基于BP网络的船舶动力定位控制技术仿真研究  

Simulation of Dynamic Positioning System Based on BP Neural Networks

作　　者：丁福光[1] 朱增帅 王奇[1] 

机构地区：[1]哈尔滨工程大学自动化学院,黑龙江哈尔滨150001

出　　处：《计算机仿真》2014年第10期405-409,449,共6页Computer Simulation

摘　　要：在船舶动力定位控制优化问题的研究中,为了解决第一代动力定位系统中PID控制器参数整定困难的缺点,提高动力定位的实时性、稳定性和可靠性,提出了将BP网络与PID控制器相结合的控制策略,使系统具有PID控制器简单可靠等优点及BP网络的自学习和自适应能力等特性。采用三层BP网络,根据梯度最速下降法,不断修正各层神经元之间的连接权值,在线调整PID控制器参数,得到满足定位要求的最佳参数组合,提高系统的快速性和定位精度。最后以某动力定位船为研究对象,在MATLAB平台上对控制器进行定位功能的仿真验证,仿真结果表明所设计的控制器具有较好的控制性能。To slove the difficulty of parameter tunning in PID controller for the ship dynamic positioning system and to improve the performance of real - time, stability and reliability, a new controller is presented which combines PID controller with BP neural networks. The proposed controller has the characteristics of both PID controller and BP neural networks, such as simple structure, high reliability, self - learning and self - adaptive. A three - layer BP network is applied in this paper. By using gradient steepest descent method, the weight coefficients of the neural net- work are modified to adjust the parameters of PID controller online. It can make the controller achieve the best pa- rameters to improve the real - time performance and positioning precision. Finally, the PID controller based on BP neural network is simulated in MATLAB platform for a dynamic positioning ship. The results verify that the proposed controller has good control performance.

关 键 词：动力定位 比例-积分-微分控制 神经网络 梯度最速下降法 

分 类 号：TP183[自动化与计算机技术—控制理论与控制工程]