基于串联深度神经网络的Chl-a浓度短期预报方法研究-海洋预报

基于串联深度神经网络的Chl-a浓度短期预报方法研究

单位：1. 国家海洋环境预报中心自然资源部海洋灾害预报技术重点实验室, 北京 100081;
2. 浙江省海洋监测预报中心, 浙江杭州 310007

分类号：X55

出版年·卷·期（页码）：2021·38·第四期（1-10）

摘要：

以浙江海洋保护区2019年5月生态浮标监测数据为基础，对叶绿素a （Chl-a）与各理化因子进行Pearson相关性分析，发现研究海域的Chl-a与溶解氧和pH呈显著正相关（P=0.01），与硝氮和磷酸盐呈显著负相关（P=0.05）。在此基础上，建立了一种串联深度神经网络（DNN）的Chl-a短期预报模型，该模型以5层神经网络为基本单元，采用前后串联方式构建了拥有6个隐层的DNN。实验结果显示：DNN模型能够较为准确地预测Chl-a浓度短期变化趋势，24 h和48 h预报结果的RMSE分别为1.25 μg/L和2.43 μg/L，MAE分别为1.03 μg/L和1.99 μg/L，相比于浅层网络预测精度更高。

Based on the monitoring data of ecobuoys in Zhejiang marine protected area in May 2019, this paper analyses the correlation between Chl-a and physicochemical factors. Statistics shows that Chl-a is positively correlated with dissolved oxygen and pH at the level of P=0.01, while it is negatively correlated with nitrate and phosphonate at the level of P=0.05. In addition, a Chl-a short-term prediction model is established, which constructs a cascade deep neural network (DNN) with 6 hidden layers in series by taking 5-layer neural network as the basic unit. The experimental results show that the cascade DNN model can accurately predict the shortterm variation trend of Chl-a with higher prediction accuracy compared to the shallow neural network. The RMSE of 24 h and 48 h prediction is 1.25 μg/L and 2.43 μg/L, respectively. The MAE of 24 h and 48 h prediction is 1.03 μg/L and 1.99 μg/L, respectively.

参考文献：

[1] 张雪, 郑小慎. 基于BP神经网络渤海湾表层叶绿素浓度反演方法探讨[J]. 海洋技术学报, 2018, 37(6):79-87.
[2] 郭文景, 符志友, 汪浩, 等. 水华过程水质参数与浮游植物定量关系的研究——以太湖梅梁湾为例[J]. 中国环境科学, 2018, 38(4):1517-1525.
[3] 阮华杰, 马骏, 何志强. 生态浮标预测赤潮暴发的分析[J]. 声学与电子工程, 2014(2):44-46, 49.
[4] 金衍健, 卓丽飞. 舟山沿岸海域叶绿素a时空分布及与水质因子的相关分析[J]. 浙江海洋大学学报(自然科学版), 2017, 36(5):389-395.
[5] 林祥. 诏安湾养殖区叶绿素a与水质因子的主成分线性多元回归分析[J]. 环境影响评价, 2018, 40(5):88-90, 96.
[6] 杨德周, 尹宝树, 俞志明, 等. 长江口叶绿素分布特征和营养盐来源数值模拟研究[J]. 海洋学报, 2009, 31(1):10-19.
[7] 崔玉洁. 三峡水库香溪河藻类生长敏感生态动力学过程及其模拟[D]. 武汉:武汉大学, 2017.
[8] 张娣, 景元书, 李亚春, 等. 基于ARMA-BP集成的藻类叶绿素a预测研究[J]. 气象科学, 2015, 35(3):312-316.
[9] 石绥祥, 王蕾, 余璇, 等. 长短期记忆神经网络在叶绿素a浓度预测中的应用[J]. 海洋学报, 2020, 42(2):134-142.
[10] 陈祥光, 裴旭东. 人工神经网络技术及应用[M]. 北京:中国电力出版社, 2003, 9:22-31.
[11] Taylor J G. On Intelligence, Jeff Hawkins, Sandra Blakeslee, Times Books (2004)[J]. Artificial Intelligence, 2005, 169(2):192-195.
[12] Hinton G E, Salakhutdinov R R. Reducing the dimensionality of data with neural networks[J]. Science, 2006, 313(5786):504-507.
[13] 陈旭伟, 傅刚, 陈环. 基于串联BP神经网络多函数拟合的研究设计[J]. 现代电子技术, 2013, 36(22):14-16.
[14] Sutskever I, Martens J, Dahl G, et al. On the importance of initialization and momentum in deep learning[C]//Proceedings of the 30th International Conference on Machine Learning. Atlanta:JMLR. org, 2013:1139-1147.
[15] Wang R Q, Jiang Y L, Lou J G. TDR:Two-stage deep recommendation model based on mSDA and DNN[J]. Expert Systems with Applications, 2020, 145:113116.
[16] 王瑞琴, 吴宗大, 蒋云良, 等. 一种基于两阶段深度学习的集成推荐模型[J]. 计算机研究与发展, 2019, 56(8):1661-1669.
[17] JayaLakshmi A N M, Kishore K V K. Performance evaluation of DNN with other Machine learning techniques in a cluster using Apache Spark and MLlib[J]. Journal of King Saud University-Computer and Information Sciences, 2018. Doi:https://doi.org/10.1016/j.jksuci.2018.09.022.
[18] 刘胜辉, 张人敬, 张淑丽, 等. 基于深度神经网络的切削刀具剩余寿命预测[J]. 哈尔滨理工大学学报, 2019, 24(3):1-8.
[19] 卢勇夺, 王朝阳, 王豹, 等. 我国海洋锚系浮标数据异常值检测方法研究——以QF110和QF306为例[J]. 海洋预报, 2019, 36(6):37-43.
[20] 祖子清, 朱学明, 王辉, 等. Argo数据处理系统设计与应用[J]. 海洋预报, 2019, 36(4):1-12.
[21] 何沁波. 龙景湖叶绿素a浓度预测模型敏感性分析[D]. 重庆:重庆大学, 2015.
[22] Kudryavtseva E, Aleksandrov S, Bukanova T, et al. Relationship between seasonal variations of primary production, abiotic factors and phytoplankton composition in the coastal zone of the south-eastern part of the Baltic Sea[J]. Regional Studies in Marine Science, 2019, 32:100862.
[23] Liu L L, Dong Y C, Kong M, et al. Towards the comprehensive water quality control in Lake Taihu:Correlating chlorphyll a and water quality parameters with generalized additive model[J]. Science of the Total Environment, 2020, 705:135993.
[24] 王嵘冰, 徐红艳, 李波, 等. BP神经网络隐含层节点数确定方法研究[J]. 计算机技术与发展, 2018, 28(4):31-35.
[25] Yu X Y, Xu J, Long A M, et al. Carbon-to-chlorophyll ratio and carbon content of phytoplankton community at the surface in coastal waters adjacent to the Zhujiang River Estuary during summer[J]. Acta Oceanologica Sinica, 2020, 39(2):123-131.
[26] Lu Z B, Liu D D, Liao J S, et al. Characterizing spatial distribution of chlorophyll a in the Southern Ocean on a circumpolar cruise in summer[J]. Science of the Total Environment, 2020, 708:134833.
[27] Yang P P, Fong D A, Lo E Y M, et al. Circulation patterns in a shallow tropical reservoir:Observations and modeling[J]. Journal of Hydro-environment Research, 2019, 27:75-86.
[28] Gao S, Wang H, Liu G M, et al. Spatio-temporal variability of chlorophyll a and its responses to sea surface temperature, winds and height anomaly in the western South China Sea[J]. Acta Oceanologica Sinica, 2013, 32(1):48-58.
[29] Maas A L, Qi P, Xie Z A, et al. Building DNN acoustic models for large vocabulary speech recognition[J]. Computer Speech & Language, 2017, 41:195-213.

服务与反馈：

【文章下载】【发表评论】【查看评论】【加入收藏】