A Framework for Monte Carlo Power-Plant Parameter Estimation

Power-plant model validation using phasor measurement unit recorded disturbance data has many challenges, such as quantitatively determining a parameter’s identifiability and estimation accuracy. Monte Carlo methods could answer these questions but are unrealistic to implement on an active power-plant. This paper presents a method for generating data sets for a Monte Carlo simulation from an initial event recording along with its preceding ambient data. Coherency analysis of the ambient data is performed to characterize the system noise at the time of the event. Recreated noise is repeatedly injected into a play-in simulation to create the Monte Carlo experiment. Each simulation is calibrated for identical parameters to develop a distribution of parameter estimates. The approach is demonstrated using an actual-system data set to evaluate: the impacts of parameter fitting using the output current vice the output power, the correlation of parameters, and the distribution of parameter estimates.