Modelling and Adaptive Control of a DC-DC Buck Converter

Abstract

With the advancement of electronic industry the requirement of low power supply is essential as numerous industrial and commercial devices rely on power converters for regulated and reliable DC power source. The demands of DC-DC converters are increasing exponentially because of their high efficiency, small size as well as simple architecture. The complexity in modelling of DC –DC converter mainly depends on its usage and its sophistication as it ranges from simple analogue design for low cost application to digital and self-adaptive model for better performance. This paper comprises of method for obtaining the small signal model of DC-DC buck converter by linearizing it using state space averaging technique. Both state space as well as non- linear model of Buck converter is the simulated in MATLAB and desired response is observed. This paper also discuss the methods of design and implementation of controller for Buck converter .The purpose of the compensation is to modify the dynamic characteristics of the converter in order to satisfy the performance specifications of the Buck converter. The performance specifications of the converter are maximum peak overshoot, settling time and steady state requirements and should be stated precisely so that the optimal control of the converter can be obtained. In this research we are interested in two approaches that are commonly used in the digitally controlled design of buck converter, the pole-zero matching approach, which provides a simple discrete time difference equation, and the systematic pole placement method. This thesis also focuses on a new alternative adaptive schemes that do not depend entirely on estimating the plant parameters is embedded with LMS algorithm. The proposed technique is based on a simple adaptive filter method and uses a one-tap finite impulse response (FIR) prediction error filter (PEF).Simulation results clearly show the LMS technique can be optimized to achieve comparable performance to classic algorithms. However, it is computationally superior; thus making it an ideal candidate technique for low cost microprocessor based applications.