This paper presents the design of an artificial neural network (ANN) controller for static var compensator (SVC) to improve the stability and the performance of a superconducting generator (SCG). The SVC is also equipped with conventional proportional integral derivative (PID) controller designed according to the pole placement technique. The ANN controller patterns are gathered from unsupervised (self learning) ANNs-PID using an optimization technique. These pattern groups are chosen to cover most operating conditions based on P-Q plane. The present control strategy is tested on a SCG connected to an infinite bus system. To achieve a high degree of accuracy, the system is represented by a detailed non-linear model including the SCG, SVC, the transmission system and non-linear constraints. The simulation results reveal that the proposed ANN-SVC controller achieves further enhancement of the system performance and stability over a wide range of operating conditions and under sever disturbances compared with conventional PID-SVC controller.