Periodical validation and calibration to microgrid models are necessary for investigating the behaviors of microgrids and evaluating the stability of connected power grids. However, instead of calibrating the problematic parameters that vary from their actual values, the commonly used methods normally adjust the selected sensitive parameters. The performance of the calibrated models is still in doubt. Therefore, this article proposes a generic approach for microgrid model calibration, which aims to accurately locate the indeed 'ill-conditioned' problematic parameters, as well as address the multiple solutions issue. With the application of sensitivity and correlation analysis, the potentially problematic parameters (PPPs), which are sensitive and independent, are selected to ensure the accuracy and uniqueness of parameter calibration results. Then, the problematic parameters are further screened out by comparing the energy distribution of PPPs and model errors in the frequency domain. Also, hybrid dynamic simulation based parameter estimation is employed to adjust the problematic parameters, by which each device in microgrids can be calibrated independently. Finally, the simulation results demonstrate the outstanding performance in parameter calibration of a real microgrid.