Strategy for ship energy efficiency based on optimization model and data-driven approach

This paper proposes an innovative approach to evaluate and enhance the ship's energy efficiency using a machine learning (ML) technique. A methodology is proposed combining an engine optimization model and a data-driven approach to monitor and predict the performance of the ship along the route. The starting point is a dataset of noon report data covering significant operational information corresponding to the main engine. Then, an engine model is developed, calibrated and validated in a 1D engine simulation software to compute the engine performance for different speeds and loads. After the simulation process is performed, operational parameters are derived from the developed engine model such as volumetric efficiency, scavenging ratio, brake mean effective pressure, and brake power according to the noon report of the ship. Lastly, a data-driven adaptive neuro-fuzzy inference system (ANFIS) model is developed to estimate brake power as a main parameter affecting fuel consumption. Various ANFIS configurations are created and evaluated by error metrics to determine the best structure of the ML model. The proposed strategy provides an effective approach to maritime companies for monitoring, controlling, and improving ships' energy efficiency status based on the noon report data that is already collected onboard.