Document Type

Thesis - Open Access

Award Date

2016

Degree Name

Master of Science (MS)

Department

Electrical Engineering and Computer Science

First Advisor

Reinaldo Tonkoski

Keywords

adaptive dynamic programming, energy storage systems, frequency stability, microgrids, PID controller, virtual inertia

Abstract

Due to low inertia and the intermittent nature of photovoltaic systems, dynamic frequency stability issues arise in microgrids with large photovoltaic systems. This limits the maximum amount of photovoltaic systems that can be penetrated in the microgrid. In order to increase the penetration of photovoltaic systems, the dynamic frequency controller, that is faster than the primary frequency controller (governor control) needs to be added in the microgrid system. For dynamic frequency control, inertial response can be provided from the energy storage system (such as battery, ultra-capacitor, photovoltaic system, etc.), which is termed as virtual inertia. A virtual inertia can be defined as the combination of an energy storage system, a power electronics converter and a proper control algorithm that improves the dynamic frequency stability of the microgrid. A virtual inertia supplies or absorbs the active power to and from the energy storage system to improve the dynamic frequency stability. This thesis presents the design and implementation of a hardware prototype of 1 kW virtual inertia in a microgrid with a real diesel generator and a load. For a step change in load, the virtual inertia improved the frequency response of the system from 57.39 Hz to 58.03 Hz. This improvement in frequency response proves the concept of existing proportional derivative based virtual inertia experimentally. With the addition of virtual inertia, the frequency of the system returns to the nominal frequency slower. Once the primary controller (governor control) of the system takes the action to regulate the frequency, virtual inertia no longer needs to add inertia to the system. So the dynamics of the VI needs to be improved so that the frequency returns to nominal frequency faster. This thesis also proposes an online learning controller based virtual inertia using adaptive dynamic programming that learns online and improves the dynamics of the controller of existing VI. The output of this controller supplements the output of the existing proportional derivative controller of virtual inertia. The supplementary controller is trained to increase the dynamics of the outer controller and to bring the system frequency to nominal frequency faster. Due to faster dynamics, the net energy delivered by the VI can be reduced significantly and improve the total possible discharge cycles from the battery. For performance evaluation, the proposed controller was implemented in a microgrid with a photovoltaic system, a diesel generator and a variable load. With the proposed controller, the frequency of the system returned to nominal frequency faster. The net energy delivered by the proposed controller in a photovoltaic diesel generator microgrid was 46.14% of the net energy delivered by the existing virtual inertia. Due to the decrement in the total energy delivered, the total number of possible battery discharge cycles with ADP based VI was 2.17 times of the total number of possible battery discharge cycles from VI.

Description

Includes bibliographical references (pages 100-103)

Format

application/pdf

Number of Pages

120

Publisher

South Dakota State University

Rights

Copyright © 2016 Dipesh Shrestha

Available for download on Tuesday, December 12, 2017

Share

COinS