Abstract
With the pressing net-zero objectives, renewable energy sources (RESs) and energy storage systems (ESSs), which serve as the main force in providing low-carbon energy and flexibility, are being integrated into the power systems at an unprecedented rate. The rising penetration trend of RESs and ESSs brings both opportunities and challenges. In the short term, the inherent intermittency of RESs and high capital costs of ESSs introduce obstacles to revenue maximization for hybrid power plants and complicate the provision of privacy-ensured, high-fidelity network modeling. Long-term planning faces the hurdle of expanding ESS capacity in microgrids while maintaining power supply resilience with specific frequency characteristics. The recent surge in machine learning and quantum computing offers potential solutions to address these key challenges. Aiming to achieve secure and economic operation through capacity scheduling of investor-owned photovoltaic-battery storage systems (PV-BSS), this dissertation utilizes a Proximal Policy Optimization (PPO)-based deep reinforcement learning (DRL) agent that can handle continuous action spaces and ensure safety constraints, addressing challenges in adapting to volatile market signals and PV generation profiles. To enhance the accuracy of equivalent models for active distribution networks (EMADNs) in the presence of high RES penetration and associated management techniques, an adaptive EMADN with tunable scales and parameters is developed, featuring a leaves-trimming topological reduction method and a distributed PPO-based agent. Then, this dissertation seeks to strategize BSS expansion planning for microgrids by introducing a two-stage multi-period framework that combines the quantile regression DRL algorithm with linear programming, ensuring adaptive planning for long-term variations in RES, load, and battery pricing. Finally, this dissertation aims to optimize energy arbitrage tasks in ESSs by introducing a quantum policy learning algorithm that involves a strategically structured variational quantum circuit, offering a novel approach to action exploration and maintaining operational safety within energy markets. Case studies based on real electricity market data and RES profiles validate the effectiveness and benefits of the proposed methodologies compared to state-of-the-art techniques.
Degree Date
Fall 12-16-2023
Document Type
Dissertation
Degree Name
Ph.D.
Department
Electrical and Computer Engineering
Advisor
Jianhui Wang
Number of Pages
172
Format
Creative Commons License
This work is licensed under a Creative Commons Attribution-Noncommercial 4.0 License
Recommended Citation
Huang, Bin, "Quantum and Classical Learning Algorithms for Grid Integration of Energy Storage and Renewables: Operation, Modelling, and Planning" (2023). Electrical Engineering Theses and Dissertations. 70.
https://scholar.smu.edu/engineering_electrical_etds/70
