Physics Informed Data-Driven Oscillation Stabilization Strategy for Renewable-Dominant Power Systems Based on Koopman Operator

Abstract

With the high penetration of volatile renewable energy generations (REGs), oscillations have emerged frequently worldwide. Unlike the low-frequency oscillations in conventional power systems, oscillations in renewable-dominant power systems exhibit higher frequencies, involve more nonlinearities, and seriously threaten the stable operation. The main technical challenges of oscillation stabilization design are the nonlinearity, the complexity and the difficulty in obtaining models of renewable-dominant power systems. To cope with the paradigm shift, this paper proposes a physics informed data-driven oscillation stabilization (PDOS) strategy based on Koopman operator (KO), which offers the merits of strong interpretability and high computational efficiency. Firstly, the global linearization of the nonlinear dynamics is captured based on KO. Subsequently, a physics informed data-driven KO predictor is proposed, with the construction of basis functions adapted to power systems and introduction of physical constraints. This predictor can more accurately characterize the nonlinear oscillation behavior. Finally, using the KO-based linear quadratic regulator informed by physics, PDOS can online stabilize oscillations by regulating the control references of REG inverters. Case studies illustrate that the proposed strategy can stabilize oscillations effectively, adaptively and robustly, and the introduction of physics can enhance the control performance.