秦皇岛湾人工智能赤潮预警研究  

Research on AI Early Warning Model of Red Tide in the Qinhuangdao Bay

作　　者：杨雯 张勇[2] 李微微 王众 马新[2] YANG Wen;ZHANG Yong;LI Weiwei;WANG Zhong;MA Xin(Qinhuangdao Marine Environmental Monitoring Central Station,Qinhuangdao 066001,China)

机构地区：[1]国家海洋局秦皇岛海洋环境监测中心站综合化验室,秦皇岛066000 [2]国家海洋局秦皇岛海洋环境监测中心站,秦皇岛066000

出　　处：《海洋信息技术与应用》2022年第1期56-64,共9页JOURNAL OF MARINE INFORMATION TECHNOLOGY AND APPLICATION

摘　　要：本文提出了多指标阈值和基于机器学习的人工智能赤潮预警模型两种赤潮灾害预警方法。其中,多指标阈值法根据赤潮生消过程中环境因子的变化规律,基于已知藻类密度超过基准密度或发生赤潮事件数据极少的现状,结合秦皇岛湾海域的地域特色,找出主要影响因子并基于这些影响因子的数值分布分别设置阈值,预测赤潮现象的发生;基于机器学习的赤潮综合预警模型是将发生赤潮与未发生赤潮的数据集建立对照组,从数据中学习并掌握赤潮发生规律,继而对赤潮现象做出预测。实验结果表明本文提出的赤潮预警方法具有较准确的预测结果。Two red tide risk warning methods,the multi-index threshold method and the machine learning-based red tide AI warning model are proposed in this paper.Combined with the regional characteristics of the Qinhuangdao Bay,the multi-index threshold method finds out the main influencing factors according to the changing law of environmental factors during the process of red tide generation and elimination,and sets thresholds based on the distribution of these factors to predict the occurrence of harmful red tides.The machine learning-based red tide comprehensive warning model is to establish a control group based on the data sets of occurrence and non-occurrence of red tide,learn and master the occurrence rules of red tide from the data,and then make predictions of harmful red tide events.

关 键 词：海湾 多指标阈值法 机器学习 赤潮预警 

分 类 号：P76[天文地球—海洋科学]