A Swarm-Based Adaptive Neural Network SMES Control for a Permanent Magnet Wind Generator

Permanent magnet synchronous generators are becoming increasingly popular as utility-scale wind generators. While their performances are satisfactorily under normal conditions, they may be degraded under wind gusts as well as in extremely low grid voltage conditions. An adaptive control of superconducting magnetic energy storage (SMES) system for efficient wind energy transfer as well as dynamic performance improvement is proposed in this article. A radial basis function neural network has been employed to determine the controller parameter values. The nominal weights of the neural network are obtained from training of a large input-output data set generated through an improved swarm optimization procedure. These weights are then updated through a novel method of tracking of system outputs in time domain. Tests carried out with the adaptive controller show that the improved particle swarm-based radial basis network SMES controller delivers wind energy to grid efficiently and at the same time exhibit very good damping profile.