Abstract: Existing methods for starting multiple induction motors mainly focus on starting timing and batch capacity, while motor starting sequences still rely heavily on production experience. Moreover, the complex network structures of mine power grids limit the applicability of traditional methods. To address these problems, a method for optimizing starting strategies of multiple induction motors in mine power grids based on an improved Decision-Making Trial and Evaluation Laboratory (DEMATEL) was proposed to objectively quantify causal relationships among motors and optimize their starting sequences accordingly. First, a simulation model for starting multiple induction motors in a regional mine power grid was established. Simulations were conducted for each pair of motors over a range of starting intervals to obtain four evaluation indices: motor starting time, voltage sag energy, minimum per-unit voltage, and maximum inrush current. These indices were then aggregated and weighted to construct a direct influence matrix based on objective simulation data. A total influence matrix and each motor's centrality and causal degree were subsequently calculated. Finally, motor starting sequences for emergency and routine starting scenarios were optimized based on centrality and causal degree, with further adjustments accounting for process constraints and key motors. The case study results showed that, compared with the experience-based starting strategy, the emergency starting strategy optimized using improved DEMATEL reduced the total starting time from 33.22 s to 30.39 s, a reduction of 8.5%, and voltage sag energy from 4.069 4 s to 3.108 9 s, a reduction of 23.6%. The routine starting strategy optimized using improved DEMATEL increased the minimum per-unit voltage from 0.848 3 pu to 0.883 8 pu, an increase of 4.2%, and reduced the maximum inrush current from 1 761.83 A to 1 569.09 A, a reduction of 10.9%. These results demonstrate the effectiveness of improved DEMATEL in optimizing multiple motor starting strategies.