Advanced Battery Management Systems

Online State of Charge (SoC), State of Power (SoP), and State of Health (SoH) estimation is essential for the efficient, safe, and reliable operation of Lithium-ion batteries. Although the battery model and estimation algorithm are essential, the nature of the battery excitations also significantly influences the estimation performance. When the input-output data, i.e., the input current and output voltage, is insufficiently rich in frequency content, the estimation performance is poor. Both simulation and experimental results verify the analysis results and show that a current profile with optimal frequency components achieves the best estimation performance, thereby providing guidelines for designing battery current profiles for improved SoC and SoH estimation performance.

Recently, our group has particular interest in leveraging both electrochemical characterization and machine learning to improve battery condition monitoring and performance prediction. Several papers are in preparation in this thread.

Reusing Second-life Batteries in Renewable Energy Systems

The Lithium-ion battery is currently the most widely used solution for energy storage systems. However, its high cost is considered one of the major barriers hampering the integration of renewable energy and the adoption of EVs. Reusing EV batteries seems a promising solution to the aforementioned problem. Based on a dynamic degradation model of Lithium-ion batteries, this paper first compares the profits that second-life and fresh batteries can bring to a wind farm. Model predictive control is adopted to solve an hourly optimal wind scheduling problem and maximize the profit of wind farm owners. The optimal size of the battery is determined, and then the comparison of second-life and fresh batteries is conducted, taking the battery degradation, the profit of the wind farm owner, and various remaining capacities of the battery into account.

Integration of Electrified Transportation and Renewable Energy Systems

Given the fast electrification rate in the transportation system, its influence on the grid will become more and more significant. Through the grid-to-vehicle (G2V) and vehicle-to-grid (V2G) techniques, the electrified vehicles can be optimally scheduled and controlled as a demand response resource to increase grid resilience and integrate more renewable energy. My research plan is to harmoniously integrate the electrified vehicles into the grid and make the entire system cleaner. I will focus on the G2V/V2G in micro-grids such as a house, an independent building, and a community, including a large number of renewable energy generations and electrified vehicles. Due to the complicated and interdisciplinary nature of these problems, I will establish a multi-disciplinary framework using data mining and machine learning from a control perspective.

Eco-Driving Strategies for Mixed Traffic Flow

Given that connected and automated vehicles (CAVs) and human-driven vehicles (HDVs) will coexist for a long period, the design of cooperative adaptive cruise control (CACC) in mixed traffic flow (see Fig. 3) is critical. To handle the uncertainty of HDVs, we proposed a robust CACC, including a time-triggered feedback control and an event-triggered feedforward control, to tackle the disturbances with various amplitudes. Tube methods are utilized to characterize the disturbances and design the controller. Theoretical analysis and numerical experiments validate the robustness, the computational efficiency, and the scalability of the proposed algorithm.