The rapid proliferation of DC-based distribution systems, driven by the inherent DC nature of renewable sources (PV, fuel cells), storage (batteries, supercapacitors), and modern loads (EVs, data centers, LED lighting), calls for a paradigm shift in how we design and control microgrids. Traditional rigid control architectures often struggle with the complexity, uncertainty, and dynamic interactions of future energy networks. This keynote explores the frontier of "Bio-Inspired Microgrids," where principles derived from neuroscience and biological systems—such as neural networks, self-organization, cognitive adaptability, and immune-system-like response—are translated into robust, resilient, and autonomous control architectures for DC microgrids. Drawing on recent advancements in distributed control, digital twins, and cybersecurity, the talk will demonstrate how energy systems can be designed to "learn, heal, and adapt" like living organisms. Furthermore, the presentation will highlight real-world implementations and case studies extending beyond traditional terrestrial grids into emerging extreme environments, including electrical ships, seaports, more-electric aircraft, and space (lunar/Mars) habitats. Attendees will gain insights into how the convergence of biology, artificial intelligence, and power electronics is paving the way for the next generation of self-sustaining DC microgrids that are critical for global decarbonization and energy resilience.

The growing adoption of renewable energy sources, energy storage systems, electric vehicles, and power electronic loads is accelerating the deployment of DC microgrids. While DC microgrids offer significant advantages in efficiency, flexibility, and integration of distributed energy resources, they also introduce complex control challenges arising from nonlinear system dynamics, operational uncertainties, and rapidly changing operating conditions. Conventional control methods, although widely used in practice, often rely on accurate system models and extensive tuning to maintain satisfactory performance under varying conditions. Reinforcement Learning (RL) has recently emerged as a promising artificial intelligence technique capable of learning control policies directly through interaction with the environment, offering adaptive and model-free decision-making capabilities. However, practical deployment of fully RL-based controllers in power systems remains challenging due to concerns related to reliability, stability, safety, and industrial acceptance. This keynote will present recent advances in applying reinforcement learning to DC microgrid control, with a particular focus on hybrid control architectures that combine the robustness of conventional controllers with the adaptability of learning-based methods. The talk will discuss RL fundamentals, voltage regulation and energy management applications, and the role of supervisory control frameworks that coordinate conventional and RL-based controllers. Future opportunities, including multi-agent reinforcement learning, digital twins, and autonomous energy management systems, will also be explored as key enablers of next-generation intelligent and resilient DC microgrids.

Data centers currently consume over 2% of the world's electricity. The growing demand for computing power driven by artificial intelligence has ushered data centers into a new era of development. The substantial power consumption of AI racks poses significant challenges to power system design. This presentation provides an overview of data center power supply systems, covering the entire power delivery path from the grid to the chip. It includes the evolution of data center power supply system architectures, power conversion technologies, and the application of wide-bandgap (WBG) devices. This talk aims to inspire researchers to explore innovative power supply technologies that can enable the sustainable development of data centers.

Over a century ago, there was a war of currents between alternating current (AC) and direct current (DC). AC was the right choice back then since it could be stepped up to high voltages for efficient power transmission over long-distance lines. However, today, many sources and loads, including solar PV, batteries, and computers, are inherently DC. Is AC still the optimal choice for the future energy landscape? The talk presents distinct advantages of adopting DC for various commercial and industrial use cases. The transition from AC to DC will be gradual and may take decades. The experience and lessons learned from well-established DC applications, such as shipboard power systems, can facilitate the ongoing and future transition of AI datacenter power supply systems. This presentation explores some critical challenges brought by this paradigm shift. The absence of zero-crossing and steep current rise rates impose new requirements on fault interruption devices demanding extra energy dissipation and high turn-off speed. While solid-state protection devices significantly reduce fault interruption time from a few milliseconds to a few microseconds, such protection devices face major difficulties in meeting the selectivity requirements of system protection schemes. DC systems normally contain many converters and feature high penetration of renewables. While grid-tied DC microgrids expand in scale, the interactions between grids and DC microgrids should be properly coordinated to benefit efficient and reliable operations of both systems. The strong nonlinearity and high dynamics stemming from converter operations and mode transitions complicate the secure and stable operations of the entire system. DC microgrids require innovative approaches to assess system stability and design stability controllers capable of handling both small and large disturbances. The challenges and potential solutions overviewed and discussed in this talk are crucial for formulating international standards and guiding engineering practices.