To acquire raw materials and distribute products, agri-food processing companies need to operate at long distances, which is associated with high energy costs and an increased negative impact on the natural environment. Many methods have been developed to reduce energy consumption in transport. One approach that stands out is optimization of transport routes, which provides benefits while incurring hardly any extra costs. In this paper, an attempt was made to modify the ILS-RVND metaheuristic by adjusting its components to the specific character of the problem being solved so that they could yield the expected results with regard to the quality of returned solutions and the time in which they were generated. A group of local search algorithms, Swap (2-1), Cross-exchange (2-1), 3-opt, Or-opt (2) and Displacement, were analyzed. The results of this analysis were used to formulate the final version of the metaheuristic. The efficiency of the algorithm was evaluated using test cases. The solutions generated by the metaheuristic produced considerable improvement in the objective function (70.99% on average) and were obtained within an acceptable time (on average 24.66 CPU seconds).