基于LightGBM的船舶航速预测模型

Ship speed prediction model based on LightGBM

针对现有基于机器学习算法的船舶航速预测模型无法兼顾计算精度高、泛化能力强及计算速度快的问题,提出基于LightGBM的船舶航速预测模型,并以一艘安装有能效监测系统的内河船舶为研究对象,运用LightGBM算法建立以实时风速、风向、水深、水流速度、尾轴转速、轴功率和主机油耗为输入的船舶航速预测模型,并同时与RR、SVR、DT、BPNN、RF、GBDT和XGBoost七种机器学习算法的航速预测结果进行比较。结果表明:基于LightGBM建立的船舶航速预测模型的精度、泛化能力、运算速度均排名第二,综合性能最好,可在保证较高预测精度和较强泛化能力前提下,实现对船舶航速的快速预测。

Aiming at the problem that the existing ship speed prediction model based on machine learning algorithm was impossible to balance the high calculation accuracy,strong generalization ability and fast calculation speed,a ship speed prediction model based on LightGBM(Light Gradient Boosting Machine)was proposed.Taking an inland river ship equipped with an energy efficiency monitoring system as the research object,the ship speed prediction model with real-time wind speed,wind direction,water depth,water speed,tail shaft speed,shaft power and main engine fuel consumption as inputs was established by using LightGBM algorithm,and compared with the speed prediction results of seven machine learning algorithms as RR(Ridge Region),SVR(Support Vector Region),DT(Decision Tree),BPNN(Back Propagation Neural Network),RF(Random Forest),GBDT(Gradient Boosting Decision Tree)and XGBoost(Extreme Gradient Boosting)at the same time.The results show that the ship speed prediction model based on LightGBM ranks second in accuracy,generalization ability and calculation speed with best comprehensive performance,which can realize fast prediction of ship speed on the premise of ensuring high prediction accuracy and strong generalization ability.