This paper presents a bi-level optimal allocation for voltage-sag monitors with the consideration of fault locating and disturbance tolerance ability. In the first level, a binary linear programming model is proposed for both symmetrical and asymmetrical faults. A binary gravity search algorithm (BGSA) is applied for solving the binary optimization problem. In the second level, to select the optimal allocation among all feasible solutions, the disturbance tolerance ability is modeled and quantified based on fuzzy inference. Allocation-level disturbance tolerance indexes are then obtained to determine the ultimate monitoring allocation. The IEEE 39-bus test system is used for the validation of proposed method.