Capturing Dynamics in the Power Grid: Formulation of Dynamic State Estimation through Data Assimilation

iii Executive Summary With the increasing complexity resulting from uncertainties and stochastic variations introduced by intermittent renewable energy sources, responsive loads, mobile consumption of plug-in vehicles, and new market designs, more and more dynamic behaviors are observed in everyday power system operation. To operate a power system efficiently and reliably, it is critical to adopt a dynamic paradigm so that effective control actions can be taken in time. The dynamic paradigm needs to include three fundamental components: dynamic state estimation; look-ahead dynamic simulation; and dynamic contingency analysis (Figure 1.1). These three components answer three basic questions: where the system is; where the system is going; and how secure the system is against accidents. The dynamic state estimation provides a solid cornerstone to support the other 2 components and is the focus of this study. Dynamic states (e.g., rotor angle and generator speed) are the minimum set of variables that can determine the status of a dynamic system. A dynamic model with accurate states can faithfully reveal system responses. Therefore, dynamic state estimation can provide a full dynamic view of a power grid and generate critical inputs for other operational tools. To estimate the dynamic states of a power grid in real time, we developed and evaluated data assimilation methods to fuse phasor measurement unit (PMU) data with power system dynamic models. In this study, we defined a general dynamic state estimation problem for a power system and performance evaluation criteria. A problem is formulated for estimating the dynamic states of synchronous generators. As an initial effort, the following four data assimilation algorithms are developed, implemented and applied to estimate the dynamic states of a synchronous generator:  Ensemble Kalman filter (EnKF)  Extended Kalman filter (EKF)  Unscented Kalman filter (UKF)  Particle filter (PF) By comparing their performance under statistical framework using Monte Carlo methods, it was found that  The EnKF algorithm outperforms other algorithms when the typical PMU sampling rate is used for estimation.  Measurement interpolation methods can improve the estimation accuracy of the EKF, UKF, and PF. The interpolation does not show significant influence on the performance of the EnKF.  Increasing the number of samples can improve the estimation and convergence of the PF.  All four algorithms are robust to missing data.  The outliers cause some significant errors for all algorithms if the outliers are processed as normal data. The EKF, UKF, …