Predictive analysis of engine power limitations for fuel reduction in a tanker ship using a rule-based machine learning technique

Abstract

To ascertain the most optimal engine power limitation (EPL) rate to meet future emission standards, as well as the latest energy efficiency indexes, this research intends to construct a hybrid M5 rules- and linear-regression fuel prediction model This research introduces a novel machine learning approach that combines the computation of the Energy Efficiency Existing Index (EEXI) and the Carbon Intensity Indicator (CII) to evaluate potential Engine Power Limitation (EPL) applications on tanker ships, contributing novelty to current literature. The data required to train the model were gathered from the logbooks and noon reports of an oceangoing tanker ship currently employing a slow-steaming procedure. The M5 rules algorithm has been trained and optimized to predict fuel consumption with EPL-applied scenarios. Regression models have been built to determine the parameters that change concerning the vessel's speed. Various EPL ratios have been examined and compared to scenarios without EPL or slow-steaming. Results demonstrate that the model predicts fuel usage satisfactorily, indicating that the algorithm's structure is appropriate for this case. Slow steaming alone cannot meet the EEXI restrictions. Additional application of EPL rates, starting at 32 %, has ensured compliance with EEXI requirements and improved CII ratings.