摘要
为取得更有效的船舶运动预报效果,提出了一种利用遗传算法(GA)优化单输出三层反向传播(BP)神经网络辨识Volterra级数核的算法。在船舶航行姿态时间序列的混沌特性识别基础上,分析了GA、BP神经网络和Volterra级数模型的特征。利用GA优化BP神经网络获得最优的初始权值和阈值,根据BP神经网络算法求得最终的最优权值和阈值。进行Taylor级数分解,得到Volterra级数各阶核,对船舶的横摇运动时间序列进行多步预报。仿真实验表明:所提方法预报精度高、时间长,具有有效性和适应性。
In order to obtain more effective prediction results of ship motion, a method is proposed using the genetic algorithm (GA) optimized single-output three-layer back propagation (BP) neural network to identify Volterra series kernels. The GA, the BP neural network and the features of the Volterra series model are further analyzed based on the chaos characteristic identification of ship motion attitude time series. The best initial weights and thresholds are obtained by using the GA optimized BP neural network. The final optimal weights and thresholds of model parameters are obtained by the BP neural network algorithm. The multi-step prediction of the time series of a ship roll motion is done by making Taylor series decomposition to obtain Volterra series kernels of each order. The simulation experiments show that the proposed algorithm has high precision and long prediction time and effectiveness and adaptability.
出处
《南京理工大学学报》
EI
CAS
CSCD
北大核心
2012年第6期962-967,共6页
Journal of Nanjing University of Science and Technology
基金
黑龙江省科学基金(QC2011C011)
关键词
遗传算法
反向传播神经网络
混沌特性识别
船舶运动
多步预报
genetic algorithm
back propagation neural network
chaos characteristic identification
ship motion
multi-step prediction