Design and real-time implementation of a self-tuning flexible ac transmission system (FACTS) controller is illustrated for power oscillation damping. Although the system model is not required for self-tuning control design, it is shown to perform similar to a model-based design. For parameter estimation, the classical recursive least square (RLS) is supplemented by a random walk (RW) with a switched structure and compared to standard variable forgetting factor (VFF) approach. It is shown that the RW improves the accuracy and convergence of the estimated system parameters, which is critical to self-tuning control following large disturbances. The performance is validated in real time using a commercial real-time simulation platform. The control computation time is shown to be considerably less than the typical sampling time used for power oscillation damping applications demonstrating the feasibility of self-tuning FACTS controllers in practice.