A primal-dual-quasi-newton-based decentralized SCOPF approach for coordinated corrective control of interconnected power grids

Abstract

Coordinated corrective control (CC) is indispensable for protecting multi-area interconnected power grids operated by independent entities against post-contingency overloads, and even subsequent cascading failures. Generally, CC strategies are generated online by solving a security-constrained optimal power flow (SCOPF) problem. Region-based decomposition and coordination through auxiliary problem principle (APP) allows distributed and parallel computations among subsystems without a central coordinator, eliminating the restriction of obtaining global data from independent entities. However, by taking a fixed core to form the auxiliary problems, the APP exhibits slow convergence in an ill-conditioned setting, which is inevitable for CC due to the penalty on undesirable actions, e.g., load shedding, in the objective function. In this paper, based on the in-depth theoretical analysis of the APP’s deficiency, a primal–dual quasi-Newton (PDQN)-APP algorithm is proposed by generalizing the fixed core to an adaptive core. And consequently, a fully decentralized SCOPF approach is developed for the coordinated CC of interconnected power grids via the proposed PDQN-APP algorithm. In particular, the proximal-gradient updates are extended to proximal-Newton updates via the state-of-the-art PDQN method. Through the generalization, adaptive curvature estimations relevant to boundary variables are contained in both the primal and dual iterative updates to correct the ill-conditioning, which significantly improves the convergence performance. Numerical studies demonstrate the superiority of the proposed PDQN-APP algorithm in convergence rate and accuracy. The effectiveness and superiority of the developed decentralized SCOPF approach and resulting CC strategies are also verified.