In response to the problems of slow solving speed and difficult parameter selection in the current proposed dynamic reactive power compensation device location and capacity determination strategies, a dynamic reactive power compensation device location and capacity determination strategy was proposed based on mixed integer second order cone programming (MISOCP). This strategy first established a mixed integer nonlinear programming (MINLP) optimization model with the objective function of minimizing active power loss and node voltage deviation during the optimization cycle of the distribution network; Secondly, the MINLP model was transformed into a MISOCP model through a two-step relaxation method of phase angle relaxation and second-order cone relaxation; Then, the MISOCP model was transformed into a mixed integer linear programming (MILP) model by using theε-relaxation method, and a commercial solver was called to solve the problem; Finally, tests were conducted in the distribution systems of IEEE 33 and IEEE 69 nodes, using model solving time, active power loss, and node voltage deviation values as evaluation indicators. The results were compared with the methods of directly solving the MISOCP model using solvers, particle swarm optimization (PSO), and simulated annealing particle swarm optimization (SA-PSO) to directly solve the MINLP model. The simulation results show that the proposed method has significantly better model solving time and effectiveness than other methods, verifying the feasibility and effectiveness of the proposed method. The proposed multi part relaxation method simplifies the difficulty of model solving while ensuring the optimal solution, and shortens the time of model solving, which provides an effective basis for the installation of reactive power compensators.