To deal with extreme overvoltage scenarios with small probabilities in regional power grids, the traditional reactive power planning model requires a huge VAR compensator investment. Obviously, such a decision that makes a large investment to cope with a small probability event is not economic. Therefore, based on the scenario analysis of power outputs of distributed generations and load consumption, a novel reactive power planning model considering the active and reactive power adjustments of distributed generations is proposed to derive the optimal allocation of VAR compensators and ensure bus voltages within an acceptable range under extreme overvoltage scenarios. The objective of the proposed reactive power planning model is to minimize the VAR compensator investment cost and active power adjustment cost of distributed generations. Moreover, since the proposed reactive power planning model is formulated as a mixed-integer nonlinear programming problem, a primal-dual interior point method-based particle swarm optimization algorithm is developed to effectively solve the proposed model. Simulation results were conducted with the modified IEEE 30-bus system to verify the effectiveness of the proposed reactive power planning model.