计算机科学
人工智能
集成学习
可预测性
阿达布思
相空间
支持向量机
机器学习
模式识别(心理学)
数学
统计
物理
热力学
作者
Qin Zhang,He Daijing,Xiong Hu
标识
DOI:10.1109/iccsnt58790.2023.10334540
摘要
Prediction of ship heave motion is of great significance to wave compensation, and high-precision prediction is the most important prerequisite for wave compensation control. Therefore, a prediction analysis of ship heave motion and ensemble learning prediction strategy are proposed in this paper. First, Hurst index is calculated to analyze the predictability of time series. Secondly, the one-dimensional sequence is reconstructed into a multidimensional phase space. Finally, the number of Long Short-term memory models in phase space involved in ensemble learning is determined, and multiple phase space LSTM models with different weights are used for AdaBoost ensemble prediction to obtain multi-step heave motion prediction data. The experimental results show that the data with large Hurst index under the same sea state level are more predictable, and the multi-step prediction effect of LSTM and AdaBoost ensemble learning through phase space reconstruction is better than ARIMA, SVM, BP, and LSTM models, which can provide high-precision prediction for wave compensation of ship heave motion.
科研通智能强力驱动
Strongly Powered by AbleSci AI