改进二阶灰色极限学习机在船舶运动预报中的应用

Application of Improved Second-Order Grey Extreme Learning Machine in Ship Motion Forecasting

为提高船舶运动预报的精度,基于海上船舶运动姿态具有灰色特性和周期性振荡特性的特点,提出一种以误差平方和最小为准则的改进二阶灰色极限学习机组合预测模型,对船舶运动姿态进行预报。该方法利用五点三次平滑算法对船舶运动姿态序列进行平滑降噪,采用余弦函数变换构建GM(2,1)预测模型;利用自适应粒子群算法(Adaptive Particle Swarm Optimization,APSO)优化极限学习机权值和阈值参数,对不同模型预测结果进行加权求和,构建改进二阶灰色极限学习机组合预测模型。对2组船模水池试验纵摇时历进行预报,并将其与其他传统的预测方法相比较,结果表明,建立的组合预测模型具有更好的预测精度和泛化能力。

Ship motion attitude has characteristics of grey and periodic oscillation, based on which an improved second-order grey extreme learning machine combination prediction model in minimum error sum of squares for precise ship motion prediction is proposed. The ship motion attitude sequence is smoothed by means of five-point cubic smoothing algorithm to reduce the noise and the GM(2,1) prediction model is constructed through cosine function transformation. The weight and threshold parameters of the extreme learning machine are optimized through APSO(Adaptive Particle Swarm Optimization). An improved second-order grey extreme learning machine combination prediction model is constructed by weighted summation of predicted results from different models. Experiments of pitching prediction in two model tests show that the combined prediction model has better prediction accuracy and generalization ability than traditional prediction methods.