基于神经网络的船舶运动建模及随机最优控制

Modeling of Ship Motion Based on Artificial Neural Network and Stochastic Optimal Control

以某水面舰艇为研究对象,利用径向基函数神经网络算法和标况下的水池实验数据,建立了基于航速、航向角和海情自适应变化的船舶横向运动非线性参数模型。然后对海浪随机干扰建模,利用成形滤波器将随机扰动白化处理,最后根据分离原理提出一种用于船舶减横摇运动的随机最优控制算法。仿真表明,径向基函数神经网络所建船舶横向运动模型误差低于2%,扰动建模及相应成形滤波器构造符合实际工况且方法简单可行,随机最优控制可使横摇角均方误差达1.42°,减摇效果达46%-74%;艏摇角均方误差达1.68°。

An intelligent model for a ship＇s horizontal motion, which can self-adapt with navigating speed, ocean condition and course, was established based on the Radial Basis Function （RBF） neural network. The detail training process of RBF neural network was proposed, and the model of random disturbance caused by ocean wave was set up and the colored disturbance with a shape filter was whitened. Finally, based on separation theory, a stochastic optimal algorithm was provided to reduce the horizontal motion. The simulation indicates that modeling with radial basis function network is fleet and accurate, and the model＇s error is below 2%; the method of disturbance modeling and the process in shape filter＇s constructing are practical and fit the real working condition; the mean square error of roll angle reduces to 1.42°; the control effect is about 46%-74%; the mean square error of yawing angle reduces to 1.68°.