computer-implemented process, thereby The instructions provide steps for implementing the functions specified in the flow chart or blocks of the flowchart and/or the block or blocks of the block diagrams.
以上所述仅是本发明的优选实施方式，应当指出，对于本技术领域的普通技术人员来说，在不脱离本发明技术原理的前提下，还可以做出若干改进和变形，这些改进和变形也应视为本发明的保护范围。The above is only a preferred embodiment of the present invention, it should be pointed out that for those of ordinary skill in the art, without departing from the technical principle of the present invention, some improvements and modifications can also be made. It should also be regarded as the protection scope of the present invention.

Claims (22)

  1. 一种超短时风速预测方法，其特征在于，包括：A method for ultra-short-term wind speed forecasting, characterized in that it comprises:
    采集当前风电场待预测时刻的前若干个单位时间的风速值；Collect the wind speed values of several units of time before the current wind farm to be predicted;
    将所述前若干个单位时间的风速值输入到预先训练好的基于不同神经网络的并行神经网络预测模型，得到从待预测时刻起的超短时风速预测结果。The wind speed values of the first several units of time are input into the pre-trained parallel neural network prediction model based on different neural networks, and the ultra-short-term wind speed prediction results from the moment to be predicted are obtained.
  2. 根据权利要求1所述的超短时风速预测方法，其特征在于，所述单位时间为秒。The ultra-short-term wind speed prediction method according to claim 1, wherein the unit time is second.
  3. 根据权利要求1所述的超短时风速预测方法，其特征在于，所述基于不同神经网络的并行神经网络预测模型为基于NARX神经网络与LSTM神经网络的并行神经网络预测模型。The ultra-short-term wind speed prediction method according to claim 1, wherein the parallel neural network prediction model based on different neural networks is a parallel neural network prediction model based on NARX neural network and LSTM neural network.
  4. 根据权利要求3所述的超短时风速预测方法，其特征在于，所述基于NARX神经网络与LSTM神经网络的并行神经网络预测模型的训练过程包括：The ultra-short-term wind speed prediction method according to claim 3, wherein, the training process of the parallel neural network prediction model based on NARX neural network and LSTM neural network comprises:
    采用插值法对采集的历史风速数据进行修正，得到适用于风机前馈控制风速预测的风速时间序列；The interpolation method is used to correct the collected historical wind speed data to obtain the wind speed time series suitable for the wind speed prediction of wind turbine feedforward control;
    设置输入数据维度n与输出数据维度m；Set input data dimension n and output data dimension m;
    根据输入数据维度n与输出数据维度m划分风速时间序列得到若干个风速样本，所述风速样本包括连续n+m个单位时间的风速值，前n个单位时间的风速值作为输入，后m个单位时间的风速值作为输出；Divide the wind speed time series according to the input data dimension n and the output data dimension m to obtain several wind speed samples. The wind speed samples include continuous n+m wind speed values per unit time. The wind speed value per unit time is used as output;
    根据划分的风扇样本确定风速预测的训练集、验证集和测试集；Determine the training set, verification set and test set for wind speed prediction according to the divided fan samples;
    根据风速预测的训练集、验证集和测试集对基于NARX神经网络与LSTM神经网络的并行神经网络预测模型进行训练、验证以及测试，得到训练好的基于NARX神经网络与LSTM神经网络的并行神经网络预测模型。According to the training set, verification set and test set of wind speed prediction, the parallel neural network prediction model based on NARX neural network and LSTM neural network is trained, verified and tested, and the trained parallel neural network based on NARX neural network and LSTM neural network is obtained. predictive model.
  5. 根据权利要求3所述的超短时风速预测方法，其特征在于，所述划分风速时间序列得到风速样本，包括：The ultra-short-term wind speed prediction method according to claim 3, wherein the wind speed samples obtained by dividing the wind speed time series include:
    将风速时间序列划分为K个长度为N的，有一定重叠的数据段，每一个数据段看作一个样本，得到K个样本，K值通过下式确定：Divide the wind speed time series into K data segments with a length of N and a certain overlap. Each data segment is regarded as a sample, and K samples are obtained. The K value is determined by the following formula:
    K＝L-nK=L-n
    其中，L为风速时间序列的数据长度。Among them, L is the data length of wind speed time series.
  6. 根据权利要求3所述的超短时风速预测方法，其特征在于，The ultra-short-time wind speed prediction method according to claim 3, wherein,
    所述将所述前若干个单位时间的风速值输入到预先训练好的基于不同神经网络的并行神经网络预测模型，得到从待预测时刻起的超短时风速预测结果，包括：The wind speed values of the first several units of time are input to the pre-trained parallel neural network prediction model based on different neural networks, and the ultra-short-term wind speed prediction results from the moment to be predicted are obtained, including:
    将所述前若干个单位时间的风速值输入到所述NARX神经网络，得到从待预测时刻起的第一预测结果；The wind speed values of the first several unit times are input to the NARX neural network to obtain the first forecast result from the moment to be predicted;
    将所述前若干个单位时间的风速值输入到所述LSTM神经网络，得到从待预测时刻起的第二预测结果；The wind speed values of the first several units of time are input to the LSTM neural network to obtain the second forecast result from the moment to be predicted;
    计算所述第一预测结果与所述第二预测结果的平均值，得到从待预测时刻起的超短时风速预测结果。An average value of the first prediction result and the second prediction result is calculated to obtain an ultra-short-term wind speed prediction result from the moment to be predicted.
  7. 根据权利要求6所述的超短时风速预测方法，其特征在于，所述将所述前若干个单位时间的风速值输入到所述NARX神经网络，得到从待预测时刻起的第一预测结果，包括：The ultra-short-term wind speed prediction method according to claim 6, wherein the wind speed values of the first several units of time are input to the NARX neural network to obtain the first prediction result from the moment to be predicted ,include:
    将前若干个单位时间的风速值输入到训练好的NARX神经网络模型，并采用迭代法进行P步预测，得到基于NARX神经网络模型的超短时风速预测结果
    Figure PCTCN2021140648-appb-100001
    表示利用NARX神经网络模型预测的第p次预测值，p＝1,2…,P。     Input the wind speed values of the previous several units of time into the trained NARX neural network model, and use the iterative method for P-step prediction, and obtain the ultra-short-term wind speed prediction results based on the NARX neural network model
    Figure PCTCN2021140648-appb-100001
    Indicates the p-th predicted value predicted by the NARX neural network model, p=1,2...,P.
  8. 根据权利要求7所述的超短时风速预测方法，其特征在于，所述将所述前若干个单位时间的风速值输入到所述LSTM神经网络，得到从待预测时刻起的第二预测结果，包括：The ultra-short-term wind speed prediction method according to claim 7, wherein the wind speed values of the first several units of time are input into the LSTM neural network to obtain the second prediction result from the moment to be predicted ,include:
    将前若干个单位时间的风速值输入到训练好的LSTM神经网络模型，并采用迭代法进行P步预测，得到基于LSTM神经网络模型的超短时风速预测结果
    Figure PCTCN2021140648-appb-100002
    表示利用LSTM神经网络模型预测的第p步预测值，p＝1,2…,P。     Input the wind speed values of the previous several units of time into the trained LSTM neural network model, and use the iterative method for P-step prediction, and obtain the ultra-short-term wind speed prediction results based on the LSTM neural network model
    Figure PCTCN2021140648-appb-100002
    Indicates the predicted value of the pth step predicted by the LSTM neural network model, p=1,2...,P.
  9. 根据权利要求8所述的超短时风速预测方法，其特征在于，所述计算所述第一预测结果与所述第二预测结果的平均值，得到从待预测时刻起的超短时 风速预测结果，包括：The ultra-short-term wind speed prediction method according to claim 8, wherein the calculation of the average value of the first prediction result and the second prediction result obtains the ultra-short-time wind speed prediction from the moment to be predicted Results, including:
    取
    Figure PCTCN2021140648-appb-100003
    和
    Figure PCTCN2021140648-appb-100004
    平均值得到最终的超短时风速预测结果[y 1,y 2,…,y P]，表示为：     Pick
    Figure PCTCN2021140648-appb-100003
    and
    Figure PCTCN2021140648-appb-100004
    The final ultra-short-time wind speed prediction results [y 1 ,y 2 ,…,y P ] are obtained by the average value, expressed as:
    Figure PCTCN2021140648-appb-100005
    Figure PCTCN2021140648-appb-100005
    y p表示利用并行神经网络预测模型预测的第p步预测值，p＝1,2…,P。 y p represents the p-th step prediction value predicted by the parallel neural network prediction model, p=1,2...,P.
  10. 根据权利要求9所述的超短时风速预测方法，其特征在于，所述采用迭代法进行P步预测，包括：The ultra-short-term wind speed prediction method according to claim 9, wherein, said adopting an iterative method to carry out P-step prediction includes:
    利用输入的前n个单位时间的风速值，预测后m个单位时长的风速值，利用前n-m个单位时间的风速值和所述后m个单位时长的风速值，预测上一轮预测之后的m个单位时长的风速值，直到最后预测得到第P个单位时长的风速值。Use the input wind speed value of the first n unit time to predict the wind speed value of the next m unit time, and use the wind speed value of the first n-m unit time and the wind speed value of the last m unit time to predict the wind speed after the last round of prediction The wind speed value of the m unit time length until the wind speed value of the Pth unit time length is finally predicted.
  11. 一种超短时风速预测系统，其特征在于，包括：An ultra-short-term wind speed forecasting system is characterized in that it includes:
    采集模块，用于采集当前风电场待预测时刻的前若干个单位时间的风速值；The collection module is used to collect the wind speed values of several units of time before the moment when the current wind farm is to be predicted;
    处理模块，用于将所述前若干个单位时间的风速值输入到预先训练好的基于不同神经网络的并行神经网络预测模型，得到从待预测时刻起的超短时风速预测结果。The processing module is used to input the wind speed values of the first several units of time into the pre-trained parallel neural network prediction model based on different neural networks, and obtain the ultra-short-term wind speed prediction result from the moment to be predicted.
  12. 根据权利要求11所述的超短时风速预测系统，其特征在于，所述单位时间为秒。The ultra-short-term wind speed prediction system according to claim 11, wherein the unit time is seconds.
  13. 根据权利要求11所述的超短时风速预测系统，其特征在于，所述基于不同神经网络的并行神经网络预测模型为基于NARX神经网络与LSTM神经网络的并行神经网络预测模型。The ultra-short-term wind speed prediction system according to claim 11, wherein the parallel neural network prediction model based on different neural networks is a parallel neural network prediction model based on NARX neural network and LSTM neural network.
  14. 根据权利要求13所述的超短时风速预测系统，其特征在于，所述处理模块包括：The ultra-short-term wind speed prediction system according to claim 13, wherein the processing module includes:
    采集单元，用于采用插值法对采集的历史风速数据进行修正，得到适用于风机前馈控制风速预测的风速时间序列；The acquisition unit is used to correct the collected historical wind speed data by interpolation method to obtain a wind speed time series suitable for wind speed prediction of fan feed-forward control;
    设置单元，用于设置输入数据维度n与输出数据维度m；The setting unit is used to set the input data dimension n and the output data dimension m;
    划分单元，用于根据输入数据维度n与输出数据维度m划分风速时间序列 得到若干个风速样本，所述风速样本包括连续n+m个单位时间的风速值，前n个单位时间的风速值作为输入，后m个单位时间的风速值作为输出；The division unit is used to divide the wind speed time series according to the input data dimension n and the output data dimension m to obtain several wind speed samples, the wind speed samples include continuous n+m wind speed values per unit time, and the wind speed values of the first n unit times are used as Input, the wind speed value of the next m unit time is output;
    确定单元，用于根据划分的风扇样本确定风速预测的训练集、验证集和测试集；A determination unit is used to determine a training set, a verification set and a test set for wind speed prediction according to the divided fan samples;
    训练单元，用于根据风速预测的训练集、验证集和测试集对基于NARX神经网络与LSTM神经网络的并行神经网络预测模型进行训练、验证以及测试，得到训练好的基于NARX神经网络与LSTM神经网络的并行神经网络预测模型。The training unit is used to train, verify and test the parallel neural network prediction model based on NARX neural network and LSTM neural network according to the training set, verification set and test set of wind speed prediction, and obtain the trained neural network based on NARX neural network and LSTM neural network. A Parallel Neural Network Predictive Model for the Web.
  15. 根据权利要求13所述的超短时风速预测系统，其特征在于，所述划分单元，The ultra-short-term wind speed prediction system according to claim 13, wherein the division unit,
    用于将风速时间序列划分为K个长度为N的，有一定重叠的数据段，每一个数据段看作一个样本，得到K个样本，K值通过下式确定：It is used to divide the wind speed time series into K data segments with a length of N and a certain overlap. Each data segment is regarded as a sample, and K samples are obtained. The K value is determined by the following formula:
    K＝L-nK=L-n
    其中，L为风速时间序列的数据长度。Among them, L is the data length of wind speed time series.
  16. 根据权利要求13所述的超短时风速预测系统，其特征在于，所述处理模块包括：The ultra-short-term wind speed prediction system according to claim 13, wherein the processing module includes:
    第一预测单元，用于将所述前若干个单位时间的风速值输入到所述NARX神经网络，得到从待预测时刻起的第一预测结果；The first prediction unit is used to input the wind speed values of the first several units of time into the NARX neural network to obtain the first prediction result from the moment to be predicted;
    第二预测单元，用于将所述前若干个单位时间的风速值输入到所述LSTM神经网络，得到从待预测时刻起的第二预测结果；The second prediction unit is used to input the wind speed values of the first several units of time into the LSTM neural network to obtain the second prediction result from the moment to be predicted;
    计算单元，用于计算所述第一预测结果与所述第二预测结果的平均值，得到从待预测时刻起的超短时风速预测结果。A calculation unit, configured to calculate the average value of the first forecast result and the second forecast result to obtain an ultra-short-term wind speed forecast result from the moment to be predicted.
  17. 根据权利要求16所述的超短时风速预测系统，其特征在于，所述第一预测单元，The ultra-short-term wind speed prediction system according to claim 16, wherein the first prediction unit,
    用于将前若干个单位时间的风速值输入到训练好的NARX神经网络模型，并采用迭代法进行P步预测，得到基于NARX神经网络模型的超短时风速预测结果
    Figure PCTCN2021140648-appb-100006
    表示利用NARX神经网络模型预测的第p次预测值，p＝1,2…,P。     It is used to input the wind speed values of the previous several units of time into the trained NARX neural network model, and use the iterative method for P-step prediction to obtain ultra-short-term wind speed prediction results based on the NARX neural network model
    Figure PCTCN2021140648-appb-100006
    Indicates the p-th predicted value predicted by the NARX neural network model, p=1,2...,P.
  18. 根据权利要求17所述的超短时风速预测系统，其特征在于，所述第二预测单元，The ultra-short-term wind speed prediction system according to claim 17, wherein the second prediction unit,
    用于将前若干个单位时间的风速值输入到训练好的LSTM神经网络模型，并采用迭代法进行P步预测，得到基于LSTM神经网络模型的超短时风速预测结果
    Figure PCTCN2021140648-appb-100007
    表示利用LSTM神经网络模型预测的第p步预测值，p＝1,2…,P。     It is used to input the wind speed values of the previous several units of time into the trained LSTM neural network model, and use the iterative method for P-step prediction to obtain the ultra-short-term wind speed prediction results based on the LSTM neural network model
    Figure PCTCN2021140648-appb-100007
    Indicates the predicted value of the pth step predicted by the LSTM neural network model, p=1,2...,P.
  19. 根据权利要求18所述的超短时风速预测系统，其特征在于，所述计算单元，The ultra-short-term wind speed prediction system according to claim 18, wherein the calculation unit,
    用于取
    Figure PCTCN2021140648-appb-100008
    和
    Figure PCTCN2021140648-appb-100009
    平均值得到最终的超短时风速预测结果[y 1,y 2,…,y P]，表示为：     used to fetch
    Figure PCTCN2021140648-appb-100008
    and
    Figure PCTCN2021140648-appb-100009
    The final ultra-short-time wind speed prediction results [y 1 ,y 2 ,…,y P ] are obtained by the average value, expressed as:
    Figure PCTCN2021140648-appb-100010
    Figure PCTCN2021140648-appb-100010
    y p表示利用并行神经网络预测模型预测的第p步预测值，p＝1,2…,P。 y p represents the p-th step prediction value predicted by the parallel neural network prediction model, p=1,2...,P.
  20. 根据权利要求19所述的超短时风速预测系统，其特征在于，所述第一预测单元和第二预测单元均包括迭代处理单元，The ultra-short-term wind speed prediction system according to claim 19, wherein the first prediction unit and the second prediction unit both include an iterative processing unit,
    用于利用输入的前n个单位时间的风速值，预测后m个单位时长的风速值，利用前n-m个单位时间的风速值和所述m个单位时长的风速值，预测上一轮预测之后的m个单位时长的风速值，直到最后预测得到第P个单位时长的风速值。It is used to predict the wind speed value of the next m unit time using the wind speed value of the first n unit time input, and use the wind speed value of the first n-m unit time and the wind speed value of the m unit time to predict after the last round of prediction The wind speed value of the m unit time length until the wind speed value of the Pth unit time length is finally predicted.
  21. 一种电子设备，其特征在于，包括，一个或多个处理器、存储器以及一个或多个程序，其中一个或多个程序存储在所述存储器中并被配置为由所述一个或多个处理器执行，所述一个或多个程序包括用于执行根据权利要求1至10所述的方法中的任一方法的指令。An electronic device, characterized by comprising, one or more processors, memory and one or more programs, wherein one or more programs are stored in the memory and configured to be processed by the one or more The one or more programs include instructions for performing any one of the methods according to claims 1-10.
  22. 一种可读存储介质，其特征在于，所述一个或多个程序包括指令，所述指令当由计算设备执行时，使得所述计算设备执行根据权利要求1至10所述的方法中的任一方法。A readable storage medium, wherein the one or more programs include instructions that, when executed by a computing device, cause the computing device to perform any of the methods according to claims 1 to 10 One method.
PCT/CN2021/140648 2021-12-20 2021-12-23 Ultra-short-time wind speed prediction method and system WO2023115425A1 (en)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
CN202111563651.8A CN116306798A (en) 2021-12-20 2021-12-20 Ultra-short time wind speed prediction method and system
CN202111563651.8 2021-12-20

Publications (1)

Publication Number Publication Date
WO2023115425A1 true WO2023115425A1 (en) 2023-06-29

Family

ID=86824425

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2021/140648 WO2023115425A1 (en) 2021-12-20 2021-12-23 Ultra-short-time wind speed prediction method and system

Country Status (2)

Country Link
CN (1) CN116306798A (en)
WO (1) WO2023115425A1 (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117313927A (en) * 2023-09-19 2023-12-29 华能澜沧江水电股份有限公司 Wind power generation power prediction method and system based on wavelet neural network
CN117494573A (en) * 2023-11-16 2024-02-02 中山大学 Wind speed prediction method and system and electronic equipment
CN117691630A (en) * 2024-02-04 2024-03-12 西安热工研究院有限公司 Novel power system frequency modulation method and system based on VMD-CEEMD
CN117744893A (en) * 2024-02-19 2024-03-22 西安热工研究院有限公司 Wind speed prediction method and system for energy storage auxiliary black start

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN110210660A (en) * 2019-05-27 2019-09-06 河海大学 A kind of ultra-short term wind speed forecasting method
CN110378518A (en) * 2019-06-24 2019-10-25 浙江大学 A kind of underwater trend prediction technique using LSTM-NARX mixed model
US20210064034A1 (en) * 2019-05-06 2021-03-04 Florida Atlantic University Board Of Trustees Hybrid aerial/underwater robotics system for scalable and adaptable maintenance of aquaculture fish farms

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20210064034A1 (en) * 2019-05-06 2021-03-04 Florida Atlantic University Board Of Trustees Hybrid aerial/underwater robotics system for scalable and adaptable maintenance of aquaculture fish farms
CN110210660A (en) * 2019-05-27 2019-09-06 河海大学 A kind of ultra-short term wind speed forecasting method
CN110378518A (en) * 2019-06-24 2019-10-25 浙江大学 A kind of underwater trend prediction technique using LSTM-NARX mixed model

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
WANG, YU: "Combined Model of Short-Term Wind Speed Prediction for Wind Farms Based on Deep Learning", IEEE TRANSACTIONS ON IMAGE PROCESSING, IEEE, USA, vol. 09, no. 08, 1 April 2021 (2021-04-01), CN, pages 1 - 79, XP009546915, DOI: 10.27517/d.cnki.gzkju.2021.002483 *

Cited By (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN117313927A (en) * 2023-09-19 2023-12-29 华能澜沧江水电股份有限公司 Wind power generation power prediction method and system based on wavelet neural network
CN117494573A (en) * 2023-11-16 2024-02-02 中山大学 Wind speed prediction method and system and electronic equipment
CN117691630A (en) * 2024-02-04 2024-03-12 西安热工研究院有限公司 Novel power system frequency modulation method and system based on VMD-CEEMD
CN117691630B (en) * 2024-02-04 2024-04-30 西安热工研究院有限公司 VMD-CEEMD-based power system frequency modulation method and system
CN117744893A (en) * 2024-02-19 2024-03-22 西安热工研究院有限公司 Wind speed prediction method and system for energy storage auxiliary black start
CN117744893B (en) * 2024-02-19 2024-05-17 西安热工研究院有限公司 Wind speed prediction method and system for energy storage auxiliary black start

Also Published As

Publication number Publication date
CN116306798A (en) 2023-06-23

Similar Documents

Publication Publication Date Title
WO2023115425A1 (en) Ultra-short-time wind speed prediction method and system
Kisvari et al. Wind power forecasting–A data-driven method along with gated recurrent neural network
CN103279804B (en) The Forecasting Methodology of super short-period wind power
CN103117546A (en) Ultrashort-term slide prediction method for wind power
CN110365053B (en) Short-term wind power prediction method based on delay optimization strategy
CN113822418A (en) Wind power plant power prediction method, system, device and storage medium
CN103984986B (en) The self study arma modeling ultrashort-term wind power prediction method of real time correction
CN111191832A (en) Typhoon disaster power distribution network tower fault prediction method and system
CN113988481A (en) Wind power prediction method based on dynamic matrix prediction control
CN116960941A (en) Power climbing event prediction method, device, equipment and storage medium
CN114819385A (en) Wind power prediction method and device, electronic equipment and storage medium
Wang et al. Probabilistic power curve estimation based on meteorological factors and density LSTM
Sun et al. Short-term power load prediction based on VMD-SG-LSTM
WO2023216576A1 (en) Photovoltaic power generation short-term power prediction method and system
CN116777109A (en) Wind generating set equipment state evaluation method and system
CN115713029A (en) Wind power plant stochastic model prediction optimization control method considering delay
CN116663727A (en) Photovoltaic power prediction method and system
CN116306236A (en) Wind power plant dynamic equivalent modeling method and system based on GRU-LSTM-FC combined network
CN116123029A (en) Pitch angle control method and device, electronic equipment and wind turbine generator
Liu et al. A Bayesian deep learning-based probabilistic risk assessment and early-warning model for power systems considering meteorological conditions
Jiangxia et al. A combination prediction model for wind farm output power
CN111242359A (en) Solar radiation online dynamic prediction method based on data drift
Chuang et al. Short-term wind power forecasting method based on mode decomposition and feature extraction
CN112036672B (en) New energy power generation ultra-short term power prediction method and system based on iterative correction
Nguyen et al. An Optimizing Method Based On Combining Edge Computer And Long Short-Term Memory Networks Applied To Solar Power Forecasting

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: 21968574

Country of ref document: EP

Kind code of ref document: A1