MAO Kaiyin,ZHAO Changming,HE Jia.A Research for 10 m Wind Speed Prediction based on XGBoost[J].Journal of Chengdu University of Information Technology,2020,35(06):604-609.[doi:10.16836/j.cnki.jcuit.2020.06.004]
基于XGBoost的10 m风速订正研究
- Title:
- A Research for 10 m Wind Speed Prediction based on XGBoost
- Keywords:
- machine learning; wind speed correction; clustering; XGBoost
- 分类号:
- TP301.6
- 文献标志码:
- A
- 摘要:
- 基于当前气象预报模式,风速预报的精确度存在一定的误差,国内外研究者对风速的预报订正做了大量的研究。提出一种CD-XGBoost(clustering and double XGBoost)算法,该算法针对现有的机器学习风速预报订正算法进行改进,主要包含以下3个改进方向: 第一,提出利用天气元素与订正元素之间的相关性进行聚类的思路,通过对簇间站点进行基于不同机器学习模型的训练,提高风速订正结果的准确性; 第二,算法突出空间因素对风速预报的影响,选取气象观测站点的K个邻近预报网格点的预报元素构建数据集,相较传统插值进行订正的方式更多地考虑站点与网格点之间的空间因素。第三,提出使用2个不同起报时刻数据独立训练XGBoost模型,对模型的输出再进行线性权重加和得到最终订正结果的算法。仿真实验中,采用中国2552个气象观测站的逐3 h观测数据与欧洲中期气象预报中心(ECWMF)的数值模式的逐3 h预报的数据,对ECWMF预报的地面10 m风速进行观测站点订正。使用改进后的算法构建的模型与目前算法构建的模型进行比较,结果证明文中算法在风速预测准确度明显优于目前的订正算法,其中3 h预报时效时准确率高于85%,168 h预报时效是其准确率高于60%,具有很好的应用前景。
- Abstract:
- Based on the current meteorological forecasting model, there is a certain error in the accuracy of wind speed forecasting, and domestic and foreign researchers have done a lot of work in revising wind speed forecasting. This paper proposes a CD-XGBoost(Clustering and Double XGBoost)algorithm, which is an improvement to the existing algorithm of machine learning wind speed forecast correction. It mainly includes the following three improvement directions: First, propose the idea of clustering the correlation between weather elements and correction elements and training the inter-cluster sites based on different machine learning models to improve the accuracy of wind speed correction results. Second, the algorithm highlights the impact of spatial factors on wind speed forecasting, and selects the forecast elements of K nearby forecast grid points of the meteorological observation station to build a data set. Compared with the traditional interpolation correction method, more consideration is given to the spatial factors between the stations and grid points. Third, an algorithm is proposed to independently train the XGBoost model using data from two different reporting points, and then add linear weights to the outputs of the two models to obtain the final correction result. In this paper’s simulation experiments, three-hour observation data from 2552 meteorological observatories in China and 3 h forecast data from the numerical model of the European Medium-Term Weather Forecasting Center(ECWMF)are used to revise the ECWMF’s 10 m surface prediction. The model constructed using the improved algorithm is compared with the model constructed by the current algorithm, and the results show that the accuracy of the wind speed prediction algorithm proposed in this paper is significantly better than that of the current revised algorithm, in which the accuracy rate of 3 h forecast time is over 85%.The accuracy rate within 168 h is higher than 60%, which has a good application prospect.
参考文献/References:
[1] 肖寅.基于WRF模式的短期风速预测及订正方法研究[D].南京:南京信息工程大学,2016.
[2] N D Bokde,A Feijóo,N Al-Ansari,et al.A Comparison Between Reconstruction Methods for Generationof Synthetic Time Series Applied to Wind Speed Simulation[J].IEEE Access,2019,7:135386-135398.
[3] R L Kashyap.Optimal choice of AR and MA parts in autoregressive moving average models[J].IEEE Trans.Pattern Anal.Mach.Intell.,1982,2(2):99-104.
[4] G P Zhang.Time series forecasting using a hybrid ARIMA and neural network model[J].Neurocomputing,2003,50:159-175.
[5] Aaron M R Culver,Adam H Monahan.The Statistical Predictability of Surface Winds over Western and Central Canada[J].Journal of Climate,2013,26(21):8305-8322.
[6] Monahan,A H,The Gaussian Statistical Predictability of Wind Speeds[J].Climate,2013,26:5563-5577.
[7] Galanis,George,Papageorgiou,et al.A hybrid Bayesian Kalman filter and applications to numerical wind speed modeling[J].Journal of Wind Engineering & Industrial Aerodynamics,1997,56(167):1-22.
[8] 孙全德,焦瑞莉,夏江江,等.基于机器学习的数值天气预报风速订正研究[J].气象,2019,45(3):426-436.
[9] Gang Zhang,Lei Zhang,Tuo Xie.Prediction of short-term wind power in wind power plant based on BP-ANN[C].2016 IEEE Advanced Information Management,Communicates,Electronic and Automation Control Conference(IMCEC),Xi’an,2016:75-79.
[10] J Jin,B Wang,M Yu,et al.The short-term wind speed prediction based on HF-EMD and BP neural network model[C].2017 12th IEEE Conference on Industrial Electronics and Applications(ICIEA),Siem Reap,2017:489-494.
[11] Gang Zhang,Lei Zhang,Tuo Xie.Prediction of short-term wind power in wind power plant based on BP-ANN[C].2016 IEEE Advanced Information Management,Communicates,Electronic and Automation Control Conference(IMCEC),Xi’an,2016:75-79.
[12] A Prasetyowati,D.Sudiana,H Sudibyo.Comparison Accuracy W-NN and WD-SVM Method In Predicted Wind Power Model on Wind Farm Pandansimo[J].2018 4th International Conference on Nano Electronics Research and Education(ICNERE),Hamamatsu,Japan,2018:1-4.
[13] H Xue,L Li,K Chao,et al.Short-Term Wind Power Prediction Based on Improved Chicken Algorithm and Support Vector Machine[J].2018 International Symposium on Computer,Consumer and Control(IS3C),Taichung,Taiwan,2018:137-140.
[14] 张颖超,肖寅,邓华.基于ELM的风电场短期风速订正技术研究[J].气象,2016,42(4):466-471.
[15] D Kang,Y Su,X Liu,et al.Short-term wind speed forecasting in wind farm based on C-C and ELM[C].2017 3rd IEEE International Conference on Computer and Communications(ICCC), Chengdu,2017:2832-2836.
[16] X Luo.Short-Term Wind Speed Forecasting via Stacked Extreme Learning Machine With Generalized Correntropy[J].in IEEE Transactions on Industrial Informatics,2018,14(11):4963-4971.
[17] 邓华,张颖超,顾荣,等.基于PCA-RBF的风电场短期风速订正方法研究[J].气象科技,2018,46(1):10-15.
[18] Y Zhang,B Chen,Y Zhao,et al.Wind Speed Prediction of IPSO-BP Neural Network Based on Lorenz Disturbance[J].in IEEE Access,2018,6:53168-53179.
[19] W Teng,X Wang,Y Meng,et al.An improved support vector clustering approach to dynamic aggregation of large wind farms[J].in CSEE Journal of Power and Energy Systems,2019,5(2):215-223.
[20] Chen T,Tong H.Higgs boson discovery with boosted trees[C].International Conferenceon High-energy Physics & Machine Learning.2014.
[21] Yang A,Zhang J,Pan L,et al.Enhanced Twitter Sentiment Analysis by Using Feature Selection and Combination[C].2015 International Symposium on Security and Privacy in Social Networks and Big Data(SocialSec).IEEE,2015.
相似文献/References:
[1]李宝林,周 坤,李仕伟.一种基于M-Bisearch的最大频繁项集挖掘算法研究[J].成都信息工程大学学报,2016,(05):463.
LI Bao-lin,ZHOU Kun,LI Shi-wei.Research on Mining Algorithm of Maximal Frequent Itemsets based on M-blsearch[J].Journal of Chengdu University of Information Technology,2016,(06):463.
[2]唐雨奇,李则辰,杨东东,等.基于机器学习的面部运动神经传导检查数据的研究及应用[J].成都信息工程大学学报,2020,35(05):519.[doi:10.16836/j.cnki.jcuit.2020.05.007]
TANG Yuqi,LI Zechen,YANG Dongdong,et al.Research and Application of Facial Motor Nerve Conduction Examination Data based on Machine Learning[J].Journal of Chengdu University of Information Technology,2020,35(06):519.[doi:10.16836/j.cnki.jcuit.2020.05.007]
[3]李孝涌,陈科艺,李熙晨.卷积神经网络在ENSO预报中的应用[J].成都信息工程大学学报,2022,37(01):81.[doi:10.16836/j.cnki.jcuit.2022.01.014]
LI Xiaoyong,CHEN Keyi,LI Xichen.Application of Convolutional Neural Network in ENSO Prediction[J].Journal of Chengdu University of Information Technology,2022,37(06):81.[doi:10.16836/j.cnki.jcuit.2022.01.014]
备注/Memo
收稿日期:2019-12-13 基金项目:国家重大专项资助项目(2017YFG502203); 国家重点研发资助项目(2019YFG0212); 四川省科技计划资助项目(2019YFG0212); 四川省科技计划资助项目(2018GZ0814)