| 摘 要: 涂装是修造船企业最大的能耗单元，能耗预测是船舶智能能效优化中的一项重要任务。应用XGBoost模型对船舶特涂工序能耗进行分析。采用基于博弈论的SHAP(解释机器学习模型输出)方法解释变量因子对目标准确预测的影响。利用粒子群算法(PSO)优化XGBoost模型超参数，从而构建PSO-XGBoost模型对船舶特涂工序能耗历史数据进行训练拟合，并与其他能耗预测模型进行对比实验。结果表明，基于PSO-XGBoost的船舶特涂工序能耗预测模型的预测结果误差MAPE仅为12.21%，效果优于XGBoost、LR、KNN、RF模型。 |
| 关键词: 船舶特涂；能耗预测；SHAP；PSO-XGBoost模型 |
|
中图分类号: TP31
文献标识码: A
|
| 基金项目: 深圳市技术攻关面上项目(JCYJ20190809145407809)；深圳信息职业技术学院校级创新科研团队项目(TD2020E001). |
|
| Energy Consumption Prediction of Tank Coating Process of Ship based on PSO-XGBoost |
|
JIANG Qinghua1, REN Xinmin1, JIANG Jun1, OUYANG Bin2, PENG Bao3
|
( 1.You Lian Dockyards (Shekou) Co., Ltd., Shenzhen 518067, China ;
2.Hunan University of Technology and Business, Changsha 410205, China ;
3.Shenzhen Institute of Information Technology, Shenzhen 518172, China)
renxinmin@cmhk.com; jiangqinghua0115@163.com; jiangjun8880@163.com; 2248918560@qq.com; pengb@sziit.edu.cn
|
| Abstract: Tank coating is the largest energy consumption unit in shipbuilding enterprises, and energy consumption prediction is an important task in intelligent energy efficiency optimization of ships. This paper proposes to use XGBoost model to analyze the energy consumption of ship tank coating process. SHAP (SHapley Additive exPlanations) method based on game theory is used to explain the influence of variable factors on accurate target prediction. Particle swarm optimization (PSO) is used to optimize the super-parameters of XGBoost model, and then PSO-XGBoost is constructed to train and fit the historical energy consumption data of ship tank coating process. The comparison experiment is conducted with other energy consumption prediction models. Experiment results show that the prediction error MAPE of energy consumption model of ship tank coating process based on PSO-XGBoost is only 12.21%, which is better than XGBoost、LR、KNN、RF models. |
| Keywords: tan coating of ships; energy consumption prediction; SHAP; PSO-XGBoost model |
