On Some Low Cost Hybrid Pi-Neuro-Fuzzy Controllers For Non-Minimum Phase Systems In Power Plants

Abstract

The non-minimum phase systems with transfer functions having "unstable" zeros are subject to difficult stabilizing and controlling due to the unusual behavior with respect to both the time domain (characterized by the presence of down shoot) and the frequency domain (the shapes of the open-loop bode plots are characterized by the fact that the phase plot is decreasing for increasing value of the frequency). The paper proposes two hybrid PI-neuro-fuzzy controllers meant that cope with this class of systems. The controllers are based on self-tuning the free parameter of the standard PI-fuzzy controllers by introducing a single neuron with a simplified back-propagation learning algorithm. Digital simulation results that can correspond to the speed control of a hydro-electric power plant prove that one of the proposed controllers provides very good control system performance in comparison with a classical PI controller and with the standard PI-fuzzy controllers.