WO2024031449A1 - Grid-clustering-based short-term power prediction method for canyon wind power - Google Patents

Grid-clustering-based short-term power prediction method for canyon wind power Download PDF

Info

Publication number
WO2024031449A1
WO2024031449A1 PCT/CN2022/111510 CN2022111510W WO2024031449A1 WO 2024031449 A1 WO2024031449 A1 WO 2024031449A1 CN 2022111510 W CN2022111510 W CN 2022111510W WO 2024031449 A1 WO2024031449 A1 WO 2024031449A1
Authority
WO
WIPO (PCT)
Prior art keywords
wind power
power
data
grid
canyon
Prior art date
Application number
PCT/CN2022/111510
Other languages
French (fr)
Chinese (zh)
Inventor
唐冬来
张捷
李玉
胡州明
宋卫平
郝建维
付世峻
黄璞
刘秋辉
杨俏
Original Assignee
四川中电启明星信息技术有限公司
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 四川中电启明星信息技术有限公司 filed Critical 四川中电启明星信息技术有限公司
Priority to PCT/CN2022/111510 priority Critical patent/WO2024031449A1/en
Publication of WO2024031449A1 publication Critical patent/WO2024031449A1/en

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
    • G06Q10/00—Administration; Management
    • G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"

Definitions

  • the present invention relates to the technical field of wind power power prediction. Specifically, it is a short-term power prediction method for canyon wind power based on grid clustering, which is used to solve the problem of low short-term power prediction accuracy of canyon wind power caused by canyon wakes. question.
  • wind energy is a renewable resource and has received widespread attention and utilization. People have done a lot of research on canyon wind, especially on short-term power prediction of canyon wind power. The research is mainly divided into error correction categories. and multi-modal decomposition wind power short-term power prediction method.
  • the short-term wind power power prediction results are corrected by analyzing the deviation of historical prediction data.
  • a short-term wind power power prediction method based on core density diffusion is proposed, and the accuracy of short-term wind power power prediction is improved through Gaussian function conversion error analysis.
  • This paper proposes a short-term wind power power prediction method based on long short-term memory network, which improves the accuracy of wind power short-term power prediction through chaos analysis of wind field data and historical error correction prediction of long short-term memory network.
  • a short-term wind power power prediction method based on random forest is proposed.
  • the short-term wind power power prediction error is extracted through a two-way gating model, and the wind power short-term power prediction error is corrected through random forest, thereby improving the accuracy of wind power short-term power prediction.
  • This paper proposes a short-term wind power power prediction method based on robust regression, which improves the accuracy of short-term wind power power prediction through robust regression of noise elimination and prediction error.
  • the above research method analyzes the wind power station as a whole, cannot decompose the power composition of the wind power station, and the prediction accuracy is insufficient.
  • the wind power station data is decomposed according to time series, and then each component is predicted separately, and finally superimposed to form the short-term power prediction data of the wind power station.
  • a short-term wind power power prediction method based on singular spectrum analysis is proposed, which improves the accuracy of short-term wind power power prediction through power time series decomposition and iterative prediction.
  • the time series adaptive decomposition of wind power power is performed, and then the component data is predicted and reconstructed to form the short-term wind power power prediction result.
  • the short-term wind power prediction method does not take into account the changeable meteorological information of the canyon. Due to the changeable wind wake in the canyon, the wind power station has the situation of adjacent wind turbines turning in the opposite direction.
  • the error correction method is similar to that of wind power.
  • the short-term wind power power forecast method the short-term wind power power forecast results are corrected by analyzing the deviation of historical forecast data. In this method, the wind power station is analyzed as a whole, and the power composition of the wind power station cannot be decomposed, so the prediction accuracy is insufficient.
  • the wind power station data is decomposed according to time series, and then each component is predicted separately, and finally superimposed to form the wind power station short-term power prediction data.
  • This method does not take into account the changeable meteorological information of the canyon. Due to the changeable wind wake in the canyon, adjacent wind turbines in the wind power station may turn in opposite directions. Therefore, the above method is insufficient to improve the accuracy of short-term power prediction of canyon wind power.
  • the present invention proposes a canyon based on grid clustering. Short-term wind power power forecasting method.
  • the purpose of the present invention is to provide a short-term power prediction method for canyon wind power based on grid clustering, which has the effect of solving the problem of low short-term power prediction accuracy of canyon wind power caused by canyon wakes.
  • a short-term power prediction method for canyon wind power based on grid clustering including the following steps:
  • Step S1 According to the characteristics of canyon wind power, the short-term power prediction method is divided into data cleaning stage, grid decomposition stage and power prediction stage;
  • Step S2 In the data cleaning process, perform anomaly detection, data correction and smoothing processing on the historical meteorology, wind power power, and wind power power prediction data learned by the LSTM network to obtain complete historical data;
  • Step S3. In the grid decomposition process, perform regional grid clustering of canyon wind power stations according to latitude, and obtain wind power power and characteristic grids related to meteorology;
  • Step S4 In the power prediction process, use the LSTM network to predict the power of each grid, and perform power superposition and error correction.
  • step S2 includes:
  • Step S2.1 Perform abnormal data detection on historical meteorological and wind power prediction data, historical wind power power data, meteorological monitoring data and wind turbine geographical location data, and eliminate outliers through outlier detection;
  • Step S2.2 Use linear interpolation method to build a missing data correction framework to achieve missing data repair
  • Step S2.3 Smooth the wind power power and meteorological observation data through Kalman filtering to eliminate useless components in the wind power power data.
  • step S3 includes:
  • Step S3.1 First, divide the wind farm into grids according to geographical longitude and latitude;
  • Step S3.2 conduct grid clustering analysis according to weather types and wind power power curve characteristic indicators to obtain grid wind power characteristics
  • Step S3.3 Finally, decompose to form a grid wind power prediction unit.
  • step S4 includes:
  • Step S4.1 First, obtain the short-term weather forecast information of the grid through digital weather forecast
  • Step S4.2 Analyze the grid wind power fluctuation trend based on the long short-term memory network to predict the short-term power of the canyon wind power in the next 10 days;
  • Step S4.3 Use LSTM to conduct short-term power prediction of canyon wind power, and use the least squares method to correct the error;
  • Step S4.4 Finally generate the short-term power prediction results of canyon wind power.
  • step S2.1 includes:
  • the method for correcting missing data in step S2.2 includes:
  • the data smoothing method in step S2.3 includes:
  • the method of geographical grid division in step S3.1 includes:
  • the normalized segmentation method is used to divide the wind farm grid according to the geographical longitude and latitude.
  • the method of performing grid clustering analysis in step S3.2 includes:
  • Grid cluster analysis was performed using hierarchical agglomerative clustering.
  • the present invention has the following advantages and beneficial effects:
  • the present invention proposes a grid clustering analysis method to obtain the characteristics of the canyon wind power station grid through grid clustering, cluster the grids according to weather type, wind speed, and power generation indicators to obtain the grid characteristics of the canyon wind power station. personality characteristics;
  • the present invention conducts geographical grid division according to the geomorphological characteristics of the canyon area and the geographical location information of the wind turbines.
  • the canyon wind power station is divided into grids according to the geographical longitude and latitude coordinates to make predictions in each grid area, thereby improving the canyon wind power generation.
  • Predictive decomposition is fine-grained.
  • Figure 1 is a flow chart of a canyon wind power short-term power prediction method based on grid clustering provided by the present invention.
  • Figure 2 is a schematic diagram of the LSTM grid power prediction provided by the present invention.
  • a canyon In the canyon wind power generation of the present invention, a canyon is a valley with steep slopes on both sides and a width smaller than its depth. In the canyon, it is built along the geological structural fissures. The slopes on both sides of the canyon are cliffs and the bottom of the canyon is flat. When the air enters the mouth of the canyon from a flat area, the cross-section of the airflow becomes smaller, and affected by the air pressure, the speed of the airflow increases, forming a strong wind in the canyon. Therefore, canyon wind is suitable for the establishment of wind power stations, but canyon wind power stations are built according to the terrain of the canyon and generally stretch for tens of kilometers. The valley bottom landform is tortuous, and the wind direction and wind force in the canyon are changeable, making it difficult to predict the short-term power of wind power.
  • Grid in informatics, is a method used to integrate or share various geographically distributed resources (including computer systems, storage systems, communication systems, files, databases, programs, etc.) to make them an organic whole , a mechanism for jointly completing various required tasks.
  • Wind turbines are electrical equipment that convert wind energy into mechanical work, which drives the rotor to rotate and ultimately outputs alternating current.
  • Wind turbines generally consist of components such as a wind wheel, a generator (including devices), a deflector (tail), a tower, a speed limiting safety mechanism, and an energy storage device.
  • the working principle of a wind turbine is relatively simple.
  • the wind rotor rotates under the influence of wind. It converts the kinetic energy of the wind into the mechanical energy of the wind rotor shaft.
  • the generator rotates driven by the wind rotor shaft to generate electricity.
  • FIG. 1 A method of short-term power prediction of canyon wind power based on grid clustering in this embodiment is shown in Figure 1.
  • a short-term power prediction method of canyon wind power based on grid clustering is proposed. This method first performs anomaly detection, data correction and smoothing on the historical meteorology, wind power power, and wind power power prediction data learned by the LSTM network. Then the regional grids of canyon wind power stations are clustered according to longitude and latitude to obtain wind power power and characteristic grids related to meteorology. On this basis, the LSTM network is used to predict the power of each grid, and perform power superposition and error correction.
  • Embodiment 1 is further optimized on the basis of Embodiment 1.
  • historical meteorological and wind power prediction data, historical wind power power data, meteorological monitoring data and wind turbine geographical location data are Carry out abnormal data detection and eliminate outliers through outlier detection; then, use linear interpolation method to build a missing data correction framework to achieve missing data repair; finally, use Kalman filtering method to smooth the wind power power and meteorological observation data. Eliminate useless components in wind power data.
  • Embodiment 1 or 2 This embodiment is further optimized on the basis of Embodiment 1 or 2.
  • the wind farm is first divided into grids according to geographical longitude and latitude, and then grid clustering is performed according to indicators such as weather type and wind power curve characteristics. Grid wind power characteristics are obtained, and finally decomposed to form a grid wind power power prediction unit.
  • Grid wind power power characteristics grid wind power power characteristics, grid wind power power prediction unit refers to meteorologically related characteristic grids.
  • the short-term weather forecast information of the grid is first obtained through digital weather forecast, and then based on long short-term memory
  • the network Long Short-Term Memory, LSTM
  • LSTM Long Short-Term Memory
  • the network analyzes the fluctuation trend of grid wind power to predict the short-term power of canyon wind power in the next 10 days. Again, the wind power power of all grids is superimposed to obtain the overall short-term predicted power of the wind power station.
  • the LSTM network is used to correct the error, and finally the short-term power prediction results of canyon wind power are generated.
  • digital weather forecast is the key to grid power prediction.
  • the digital weather forecast (numerical weather prediction, NWP) in this article comes from Spanish analytical data and has high accuracy. Due to space limitations, the prediction of digital weather forecast No more details will be given in the article.
  • LSTM is a neural network that loops in time, which can solve the problem of insufficient dependence on different time scales in Recurrent Neural Network (RNN).
  • RNN Recurrent Neural Network
  • the LSTM network has input, output and forget gates and can handle different time scales of canyon wind power. Problems with wind power forecasting. Therefore, this paper uses LSTM for short-term power prediction of canyon wind power.
  • th be the current moment of the canyon wind power grid prediction
  • the previous moment is th-1, th-1, Ph-1 and Ph
  • the distribution is the input grid historical power and grid predicted power of the LSTM network.
  • the input and output digital grid weather forecast data are Sh-1 and Sh.
  • the Zg, Zh and Zo distributions are the forgetting, input and output gates of the LSTM network.
  • the current input state is Pw.
  • the sigmiod function be ⁇ and the hyperbolic tangent function be tanh
  • the LSTM prediction structure is shown in Figure 2.
  • all the short-term power predictions of the canyon wind power station are superimposed to form the overall power of the canyon wind power station.
  • the effect achieved by using the LSTM network for short-term power prediction of canyon wind power is most in line with expectations. It should be noted that the LSTM network selected in this embodiment is the most expected method among all current methods. If other methods appear later, Achieving better prediction effects also falls within the protection scope of the present invention.
  • this method uses error prediction correction to update the model. Assume that the number of collection points in the 10 days (period) before the short-term prediction power is input is nx, and the different wind power power data is Pri, The different fitting coefficients of canyon wind power power using the least squares method are Bri, the data before the canyon wind power short-term power prediction is corrected is Pk, then the corrected short-term power prediction data Po is:
  • error prediction correction method selected in this embodiment is the most expected method among all methods currently. If other methods that appear later achieve better or equivalent correction, they will also fall within the protection scope of the present invention.
  • the collected data of the wind power station includes 96 points of real-time power of a single wind turbine unit, weather forecast data, wind power forecast data, etc.
  • the wind power station monitoring database is susceptible to interference from transmission noise, and is prone to missing data and data exceeding the normal range. Therefore, before wind power power prediction, abnormal data in historical data must first be detected.
  • the outlier local outlier factor (LOF) is an outlier detection method based on distance analysis of wind power power and meteorological data. This method can detect abnormal data that is significantly different from the normal attributes of wind power data through distance analysis between data.
  • the outlier point local abnormal factor detection method has the characteristics of fast speed and high efficiency. Therefore, this method is adopted as the abnormal data detection method in the present invention.
  • the s-th distance of wind power collection data point j is hs(j)
  • the number of areas of point j is Ls(j)
  • the reachable distance from the o-th point to the j point is Re
  • the local density lrd s (j) of point j can be obtained as:
  • the LOF method Through the LOF method, outliers in 96 points of real-time power of a single wind turbine, weather forecast data, wind power forecast data and other data can be detected and eliminated. It should be noted that the LOF method selected in this embodiment is the most expected method among all methods currently. If other methods that appear later achieve better or equivalent detection and elimination effects, they will also fall within the protection scope of the present invention. Inside.
  • This embodiment is further optimized based on any one of the above-mentioned embodiments 1-5.
  • the purpose of missing data correction is to enable the canyon wind power short-term power prediction model to accurately learn history when analyzing historical data. Change patterns and reduce the impact of missing data on model training.
  • the linear interpolation method is a method of using two known power and meteorological data to determine the unknown power and meteorological data between two points in historical data such as wind power power and meteorology. This method has small correction errors for parabolic, straight-line and other power and meteorological curve data. Therefore, the linear interpolation method is used in this paper to build a missing data correction framework to achieve missing data repair.
  • the canyon wind power power and meteorological data sequence be k, and the statistical data time period be n a .
  • the data before and after the missing point k h in the sequence be k a and k b respectively.
  • the interpolation distribution before and after the pre-missing point be e a and e b , missing point insertion value ec , missing point k h satisfies:
  • linear interpolation method selected in this embodiment is the most expected method among all methods currently. If other methods that appear later achieve better or equivalent missing data repair effects, they will also fall within the scope of the present invention. within.
  • the purpose of data smoothing is to solve the problem of noise data contaminating real data caused by wind power historical measurement data being affected by equipment accuracy. .
  • the canyon wind power measurement data can be made as close as possible to the actual measurement value of the equipment.
  • Kalman filter is a linear state filtering algorithm that makes the best estimate of the filtering state by observing the input of the Kalman filter, the output historical power of wind power, and meteorological data. This method can eliminate the impact of noise and interfering data on the model. Therefore, Kalman filter is selected for data smoothing in this article.
  • the state transition matrix of the Kalman filter is Ra
  • the matrix converted from input to state is Ga
  • the state matrix of the Kalman filter at the previous moment is Ba-1
  • the noise of the Kalman filter conversion process is Oa-1
  • a- The control amount of the system at time 1 is Ca-1
  • the Kalman filter state matrix is:
  • the filter state quantity observation matrix is:
  • v b R b ⁇ B a + z a ;
  • E a R a ⁇ E a-1 +A a ;
  • v e (H a -E a ⁇ W a ) ⁇ v b ;
  • Kalman filter method selected in this embodiment is the most expected method among all methods currently. If other methods that appear later achieve better or equivalent effects, they will also fall within the protection scope of the present invention. .
  • the canyon is a valley with steep slopes on both sides and a width smaller than the depth.
  • the canyon is built along the geological structural fissures.
  • the slopes on both sides of the canyon are cliffs and the bottom of the canyon is flat.
  • canyon wind is suitable for the establishment of wind power stations, but canyon wind power stations are built according to the terrain of the canyon and generally stretch for tens of kilometers.
  • the valley bottom landform is tortuous, and the wind direction and wind force in the canyon are changeable, making it difficult to predict the short-term power of wind power.
  • the canyon wind power station is divided into grids according to the geographical longitude and latitude coordinates to conduct predictions within each grid area, thereby improving the fine-grained decomposition of canyon wind power prediction.
  • N-cut Normalized Cuts
  • the canyon wind power topographic map is classified according to the graph theory. Divide.
  • the topographic map of the entire canyon wind power station is Qall, and any two sets of topography of the canyon wind power station are Qa and Qb.
  • the sum of the weights for dividing the above two sets is cut(Qa,Qb).
  • the coordinate points in Qa and Qb are related to the entire canyon wind power station.
  • the sum of the connected edge weights between the topographic maps of the canyon wind power station is as(Qa,Qall) and as(Qb,Qall), then the minimum division of the canyon wind power station is:
  • the purpose of grid clustering is to obtain the characteristics of the canyon wind power station grid, through weather type, wind speed, and power generation indicators.
  • the grids are clustered to obtain the grid characteristics of the canyon wind power station.
  • Hierarchical Agglomerative Clustering (HAC) is a single-chain clustering method that clusters a sample data in the canyon wind power grid, and then condenses two adjacent cluster sets and iterates . This method has strong adaptability in canyon wind power grids, so this method is selected for clustering in this paper.
  • Euclidean distance is used to calculate the similarity between two categories.
  • the objects of two different classes of clustering are Yai and Ybi
  • the number of clustering features is nf
  • the similarity between the two classes is
  • the classification of canyon wind power grids can be determined. After clustering is completed, a grid wind power prediction unit is formed.
  • the hierarchical agglomerative clustering method selected in this embodiment is the most expected method among all methods currently. If other methods that appear later achieve better or equivalent clustering effects, they will also fall under the protection of the present invention. within the range.

Landscapes

  • Engineering & Computer Science (AREA)
  • Business, Economics & Management (AREA)
  • Human Resources & Organizations (AREA)
  • Economics (AREA)
  • Strategic Management (AREA)
  • Marketing (AREA)
  • Game Theory and Decision Science (AREA)
  • Entrepreneurship & Innovation (AREA)
  • Development Economics (AREA)
  • Operations Research (AREA)
  • Quality & Reliability (AREA)
  • Tourism & Hospitality (AREA)
  • Physics & Mathematics (AREA)
  • General Business, Economics & Management (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Management, Administration, Business Operations System, And Electronic Commerce (AREA)

Abstract

The present invention relates to the technical field of wind power prediction. Disclosed is a grid-clustering-based short-term power prediction method for canyon wind power. The method comprises the following steps: step S1, according to canyon wind power features, dividing the short-term power prediction method into a data cleaning stage, a grid decomposition stage and a power prediction stage; step S2, in the data cleaning stage, performing abnormality detection, data supplementation and correction, and smoothing processing on historical meteorological data, historical wind power data and historical wind power prediction data which are learned by means of an LSTM network, so as to acquire complete historical data; step S3, in the grid decomposition stage, performing grid clustering of a canyon wind power station area according to latitude, so as to acquire wind power, and feature grids associated with meteorology; and step S4, in the power prediction stage, performing power prediction on each grid by using the LSTM network, and performing power superposition and error correction. The present invention is used for solving the problem of the accuracy of short-term wind power prediction being low due to canyon wind power being affected by a canyon wake flow.

Description

一种基于网格聚类的峡谷风电短期功率预测方法A short-term power prediction method for canyon wind power based on grid clustering 技术领域Technical field
本发明涉及风电功率预测技术领域，具体地说，是一种基于网格聚类的峡谷风电短期功率预测方法，用于解决峡谷风电受峡谷尾流影响，造成的风电短期功率预测准确率低的问题。The present invention relates to the technical field of wind power power prediction. Specifically, it is a short-term power prediction method for canyon wind power based on grid clustering, which is used to solve the problem of low short-term power prediction accuracy of canyon wind power caused by canyon wakes. question.
背景技术Background technique
风能作为一种有效的清洁能源，属于可再生资源，得到了人们普遍的重视与利用，针对峡谷风人们又做了大量研究，尤其是针对峡谷风电短期功率预测方面，研究主要分为误差修正类和多模态分解类风电短期功率预测方法。As an effective clean energy source, wind energy is a renewable resource and has received widespread attention and utilization. People have done a lot of research on canyon wind, especially on short-term power prediction of canyon wind power. The research is mainly divided into error correction categories. and multi-modal decomposition wind power short-term power prediction method.
误差修正类风电短期预测功率预测方法中，通过对历史预测数据偏差分析，从而修正风电短期功率预测结果。提出了一种基于核心密度扩散的风电短期功率预测方法，通过高斯函数转换误差分析，提高风电短期功率预测准确性。提出了一种基于长短期记忆网络的风电短期功率预测方法，通过风场数据混沌分析与长短期记忆网络历史误差修正预测，从而提高风电短期功率预测准确性。提出了一种基于随机森林的风电短期功率预测方法，通过双向门控模型提取风电短期功率预测误差，通过随机森林修正风电短期功率预测误差，从而提高风电短期功率预测准确性。提出了一种基于鲁棒回归的风电短期功率预测方法，通过对噪声消除与预测误差鲁棒回归，从而提高风电短期功率预测准确性。但上述研究的方法中，将风电站作为一个整体进行分析，不能对风电站的功率构成进行分解，预测精度不足。In the error correction short-term wind power prediction power prediction method, the short-term wind power power prediction results are corrected by analyzing the deviation of historical prediction data. A short-term wind power power prediction method based on core density diffusion is proposed, and the accuracy of short-term wind power power prediction is improved through Gaussian function conversion error analysis. This paper proposes a short-term wind power power prediction method based on long short-term memory network, which improves the accuracy of wind power short-term power prediction through chaos analysis of wind field data and historical error correction prediction of long short-term memory network. A short-term wind power power prediction method based on random forest is proposed. The short-term wind power power prediction error is extracted through a two-way gating model, and the wind power short-term power prediction error is corrected through random forest, thereby improving the accuracy of wind power short-term power prediction. This paper proposes a short-term wind power power prediction method based on robust regression, which improves the accuracy of short-term wind power power prediction through robust regression of noise elimination and prediction error. However, the above research method analyzes the wind power station as a whole, cannot decompose the power composition of the wind power station, and the prediction accuracy is insufficient.
多模态分解类风电短期预测功率预测方法中，将风电站的数据按时间序列进行分解，再单独对每个分量进行预测，最后叠加形成风电站短期功率预测数据。提出了一种基于奇异谱分析的风电短期功率预测方法，通过功率时间序列分解和迭代预测，从而提高风电短期功率预测准确性。首先对风电的功率进行时间序列自适应分解，然后进行分量数据预测与重构，形成风电短期功率预测结果。In the multi-modal decomposition wind power short-term prediction power prediction method, the wind power station data is decomposed according to time series, and then each component is predicted separately, and finally superimposed to form the short-term power prediction data of the wind power station. A short-term wind power power prediction method based on singular spectrum analysis is proposed, which improves the accuracy of short-term wind power power prediction through power time series decomposition and iterative prediction. First, the time series adaptive decomposition of wind power power is performed, and then the component data is predicted and reconstructed to form the short-term wind power power prediction result.
由此可见，峡谷风电短期预测功率预测方法多样，但上述方法中，未考虑峡谷多变的气象信息，受峡谷风力尾流多变，风电站存在相邻风机转向相反的情况，误差修正类风电短期预测功率预测方法中，通过对历史预测数据偏差分析，从而修正风电短期功率预测结果。该方法中，将风电站作为一个整体进行分析，不能对风电站的功率构成进行分解，预测精度不足。多模态分解类风电短期预测功率预测方法中，将风电站的数据按时间序列进 行分解，再单独对每个分量进行预测，最后叠加形成风电站短期功率预测数据。该方法未考虑峡谷多变的气象信息，受峡谷风力尾流多变，风电站存在相邻风机转向相反的情况，因此，上述方法对峡谷风电短期功率预测准确率提高不足。It can be seen that there are various short-term power prediction methods for canyon wind power. However, the above methods do not take into account the changeable meteorological information of the canyon. Due to the changeable wind wake in the canyon, the wind power station has the situation of adjacent wind turbines turning in the opposite direction. The error correction method is similar to that of wind power. In the short-term forecast power forecast method, the short-term wind power power forecast results are corrected by analyzing the deviation of historical forecast data. In this method, the wind power station is analyzed as a whole, and the power composition of the wind power station cannot be decomposed, so the prediction accuracy is insufficient. In the multi-modal decomposition wind power short-term prediction power prediction method, the wind power station data is decomposed according to time series, and then each component is predicted separately, and finally superimposed to form the wind power station short-term power prediction data. This method does not take into account the changeable meteorological information of the canyon. Due to the changeable wind wake in the canyon, adjacent wind turbines in the wind power station may turn in opposite directions. Therefore, the above method is insufficient to improve the accuracy of short-term power prediction of canyon wind power.
因此，上述方法对峡谷风电短期功率预测准确率提高不足，为了解决峡谷风电受峡谷尾流影响，造成的风电短期功率预测准确率低的问题，本发明提出了一种基于网格聚类的峡谷风电短期功率预测方法。Therefore, the above method is not enough to improve the accuracy of short-term power prediction of canyon wind power. In order to solve the problem of low accuracy of short-term power prediction of canyon wind power caused by the influence of canyon wake, the present invention proposes a canyon based on grid clustering. Short-term wind power power forecasting method.
发明内容Contents of the invention
本发明的目的在于提供一种基于网格聚类的峡谷风电短期功率预测方法，具有解决峡谷风电受峡谷尾流影响，造成的风电短期功率预测准确率低的问题的效果。The purpose of the present invention is to provide a short-term power prediction method for canyon wind power based on grid clustering, which has the effect of solving the problem of low short-term power prediction accuracy of canyon wind power caused by canyon wakes.
本发明通过下述技术方案实现：一种基于网格聚类的峡谷风电短期功率预测方法，包括以下步骤：The present invention is implemented through the following technical solution: a short-term power prediction method for canyon wind power based on grid clustering, including the following steps:
步骤S1.根据峡谷风电特征将短期功率预测方法分为数据清洗环节、网格分解环节和功率预测环节；Step S1. According to the characteristics of canyon wind power, the short-term power prediction method is divided into data cleaning stage, grid decomposition stage and power prediction stage;
步骤S2.在数据清洗环节对LSTM网络学***滑处理，获取完整的历史数据；Step S2. In the data cleaning process, perform anomaly detection, data correction and smoothing processing on the historical meteorology, wind power power, and wind power power prediction data learned by the LSTM network to obtain complete historical data;
步骤S3.在网格分解环节根据纬度进行峡谷风电站地域网格聚类，获取风电功率和与气象关联的特征网格；Step S3. In the grid decomposition process, perform regional grid clustering of canyon wind power stations according to latitude, and obtain wind power power and characteristic grids related to meteorology;
步骤S4.在功率预测环节使用LSTM网络对每个网格进行功率预测，并进行功率叠加与误差修正。Step S4. In the power prediction process, use the LSTM network to predict the power of each grid, and perform power superposition and error correction.
为了更好地实现本发明，进一步地，步骤S2包括：In order to better implement the present invention, further, step S2 includes:
步骤S2.1.对历史的气象与风电预测数据、历史风电功率数据、气象监测数据和风机地理位置数据进行异常数据检测，通过离群点检测方式剔除异常值；Step S2.1. Perform abnormal data detection on historical meteorological and wind power prediction data, historical wind power power data, meteorological monitoring data and wind turbine geographical location data, and eliminate outliers through outlier detection;
步骤S2.2.采用线性插值方法搭建缺失数据补正框架，实现缺失数据修复；Step S2.2. Use linear interpolation method to build a missing data correction framework to achieve missing data repair;
步骤S2.3.通过卡尔曼滤波方式对风电功率、气象观测数据进行平滑处理，消除风电功率数据中的无用分量。Step S2.3. Smooth the wind power power and meteorological observation data through Kalman filtering to eliminate useless components in the wind power power data.
为了更好地实现本发明，进一步地，步骤S3包括：In order to better implement the present invention, further, step S3 includes:
步骤S3.1.首先，按照地理经纬度进行风电场网格划分；Step S3.1. First, divide the wind farm into grids according to geographical longitude and latitude;
步骤S3.2.然后，按照天气类型和风电功率曲线特性指标进行网格聚类分析，获得网格风电特征；Step S3.2. Then, conduct grid clustering analysis according to weather types and wind power power curve characteristic indicators to obtain grid wind power characteristics;
步骤S3.3.最后，分解形成网格风电功率预测单元。Step S3.3. Finally, decompose to form a grid wind power prediction unit.
为了更好地实现本发明，进一步地，步骤S4包括：In order to better implement the present invention, further, step S4 includes:
步骤S4.1.首先通过数字天气预报获得网格的短期天气预报信息；Step S4.1. First, obtain the short-term weather forecast information of the grid through digital weather forecast;
步骤S4.2.基于长短期记忆网络分析网格风电波动趋势从而预测网格未来10天的峡谷风电短期功率；Step S4.2. Analyze the grid wind power fluctuation trend based on the long short-term memory network to predict the short-term power of the canyon wind power in the next 10 days;
步骤S4.3.采用LSTM进行峡谷风电短期功率预测，采用最小二乘法修正误差；Step S4.3. Use LSTM to conduct short-term power prediction of canyon wind power, and use the least squares method to correct the error;
步骤S4.4.最后生成峡谷风电短期功率预测结果。Step S4.4. Finally generate the short-term power prediction results of canyon wind power.
为了更好地实现本发明，进一步地，步骤S2.1中异常数据检测的方法包括：In order to better implement the present invention, further, the method for detecting abnormal data in step S2.1 includes:
根据离群点局部异常因子和数据间的距离分析，检测出与风电数据正常属性差异大的异常数据。Based on the local anomaly factor of outliers and the distance analysis between data, abnormal data that is significantly different from the normal attributes of wind power data is detected.
为了更好地实现本发明，进一步地，步骤S2.2中缺失数据补正的方法包括：In order to better implement the present invention, further, the method for correcting missing data in step S2.2 includes:
使用线性插值法搭建缺失数据补正框架，实现缺失数据修复。Use linear interpolation method to build a missing data correction framework to achieve missing data repair.
为了更好地实现本发明，进一步地，步骤S2.3中数据平滑处理的方法包括：In order to better implement the present invention, further, the data smoothing method in step S2.3 includes:
使用卡尔曼滤波对滤波状态进行最佳估算，消除噪声和干扰数据对模型的影响。Use Kalman filtering to make the best estimate of the filtering state and eliminate the impact of noise and interfering data on the model.
为了更好地实现本发明，进一步地，步骤S3.1中地理网格划分的方法包括：In order to better implement the present invention, further, the method of geographical grid division in step S3.1 includes:
采用归一化分割的方式按照地理经纬度进行风电场网格划分。The normalized segmentation method is used to divide the wind farm grid according to the geographical longitude and latitude.
为了更好地实现本发明，进一步地，步骤S3.2中进行网格聚类分析的方法包括：In order to better implement the present invention, further, the method of performing grid clustering analysis in step S3.2 includes:
使用层次凝聚聚类法进行网格聚类分析。Grid cluster analysis was performed using hierarchical agglomerative clustering.
本发明与现有技术相比，具有以下优点及有益效果：Compared with the existing technology, the present invention has the following advantages and beneficial effects:
(1)本发明提出了网格聚类分析方法，通过网格聚类获取峡谷风电站网格的特性，通过天气类型、风速、发电功率指标对网格进行聚类，获得峡谷风电站的网格特征；(1) The present invention proposes a grid clustering analysis method to obtain the characteristics of the canyon wind power station grid through grid clustering, cluster the grids according to weather type, wind speed, and power generation indicators to obtain the grid characteristics of the canyon wind power station. personality characteristics;
(2)本发明针对峡谷地区的地貌特点和风机地理位置信息进行了地理网格划分，按照地理经纬度坐标进行峡谷风电站网格划分，以在每个网格区域内进行预测，从而提高峡谷风电预测分解细粒度。(2) The present invention conducts geographical grid division according to the geomorphological characteristics of the canyon area and the geographical location information of the wind turbines. The canyon wind power station is divided into grids according to the geographical longitude and latitude coordinates to make predictions in each grid area, thereby improving the canyon wind power generation. Predictive decomposition is fine-grained.
附图说明Description of drawings
本发明结合下面附图和实施例做进一步说明，本发明所有构思创新应视为所公开内容和本发明保护范围。The present invention will be further described with reference to the following drawings and examples. All innovative ideas and innovations of the present invention shall be regarded as the disclosed content and the protection scope of the present invention.
图1为本发明所提供的一种基于网格聚类的峡谷风电短期功率预测方法的流程图。Figure 1 is a flow chart of a canyon wind power short-term power prediction method based on grid clustering provided by the present invention.
图2为本发明所提供的LSTM网格功率预测示意图。Figure 2 is a schematic diagram of the LSTM grid power prediction provided by the present invention.
具体实施方式Detailed ways
为了更清楚地说明本发明实施例的技术方案，下面将结合本发明实施例中的附图，对本发明实施例中的技术方案进行清楚、完整地描述，应当理解，所描述的实施例仅仅是本发明的一部分实施例，而不是全部的实施例，因此不应被看作是对保护范围的限定。基于 本发明中的实施例，本领域普通技术工作人员在没有做出创造性劳动前提下所获得的所有其他实施例，都属于本发明保护的范围。In order to explain the technical solutions in the embodiments of the present invention more clearly, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. It should be understood that the described embodiments are only Some of the embodiments of the present invention are not all embodiments, and therefore should not be regarded as limiting the scope of protection. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without making creative efforts fall within the scope of protection of the present invention.
在本发明中峡谷风电，峡谷是两侧谷坡陡峻，宽度比深度小的山谷。在峡谷中，沿地质构造裂隙构建，峡谷两侧坡为悬崖，谷底平坦。当空气由平坦区域进入峡谷口时，气流横切面变小，受空气压力影响，提高气流运动速度，形成峡谷强风。因此，峡谷风适合建立风电站，但峡谷风电站依峡谷的地形建设，普遍绵延数十公里。峡谷的谷底地貌曲折，峡谷风向、风力多变，造成风电短期功率预测难度大。In the canyon wind power generation of the present invention, a canyon is a valley with steep slopes on both sides and a width smaller than its depth. In the canyon, it is built along the geological structural fissures. The slopes on both sides of the canyon are cliffs and the bottom of the canyon is flat. When the air enters the mouth of the canyon from a flat area, the cross-section of the airflow becomes smaller, and affected by the air pressure, the speed of the airflow increases, forming a strong wind in the canyon. Therefore, canyon wind is suitable for the establishment of wind power stations, but canyon wind power stations are built according to the terrain of the canyon and generally stretch for tens of kilometers. The valley bottom landform is tortuous, and the wind direction and wind force in the canyon are changeable, making it difficult to predict the short-term power of wind power.
网格，在信息学中，网格是一种用于集成或共享地理上分布的各种资源(包括计算机***、存储***、通信***、文件、数据库、程序等)，使之成为有机的整体，共同完成各种所需任务的机制。Grid, in informatics, is a method used to integrate or share various geographically distributed resources (including computer systems, storage systems, communication systems, files, databases, programs, etc.) to make them an organic whole , a mechanism for jointly completing various required tasks.
风力发电机是将风能转换为机械功，机械功带动转子旋转，最终输出交流电的电力设备。风力发电机一般有风轮、发电机(包括装置)、调向器(尾翼)、塔架、限速安全机构和储能装置等构件组成。风力发电机的工作原理比较简单，风轮在风力的作用下旋转，它把风的动能转变为风轮轴的机械能，发电机在风轮轴的带动下旋转发电。Wind turbines are electrical equipment that convert wind energy into mechanical work, which drives the rotor to rotate and ultimately outputs alternating current. Wind turbines generally consist of components such as a wind wheel, a generator (including devices), a deflector (tail), a tower, a speed limiting safety mechanism, and an energy storage device. The working principle of a wind turbine is relatively simple. The wind rotor rotates under the influence of wind. It converts the kinetic energy of the wind into the mechanical energy of the wind rotor shaft. The generator rotates driven by the wind rotor shaft to generate electricity.
实施例1：Example 1:
本实施例的一种基于网格聚类的峡谷风电短期功率预测方法，如图1所示，在本实施例中，为解决峡谷风电受峡谷尾流影响，造成的风电短期功率预测准确率低的问题，提出了一种基于网格聚类的峡谷风电短期功率预测方法。该方法首先对LSTM网络学***滑处理。然后按照经纬度进行峡谷风电站地域网格聚类，获得风电功率，与气象关联的特征网格，在此基础上，采用LSTM网络对每个网格进行功率预测，并进行功率叠加与误差修正。A method of short-term power prediction of canyon wind power based on grid clustering in this embodiment is shown in Figure 1. In this embodiment, in order to solve the problem of low accuracy of short-term power prediction of canyon wind power caused by the canyon wake, To solve the problem, a short-term power prediction method of canyon wind power based on grid clustering is proposed. This method first performs anomaly detection, data correction and smoothing on the historical meteorology, wind power power, and wind power power prediction data learned by the LSTM network. Then the regional grids of canyon wind power stations are clustered according to longitude and latitude to obtain wind power power and characteristic grids related to meteorology. On this basis, the LSTM network is used to predict the power of each grid, and perform power superposition and error correction.
实施例2：Example 2:
本实施例在实施例1的基础上做进一步优化，在本实施例中，在数据清洗环节中，首先，对历史的气象与风电预测数据、历史风电功率数据、气象监测数据和风机地理位置数据进行异常数据检测，通过离群点检测方式剔除异常值；然后，采用线性插值方法搭建缺失数据补正框架，实现缺失数据修复；最后，通过卡尔曼滤波方式对风电功率、气象观测数据进行平滑处理，消除风电功率数据中的无用分量。This embodiment is further optimized on the basis of Embodiment 1. In this embodiment, in the data cleaning process, first, historical meteorological and wind power prediction data, historical wind power power data, meteorological monitoring data and wind turbine geographical location data are Carry out abnormal data detection and eliminate outliers through outlier detection; then, use linear interpolation method to build a missing data correction framework to achieve missing data repair; finally, use Kalman filtering method to smooth the wind power power and meteorological observation data. Eliminate useless components in wind power data.
本实施例的其他部分与实施例1相同，故不再赘述。The other parts of this embodiment are the same as those of Embodiment 1, so they will not be described again.
实施例3：Example 3:
本实施例在实施例1或2的基础上做进一步优化，在网格分解中，首先按照地理经纬度进行风电场网格划分，然后按照天气类型、风电功率曲线特性等指标进行网格聚类，获 得网格风电特征，最后分解形成网格风电功率预测单元。网格风电功率特征网格风电功率特征，网格风电功率预测单元指气象关联的特征网格。This embodiment is further optimized on the basis of Embodiment 1 or 2. In the grid decomposition, the wind farm is first divided into grids according to geographical longitude and latitude, and then grid clustering is performed according to indicators such as weather type and wind power curve characteristics. Grid wind power characteristics are obtained, and finally decomposed to form a grid wind power power prediction unit. Grid wind power power characteristics grid wind power power characteristics, grid wind power power prediction unit refers to meteorologically related characteristic grids.
本实施例的其他部分与上述实施例1或2相同，故不再赘述。The other parts of this embodiment are the same as those of the above-mentioned Embodiment 1 or 2, so they will not be described again.
实施例4：Example 4:
本实施例在上述实施例1-3任一项的基础上做进一步优化，在本实施例中，在功率预测环节，首先通过数字天气预报获得网格的短期天气预报信息，然后基于长短期记忆网络(Long Short-Term Memory,LSTM)分析网格风电波动趋势从而预测网格未来10天的峡谷风电短期功率。再次，将所有网格的风电功率进行叠加，获得风电站的整体短期预测功率。并采用LSTM网络修正误差，最后生成峡谷风电短期功率预测结果。This embodiment is further optimized based on any one of the above-mentioned Embodiments 1-3. In this embodiment, in the power prediction link, the short-term weather forecast information of the grid is first obtained through digital weather forecast, and then based on long short-term memory The network (Long Short-Term Memory, LSTM) analyzes the fluctuation trend of grid wind power to predict the short-term power of canyon wind power in the next 10 days. Again, the wind power power of all grids is superimposed to obtain the overall short-term predicted power of the wind power station. The LSTM network is used to correct the error, and finally the short-term power prediction results of canyon wind power are generated.
如图2所示，数字天气预报是网格功率预测的关键，文中的数字天气预报(numerical weather prediction，NWP)来源于西班牙分析数据，具有较高的准确性，限于篇幅，数字天气预报的预测文中不再累述。LSTM是一种在时间上循环的神经网络，可以解决循环神经网络(Recurrent Neural Network,RNN)中不同时间尺度依赖不足的问题，LSTM网络有输入、输出和遗忘门，可处理峡谷风电不同时间尺度风电预测的问题。因此，文中采用LSTM进行峡谷风电短期功率预测。设th为峡谷风电网格预测的当前时刻，上一时刻为th-1，th-1，Ph-1和Ph,分布为LSTM网络的输入网格历史功率和网格预测功率。输入输出的数字网格天气预报数据为Sh-1和Sh，Zg、Zh和Zo分布为LSTM网络的遗忘、输入和输出门，当前输入状态为Pw，设sigmiod函数为β，双曲正切函数为tanh，LSTM预测结构如图2所示。在峡谷风电网格短期功率预测的基础上，将峡谷风电站所有的网格短期预测功率进行叠加，叠加后形成峡谷风电站的整体功率。本实施例中采用LSTM网络进行峡谷风电短期功率预测达到的效果最符合预期，需要说明的是，本实施例中选用的LSTM网络是目前所有方法中最符合预期的方法，如果之后出现的其他方法达到更好的预测效果也落入本发明的保护范围之内。As shown in Figure 2, digital weather forecast is the key to grid power prediction. The digital weather forecast (numerical weather prediction, NWP) in this article comes from Spanish analytical data and has high accuracy. Due to space limitations, the prediction of digital weather forecast No more details will be given in the article. LSTM is a neural network that loops in time, which can solve the problem of insufficient dependence on different time scales in Recurrent Neural Network (RNN). The LSTM network has input, output and forget gates and can handle different time scales of canyon wind power. Problems with wind power forecasting. Therefore, this paper uses LSTM for short-term power prediction of canyon wind power. Let th be the current moment of the canyon wind power grid prediction, the previous moment is th-1, th-1, Ph-1 and Ph, and the distribution is the input grid historical power and grid predicted power of the LSTM network. The input and output digital grid weather forecast data are Sh-1 and Sh. The Zg, Zh and Zo distributions are the forgetting, input and output gates of the LSTM network. The current input state is Pw. Let the sigmiod function be β and the hyperbolic tangent function be tanh, the LSTM prediction structure is shown in Figure 2. On the basis of the short-term power prediction of the canyon wind power grid, all the short-term power predictions of the canyon wind power station are superimposed to form the overall power of the canyon wind power station. In this embodiment, the effect achieved by using the LSTM network for short-term power prediction of canyon wind power is most in line with expectations. It should be noted that the LSTM network selected in this embodiment is the most expected method among all current methods. If other methods appear later, Achieving better prediction effects also falls within the protection scope of the present invention.
本实施例中为了降低峡谷风电短期功率预测的误差，该方法采用误差预测修正进行模型更新，设输入短期预测功率前10天(期时间内)的采集点数为nx，不同风电功率数据为Pri，采用最小二乘法进行峡谷风电功率的不同拟合系数为Bri，峡谷风电短期功率预测修正前的数据为Pk，则修正后的短期功率预测数据Po为：In this embodiment, in order to reduce the error of short-term power prediction of canyon wind power, this method uses error prediction correction to update the model. Assume that the number of collection points in the 10 days (period) before the short-term prediction power is input is nx, and the different wind power power data is Pri, The different fitting coefficients of canyon wind power power using the least squares method are Bri, the data before the canyon wind power short-term power prediction is corrected is Pk, then the corrected short-term power prediction data Po is:
Figure PCTCN2022111510-appb-000001
Figure PCTCN2022111510-appb-000001
需要说明的是，本实施例中选用的误差预测修正法是目前所有方法中最符合预期的方 法，如果之后出现的其他方法达到更好或同等的修正也落入本发明的保护范围之内。It should be noted that the error prediction correction method selected in this embodiment is the most expected method among all methods currently. If other methods that appear later achieve better or equivalent correction, they will also fall within the protection scope of the present invention.
本实施例的其他部分与上述实施例1-3相同，故不再赘述。The other parts of this embodiment are the same as the above-mentioned Embodiments 1-3, so they will not be described again.
实施例5：Example 5:
本实施例在上述实施例1-4任一项基础上做进一步优化，在本实施例中，风电站的采集数据包括96点的单台风电机组实时功率、天气预报数据、风电预测数据等，数量类型多，数量大。风电站监测数据库易受到传输噪声干扰，容易出现缺失和超过正常范围的数据。因此，在风电功率预测前，需首先对历史数据的的异常数据进行检测。离群点局部异常因子(Local Outlier Factor,LOF)是基于风电功率及气象数据距离分析的异常值检测方法。该方法通过数据间的距离分析，可以检测出与风电数据正常属性差异大的异常数据。离群点局部异常因子检测方法具有速度快，效率高的特点，因此，采用该方法作为本发明中的异常数据检测方法。This embodiment is further optimized based on any one of the above-mentioned embodiments 1-4. In this embodiment, the collected data of the wind power station includes 96 points of real-time power of a single wind turbine unit, weather forecast data, wind power forecast data, etc., There are many types of quantities and large quantities. The wind power station monitoring database is susceptible to interference from transmission noise, and is prone to missing data and data exceeding the normal range. Therefore, before wind power power prediction, abnormal data in historical data must first be detected. The outlier local outlier factor (LOF) is an outlier detection method based on distance analysis of wind power power and meteorological data. This method can detect abnormal data that is significantly different from the normal attributes of wind power data through distance analysis between data. The outlier point local abnormal factor detection method has the characteristics of fast speed and high efficiency. Therefore, this method is adopted as the abnormal data detection method in the present invention.
设风电采集数据点j的第s个距离为hs(j)，点j的领域个数为Ls(j)，且Ls(j)≥s，第o点到j点的可达距离为Re，则点j的局部密度lrd s(j)可达为： Assume that the s-th distance of wind power collection data point j is hs(j), the number of areas of point j is Ls(j), and Ls(j)≥s, the reachable distance from the o-th point to the j point is Re, Then the local density lrd s (j) of point j can be obtained as:
Figure PCTCN2022111510-appb-000002
Figure PCTCN2022111510-appb-000002
设o点局部密度可达为lrds(o)，则j点的离群点因子F s(j)为： Assuming that the local density of point o can be lrds(o), then the outlier factor F s (j) of point j is:
Figure PCTCN2022111510-appb-000003
Figure PCTCN2022111510-appb-000003
通过LOF方法，可对括96点的单台风电机组实时功率、天气预报数据、风电预测数据等数据的异常值进行检测和剔除。需要说明的是，本实施例中选用的LOF方法是目前所有方法中最符合预期的方法，如果之后出现的其他方法达到更好或同等的检测和剔除效果，也落入本发明的保护范围之内。Through the LOF method, outliers in 96 points of real-time power of a single wind turbine, weather forecast data, wind power forecast data and other data can be detected and eliminated. It should be noted that the LOF method selected in this embodiment is the most expected method among all methods currently. If other methods that appear later achieve better or equivalent detection and elimination effects, they will also fall within the protection scope of the present invention. Inside.
本实施例的其他部分与上述实施例1-4任一项相同，故不再赘述。The other parts of this embodiment are the same as any one of the above-mentioned Embodiments 1-4, and therefore will not be described again.
实施例6：Example 6:
本实施例在上述实施例1-5任一项基础上做进一步优化，在本实施例中，缺失数据补正的目的是为峡谷风电短期功率预测模型在对历史数据分析时，能精确的学习历史变化规律，减少丢失数据对模型训练带来的影响。This embodiment is further optimized based on any one of the above-mentioned embodiments 1-5. In this embodiment, the purpose of missing data correction is to enable the canyon wind power short-term power prediction model to accurately learn history when analyzing historical data. Change patterns and reduce the impact of missing data on model training.
线性插值法是在风电功率、气象等历史数据中，使用两个已知的功率、气象数据确定两点之间未知功率、气象数据的方法。该方法对于抛物线、直线等功率、气象曲线数据具有较小的补正误差，因此，文中采用线性插值法搭建缺失数据补正框架，实现缺失数据修复。The linear interpolation method is a method of using two known power and meteorological data to determine the unknown power and meteorological data between two points in historical data such as wind power power and meteorology. This method has small correction errors for parabolic, straight-line and other power and meteorological curve data. Therefore, the linear interpolation method is used in this paper to build a missing data correction framework to achieve missing data repair.
设峡谷风电功率、气象数据序列为k，统计的数据时间周期为n a，设序列中缺失点k h前后的数据分别为k a和k b，设预缺失点前后的插值分布为e a和e b，缺失点***值e c，缺失点k h满足： Let the canyon wind power power and meteorological data sequence be k, and the statistical data time period be n a . Let the data before and after the missing point k h in the sequence be k a and k b respectively. Let the interpolation distribution before and after the pre-missing point be e a and e b , missing point insertion value ec , missing point k h satisfies:
Figure PCTCN2022111510-appb-000004
Figure PCTCN2022111510-appb-000004
通过上式的线性插值，可获得峡谷风电功率、气象数据缺失点预测值，消除缺失数据对峡谷风电短期预测数据的影响。Through the linear interpolation of the above formula, the prediction values of missing points of canyon wind power power and meteorological data can be obtained, and the impact of missing data on the short-term prediction data of canyon wind power can be eliminated.
需要说明的是，本实施例中选用的线性插值法是目前所有方法中最符合预期的方法，如果之后出现的其他方法达到更好或同等的缺失数据修复效果，也落入本发明的保护范围之内。It should be noted that the linear interpolation method selected in this embodiment is the most expected method among all methods currently. If other methods that appear later achieve better or equivalent missing data repair effects, they will also fall within the scope of the present invention. within.
本实施例的其他部分与上述实施例1-5任一项相同，故不再赘述。The other parts of this embodiment are the same as any one of the above-mentioned Embodiments 1-5, and therefore will not be described again.
实施例7：Example 7:
本实施例在上述实施例1-6任一项基础上做进一步优化，在本实施例中，数据平滑处理的目的是解决风电历史量测数据受设备精度影响造成的噪声数据污染真实数据的问题。通过数据平滑处理可以使得峡谷风电量测数据最大限度的接近设备真实的量测值。卡尔曼滤波是一种线性状态滤波算法，通过对卡尔曼滤波器的输入、输出的风电历史功率、气象数据进行观测，从而对滤波状态进行最佳估算。该方法可以消除噪声和干扰数据对模型的影响。因此，文中选用卡尔曼滤波进行数据平滑处理。设卡尔曼滤波器的状态转移矩阵为Ra，输入转换成状态的矩阵为Ga，上一个时刻的卡尔曼滤波器状态矩阵为Ba-1，卡尔曼滤波器转换过程噪声为Oa-1，a-1时刻对***的控制量为Ca-1，则卡尔曼滤波器状态矩阵为：This embodiment is further optimized based on any one of the above-mentioned embodiments 1-6. In this embodiment, the purpose of data smoothing is to solve the problem of noise data contaminating real data caused by wind power historical measurement data being affected by equipment accuracy. . Through data smoothing processing, the canyon wind power measurement data can be made as close as possible to the actual measurement value of the equipment. Kalman filter is a linear state filtering algorithm that makes the best estimate of the filtering state by observing the input of the Kalman filter, the output historical power of wind power, and meteorological data. This method can eliminate the impact of noise and interfering data on the model. Therefore, Kalman filter is selected for data smoothing in this article. Suppose the state transition matrix of the Kalman filter is Ra, the matrix converted from input to state is Ga, the state matrix of the Kalman filter at the previous moment is Ba-1, and the noise of the Kalman filter conversion process is Oa-1, a- The control amount of the system at time 1 is Ca-1, then the Kalman filter state matrix is:
B a＝R a×B a-1+G a×c a-1+o a-1； B a =R a ×B a-1 +G a ×c a-1 +o a-1 ;
设卡尔曼滤波器的状态观测矩阵为Rb，卡尔曼滤波器的测量噪声为Za构成，噪声服 从高斯分布，则滤波器状态量观测阵为：Suppose the state observation matrix of the Kalman filter is Rb, the measurement noise of the Kalman filter is composed of Za, and the noise obeys Gaussian distribution, then the filter state quantity observation matrix is:
v b＝R b×B a+z a； v b = R b × B a + z a ;
设卡尔曼滤波器的上一时刻先验斜方差为Ea-1，滤波器转换过程的噪声方差为Aa，则滤波器中间输出结果a时刻的协方差Ea为：Assume that the prior slope variance of the Kalman filter at the previous moment is Ea-1, and the noise variance of the filter conversion process is Aa. Then the covariance Ea of the filter's intermediate output result a at moment a is:
E a＝R a×E a-1+A a； E a =R a ×E a-1 +A a ;
设测量噪声协方差为Zb，状态转换为观测值的矩阵为Wa，则卡尔曼增益滤波矩阵Ua为：Assume that the covariance of the measurement noise is Zb, and the matrix that converts the state into an observation value is Wa, then the Kalman gain filter matrix Ua is:
Figure PCTCN2022111510-appb-000005
Figure PCTCN2022111510-appb-000005
设滤波转换矩阵为Ha，后验时刻的滤波结果协方差Ve为：Suppose the filtering transformation matrix is Ha, and the covariance Ve of the filtering result at the posterior moment is:
v e＝(H a-E a×W a)×v b； v e =(H a -E a ×W a )×v b ;
需要说明的是，本实施例中选用的卡尔曼滤波法是目前所有方法中最符合预期的方法，如果之后出现的其他方法达到更好或同等的效果，也落入本发明的保护范围之内。It should be noted that the Kalman filter method selected in this embodiment is the most expected method among all methods currently. If other methods that appear later achieve better or equivalent effects, they will also fall within the protection scope of the present invention. .
本实施例的其他部分与上述实施例1-6任一项相同，故不再赘述。The other parts of this embodiment are the same as any one of the above-mentioned Embodiments 1-6, and therefore will not be described again.
实施例8：Example 8:
本实施例在上述实施例1-7任一项基础上做进一步优化，在本实施例中，峡谷是两侧谷坡陡峻，宽度比深度小的山谷。在峡谷中，沿地质构造裂隙构建，峡谷两侧坡为悬崖，谷底平坦。当空气由平坦区域进入峡谷口时，气流横切面变小，受空气压力影响，提高气流运动速度，形成峡谷强风。因此，峡谷风适合建立风电站，但峡谷风电站依峡谷的地形建设，普遍绵延数十公里。峡谷的谷底地貌曲折，峡谷风向、风力多变，造成风电短期功率预测难度大。针对峡谷地区的地貌特点和风机地理位置信息，按照地理经纬度坐标进行峡谷风电站网格划分，以在每个网格区域内进行预测，从而提高峡谷风电预测分解细粒度。This embodiment is further optimized based on any one of the above-mentioned embodiments 1-7. In this embodiment, the canyon is a valley with steep slopes on both sides and a width smaller than the depth. In the canyon, it is built along the geological structural fissures. The slopes on both sides of the canyon are cliffs and the bottom of the canyon is flat. When the air enters the mouth of the canyon from a flat area, the cross-section of the airflow becomes smaller, and affected by the air pressure, the speed of the airflow increases, forming a strong wind in the canyon. Therefore, canyon wind is suitable for the establishment of wind power stations, but canyon wind power stations are built according to the terrain of the canyon and generally stretch for tens of kilometers. The valley bottom landform is tortuous, and the wind direction and wind force in the canyon are changeable, making it difficult to predict the short-term power of wind power. Based on the geomorphological characteristics of the canyon area and the geographical location information of the wind turbines, the canyon wind power station is divided into grids according to the geographical longitude and latitude coordinates to conduct predictions within each grid area, thereby improving the fine-grained decomposition of canyon wind power prediction.
归一化分割(Normalized Cuts，N-cut)是一种基于图论的最小化峡谷地形分割方法，通过将地形图映射为带权值的峡谷无向图，即将峡谷风电地形图按图论中进行划分。设整个峡谷风电站的地形图为Qall，峡谷风电站地形任意两个集合为Qa和Qb，分割上述两个集合的权值和为cut(Qa,Qb)，Qa和Qb中的坐标点与整个峡谷风电站的地形图之间的相连边权值之和为as(Qa,Qall)和as(Qb,Qall)，则峡谷风电站最小分割为：Normalized Cuts (N-cut) is a minimizing canyon terrain segmentation method based on graph theory. By mapping the topographic map into a weighted canyon undirected graph, the canyon wind power topographic map is classified according to the graph theory. Divide. Suppose the topographic map of the entire canyon wind power station is Qall, and any two sets of topography of the canyon wind power station are Qa and Qb. The sum of the weights for dividing the above two sets is cut(Qa,Qb). The coordinate points in Qa and Qb are related to the entire canyon wind power station. The sum of the connected edge weights between the topographic maps of the canyon wind power station is as(Qa,Qall) and as(Qb,Qall), then the minimum division of the canyon wind power station is:
Figure PCTCN2022111510-appb-000006
Figure PCTCN2022111510-appb-000006
需要说明的是，本实施例中选用的归一化分割法是目前所有方法中最符合预期的方法，如果之后出现的其他方法达到更好或同等的划分效果，也落入本发明的保护范围之内。It should be noted that the normalized segmentation method selected in this embodiment is the most expected method among all methods currently. If other methods that appear later achieve better or equivalent segmentation effects, they will also fall within the protection scope of the present invention. within.
本实施例的其他部分与上述实施例1-7任一项相同，故不再赘述。The other parts of this embodiment are the same as any one of the above-mentioned Embodiments 1-7, and therefore will not be described again.
实施例9：Example 9:
本实施例在上述实施例1-8任一项基础上做进一步优化，在本实施例中，网格聚类的目的是获取峡谷风电站网格的特性，通过天气类型、风速、发电功率指标对网格进行聚类，获得峡谷风电站的网格特征。层次凝聚聚类法(Hierarchical Agglomerative Clustering,HAC)，是一种单链聚类方法，通过把峡谷风电网格中的一个样本数据进行聚类，然后将相邻的两个聚类集合凝聚并迭代。该方法在峡谷风电网格中适应性强，因此，文中选用该方法进行聚类。在对峡谷风电站网格进行HAC聚类时，采用欧式距离计算两个类别之间的相似程度。设聚类的两个不同类的对象为Yai和Ybi，聚类的特征个数为nf个，两个类的相似度为This embodiment is further optimized based on any one of the above-mentioned Embodiments 1-8. In this embodiment, the purpose of grid clustering is to obtain the characteristics of the canyon wind power station grid, through weather type, wind speed, and power generation indicators. The grids are clustered to obtain the grid characteristics of the canyon wind power station. Hierarchical Agglomerative Clustering (HAC) is a single-chain clustering method that clusters a sample data in the canyon wind power grid, and then condenses two adjacent cluster sets and iterates . This method has strong adaptability in canyon wind power grids, so this method is selected for clustering in this paper. When performing HAC clustering on canyon wind power station grids, Euclidean distance is used to calculate the similarity between two categories. Suppose the objects of two different classes of clustering are Yai and Ybi, the number of clustering features is nf, and the similarity between the two classes is
Figure PCTCN2022111510-appb-000007
通过迭代计算峡谷风电网格的相似层度，即可判断峡谷风电网格的分类。在聚类完成后，形成网格风电功率预测单元。
Figure PCTCN2022111510-appb-000007
By iteratively calculating the similarity levels of canyon wind power grids, the classification of canyon wind power grids can be determined. After clustering is completed, a grid wind power prediction unit is formed.
需要说明的是，本实施例中选用的层次凝聚聚类法是目前所有方法中最符合预期的方法，如果之后出现的其他方法达到更好或同等的聚类效果，也落入本发明的保护范围之内。It should be noted that the hierarchical agglomerative clustering method selected in this embodiment is the most expected method among all methods currently. If other methods that appear later achieve better or equivalent clustering effects, they will also fall under the protection of the present invention. within the range.
本实施例的其他部分与上述实施例1-8任一项相同，故不再赘述。The other parts of this embodiment are the same as any one of the above-mentioned Embodiments 1-8, and therefore will not be described again.
以上所述，仅是本发明的较佳实施例，并非对本发明做任何形式上的限制，凡是依据本发明的技术实质对以上实施例所做的任何简单修改、等同变化，均落入本发明的保护范围之内。The above are only preferred embodiments of the present invention and do not limit the present invention in any form. Any simple modifications or equivalent changes made to the above embodiments based on the technical essence of the present invention fall within the scope of the present invention. within the scope of protection.

Claims (9)

  1. 一种基于网格聚类的峡谷风电短期功率预测方法，其特征在于，包括以下步骤：A short-term power prediction method for canyon wind power based on grid clustering, which is characterized by including the following steps:
    步骤S1.根据峡谷风电特征将短期功率预测方法分为数据清洗环节、网格分解环节和功率预测环节；Step S1. According to the characteristics of canyon wind power, the short-term power prediction method is divided into data cleaning stage, grid decomposition stage and power prediction stage;
    步骤S2.在数据清洗环节对LSTM网络学***滑处理，获取完整的历史数据；Step S2. In the data cleaning process, perform anomaly detection, data correction and smoothing processing on the historical meteorology, wind power power, and wind power power prediction data learned by the LSTM network to obtain complete historical data;
    步骤S3.在网格分解环节根据纬度进行峡谷风电站地域网格聚类，获取风电功率和与气象关联的特征网格；Step S3. In the grid decomposition process, perform regional grid clustering of canyon wind power stations according to latitude, and obtain wind power power and characteristic grids related to meteorology;
    步骤S4.在功率预测环节使用LSTM网络对每个网格进行功率预测，并进行功率叠加与误差修正。Step S4. In the power prediction process, use the LSTM network to predict the power of each grid, and perform power superposition and error correction.
  2. 根据权利要求1所述的一种基于网格聚类的峡谷风电短期功率预测方法，其特征在于，所述步骤S2包括：A short-term power prediction method for canyon wind power based on grid clustering according to claim 1, characterized in that the step S2 includes:
    步骤S2.1.对LSTM网络学习的历史气象与风电预测数据、历史风电功率数据、气象监测数据和风机地理位置数据进行异常数据检测，通过离群点检测方式剔除异常值；Step S2.1. Perform abnormal data detection on the historical meteorological and wind power prediction data, historical wind power power data, meteorological monitoring data and wind turbine geographical location data learned by the LSTM network, and eliminate outliers through outlier detection;
    步骤S2.2.采用线性插值方法搭建缺失数据补正框架，实现缺失数据修复；Step S2.2. Use linear interpolation method to build a missing data correction framework to achieve missing data repair;
    步骤S2.3.通过卡尔曼滤波方式对风电功率、气象观测数据进行平滑处理，消除风电功率数据中的无用分量。Step S2.3. Smooth the wind power power and meteorological observation data through Kalman filtering to eliminate useless components in the wind power power data.
  3. 根据权利要求1所述的一种基于网格聚类的峡谷风电短期功率预测方法，其特征在于，所述步骤S3包括：A short-term power prediction method for canyon wind power based on grid clustering according to claim 1, characterized in that step S3 includes:
    步骤S3.1.根据地理经纬度进行风电场网格划分；Step S3.1. Grid the wind farm according to geographical longitude and latitude;
    步骤S3.2.根据天气类型和风电功率曲线特性指标进行网格聚类分析，获得网格风电功率特征；Step S3.2. Conduct grid clustering analysis based on weather types and wind power curve characteristic indicators to obtain grid wind power characteristics;
    步骤S3.3.分解形成网格风电功率预测单元。Step S3.3. Decompose the grid wind power power prediction unit.
  4. 根据权利要求1所述的一种基于网格聚类的峡谷风电短期功率预测方法，其特征在于，所述步骤S4包括：A short-term power prediction method for canyon wind power based on grid clustering according to claim 1, characterized in that step S4 includes:
    步骤S4.1.通过数字天气预报获取网格的短期天气预报信息；Step S4.1. Obtain grid short-term weather forecast information through digital weather forecast;
    步骤S4.2.基于长短期记忆网络分析网格风电波动趋势，预测网格预期时间内的峡谷风电短期功率；Step S4.2. Analyze the fluctuation trend of grid wind power based on the long short-term memory network and predict the short-term power of canyon wind power within the expected grid time;
    步骤S4.3.使用LSTM网络进行峡谷风电短期功率预测，采用最小二乘法修正误差；Step S4.3. Use the LSTM network to conduct short-term power prediction of canyon wind power, and use the least squares method to correct the error;
    步骤S4.4.生成峡谷风电短期功率预测结果。Step S4.4. Generate the short-term power prediction results of canyon wind power.
  5. 根据权利要求2所述的一种基于网格聚类的峡谷风电短期功率预测方法，其特征在于，所 述步骤S2.1中异常数据检测的方法包括：A short-term power prediction method for canyon wind power based on grid clustering according to claim 2, characterized in that the method for abnormal data detection in step S2.1 includes:
    根据离群点局部异常因子和数据间的距离分析，检测出与风电数据正常属性差异大的异常数据。Based on the local anomaly factor of outliers and the distance analysis between data, abnormal data that is significantly different from the normal attributes of wind power data is detected.
  6. 根据权利要求2所述的一种基于网格聚类的峡谷风电短期功率预测方法，其特征在于，所述步骤S2.2中缺失数据补正的方法包括：A short-term power prediction method for canyon wind power based on grid clustering according to claim 2, characterized in that the method for correcting missing data in step S2.2 includes:
    选用线性插值法搭建缺失数据补正框架，实现缺失数据修复。The linear interpolation method is used to build a missing data correction framework to achieve missing data repair.
  7. 根据权利要求2所述的一种基于网格聚类的峡谷风电短期功率预测方法，其特征在于，所述步骤S2.3中数据平滑处理的方法包括：A short-term power prediction method for canyon wind power based on grid clustering according to claim 2, characterized in that the data smoothing method in step S2.3 includes:
    选用卡尔曼滤波对滤波状态进行最佳估算，消除噪声和干扰数据对模型的影响。Kalman filter is used to best estimate the filtering state and eliminate the impact of noise and interference data on the model.
  8. 根据权利要求3所述的一种基于网格聚类的峡谷风电短期功率预测方法，其特征在于，所述步骤S3.1中地理网格划分的方法包括：A short-term power prediction method for canyon wind power based on grid clustering according to claim 3, characterized in that the method of geographical grid division in step S3.1 includes:
    选用归一化分割方法，并根据地理经纬度进行风电场网格划分。The normalized segmentation method is selected and the wind farm grid is divided according to the geographical longitude and latitude.
  9. 根据权利要求3所述的一种基于网格聚类的峡谷风电短期功率预测方法，其特征在于，所述步骤S3.2中进行网格聚类分析的方法包括：A short-term power prediction method for canyon wind power based on grid clustering according to claim 3, characterized in that the method of performing grid clustering analysis in step S3.2 includes:
    先用层次凝聚聚类法进行网格聚类分析。First, hierarchical agglomerative clustering method was used to conduct grid clustering analysis.
PCT/CN2022/111510 2022-08-10 2022-08-10 Grid-clustering-based short-term power prediction method for canyon wind power WO2024031449A1 (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
PCT/CN2022/111510 WO2024031449A1 (en) 2022-08-10 2022-08-10 Grid-clustering-based short-term power prediction method for canyon wind power

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
PCT/CN2022/111510 WO2024031449A1 (en) 2022-08-10 2022-08-10 Grid-clustering-based short-term power prediction method for canyon wind power

Publications (1)

Publication Number Publication Date
WO2024031449A1 true WO2024031449A1 (en) 2024-02-15

Family

ID=89850261

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/CN2022/111510 WO2024031449A1 (en) 2022-08-10 2022-08-10 Grid-clustering-based short-term power prediction method for canyon wind power

Country Status (1)

Country Link
WO (1) WO2024031449A1 (en)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN118013300A (en) * 2024-04-08 2024-05-10 江苏海龙风电科技股份有限公司 Short-term wind power prediction method and system for wind turbine generator
CN118100171A (en) * 2024-04-19 2024-05-28 阿里巴巴（中国）有限公司 Power prediction method and system of energy conversion system and electronic equipment

Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2003021687A (en) * 2001-07-06 2003-01-24 Digital Weather Platform Kk Method for weather forecasting for narrow area, method for distribution of weather forecast for narrow area, method for commodity sales promotion, weather forecasting device, and weather forecasting information providing system
US20160169202A1 (en) * 2013-05-03 2016-06-16 State Grid Corporation Of China Short-term operation optimization method of electric power system including large-scale wind power
CN106684905A (en) * 2016-11-21 2017-05-17 国网四川省电力公司经济技术研究院 Wind power plant dynamic equivalence method with wind power forecast uncertainty considered
CN114219122A (en) * 2021-11-09 2022-03-22 三峡大学 Ultra-short-term prediction method for generating power of small hydropower station based on grid meteorological data

Patent Citations (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2003021687A (en) * 2001-07-06 2003-01-24 Digital Weather Platform Kk Method for weather forecasting for narrow area, method for distribution of weather forecast for narrow area, method for commodity sales promotion, weather forecasting device, and weather forecasting information providing system
US20160169202A1 (en) * 2013-05-03 2016-06-16 State Grid Corporation Of China Short-term operation optimization method of electric power system including large-scale wind power
CN106684905A (en) * 2016-11-21 2017-05-17 国网四川省电力公司经济技术研究院 Wind power plant dynamic equivalence method with wind power forecast uncertainty considered
CN114219122A (en) * 2021-11-09 2022-03-22 三峡大学 Ultra-short-term prediction method for generating power of small hydropower station based on grid meteorological data

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN118013300A (en) * 2024-04-08 2024-05-10 江苏海龙风电科技股份有限公司 Short-term wind power prediction method and system for wind turbine generator
CN118013300B (en) * 2024-04-08 2024-06-04 江苏海龙风电科技股份有限公司 Short-term wind power prediction method and system for wind turbine generator
CN118100171A (en) * 2024-04-19 2024-05-28 阿里巴巴（中国）有限公司 Power prediction method and system of energy conversion system and electronic equipment

Similar Documents

Publication Publication Date Title
CN107766990B (en) Method for predicting power generation power of photovoltaic power station
Gao et al. Hourly forecasting of solar irradiance based on CEEMDAN and multi-strategy CNN-LSTM neural networks
WO2024031449A1 (en) Grid-clustering-based short-term power prediction method for canyon wind power
Das et al. Forecasting of photovoltaic power generation and model optimization: A review
Yang et al. A weather-based hybrid method for 1-day ahead hourly forecasting of PV power output
Sideratos et al. Probabilistic wind power forecasting using radial basis function neural networks
CN106920007B (en) PM based on second-order self-organizing fuzzy neural network2.5Intelligent prediction method
CN103117546B (en) A kind of Ultrashort-term slide prediction method for wind power
Bilal et al. Wind turbine power output prediction model design based on artificial neural networks and climatic spatiotemporal data
Çevik et al. A new multistage short-term wind power forecast model using decomposition and artificial intelligence methods
Munshi et al. Photovoltaic power pattern clustering based on conventional and swarm clustering methods
CN111488896B (en) Distribution line time-varying fault probability calculation method based on multi-source data mining
CN111626473A (en) Two-stage photovoltaic power prediction method considering error correction
CN117390894B (en) Method for predicting extreme storm tide level
CN114186723A (en) Distributed photovoltaic power grid virtual prediction system based on space-time correlation
CN110991701A (en) Wind power plant fan wind speed prediction method and system based on data fusion
CN115169659A (en) Ultra-short-term roof photovoltaic power prediction method and system based on multiple classifiers
CN112884601B (en) Power system operation risk assessment method based on weather division strategy
CN113610285A (en) Power prediction method for distributed wind power
Ahmed et al. Comparison of memory-less and memory-based models for short-term solar irradiance forecasting
Halidah et al. Short-Term Wind Forecasting with Weather Data using Deep Learning-Case Study in Baron Techno Park
Jin et al. A hybrid prediction framework based on deep learning for wind power
CN115358303A (en) Grid clustering-based canyon wind power short-term power prediction method
Kartini et al. Development hybrid model deep learning neural network (dl-nn) for probabilistic forecasting solar irradiance on solar cells to improve economics value added
Gutiérrez et al. A Comparison of the Performance of Supervised Learning Algorithms for Solar Power Prediction. Energies 2021, 14, 4424

Legal Events

Date Code Title Description
WWE Wipo information: entry into national phase

Ref document number: 202217060161

Country of ref document: IN

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

Ref document number: 22954424

Country of ref document: EP

Kind code of ref document: A1