Date of Graduation
Doctor of Philosophy in Engineering (PhD)
Second Committee Member
Third Committee Member
Fourth Committee Member
Machine learning, Renewable energy, Smart grid
The integration of high levels of renewable energy into smart grids is crucial for achieving a sustainable and efficient energy infrastructure. However, this integration presents significant technical and operational challenges due to the intermittent nature and inherent uncertainty of renewable energy sources (RES). Therefore, the energy storage system (ESS) has always been bound to renewable energy, and its charge and discharge control has become an important part of the integration. The addition of RES and ESS comes with their complex control, communication, and monitor capabilities, which also makes the grid more vulnerable to attacks, brings new challenges to the cybersecurity. A large number of works have been devoted to the optimization integration of the RES and ESS system to the traditional grid, along with combining the ESS scheduling control with the traditional Optimal Power Flow (OPF) control. Cybersecurity problem focusing on the RES integrated grid has also gradually aroused researchers’ interest. In recent years, machine learning techniques have emerged in different research field including optimizing renewable energy integration in smart grids. Reinforcement learning (RL), which trains agent to interact with the environment by making sequential decisions to maximize the expected future reward, is used as an optimization tool. This dissertation explores the application of RL algorithms and models to achieve high renewable energy integration in smart grids. The research questions focus on the effectiveness, benefits of renewable energy integration to individual consumers and electricity utilities, applying machine learning techniques in optimizing the behaviors of the ESS and the generators and other components in the grid. The objectives of this research are to investigate the current algorithms of renewable energy integration in smart grids, explore RL algorithms, develop novel RL-based models and algorithms for optimization control and cybersecurity, evaluate their performance through simulations on real-world data set, and provide practical recommendations for implementation. The research approach includes a comprehensive literature review to understand the challenges and opportunities associated with renewable energy integration. Various optimization algorithms, such as linear programming (LP), dynamic programming (DP) and various RL algorithms, such as Deep Q-Learning (DQN) and Deep Deterministic Policy Gradient (DDPG), are applied to solve problems during renewable energy integration in smart grids. Simulation studies on real-world data, including different types of loads, solar and wind energy profiles, are used to evaluate the performance and effectiveness of the proposed machine learning techniques. The results provide insights into the capabilities and limitations of machine learning in solving the optimization problems in the power system. Compared with traditional optimization tools, the RL approach has the advantage of real-time implementation, with the cost being the training time and unguaranteed model performance. Recommendations and guidelines for practical implementation of RL algorithms on power systems are provided in the appendix.
Li, Y. (2023). Achieving High Renewable Energy Integration in Smart Grids with Machine Learning. Graduate Theses and Dissertations Retrieved from https://scholarworks.uark.edu/etd/4901