Improving ship energy efficiency: models, methods, and applications

Abstract

Maritime transportation is the backbone of global trade, as ships carry over 80% of trading goods worldwide. As the shipping industry is mainly powered by heavy fuel oil, it has an adverse environmental footprint due to the emissions of greenhouse gases and polluting substances. To comply with IMO emission regulations and optimally save on fuel costs (which can account up for 50% to 60% of the total cost of operating a ship), shipping companies are motivated to optimize energy consumption. In this study, we first develop am innovative and tailored artificial neural network-based fuel consumption prediction model. This model innovates in that it explicitly considers shipping domain knowledge by modifying and optimizing its structure and parameters, where such properties have rigorously been proven. Moreover, it considers a broad range of influence factors based on data fusion technology. Next, we optimize the ship sailing speed profile for a bulk carrier in two application scenarios using the predicted fuel consumption rates by the proposed neural network-based model as the input: one is a bi-objective model, and the other considers market-based measures. Numerical experiments show that the proposed fuel consumption prediction model outperforms other models and that the model we propose can help to improve ship energy efficiency by a considerable extent. The proposed model conforms more closely to common sense than existing models; thus, it will likely have a better potential for use in the maritime industry and other problems with similar domain knowledge possessed.