An Adaptive Neural Network-Based Controller to Stabilize Power Oscillations in Wind-integrated Power Systems

Abstract

This paper proposes an improved neuro-adaptive control scheme, based on online system identification and simultaneous control, for damping the low-frequency oscillations generated due to the wind integration in conventional power systems. A simple, linear neural identifier, with a few adjustable connection weights, is used which ensures minimal computational burden and fast learning capability in comparison with other neuro-controllers having highly complex and non-linear architecture, which may pose difficulties in their real-time implementation. It is demonstrated that a simple linear neural identifier, which approximates a local linear model of a system, by adjustment of its parameters online, is faithfully able to track the varying dynamics of the system. The proposed controller is used as a supplementary control in the rotor-side converter of the doubly-fed induction generator. Improved oscillation-damping performance over a wide range of operating conditions, in comparison with a well-designed genetic algorithm-based conventional lead-lag controller, has been validated through simulation studies carried out on wind-integrated-inter-connected power system. Moreover, the proposed control scheme is model free, easier to implement and needs local measurements only.