基于动态数据驱动模型的船舶运动实时预报与航行优化研究

The real-time perception of a ship’s status, natural and navigational environments as well as the intelligent navigation decision is an important component of ship intelligent. A real-time ship motion identification and prediction approach is proposed based on dynamic data-driven model. A sliding data window (SDW) with adaptive width is employed as online ship status observer. The variable-structure dynamic data-driven model, such as neural network and fuzzy inference system, is thus established based on the ship status observer served by a SDW whose width being tuned online. The grey-box online ship motion prediction model is established by combining the variable-structure data-driven model and ship motion mechanism model. A dynamic ship multivariate domain model is established based on the online ship status prediction and the real-time perception of natural environment and navigational environment. The multi-objections of ship’s route and sailing speed are optimized based on improved particle swarm algorithm. The online ship motion identification and prediction model are established to represent the ship motion characteristics such as nonlinearity, uncertainty and time-varying dynamics. And the ship route and sailing speed are optimized based on the intelligent ship perceptions. The project aims at improving the level of ship’s safety, efficiency and environmental protection under complex environment.

对船舶动态、自然环境及通航环境的实时感知以及智能航行决策是船舶智能化的重要组成部分。本项目提出一种基于动态数据驱动模型的船舶运动在线辨识和预报方法，利用宽度可自适应调整的滑动数据窗口作为船舶运动状态的在线观测器，在线调整模型结构和参数以构造可变结构的神经网络或模糊推理系统等动态数据驱动模型；将该动态数据驱动模型与船舶运动的机理模型结合构建灰箱预报模型，对船舶海上运动的时变动态进行在线辨识和实时预报。在对船舶运动实时预报的基础上，通过对船舶所处的时变自然环境和通航环境的实时感知，建立船舶多元动态领域模型，并利用改进的粒子群算法等优化算法调整船舶航路和航速进行多目标实时优化。本项目旨在研究适合船舶海上运动非线性、不确定性和动态时变等复杂特性的船舶运动在线辨识和预报模型，并结合智能船舶感知进行船舶航行多目标优化，提高海上复杂环境条件下船舶航行的安全、效率和环保水平。

通过对船舶动态、自然环境及通航环境的实时感知，并在此基础上对船舶运动预进行报以及对航线进行优化，可优化船舶航行性能，保证航行安全，提高航行效率和节能减排，同时对于船舶运动控制系统和动力定位系统都是不可或缺的。本项目利用宽度可自适应调整的滑动数据窗口作为船舶运动状态的在线观测器，通过时间序列在线分割算法动态调整滑动状态观测器窗口宽度，提出一种动态数据驱动的船舶海上运动在线辨识和预报方法，通过同时在线调整模型结构和参数以构造可变结构的动态数据驱动智能并行计算模型（模糊推理系统或神经网络），对船舶海上运动的时变动态进行在线辨识和实时预报。并研究对海洋环境要素的数据驱动实时预报，以提高船舶运动预报的精度。在对船舶运动实时预报的基础上，通过对船舶所处的时变自然环境和通航环境的实时感知，利用改进的智能优化算法，进行多目标实时优化。本项目研究适合船舶海上运动非线性、不确定性和动态时变等复杂特性的船舶运动在线辨识和预报模型，并结合智能船舶航行和通航态势感知进行船舶航行路径、航行速度以及船舶温室气体排放的智能化多目标优化。项目研究可为提高海上复杂环境条件下船舶航行的安全、效率和环保水平提供参考。