针对风速预报订正中的数据非平衡问题,提出了一种基于分类/回归层次结构的订正模型.该模型的核心思想是采用分治策略,逐步解决风速数据中的非平衡问题.在分类层中,使用了重加权策略来初步解决数据中的非平衡问题.在回归层中,提出了一种分组扩展的训练策略,有效纠正了受非平衡影响而被错误分类的样本,从而进一步解决数据非平衡问题.此外,还基于贪心策略设计了一种概率加权方法,目的是对有把握的样本输出更加准确的预测.该模型在山东沿海14个地区的风速数据集上进行了验证,并与相关方法进行了比较,其中,订正后的风速预报整体和极端风速事件的平均绝对误差分别降低了34.4%和69.0%,表明该模型在提高极端风速事件预测能力的同时,也保持了对稳定事件的预测性能.
To address the data imbalance problem in wind speed forecast correction, a correction model based on a classification/regression hierarchical structure is proposed. The core idea of this model is to adopt a divide-and-conquer strategy to gradually address the data imbalance in wind speed data. In the classification layer, a reweighting strategy is used to preliminarily address the data imbalance. In the regression layer, a group-wise expansion training strategy is proposed to effectively correct samples incorrectly classified due to imbalance, further addressing the data imbalance issue. Additionally, a probability weighting method based on a greedy strategy is designed to provide more accurate predictions for confident samples. The corrected wind speed forecasts showed a 34.4% reduction in overall Mean Absolute Error, with a 69.0% reduction for extreme wind speed events, indicating that the model not only improves the prediction accuracy for extreme wind speed events but also maintains performance in predicting stable events.
[1] 宗晓萍,武子瀚,刘言. 基于遗传算法寻优的SVR雾霾预测模型[J]. 河北大学学报(自然科学版), 2016, 36(3): 307-311. DOI: 10.3969/j.issn.1000-1565.2016.03.014.
[2] 姜艳波,徐宁伟,陈泰熙,等. 基于BP神经网络的土地利用智能分类识别与雨洪风险模拟[J]. 河北大学学报(自然科学版),2024, 44(2): 208-215. DOI: 10.3969/j.issn.1000-1565.2024.02.012.
[3] CHEN Y, BAI M, ZHANG Y, et al. Multivariable space-time correction for wind speed in numerical weather prediction(NWP)based on ConvLSTM and the prediction of probability interval[J]. Earth Sci Inform, 2023, 16(3): 1953-1974. DOI: 10.1007/s12145-023-01036-1.
[4] GLAHN H R, LOWRY D A. The use of model output statistics(MOS)in objective weather forecasting[J]. J Appl Meteor, 1972, 11(8): 1203-1211. DOI: 10.1175/1520-0450(1972)011<1203:TUOMOS>2.0.CO;2.
[5] WANG A, XU L, LI Y, et al. Random-forest based adjusting method for wind forecast of WRF model[J]. Comput Geosci, 2021, 155: 104842. DOI: 10.1016/j.cageo.2021.104842.
[6] YE L, DAI B, LI Z, et al. An ensemble method for short-term wind power prediction considering error correction strategy[J]. Appl Energy, 2022, 322: 119475. DOI: 10.1016/j.apenergy.2022.119475.
[7] YANG Y, ZHA K, CHEN Y, et al. Delving into deep imbalanced regression[C] //International Conference on Machine Learning, PMLR, 2021: 11842-11851.
[8] FERNáNDEZ A, GARCIA S, HERRERA F, et al. SMOTE for learning from imbalanced data: progress and challenges, marking the 15-year anniversary[J]. Journal of artificial intelligence research, 2018, 61: 863-905. DOI: 10.1613/jair.1.11192.
[9] TORGO L, RIBEIRO R P, PFAHRINGER B, et al. SMOTE for Regression[M] //CORREIA L, REIS L P, CASCALHO J. Progress in Artificial Intelligence, Berlin, Heidelberg: Springer, 2013: 378-389. DOI: 10.1007/978-3-642-40669-0_33.
[10] BRANCO P, TORGO L, RIBEIRO R P. SMOGN: a pre-processing approach for imbalanced regression[C] //First international workshop on learning with imbalanced domains: Theory and applications. PMLR, 2017: 36-50.
[11] PELáEZ-RODRíGUEZ C, PéREZ-ARACIL J, PRIETO-GODINO L, et al. A fuzzy-based cascade ensemble model for improving extreme wind speeds prediction[J]. J Wind Eng Ind Aerodyn, 2023, 240: 105507. DOI: 10.1016/j.jweia. 2023.105507.
[12] KE S W, LIN W C, TSAI C F, et al. Soft estimation by hierarchical classification and regression[J]. Neurocomputing, 2017, 234: 27-37. DOI: 10.1016/j.neucom.2016.12.037.
[13] PELÁEZ-RODRÍGUEZ C, PÉREZ-ARACIL J, FISTER D, et al. A hierarchical classification/regression algorithm for improving extreme wind speed events prediction[J]. Renew Energy, 2022, 201: 157-178. DOI: 10.1016/j.renene.2022.11.042.
[14] PROKHORENKOVA L, GUSEV G, VOROBEV A, et al. CatBoost: unbiased boosting with categorical features[EB/OL]. 2017: arXiv: 1706.09516. http://arxiv.org/abs/1706.09516.
[15] KE G, MENG Q, FINLEY T, et al. Lightgbm: A highly efficient gradient boosting decision tree[J]. Advances in neural information processing systems, 2017, 30. DOI: 10.5555/3294996.3295074.
[16] SIMIU E, HECKERT N A. Extreme wind distribution tails: A “peaks over threshold” approach[J]. Journal of Structural Engineering, 1996, 122(5): 539-547. DOI: 10.1061/(ASCE)0733-9445(1996)122:5(539).
[17] DAVISON A C. Modelling excesses over high thresholds, with an application[M] //OLIVEIRA J T. Statistical Extremes and Applications. Dordrecht: Springer Netherlands, 1984: 461-482. DOI: 10.1007/978-94-017-3069-3_34.
[18] DEL SER J, CASILLAS-PEREZ D, CORNEJO-BUENO L, et al. Randomization-based machine learning in renewable energy prediction problems: critical literature review, new results and perspectives[J]. Applied Soft Computing, 2022, 118: 108526. DOI: 10.1016/j.asoc.2022.108526.
[19] BREIMAN L. Random forests[J]. Mach Learn, 2001, 45(1): 5-32. DOI: 10.1023/A:1010933404324.
[20] AWAD M, KHANNA R. Support vector regression[J]. Efficient Learning Machines, 2015, 10: 978-1. DOI: 10.1007/978-1-4302-5990-9_4.
[21] GARDNER M W, DORLING S R. Artificial neural networks(the multilayer perceptron)—a review of applications in the atmospheric sciences[J]. Atmos Environ, 1998, 32(14-15): 2627-2636. DOI:10.1016/S1352-2310(97)00447-0.
[22] LIU Y, DONG H, WANG X, et al. Time series prediction based on temporal convolutional network[C] //2019 IEEE/ACIS 18th International Conference on Computer and Information Science(ICIS), IEEE, 2019: 300-305. DOI: 10.1109/ICIS46139. 2019.8940265.
[23] GRAVES A. Long Short-Term Memory[M] //GRAVES A. Supervised Sequence Labelling with Recurrent Neural Networks, Berlin, Heidelberg: Springer Berlin Heidelberg, 2012: 37-45. DOI: 10.1007/978-3-642-24797-2_4.
[24] VASWANI A, SHAZEER N, PARMAR N. 2017. Attention is all you need[C]. Advances in Neural Information Processing Systems. 30: 6000- 6010. Long Beach, CA, USA.
[25] ZHOU T, MA Z Q, WEN Q S, et al. FEDformer: frequency enhanced decomposed transformer for long-term series forecasting[EB/OL]. 2022: arXiv: 2201.12740. http://arxiv.org/abs/2201.12740.
[26] NIE Y Q, NGUYEN N H, SINTHONG P, et al. A time series is worth 64 words: long-term forecasting with transformers[EB/OL]. 2022: arXiv: 2211.14730. http://arxiv.org/abs/2211.14730. (
