基于NARX神经网络的极端风暴潮事件预报研究-海洋预报

基于NARX神经网络的极端风暴潮事件预报研究

单位：自然资源部第一海洋研究所, 山东青岛 266061

分类号：P731.23

出版年·卷·期（页码）：2023·40·第三期（11-18）

摘要：

通过构建采用外部输入的非线性自回归神经网络（NARX），利用1979年1月1日00时—2003年12月25日23时逐时的实测潮位数据和再分析气象数据结合调和分析预报结果搭建模型，对库克斯（Cuxhaven）港口2004—2018年中增水最大的两次风暴潮极端事件潮位进行预报和验证，同时对影响模型性能的参数进行量化分析。结果表明：在NARX神经网络延迟数为24 h时模型的精度最高，两次极端风暴潮验证下的R2分别为0.94和0.95，且在最高潮位时的误差分别为57.78 cm和26.55 cm。实验中模型在延迟数方面存在阈值，当延迟数为24 h时模型效果最佳，在延迟数达到阈值前模型的精度逐渐上升，超过该阈值后模型精度下降；输入时间数据序列的长短会影响模型的精度，序列越长模型精度越高，但影响效果会逐渐降低。

Using hourly tide level measurements from 00:00 on January 1, 1979 to 23:00 on December 25, 2003, meteorological reanalysis data, and reconciliation analysis of the forecast results, a storm surge forecasting model based on a nonlinear autoregressive model with exogenous inputs (NARX) neural network is conducted and validated in two storm surge extreme events with the largest water gain in Cuxhaven harbor from 2004 to mid-2018. The effects of the model parameters on the model's performance are quantitatively assessed. The results show that the model's accuracy is the highest when the NARX neural network's delay number is 24 hours, and the R2 is 0.94 and 0.95 for the two extreme storm surges, with errors of 57.78 cm and 26.55 cm at the highest tide level, respectively. The model's accuracy gradually increases before the delay number reaches the threshold of 24 hours, and gradually decreases after the delay number exceeds the threshold. The temporal duration of the input data also affects the model's accuracy, and longer input data series leads to higher accuracy of the model, but such relationship becomes weak as the temporal duration of the input data exceed a threshold.

参考文献：

[1] SPENCER T, BROOKS S M, EVANS B R, et al. Southern North Sea storm surge event of 5 December 2013:Water levels, waves and coastal impacts[J]. Earth-Science Reviews, 2015, 146:120-145.
[2] JENSEN J, ARNS A, WAHL T. Yet another 100Yr storm surge event:the role of individual storm surges on design water levels[J]. Journal of Marine Science and Technology, 2015, 23(6):882-887.
[3] CAVALERI L, BAJO M, BARBARIOL F, et al. The October 29, 2018 storm in Northern Italy-an exceptional event and its modeling[J]. Progress in Oceanography, 2019, 178:102178.
[4] 曾德美. 青岛港风暴潮经验统计预报[J]. 海洋预报, 1992, 9(3):66-73. ZENG D M. A statistical forecasting of storm surge in Qingdao Harbor[J]. Marine Forecasts, 1992, 9(3):66-73.
[5] 郭文云, 安佰超, 裘诚, 等. 基于多源数据的台风风暴潮概率预报研究:台风集合的构建[J]. 海洋预报, 2021, 38(1):26-33. GUO W Y, AN B C, QIU C, et al. Probabilistic forecast for typhoon storm surge based on multi-source data:creation of typhoon ensemble[J]. Marine Forecasts, 2021, 38(1):26-33.
[6] 张敏, 罗军, 胡金磊, 等. 雷州市沿海风暴潮淹没危险性评估[J]. 热带海洋学报, 2019, 38(2):1-12. ZHANG M, LUO J, HU J L, et al. Inundation risk assessment of storm surge along Lei Zhou coastal areas[J]. Journal of Tropical Oceanography, 2019, 38(2):1-12.
[7] 韩雪, 盛建明, 潘锡山, 等. 南黄海海域风暴潮精细化数值模式研究[J]. 海洋预报, 2019, 36(1):52-58. HAN X, SHENG J M, PAN X S, et al. Study on the refined storm surge numerical model in the Southern Yellow Sea[J]. Marine Forecasts, 2019, 36(1):52-58.
[8] 傅赐福, 董剑希, 刘秋兴, 等. 1409号和1415号台风风暴潮预报的数值研究[J]. 海洋预报, 2016, 33(4):26-33. FU C F, DONG J X, LIU Q X, et al. Numerical simulation study on typhoon "Rammasun" (1409) and typhoon "Kalmaegi" (1415) storm surge forecast[J]. Marine Forecasts, 2016, 33(4):26-33.
[9] 曹丛华, 白涛, 高松, 等. 胶州湾高分辨率三维风暴潮漫滩数值模拟[J]. 海洋科学, 2013, 37(2):118-125. CAO C H, BAI T, GAO S, et al. High resolution 3D storm surge and inundation numerical model used in the Jiaozhou Bay[J]. Marine Sciences, 2013, 37(2):118-125.
[10] 端义宏, 朱建荣, 秦曾灏, 等. 一个高分辨率的长江口台风风暴潮数值预报模式及其应用[J]. 海洋学报, 2005, 27(3):11-19. DUAN Y H, ZHU J R, QIN Z H, et al. A high-resolution numerical storm surge model in the Changjiang River estuary and its application[J]. Acta Oceanologica Sinica, 2005, 27(3):11-19.
[11] SCHMIDHUBER J. Deep learning in neural networks:an overview[J]. Neural Networks, 2015, 61:85-117.
[12] 张娟, 周水华, 黄宝霞, 等. 人工神经网络在台风风暴潮模拟中的解释应用[J]. 海洋预报, 2016, 33(2):60-65. ZHANG J, ZHOU S H, HUANG B X, et al. Interpretation of numerical storm surge model results using the artificial neural network[J]. Marine Forecasts, 2016, 33(2):60-65.
[13] LEE T L, JENG D S. Application of artificial neural networks in tide-forecasting[J]. Ocean Engineering, 2002, 29(9):1003-1022.
[14] PROUTY D B. Using artificial neural networks to predict storm surge in the North Sea and the Thames Estuary[D]. Southampton:University of Southampton, 2007.
[15] SHETTY R, DWARAKISH G S. Prediction of tides using neural networks at Karwar, west coast of India[J]. Development and Applications of Oceanic Engineering, 2013, 2(3):77-85.
[16] LEE J W, IRISH J L, BENSI M T, et al. Rapid prediction of peak storm surge from tropical cyclone track time series using machine learning[J]. Coastal Engineering, 2021, 170:104024.
[17] 张广平, 彭世球, 张晨晓. 一种融合多因素的MOS风暴潮灾害过程模拟研究[J]. 安全与环境工程, 2019, 26(3):50-55. ZHANG G P, PENG S Q, ZHANG C X. A simulation study of MOS storm surge disaster process with multiple factors[J]. Safety and Environmental Engineering, 2019, 26(3):50-55.
[18] 李未, 王如云, 卢长娜, 等. 神经网络在珠江口风暴潮预报中的应用[J]. 热带海洋学报, 2006, 25(3):10-13. LI W, WANG R Y, LU C N, et al. Forecast of storm surge in Zhujiang River estuary by means of artificial neural network[J]. Journal of Tropical Oceanography, 2006, 25(3):10-13.
[19] 卢君峰, 李少伟, 袁方超. 基于BP神经网络的厦门沿海风暴潮预报应用[J]. 海洋预报, 2016, 33(4):9-16. LU J F, LI S W, YUAN F C. Application of storm surge forecasting by BP artificial neural network off coast of Xiamen[J]. Marine Forecasts, 2016, 33(4):9-16.
[20] 雷森, 史振威, 石天阳, 等. 基于递归神经网络的风暴潮增水预测[J]. 智能系统学报, 2017, 12(5):640-644. LEI S, SHI Z W, SHI T Y, et al. Prediction of storm surge based on recurrent neural network[J]. CAAI Transactions on Intelligent Systems, 2017, 12(5):640-644.
[21] 吕忻, 丁骏. 基于深度学习的潮位预报订正技术研究[J]. 海洋预报, 2022, 39(2):70-79. LYU X, DING J. Study on the correction technology of tide level forecast based on deep learning[J]. Marine Forecasts, 2022, 39(2):70-79.
[22] MARQUARDT D W. An algorithm for least-squares estimation of nonlinear parameters[J]. Journal of the Society for Industrial and Applied Mathematics, 1963, 11(2):431-441.
[23] 李国民, 宿梦瑶, 朱代先. 光束平差法中的一种改进LM算法[J]. 西安科技大学学报, 2022, 42(1):152-159. LI G M, SU M Y, ZHU D X. An improved LM algorithm in bundle adjustment[J]. Journal of Xi'an University of Science and Technology, 2022, 42(1):152-159.
[24] HESTENES M R, STIEFEL E. Methods of conjugate gradients for solving linear systems[J]. Journal of Research of the National Bureau of Standards, 1952, 49(6):409-436.
[25] 苏高利, 邓芳萍. 论基于MATLAB语言的BP神经网络的改进算法[J]. 科技通报, 2003, 19(2):130-135. SU G L, DENG F P. On the improving backpropagation algorithms of the neural networks based on MATLAB language:a review[J]. Bulletin of Science and Technology, 2003, 19(2):130-135.
[26] FORESEE F D, HAGAN M T. Gauss-newton approximation to Bayesian learning[C]//Proceedings of International Conference on Neural Networks. Houston:IEEE, 1997:1930-1935.
[27] 刘墨阳, 李巧玲, 李致家, 等. 基于小波分析的NARX神经网络在水位预测中的应用[J]. 南水北调与水利科技, 2019, 17(5):56-63. LIU M Y, LI Q L, LI Z J, et al. The application of NARX neural network model based on wavelet analysis for water level prediction[J]. South-to-North Water Transfers and Water Science & Technology, 2019, 17(5):56-63.
[28] NUNNO F D, DE MARINIS G D, GARGANO R, et al. Tide prediction in the Venice lagoon using nonlinear autoregressive exogenous (NARX) neural network[J]. Water, 2021, 13(9):1173.

服务与反馈：

【文章下载】【发表评论】【查看评论】【加入收藏】