Estimation of ship operational performance degradation using deep-learning-based fuel oil consumption prediction models

This paper proposes a novel method for estimating ship operational performance degradation (SOPD) using a fuel oil consumption (FOC) prediction model based on deep neural networks with shortcut connections. The model leverages operational and environmental data from a crude oil tanker over a 21-month period to predict FOC and assess SOPD. A cumulative anchoring effect is introduced as a new feature of the FOC prediction model, capturing the impact of biofouling caused by prolonged anchoring in warm waters. In this study, SOPD is considered the additional fuel rate required for a journey leg due to degradation, which is estimated by comparing predicted FOC with and without the cumulative anchoring effect. The SOPD estimation is illustrated according to increasing journey legs based on the FOC prediction models. The proposed SOPD estimation method provides valuable insights for shipping companies to optimize operational schedules and improve fuel efficiency. • A new feature for cumulative anchoring effects is introduced. • Fuel oil consumption (FOC) is predicted using a deep neural network. • Ship operational performance degradation is estimated using the FOC prediction model. • The degradation is considered an additional fuel consumption rate.