In view of the fact that the moth-flame optimization(MFO) algorithm is poor in solving large-scale optimization problems, the direct flight model of moths is proposed based on the mutation idea of differential evolution algorithm(DE). The concept of the sight distance factor and the scaling factor is used, and the flame around the historical optimal moth and the current random moth is defined as local development and global exploration. Three different types of the sight distance factor are designed to guide the search algorithm and the global exploration and local development from the macroscopic point of view are started. The influence of different start-up time on the precision of the MFO algorithm in the large-scale problem is analyzed, and the idea that different optimization problems have different start-up opportunity for local development. The optimization efficiency of the three combined strategies is discussed on the model of direct flight and helical flight, and the better performance of the proposed algorithm is verified. Numerical experiments vprify the superiority and robustness of the improved algorithm, compared with the existing literature on large-scale optimization problems.