摘要
中国海域辽阔,海岸带面积约占全国总面积的13%,在沿海区域的交通运输及经济建设领域,都需要具备精确的潮位数据,因此实现精准快速的潮位预报具有重要的应用价值和实际意义。为了提高潮位预测精度和稳定性,提出了一种基于带外源输入的非线性自回归(nonlinear auto-regressive exogenous, NARX)神经网络的实时潮位预测方法,并在其基础上做了相应改进。首先采用了模块化潮位预测(modular tide level prediction)方法,将潮汐数据分为天文潮及非天文潮两部分,其次引入滑动时间窗(sliding time window, STW)概念构建出改进的MS-NARX神经网络预测模型。利用美国比斯坎湾(Biscayne bay)的实测潮汐值数据进行潮位预测的仿真试验,并与传统NARX神经网络及自适应粒子群算法优化的基本反向传播(SAPSO-BP)神经网络两种预测方法进行比较,结果表明在MAE、MSE及RMSE三项精度指标测算中,MS-NARX神经网络均为最小,可见其针对数据预测的精度和稳定性均优于SAPSO-BP神经网络和传统NARX神经网络,能够为提高船舶运营效率和保障船舶航行安全提供指导。
China’s sea area is vast, and the coastal zone covers about 13% of the country’s total area. In the field of transportation and economic construction in coastal areas, it is necessary to have accurate tide level data, so it is of great value and practical significance to achieve accurate and fast tide level forecasting. In order to improve the accuracy and stability of tide level prediction, a real-time tide level prediction method based on the nonlinear auto-regressive eXogenous(NARX) neural network was proposed and improved on its basis. Firstly, a modular tide level prediction method was adopted to divide the tidal data into two parts: astronomical tide and non-astronomical tide, and secondly, the sliding time window(STW) concept was introduced to construct an improved MS-NARX neural network prediction model. The tide level prediction simulation experiment was carried out by using the measured tide data of Biscayne Bay in the United States and was compared with the traditional NARX neural network and basic Back Propagation optimized by self-adaption particle swarm optimization(SAPSO-BP) neural network. The results show that among the three accuracy indexes of MAE, MSE and RMSE, MS-NARX neural network is the smallest, which shows that its accuracy and stability for data prediction are better than SAPSO-BP neural network and traditional NARX neural network, which can guide for improving ship operation efficiency and ensuring ship navigation safety.
作者
李连博
武文昊
章文俊
尹建川
朱振宇
LI Lian-bo;WU Wen-hao;ZHANG Wen-jun;YIN Jian-chuan;ZHU Zhen-yu(Navigation College,Dalian Maritime University,Dalian 116026,China;Maritime Transport College,Guangdong Ocean University,Zhanjiang 524088,China)
出处
《科学技术与工程》
北大核心
2022年第22期9728-9735,共8页
Science Technology and Engineering
基金
国家重点研发计划(2019YFB1600602)
国家自然科学基金(51879024)
辽宁省自然科学基金(20180550283,2020-HYLH-27,201801704)
大连海事大学教学改革项目(2020Y04,2020Y79)
中国交通教育研究会2020—2022年度交通教育科学研究课题(JTYB20-03,JTYB20-09)。