Author ORCID Identifier
https://orcid.org/0000-0002-0167-2170
Date Available
11-19-2025
Year of Publication
2025
Document Type
Doctoral Dissertation
Degree Name
Doctor of Philosophy (PhD)
College
Engineering
Department/School/Program
Electrical and Computer Engineering
Faculty
Professor Dan M. Ionel
Abstract
Residential digital twins are fundamental to the smart grid transition, and thus, must be both accurate and representative of existing homes and scalable for large distribution systems. Within this dissertation, new machine learning (ML) and physics-based methodologies are applied to the major individual residential loads, energy storage devices, and resources in the US to develop computationally efficient digital twins and new optimal control strategies for the virtual power plant (VPP) concept. Big data from experimental field demonstrations with dedicated metering, thousands of residential smart meter profiles, and large national human behavior surveys are employed to develop ultra-fast scalable residential load digital twins with IEEE testbench distribution systems for VPP controls.
In the first two chapters, two novel machine learning methodologies disaggregate and model the heating ventilation and air conditioning (HVAC) system because it is the largest residential load. With the long short term memory (LSTM) and a hybrid k-means and multiple linear regression (MLR) models, the HVAC load of individual experimental smart homes are separated from total smart meter power profiles and emulated from physics-based models with very low, high satisfactory error performance beyond industry energy modeling standards. These modeling contributions enable estimations of load change with new energy efficient SEER ratings and the electrification of heating and cooling in the US. Additionally, VPP co-simulation is advanced to include minutely power flow network calculations with estimations of individual home temperature and HVAC power and smart controls for grid performance benefits.
The second largest load, the electric water heater (EWH), is also assessed for VPP thermal energy storage and controls for grid benefits of peak load shifting and conservation voltage reduction in the third chapter. A thermal equivalent resistance capacitance (RC) circuit is validated against experimental profiles for an EWH, scaled to thousands of homes with unique hot water draw profiles, and evaluated in a developed distribution system simulation framework for short and long term VPP controls. The conceptualization of generalized energy storage with uniform communication through industry standards such as CTA-2045 is discussed for modeling requirements and community impacts.
Electric vehicles (EVs) add substantial energy storage capability to residential homes and, thus, are the focus of the remaining two chapters for the development of modeling procedures that may be applied in high numbers yet controlled through VPP communication signals individually or in groups following a proposed ML sensitivity study. Stochastic battery modeling for EV state of charge (SOC) upon arrival home is employed with the proposed metric “EV charging homogeneity” to quantify human behavior and decision making to charge in the same time period across nodes and transformers in a developed distribution system framework with 1,765 smart meter profiles. Coordinated and conventional controls assess grid impacts and charging threshold of a distribution system, and model predictive control VPP schemes are optimized for grid performance during vehicle-to-grid operation.
The last chapter covers the next phase of VPP and smart grid research with residential and distribution controls, hardware-in-the-loop experimentation. A review of 50+ references describes the state-of-the-art HIL facilities and studies on the topics of large-scale distribution systems, transformer and voltage regulator equipment, microgrid, renewable energy resources, and EV specific controls. Future HIL use-cases for EV fast charging and reactive power support (RPS) from EVs are simulated. This dissertation provides a comprehensive evaluation of residential modeling for the largest residential loads and energy resources for VPP development considering physical distribution system.
Digital Object Identifier (DOI)
https://doi.org/10.13023/etd.2025.513
Funding Information
This work was supported by the National Science Foundation (NSF) under the NSF Graduate Research Fellowship through Grant No. #2239063 in 2021, 2023, and 2024, and Award Nos. #1943035, #1936131, and ECCF #1936494 in 2019 and 2020. The support of the National Aeronautics and Space Administration (NASA) and the Kentucky Space Grant Consortium under the NASA award number 80NSSC20M0047 in 2022, the Leverhulme Trust under their visiting professorship scheme in 2024 and University of Kentucky through the L. Stanley Pigman Chair in Power Endowment and the Lighthouse Beacon Foundation in 2024 and 2025 is also gratefully acknowledged. Any findings and conclusions expressed herein are those of the authors and do not necessarily reflect the views of the sponsors.
Recommended Citation
Alden, Rosemary E., "Smart Homes, Grids, and Electric Vehicles Large-Scale Integration Studies Employing Machine Learning and Optimization Techniques" (2025). Theses and Dissertations--Electrical and Computer Engineering. 222.
https://uknowledge.uky.edu/ece_etds/222
