基于SSA-BP神经网络对南港水道船舶航行行为的预测分析

基于SSA-BP神经网络对南港水道船舶航行行为的预测分析摘要随着疫情后的经济复苏，经济全球化进程持续推进，海上交通业务日趋繁忙，船舶航行安全问题也日益突出，其中在南港航道北部到吴淞口一段，是整个长江口水域交通密度最大、态势最复杂的地区之一，为了能够充分利用AS信息，进一步增强船舶交通服务系统(Vessel Traffic Service,TS)的监管职能，本文针对“船舶航行的航速和航向”开展研究，对于航行信息进行预判，实现事前性预警，降低船舶进港航行时的碰撞风险，避免水域交通事故频繁发生。本文以船舶AS数据为基础，结合BP神经网络的时间序列预测开展关于船舶航速和航向的相关研究。为解决AS数据中存在的信息丢失和不准确的问题，本文在保留原有航行属性特征的基础上，对其问题数据进行清洗。原始船舶AS数据在历经采集、封装、传输、接收等多个过程后，可能会出现数据缺失，需要进行填补，为了最大可能满足预测数据的准确性和时效性，本文采用了线性插值法对缺失数据进行补充。船舶AIS在接收信息前，经历多个阶段，难免会受到噪音的影响，从而最终影响预测数据的精度，为了使得船舶数据更加准确，本文在采用神经网络预测前将会对数据进行奇异谱分析，最大程度上还原数据的真实性，去掉可能是噪音的部分，从而使预测结果更加准确。考虑到AIS数据传输过程中可能会受到噪音影响，本文通过构建SSA-BP神经网络的组合模型研究实现了对船舶航行行为的预测，并通过与不经奇异谱分析的BP神经网络预测分析结果对比，证明了去噪的对还原原始数据的必要性和基于SSA-BP神经网络预测模型的精确性。该成果对降低船舶进港航行时的碰撞风险，保障船舶通航安全具有现实意义。关键词：船舶航行：航速和航向：AIS数据：线性插值法：BP神经网络：奇异谱分析AbstractWith the economic recovery after the epidemic and the continuous advancement ofeconomic globalization,the maritime traffic business becomes increasingly busy,and theproblem of ship navigation safety becomes increasingly prominent.Among them,the sectionfrom the north of Nangang Channel to Wusongkou is one of the areas with the highest trafficdensity and the most complex situation in the whole Yangtze River Estuary.In order to makefull use of AIS information,and further enhance the supervision function of Vessel TrafficService (VTS).Based on ship AIS data and combined with BP neural network time seriesprediction,this paper carries out related research on sailing speed and heading.In order to solve the problem of information loss and inaccuracy in AIS data,thispaper cleans and repairs the problem data on the basis of retaining the original navigationattributes.After multiple processes such as acquisition,encapsulation,transmission andreception of the original ship AIS data,there may be data loss,which needs to be filled.Inorder to maximize the accuracy and timeliness of the predicted data,the linear interpolationmethod was adopted in this paper to supplement the missing data.Ship AIS goes through multiple stages before receiving information,which willinevitably be affected by noise,thus ultimately affecting the accuracy of forecast data.Inorder to make ship data more accurate,this paper will carry out singular spectrum analysis onthe data before using neural network prediction,so as to restore the authenticity of the data tothe maximum extent and remove the parts that may be noisy.So that the prediction results aremore accurate.Considering the AIS data transmission process may be affected by noise,this paperrealizes the prediction of ship navigation behavior by constructing the combined model ofSSA-BP neural network,and by comparing the prediction analysis results with BP neuralnetwork without singular spectrum analysis,The necessity of restoring original data bydenoising and the accuracy of prediction model based on SSA-BP neural network are proved.The results have practical significance for reducing the collision risk and ensuring the safetyof shipping.Key words:Ship navigation;speed and heading;AIS data;