Computational meta-heuristics based on Machine Learning to optimize fuel consumption of vessels using diesel engines

Abstract

With the expansion of river transportation, especially in the case of small and medium-sized vessels that make longer routes, the cost of fuel, if not taken as an analysis criterion for a larger profit margin, is considered a primary factor, considering that the value of fuel, specifically diesel, to power internal combustion engines is high. Therefore, the use of tools that assist in decision-making becomes necessary, as is the case of the present research, which aims to contribute with a computational model of prediction and optimization of the best speed to decrease fuel cost, considering the characteristics of the SCANIA 315 propulsion model, of a vessel from the river port of Manaus that carries out river transportation to several municipalities in Amazonas. According to the results of the simulations, the best training algorithm of the Artificial Neural Network (ANN) was the BFGS Quasi-Newton, considering the characteristics of the engine for optimization with Genetic Algorithm (GA).