Systematic SDP Search for Decentralized Stability Certificates

Abstract

Decentralized stability certificates allow inverter-based resources to be connected to a grid without repeating a system-wide small-signal study. Passivity provides such certificates but is violated by practical grid-forming (GFM) controls at low frequency, and its generalizations---transformation matrices, extended passivity indices, loop shifting, and gain--phase conditions---rely on multipliers that are constructed analytically or chosen heuristically. All of these certificates are linear matrix inequalities in their multipliers at every frequency, so the search for the least conservative multiplier is a convex semidefinite program (SDP) with a global optimum and a definite infeasibility verdict. This paper demonstrates how SDP can be applied to searching for optimal multipliers for some decentralized methods proposed in the literature. Then it illustrates the methodology by presenting a systematic search algorithm for the fully plug-and-play additive nodal-admittance certificate, which requires no assembled network model and holds for arbitrary interconnections. Only sampled frequency responses are needed, making the search compatible with black-box models and impedance scans. Chebyshev parameterization together with adaptive frequency splitting provides a piecewise-polynomial multiplier without manual partitioning. On the IEEE 39-bus system with ten droop-controlled GFMs, this multiplier certifies a larger region of droop gains than piecewise-constant multipliers, as confirmed by simulation of random inverter placements.