Accurate load modeling is essential for power systems dynamic simulation. It has been observed that different load models have different, even contradictory impacts on simulation results especially for dynamic voltage stability studies. Load modeling includes two tasks: determining a suitable model structure of a load bus and deriving parameter values for the associated model structure. Two approaches are used for identifying the load model parameters, namely, the component-based method and the measurement-based method. Nowadays, parameters estimation of the load models is based on the field measurement since it is more accurate than the component-based approach. This thesis focuses on chosing an appropriate load model for short-term voltage stability studies. In this thesis, the parameters estimation of a composite load model (CLOD) is investigated through the measurement-based approach. Different optimization algorithms for the parameter estimation of the CLOD model are presented and analyzed.‌ ‌To do so, a nonlinear optimization problem is formulated and a suitable solution is presented. For this work, an error between active and reactive power data measured from the real system and those computed from the model system is used to evaluate the accuracy of the load model. Based on the optimization results; the optimization method has an acceptable performance. ‌‌Measurement noises are also takan into account. These noises can come from different sources, such as: loads turning on and off and measurement errors of the sensors. We evaluated the parameters estimation algorithm withoisy data, it can be seen thatfor a specific noise level, the proposed method is robusttomeasurement noise.Howvevr for this thesis, a nonlinear filter was chosen to reduce measurement noises.‌ Keywords: Load Modeling, Short-Term Voltage Stability, Parameter Estimation, Measurement-Based Approach, Measurement Noise