Machine Learning-Based Ship Energy Efficiency Algorithms

This paper establishes fuel consumption prediction models using the Random Forest algorithm (RF) and Deep Neural Network (DNN) algorithm, with the models’ performance evaluated based on Mean Square Error (MSE) and the coefficient of determination (R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup>). On this basis, a speed optimization model is developed using a DQN algorithm to dynamically adjust speed for minimizing fuel consumption. The evaluation of the model is conducted through the accumulation of reward values. The results show that the fuel consumption prediction model built using the RF algorithm achieves an MSE of 0.00064 and an R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> of 0.99781. The model constructed with the DNN algorithm yields an MSE of 0.00516 and an R<sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> of 0.99799. Comparatively, the RF algorithm slightly outperforms the DNN. Finally, the speed optimization model developed using the DQN algorithm learns the optimal policy through interactions with the environment to maximize the expected accumulated reward, ultimately enhancing energy efficiency by optimizing speed in the given environmental data.