Small-signal stability of DC microgrid under data center load: Theoretical analysis and experimental validation

This paper investigates and proposes solutions to enhance the small-signal stability of DC microgrid operating under data center load profiles. For this purpose, a fundamental one-source one-cable one-load DC MG is established to capture all destabilizing elements without adding unnecessary complexity. The dynamic behavior of the data center load is fully investigated, with a specific focus on its negative incremental impedance characteristic and the severe power fluctuations induced by AI model execution. Eigenvalue analysis highlights the potential adverse effects of control interference and improper system parameter design. Complementarily, an impedance-based stability analysis relying solely on terminal measurements is conducted. It is demonstrated that both stability analysis methods can effectively determine the stability of the system. The accuracy of the analysis is validated by an experimental testbed using imperix rapid control prototyping solutions. Two stability enhancement solutions are proposed to damp the poorly damped oscillatory mode without changing the control structure. Finally, the foundational framework is extended to a multi-converter DC microgrid configuration, successfully demonstrating the scalability of the proposed stability analysis.