Analysis of energy consumption in trucks using data mining method

Fuel economy (FE) which is the inverse of fuel consumption plays an important role for firms in the transportation sector. With an increase in fuel efficiency (km/l or l/ton-km), marginal fuel consumption decreases resulting the reduced transportation cost. In this research, apart from factors that affect the FE that have to be considered by the decision tree technique, the technique of random forests is developed using the oil trucks data in order to obtain the forecasted amount of energy consumption used for each trip and in each region. The result shows that when the truck is loaded, the lowest fuel consumption rate occurs in the Southern region 3.4418 km/L, while it shows the highest fuel consumption 3.7908 km/L when the no-load truck operates across the border between Lao PDR. In this research also propose a method for a routing and scheduling of fuel trucks to minimize energy consumption. A multi-trip vehicle routing problem with a soft time window (MVRPSTW) was addressed and differential evolution (DE) algorithm was developed for the problem. The computational work revealed that the DE and current practice procedure are significantly different, with 95% confidence interval. As a result, the proposed DE algorithm can handle various realistic situations of vehicle routing problem. The heuristic performance (HP) obtaining an average of 118.05% while the relative improvement (RI) had an average of 17.88%.