Data-Driven Transient Stability Evaluation of Electric Distribution Networks Dominated by EV Supercharging Stations

Accelerated deployment of high-power electric vehicle (EV) supercharging stations is expected to alleviate EV drivers' range anxiety, while imposing stress on the electric distribution networks (EDNs) and threatening their transient stability. As a powerful transient stability evaluation (TSE) tool, the estimation of region of attraction (ROA) plays a vital role in maintaining situational awareness and stable operation of the emerging EDNs. However, EDNs dominated by EV charging stations typically involve highly nonlinear and complex system dynamics, rendering the model-based approaches for ROA estimation computationally intensive. Thus, solution accuracy is usually compromised due to simplified system modeling. This paper proposes a data-driven approach to ROA estimation of emerging EDNs based on the Koopman operator theory. Numerically stable Koopman eigenfunctions can be learned from the system measured data and then employed to establish a set of linearly parameterized Lyapunov candidate functions. Various trajectory data are then employed to establish a tight feasible polytope. Through efficient sampling and linear optimization, the union of invariant sublevel sets of the determined Lyapunov functions can constitute a tight inner approximation to the actual ROA. The proposed method is evaluated to be computationally efficient and permits real-time ROA estimation. Numerical simulations of a DC EDN interfaced to an AC grid validate the superior performance of the proposed method.