According to the characteristics of fuel transportation problem, the traditional genetic algorithm model is improved in this paper. The complexity of encoding is simplified by considering the condition of putting the distances of the tanker going halfway back and forth into the objective function. Scanning method is used to generate the initial population improving the quality of chromosomes in the initial population. Adopting the way of "interval crossover, random replacement" ensures the effectiveness and randomness of the crossover. Adding the operation of evolutionary cycle after crossover...