Modeling and design of proportional-resonant current control for grid-connected converters

Abstract

Grid-connected converters, such as the three-phase two-level rectifier or inverter topology, provide an interface for drawing power from, or feeding power into, the utility grid. This is done by switching a DC voltage across phase inductors in an alternating fashion. For the two-level converters, the resulting voltage at the the converter output is a square wave at the switching frequency, often in the kilohertz to tens of kilohertz range. This will create an alternating, sinusoidal current through the inductor. Current control is an important part of converters, and is usually implemented in the synchronous reference frame, with PI-controllers which operate on DC-valued components, transformed from sinusoidal form. In recent years, another approach, known as proportional-resonant control, has gained attraction. It is the equivalent of PI-control in the SRF, operating in the stationary reference frame, on quantities which are rotating/sinusoidal. PI-control works in the so-called dq-domain, wherein the frame of reference rotates along with the quantities at a fundamental frequency. PR-controllers are realized by inverse-rotating the PI-controller transfer function from dq- to alpha-beta-domain. PI-control works well in cases of ideal voltage and grid, where for example harmonic components of the voltage other than the fundamental are negligible. PR-control offers the significant advantage of straightforward addition of controllers which are tuned to regulate specific harmonic components, without needing additional filters or harmonic detection methods. Many rules-of-thumb have been formed by previous methods of tuning such controllers. This thesis shows how the dynamic, linearized model of grid-converters can be used in tuning; first, by finding gains for the PI-scheme from the frequency response, and second, by adapting those gains into the stationary frame equivalent. PR-based current control was simulated and compared to PI-control in various different scenarios, showing promising results and potential for improving control strategies.