Scholarlink

Accurate fuel consumption predicting can assist ship managers in making more informed decisions regarding route selection and speed adjustments, thereby enhancing shipping efficiency. Due to the characteristics of data features and the attributes of data-driven methods, it is challenging to find a single energy consumption prediction method suitable for all data samples. Therefore, it is necessary to analyze the performance of different forecasting methods based on the specific characteristics of energy efficiency data and research objectives, selecting the optimal method for fuel consumption prediction. This study utilizes shipboard noon report data and employs Least Absolute Shrinkage and Selection Operator(LASSO), Back Propagation Neural Network (BPNN), Decision Tree Regression (DTR), and Support Vector Regression (SVR) for fuel consumption predicting. Experimental results indicate that LASSO achieves the highest prediction accuracy, with a root mean square error of 1.512 Metric Ton(MT)/day, an average absolute percentage error of 2.801%, and a prediction accuracy of 97.199%. The proposed ship energy consumption prediction method provides valuable references for fuel anomaly detection, energy efficiency assessment, and energy efficiency optimization.

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Machine Learning for Ship Energy Consumption Predicting Based on Noon Report Data

  • 会议论文

  • 发表日期：

    2025年

查看全文
  • Tianrui Zhou,
  • Linghong Wang

摘要

Accurate fuel consumption predicting can assist ship managers in making more informed decisions regarding route selection and speed adjustments, thereby enhancing shipping efficiency. Due to the characteristics of data features and the attributes of data-driven methods, it is challenging to find a single energy consumption prediction method suitable for all data samples. Therefore, it is necessary to analyze the performance of different forecasting methods based on the specific characteristics of energy efficiency data and research objectives, selecting the optimal method for fuel consumption prediction. This study utilizes shipboard noon report data and employs Least Absolute Shrinkage and Selection Operator(LASSO), Back Propagation Neural Network (BPNN), Decision Tree Regression (DTR), and Support Vector Regression (SVR) for fuel consumption predicting. Experimental results indicate that LASSO achieves the highest prediction accuracy, with a root mean square error of 1.512 Metric Ton(MT)/day, an average absolute percentage error of 2.801%, and a prediction accuracy of 97.199%. The proposed ship energy consumption prediction method provides valuable references for fuel anomaly detection, energy efficiency assessment, and energy efficiency optimization.