Self-Tuning Neural Predictive Control Scheme for Ultrabattery to Emulate a Virtual Synchronous Machine in Autonomous Power Systems.

Abstract

An adaptive neural predictive controller (ANPC) is proposed for an ultrabattery energy storage system (UBESS) to enable its operation as a virtual synchronous machine (VSM) in an autonomous wind-diesel power system. The proposed VSM emulates the inertial response and oscillation damping capability of a typical synchronous machine (employed in conventional power plants) by adaptively controlling the power electronic interface of the UBESS. The control objective is to support the network frequency while ensuring efficient/economic use of the UBESS energy. During the load-generation mismatch, ANPC continuously searches for optimal VSM parameters to minimize the actual frequency variations, their rate of change of frequency (ROCOF), and the power flow through the UBESS while maintaining the state of the charge (voltage) of the ultrabattery bank to tackle subsequent disturbances. Simulations confirm that the proposed self-tuning VSM achieves similar performance as that of other VSM control schemes while substantially reducing the power flow through the UBESS and, hence, uses significantly less energy per hertz improvement (in frequency). An index is used to evaluate the performance of the proposed scheme. In addition, the self-tuning VSM has a better dynamic response (quantified as a reduction in ROCOF and settling times) while attenuating the frequency excursions for all simulated cases.