In this paper, the empirical controllability covariance (ECC), which is calculated around the considered operating condition of a power system, is applied to quantify the degree of controllability of system voltages under specific dynamic var source locations. An optimal dynamic var source placement method addressing fault-induced delayed voltage recovery (FIDVR) issues is further formulated as an optimization problem that maximizes the determinant of ECC. The optimization problem is effectively solved by the NOMAD solver, which implements the Mesh Adaptive Direct Search algorithm. The proposed method is tested on an NPCC 140-bus system and the results show that the proposed method with fault specified ECC can solve the FIDVR issue caused by the most severe contingency with fewer dynamic var sources than the Voltage Sensitivity Index (VSI) based method. The proposed method with fault unspecified ECC does not depend on the settings of the contingency and can address more FIDVR issues than VSI method when placing the same number of SVCs under different fault durations. It is also shown that the proposed method can help mitigate voltage collapse. ). His research interest includes computational optimal control, nonlinear filtering, cooperative control of autonomous vehicles, industry applications of control theory, nonlinear H∞ control, and bifurcations and normal forms. His early research includes topics on Lie groups, Lie algebras, and differential geometry. Dr. Kang is a fellow of IEEE. He was a plenary speaker in several international conferences of SIAM and IFAC. He served as an associate editor in several journals, including IEEE TAC and Automatica.