Model Estimation of Electric Power Systems by Phasor Measurement Units Data
Abstract
This
dissertation tackles the online estimation of synchronous machines’ power subsystems
electromechanical models using the output based Phasor Measurements Units (PMUs)
data while disregarding any inside data. The research develops state space
models and estimates their parameters and states. The research tests the
developed algorithms against models of a higher and of the same complexity as
the estimated models.
The
dissertation explores two estimations approaches using the PMUs data: i)non- inear
Kalman filters namely
the Extended Kalman Filter (EKF) and then the Unscented Kalman Filter (UKF) and
ii) Least Squares Estimation (LSE) with Finite Differences (FN) and then with System Identification. The EKF based research i)
establishes a decoupling technique for the subsystem the rest of the power
system ii) finds the
maximum number of parameters to estimate for classical machine model and iii)
estimates such parameters The UKF based
research i) estimates a set of electromechanical parameters and states for the flux decay model and ii) shows the
advantage of using a dual estimation filter with colored noise to solve the difficulty of some simultaneous state and parameter
estimation.
The LSE
with FN estimation i) evaluates numerically the state space differential equations and transform the
problem to an overestimated linear system whose parameters can be estimated,
ii) carries out sensitivity studies evaluating the impact of operating conditions
and iii) addresses the requirements for implementation on real data taken from the
electric grid of the United States. The System Identification method i) develops a linearized
electromechanical model, ii) completes a parameters sub-set selection study
using singular values decomposition, iii) estimates the parameters of the
proposed model and iv) validates its output versus the measured output.
