Self learning control strategies for dc dc power converter systems

Abstract

The design and output control of DC-DC power converters pose significant challenges newlinedue to factors such as sudden load shifts, input voltage variations, and system uncertainties, newlineall of which contribute to instability. To overcome these issues, nonlinear newlineself-learning controls provide a promising solution. These controls leverage adaptive newlinemechanisms, enabling real-time adjustments and robust responses to a wide range of newlineuncertainties. Such controllers enhance performance by effectively handling fluctuating newlineloads, varying input conditions, and unpredictable system behaviors. In this newlinethesis, self-learning control strategies have been designed for a DC-DC buck converter newlineand rigorously tested under diverse conditions. newlineFirstly, we present a self-learning Zernike neural network based robust control for newlineoutput voltage tracking in DC-DC buck power converters with resistive load and newlinePMDC motor load, particularly for applications demanding high precision performance newlinein face of large load uncertainties. The design involves a computationally newlinesimple online single hidden layer neural network, to rapidly estimate the unanticipated newlineload changes and exogenous disturbances over a wide range. The controller is newlinedesigned within a backstepping framework and utilizes the learnt uncertainty from newlinethe neural network for subsequent compensation; to eventually ensure an asymptotic newlinestability of the tracking error dynamics. The results obtained feature a significant improvement newlineof dynamic and steady-state performance concurrently of the output state newlinein contrast to other competent control strategies lately proposed in literature for similar newlineapplications. Extensive numerical simulations and experimentation on a developed newlinelaboratory prototype are carried out to justify the practical applicability and feasibility newlineof the proposed controller. Experimental results substantiate the claims of fast newlinedynamic performance in the settling time, besides an accurate steady state tracking. newline