Simulation-Based Optimization of a DC Microgrid - With Machine-Learning-Based Models and Hybrid Meta-Heuristic Algorithms - Doctoral Dissertation

Aug 5, 2026

Upload by Hazel

Economic dispatch (ED) aims to optimize power flow while minimizing operational costs in power systems. Simulation-based optimization (SBO) can model system interactions accurately, yet it requires many simulations and may fail to reach a global optimum. This dissertation develops a machine-learning-enhanced SBO approach for ED, combining state-reduction and neural-network observers to cut simulation time, and a hybrid genetic algorithm plus particle swarm search guided by cost-to-parameter correlation. The integrated method enables economic dispatch parameter alignment in minutes on standard computing devices.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages113

Reading Time4 h 45 min

AccessFree to download

Acknowledgments

Abstract

List of Figures

Chapter 1 Introduction

Introduction

Problem Statement

Machine Learning

RC Droop

Simulation-based Optimization

Chapter 2 Efficient Neural Network-based Observer-aided Behavioral Modeling of Energy Storage Device SoC for Bus-tied Converters

Introduction

Baseline Model DTB

Observer Feature Selection

Neural Network Structure and Feature Selection

Structure of FF-NN

Feature Exploration and Training

Observer-aided Behavioral Model

Computational Considerations

Results

University of South Carolina Scholar Commons Theses and Dissertations Fall 2023 Simulation-Based Optimization of a DC Microgrid: With Machine-Learning-Based Models and Hybrid Meta-Heuristic Algorithms Tyler Van Deese Follow this and additional works at: [https://scholarcommons.sc.edu/etd](https://scholarcommons.sc.edu/etd) Part of the Electrical and Computer Engineering Commons Recommended Citation Deese, T. V. (2023) . Simulation-Based Optimization of a DC Microgrid: With Machine-Learning-Based Models and Hybrid Meta-Heuristic Algorithms. (Doctoral dissertation) . Retrieved from [https://scholarcommons.sc.edu/etd/7643](https://scholarcommons.sc.edu/etd/7643) This Open Access Dissertation is brought to you by Scholar Commons. It has been accepted for inclusion in Theses and Dissertations by an authorized administrator of Scholar Commons. For more information, please [contact](contact digres@mailbox.sc.edu)[ digres@mailbox.sc.edu](contact digres@mailbox.sc.edu). Simulation-based Optimization of a DC Microgrid: With Machine-learning-based Models and Hybrid Meta-heuristic Algorithms by Tyler Deese . Bachelors of Science The Citadel 2017 Masters of Engineering University of South Carolina 2021 Submitted in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy in Electrical Engineering College of Engineering and Computing University of South Carolina 2023 Accepted by: Herbert Ginn, Major Professor Kristen Booth, Committee Member Austin Downey, Committee Member Roger Dougal, Committee Member Ann Vail, Dean of the Graduate School © Copyright by Tyler Deese, 2023 All Rights Reserved. ii Acknowledgments I believe that achieving a doctorate requires a significant combination of hard work and some element of luck. I feel incredibly fortunate to find myself in this position, and it’s a culmination of efforts from many remarkable individuals who have contributed to my personal growth. Notably, my parents have played an instrumental role in affording me the opportunities to reach this point. My brother, John, inspired me to challenge my epistemology from a young age, and my sister, Tina, has been a profound influence on shaping the person I am today. I must express my gratitude to my advisor, Herb, who has provided me with the guidance and support necessary to develop as both an engineer and a scientist. He consistently encourages experimentation and critical thinking. He is a good advisor, and a great man. Abstract The field of economic dispatch (ED) focuses on optimizing power flow in a power system to minimize costs. It has the potential to significantly enhance system effectiveness, and efficiency, and reduce operating costs. Various techniques have been employed to tackle this problem, each with its own strengths and weaknesses. One promising approach is simulation-based optimization (SBO), which allows for accurate modeling of system interactions and improved representation of expected results. However, SBO requires running numerous simulations to identify an optimal solution, and there is a possibility of not achieving the global optimum. This work aims to address these challenges using machine learning. The first contribution involves enhancing the computational efficiency of the SBO model by employing state-reduction techniques and neural network-based observers. This optimization reduces simulation time, thereby speeding up the search process. The second contribution involves developing a hybrid search algorithm by combining the genetic algorithm and the particle swarm method. Additionally, leveraging the cost-to-parameter correlation helps expedite the parameter search. This modified hybrid genetic algorithm reduces the number of simulations required to discover the optimum solution while providing increased confidence in the result. Finally, these two methods are applied to a system to demonstrate that, with their integration, a simulation-based optimizer

FAQs

What problem does the dissertation address?

The work addresses optimization of economic dispatch in power systems using simulation-based optimization, focusing on reducing computation time and improving solution quality.

How does the approach speed up simulation-based optimization?

It improves computational efficiency using state-reduction techniques and neural network-based observers that reduce simulation time and accelerate the search process.

How does the dissertation’s hybrid algorithm reduce the number of simulations?

It combines a genetic algorithm with particle swarm search, and uses cost-to-parameter correlation to accelerate parameter exploration, decreasing the simulations needed to find an optimum.

Aug 5, 2026

Upload by Hazel