This paper presents a novel approach to determine both the exponential and ZIP load model parameters in electrical power system substations. The proposed approach aims to estimate the polynomial and exponential parameters for active and reactive power loads. Field measurements are obtained from a rigorous testing script to impose sustained voltage variations, and the parameters are estimated by using a constrained least square optimization technique. The testing script to acquire data is applied at a distribution substation from CEMIG (Distribution Electricity Company of Minas Gerais) in Brazil. From these data set, polynomial and exponential load models are determined, and comparisons between the measured and estimated active and reactive power curves are presented to show the effectiveness of the proposed methodology.