Files in this item



application/pdfXU-DISSERTATION-2019.pdf (2MB)
(no description provided)PDF


Title:Data-driven coordination of assets in power distribution systems for ancillary service provision
Author(s):Xu, Hanchen
Director of Research:Domínguez-García, Alejandro D.
Doctoral Committee Chair(s):Domínguez-García, Alejandro D.; Sauer, Peter W.
Doctoral Committee Member(s):Veeravalli, Venugopal V.; Başar, Tamer
Department / Program:Electrical & Computer Eng
Discipline:Electrical & Computer Engr
Degree Granting Institution:University of Illinois at Urbana-Champaign
Subject(s):power distribution system
distributed energy resource
load tap changer
voltage regulation
frequency regulation
reinforcement learning
Abstract:In this dissertation, we develop several data-driven frameworks for coordinating distributed energy resources (DERs) in power distribution systems to provide ancillary services including active power provision and reactive power regulation. The proposed frameworks generally consist of three components, namely (i) an input-output (IO) model of the system describing the relation between the variables of interest to the problem, (ii) an estimator that provides estimates of the parameters that populate the model in (i), and (iii) a controller that uses the model in (i) with the parameters estimated via (ii) to determine the active and/or reactive power injection set-points of the DERs by solving the optimal DER coordination problem (ODCP), which is cast as a static optimization problem. We develop efficient estimation algorithms that utilize measurements to estimate the parameters in the IO model. Special emphasis is devoted to algorithms that address the potential collinearity issue in the measurements, and formulations that significantly reduce the number of parameters to be estimated. The idea of data-driven coordination is also applied to address the problem of coordinating load tap changers (LTCs)---an important class of assets used for voltage control in distribution networks---using only measurements of voltage magnitudes. Different from the ODCP that is cast as a static optimization problem, the optimal LTC coordination problem is cast as a multi-stage decision-making problem and formulated as a Markov decision process (MDP), in which the unknown power injections are modeled as uncertainty sources. The MDP is solved via a reinforcement learning algorithm to obtain a control policy that maps the voltage magnitude measurements to the optimal tap positions. The data-driven nature makes the proposed frameworks intrinsically adaptive and robust to changes in operating conditions and power distribution system models, which are illustrated via extensive case studies.
Issue Date:2019-05-08
Rights Information:Copyright 2019 Hanchen Xu
Date Available in IDEALS:2019-11-26
Date Deposited:2019-08

This item appears in the following Collection(s)

Item Statistics