基于Resnet50的江苏近海海面风场预报订正方法研究-海洋预报

基于Resnet50的江苏近海海面风场预报订正方法研究

单位：1. 国网江苏省电力有限公司, 江苏南京 210000;
2. 江苏方天电力技术有限公司, 江苏南京 210000

分类号：P732.1

出版年·卷·期（页码）：2024·41·第四期（57-65）

摘要：

海面风场的精确预报对于海上风能开发具有非常重要的影响。利用 2019—2021 年ERA5再分析数据系统评估了欧洲中期天气预报中心的EC细网格10 m风场预报在江苏近海区域的预报准确性，发现EC细网格对于该区域4级风的预报准确性最高，24 h（48 h）风速预报的均方根误差为2.28 m/s（2.34 m/s），但是随着风级增大，风速预报的准确性大幅降低，5～11级风24 h（48 h）预报的均方根误差（RMSE）由2.39 m/s（2.58 m/s）增加到8.67 m/s（8.51 m/s）。此外，风速预报误差存在显著的空间差异性，误差随着离岸距离的增大而增大。在此基础上，基于 Resnet50模型构建了江苏近海海面风场预报订正模型，并利用2022年的预报数据对其进行独立性检验。结果表明：订正模型可以显著改善 EC 细网格 24 h（48 h）的 10 m 风速预报结果，订正后的 RMSE 为 1.45 m/s（1.66 m/s），较订正前降低了 45%（40%）。对于 3～10级风，订正模型 24 h和 48 h预报的 RMSE 为1.13～6.67 m/s和1.21～5.68 m/s，同样明显低于订正前（2.33～7.65 m/s和2.58～9.97 m/s）。

Accurate forecasts of surface winds plays an important role in offshore wind energy development. This study uses the ERA5 reanalysis data during 2019 — 2021 to evaluate the EC fine-grid 10 m winds forecasts in Jiangsu offshore area. It is found that the EC fine-grid 10 m wind speed forecasts perform best in accuracy for wind scale 4, with a 24 hour (48 hour) forecasting RMSE of 2.28 m/s (2.34 m/s). As wind scale increases, the wind speed forecasting accuracy decreases significantly, and the 24 hour (48 hour) forecasting RMSE for wind scales 5~ 11 increases from 2.39 m/s (2.58 m/s) to 8.67 m/s (8.51 m/s). In addition, significant spatial differentiation exists in the 10 m wind speed forecast errors, and the errors grow along with the increase of offshore distance. Based on the Resnet50 model, a correction method for surface wind forecasts in Jiangsu offshore area is constructed. The independence test using the forecasting data in 2022 shows that the correction method significantly improves the accuracy of the EC 24-hour and 48-hour 10 m wind speed forecasts, with a 24 hour (48 hour) forecasting RMSE of 1.48 m/s (1.65 m/s), which is 45% (40%) lower than the original EC forecasts. For wind scales 3~10, the 24-hour and 48-hour corrected forecasting RMSE are 1.13 m/s to 6.67 m/s and 1.21 m/s to 5.68 m/s, which is also significantly lower than the original EC forecasts (2.33 m/s to 7.65 m/s and 2.58 m/s to 9.97 m/s).

参考文献：

[1] ALLEN C, METTERNICHT G, WIEDMANN T. National pathways to the sustainable development goals (SDGs): a comparative review of scenario modelling tools[J]. Environmental Science & Policy, 2016, 66: 199-207.
[2] GIELEN D, BOSHELL F, SAYGIN D, et al. The role of renewable energy in the global energy transformation[J]. Energy Strategy Reviews, 2019, 24: 38-50.
[3] 王浩, 易侃, 杜梦蛟, 等. 漂浮式激光雷达海上测风可靠性及影响因素研究[J]. 海洋预报, 2022, 39(5): 70-83. WANG H, YI K, DU M J, et al. Study on the reliability and influencing factors using floating lidar systems for offshore wind measurement[J]. Marine Forecasts, 2022, 39(5): 70-83.
[4] 宋丽莉, 周荣卫, 杨振斌, 等. 风能资源开发利用的气象技术应用和发展[J]. 中国工程科学, 2012, 14(9): 96-101. SONG L L, ZHOU R W, YANG Z B, et al. Meteorological technology application and development in wind energy resources utilization[J]. Engineering Sciences, 2012, 14(9): 96-101.
[5] 李敏, 王辉, 金啟华. 中国近海海面风场预报方法综述[J]. 海洋预报, 2009, 26(3): 114-120. LI M, WANG H, JIN Q H. A review on the forecast method of China offshore wind[J]. Marine Forecasts, 2009, 26(3): 114-120.
[6] ROZAS-LARRAONDO P, INZA I, LOZANO J. A method for wind speed forecasting in airports based on nonparametric regression[J]. Weather and Forecasting, 2014, 29(6): 1332-1342.
[7] BAUER P, THORPE A, BRUNET G. The quiet revolution of numerical weather prediction[J]. Nature, 2015, 525(7567): 47-55.
[8] BONAVITA M, HÓLM E, ISAKSEN L, et al. The evolution of the ECMWF hybrid data assimilation system[J]. Quarterly Journal of the Royal Meteorological Society, 2016, 142(694): 287-303.
[9] 毛开银, 赵长名, 何嘉. 基于XGBoost的10 m风速订正研究[J]. 成都信息工程大学学报, 2020, 35(6): 604-609. MAO K Y, ZHAO C M, HE J. A research for 10 m wind speed prediction based on XGBoost[J]. Journal of Chengdu University of Information Technology, 2020, 35(6): 604-609.
[10] RANABOLDO M. Multiple linear regression MOS for short-term wind power forecast[D]. Barcelona: Universitat de Barcelona, 2011.
[11] 肖擎曜, 胡非, 范绍佳. 风能数值预报的模式输出统计（MOS）研究[J]. 资源科学, 2017, 39(1): 116-124. XIAO Q Y, HU F, FAN S J. Model output statistics and wind power numerical prediction[J]. Resources Science, 2017, 39(1): 116-124.
[12] LI D L, FENG J L, XU Z H, et al. Statistical bias correction for simulated wind speeds over CORDEX ‐ east Asia[J]. Earth and Space Science, 2019, 6(2): 200-211.
[13] BOKDE N D, FEIJÓO A, AL-ANSARI N, et al. A comparison between reconstruction methods for generation of synthetic time series applied to wind speed simulation[J]. IEEE Access, 2019, 7: 135386-135398.
[14] MONAHAN A H. The Gaussian statistical predictability of wind speeds[J]. Journal of Climate, 2013, 26(15): 5563-5577.
[15] 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. SUN Q D, JIAO R L, XIA J J, et al. Adjusting wind speed prediction of numerical weather forecast model based on machine learning methods[J]. Meteorological Monthly, 2019, 45(3): 426-436.
[16] 孔令彬, 赵艳茹, 王聚杰, 等. 基于支持向量机风速订正方法的研究[J]. 西南大学学报（自然科学版）, 2014, 36(5): 194-200. KONG L B, ZHAO Y R, WANG J J, et al. Modification of wind speed based on SVM[J]. Journal of Southwest University (Natural Science Edition), 2014, 36(5): 194-200.
[17] ZJAVKA L. Wind speed forecast correction models using polynomial neural networks[J]. Renewable Energy, 2015, 83: 998-1006.
[18] 叶小岭, 顾荣, 邓华, 等. 基于WRF模式和PSO-LSSVM的风电场短期风速订正[J]. 电力系统保护与控制, 2017, 45(22): 48-54. YE X L, GU R, DENG H, et al. Modification technology research of short-term wind speed in wind farm based on WRF model and PSO-LSSVM method[J]. Power System Protection and Control, 2017, 45(22): 48-54.
[19] 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. DENG H, ZHANG Y C, GU R, et al. Correction method of shortterm wind speed in wind farm research based on PCA and RBF neural network[J]. Meteorological Science and Technology, 2018, 46(1): 10-15.
[20] 潘微, 邢建勇, 万莉颖. 一种基于BP神经网络方法的HY-2A散射计反演风场偏差订正方案[J]. 海洋预报, 2018, 35(2): 8-18. PAN W, XING J Y, WAN L Y. A deviation correction of HY-2A scatterometer inversed wind field based on BP neural network method[J]. Marine Forecasts, 2018, 35(2): 8-18.
[21] 付旭东. 基于机器学习的短时风场预报与订正研究[D]. 兰州: 兰州大学, 2020. FU X D. Research on short-term wind field forecast and correction based on machine learning [D]. Lanzhou: Lanzhou University, 2020.
[22] HERSBACH H, BELL B, BERRISFORD P, et al. ERA5 hourly data on single levels from 1940 to present[EB/OL]. [2023-05-01]. https://doi.org/10.24381/cds.adbb2d47.
[23] 周飞燕, 金林鹏, 董军. 卷积神经网络研究综述[J]. 计算机学报, 2017, 40(6): 1229-1251. ZHOU F Y, JIN L P, DONG J. Review of convolutional neural network[J]. Chinese Journal of Computers, 2017, 40(6): 1229- 1251.
[24] LECUN Y, BOTTOU L, BENGIO Y, et al. Gradient-based learning applied to document recognition[J]. Proceedings of the IEEE, 1998, 86(11): 2278-2324
[25] HE K M, ZHANG X Y, REN S Q, et al. Deep residual learning for image recognition[C]//2016 IEEE Conference on Computer Vision and Pattern Recognition. Las Vegas: IEEE, 2016: 770-778.
[26] IOFFE S, SZEGEDY C. batch normalization: accelerating deep network training by reducing internal covariate shift[C]// Proceedings of the 32nd International Conference on Machine Learning. Lille: JMLR.org, 2015: 448-456.
[27] HARVEY D, LEYBOURNE S, NEWBOLD P. Testing the equality of prediction mean squared errors[J]. International Journal of Forecasting, 1997, 13(2): 281-291.

服务与反馈：

【文章下载】【发表评论】【查看评论】【加入收藏】