基于变结构径向基函数网络的船舶运动预测PID控制

Ship motion predictive PID control based on variable structure radial basis function network

中文关键词  径向基函数神经网络  变结构神经网络  预测控制  序贯学习

英文关键词  radial basis function neural networks  variable structure neural network  predictive control  sequential learning

中文摘要

      针对船舶在海上运动的大时滞和动态时变等特点, 提出基于一种变结构径向基函数(RBF)神经网络的预测PID控制器. 通过建立反映系统动态变化的滑动数据窗口, 在线序贯学习窗口内的数据, 动态调整隐层节点与隐层至输出层的连接权值, 得到结构可自适应变化的RBF网络. 将该变结构RBF网络用于预测PID控制器中系统状态的在线多步预测, 通过得到的预测模型灵敏度信息在线调整PID控制器参数以控制系统的输出. 将该控制器用于船舶航向跟踪控制的仿真实验, 结果表明该控制器具有良好的的适应性和鲁棒性.

英文摘要

      To deal with the long time-delay and time-varying dynamics of the ship motion in sea, we present a predictive PID controller based on variable structure radial basis function (RBF) network. This network performs sequential learning through a sliding data window reflecting system dynamic changes, and adjusts online the hidden layer nodes and their weighting values in the connection to output layers. We thus obtain an adaptive variable structure RBF network. This variable structure RBF network is employed as a multi-step online predictor for a predictive PID controller. Parameters of the controller are online tuned based on the sensitivity information obtained from the variable RBF network predictor. The proposed predictive PID controller is applied to ship course tracking control. Simulation results demonstrate satisfactory adaptation and robustness of the controllers.