Research On Optimal Control For Dynamic Positioning Of Vessels Based On Adaptive Dynamic Programming

The sea is rich in biological,oil,gas and mineral resources.Since the 1960 s,the world began to pay attention to the exploitation of marine resources,and marine resources have become an important part of the economic development of each country.In order to effectively explore and develop the ocean,dynamic positioning of marine vessels has been proposed and widely used in the positioning operations of drilling vessels,pipe laying vessels,supply vessels,offshore platforms and other tools.Dynamic positioning of vessels means that the vessel can resist the influence of external disturbance by using its propulsion device and keep in a certain target position on the sea with a certain attitude or accurately track a target track without using the mooring system.The control strategy used in the dynamic positioning system is related to the performance of the vessel,also affects the economy and environmental protection of the vessel operation.It is a challenging problem of nonlinear dynamic positioning system optimal control to reduce energy consumption and loss on the basis of ensuring the normal operation of dynamic positioning system.Adaptive/approximate dynamic programming(ADP)optimal control method based on reinforcement learning can estimate the cost function by using a function approximation structure(such as neural networks,fuzzy logic systems,polynomial,etc.).It can effectively solve the optimal control problem of nonlinear systems.Therefore,it is of great theoretical significance and practical application value to study nonlinear optimal control for dynamic positioning systems based on ADP.Based on the ADP,this thesis studies the problems of uncertainties of model,time-varying marine environmental disturbance and unmeasured velocities,as well as the problem of reducing the time consuming of neural network-ADP.The main research contents are as follows.(1)For the problem of discrete time dynamic positioning nonlinear optimal control of fixed-point positioning with uncertain model,a model network is established by using neural network(NN),the state and control information of the vessel is used as input and output to train the model network.The dynamic positioning nonlinear system with uncertainties is identified.On this basis,the critic network and the action network are established respectively.The weight update laws of the two networks are designed by the current and historical data of the dynamic positioning system to approximate the optimal cost function and control.The proposed dynamic positioning optimal control strategy can keep the vessel’s ideal position and heading angle while guarantee the uniform ultimate boundedness of all signals in the closed-loop system.Simulation results verify the effectiveness of the proposed control scheme.(2)Further,in order to reduce the time consuming of the optimal control method of dynamic positioning based on traditional NN-ADP,firstly,the BLS is used to construct the model-critic-action structure,a BLS-based identifier is established as the model network to obtain the unknown discrete time nonlinear dynamic positioning system.Pseudo inverse method and ridge regression method are used to directly calculate the weight of the model network without iteration.Secondly,the critic network and action network based on BLS are constructed to approximate the optimal cost function and controller of the identified dynamic positioning system.Finally,the simulation results show that the BLS-ADP based dynamic positioning optimal controller can keep the vessel in an ideal attitude.Compared with the dynamic positioning optimal control based on traditional NN-ADP,the controller based on BLS-ADP can not only save time,but also minimized the performance index of the dynamic positioning control system.(3)For the continuous time dynamic positioning optimal tracking control problem with unmeasured velocity and modeling error,firstly,a NN state observer is designed to estimate the unknown velocity and approximate the modeling error.Secondly,a NN is used to approximate the modeling error of the dynamic positioning system.Then,a steady-state controller is proposed by using the vectorial backstepping method.By the designed steady-state controller,the dynamic positioning system can be transformed into a nonlinear error system.An optimal controller is designed for the error system with the ADP technology of single network adaptive critic.Based on the Lyapunov stability theory,it is proved that the proposed optimal control method can ensure that all the signals in the closed-loop dynamic positioning system are bounded.Finally,the simulation results demonstrate the effectiveness of the proposed method.(4)Further,to solve the problem of robust adaptive optimal tracking control of continuous time dynamic positioning with the unknown time-varying ocean disturbances,unmeasured velocities and the modeling errors,firstly,a disturbance observer with a NN state observer are designed to estimate unknown time-varying ocean disturbances and unmeasured velocities simultaneously.The modeling error of dynamic positioning system is approximated by a NN.Then,a steady state controller with disturbance compensation term is designed and a nonlinear error system is constructed by using vector backward inference method.An optimal controller is designed for the error system by using single network adaptive critic structure.This controller overcomes the problems of unmeasured velocities and unknown environment disturbances which are less considered in the optimal control of dynamic positioning and is more suitable for practical application.Finally,the simulation results show the effectiveness of the proposed control scheme.