一种改进的ARIMA预测模型在船舶系统中的应用  

Application of Improved ARIMA Prediction Model to Ship Systems

作　　者：郑卓 陈奥 李越曌 ZHENG Zhuo;CHEN Ao;LI Yuezhao(Marine Design&Research Institute of China,Shanghai 200011,China)

机构地区：[1]中国船舶及海洋工程设计研究院,上海200011

出　　处：《船舶与海洋工程》2023年第6期29-34,55,共7页Naval Architecture and Ocean Engineering

摘　　要：为实现船舶设备维护方式的智能化升级,用视情维修代替传统的定期巡检,提出一种基于改进粒子群算法(Particle Swarm Optimization,PSO)优化的Elman神经网络融合自回归差分移动平均模型(Auto-Regressive Integrated Moving Average,ARIMA)模型的组合预测模型,用于对设备状态参数进行预测。根据序列特征拟合ARIMA模型,利用改进的PSO算法优化Elman神经网络的权值和阈值,基于改进的PSO-Elman模型的残差预测值修正ARIMA模型预测结果。采用某船设备实际数据对该组合预测模型进行训练和验证,将其预测结果与其他模型的预测结果相对比,结果表明,该组合预测模型具有较高的预测精度和稳定性。In order to realize the intelligent upgrade of equipment maintenance and enable the condition-based maintenance to replace the traditional regular inspections,an Auto-Regressive Integrated Moving Average(ARIMA)model with Elman neural network fusion based on improved Particle Swarm Optimization(PSO)is proposed for equipment status parameters prediction.The ARIMA model is fitted according to the sequence characteristics,the weights and thresholds of the Elman neural network are optimized by the improved PSO algorithm and the ARIMA prediction results are corrected based on the residual prediction value of the improved PSO-Elman model.The data from a ship's equipment is used to train and verify the proposed prediction model and the result is compared with other models.The results show that the proposed prediction model has high accuracy and stability.

关 键 词：时间序列 ARIMA模型 粒子群算法 ELMAN神经网络 残差修正 

分 类 号：U672.7[交通运输工程—船舶及航道工程]