In this paper, an improved maximum loading point (MLP) calculation method is proposed. The calculation process is based on the properties of the normal vector to the feasibility boundary computed close to MLP, which is calculated by a load flow method with step size optimization (LFSSO). The process is characterized by obtaining consecutive approximations of the MLP within the infeasible region. Since the feasibility boundary contour in the neighborhood of the MLP may not be smooth, some of the computed approximations may fall within the feasible region. In an earlier paper, a mechanism based on binary search was used to drive the operating point back to the infeasible region. A new load curtailment method to improve the proximity towards feasibility boundary guaranteeing that the next solution will lay within the infeasibility region is the main contribution of this paper. Reactive power generation limits are taken into account. The proposed formulation requires information as the normal vector and power mismatches, and results in a better convergence path towards MLP in comparison with the original version of the method presented earlier. Simulation results for IEEE test systems are shown to validate the proposed method. © 2009 IEEE.