Zhichao Liu

Date of Award

Summer 2019

Document Type

Open Access Dissertation


Electrical Engineering

First Advisor

Bin Zhang


With the growth of electricity demand and renewable energy power source, power converter becomes a more and more significant component in electrical power systems. The requirement of the power converter controller is to produce an accurate and low-distorted voltage or current under different load conditions. Although the conventional controller can meet the requirement of some applications, it requires accurate knowledge of the system model and cannot provide a satisfactory result especially under nonlinear loads or sudden load change. Repetitive control (RC) presents an attractive solution to achieve excellent steady-state tracking error and low total harmonic distortion for periodic signals, and it is increasingly applied to power converter systems.

However, there are still some limitations or requirements of RC when it is applied to power electronics system: first, RC requires the system sampling frequency is a fixed value and needs to be an integral multiple of the reference frequency; second, low controller sampling frequency results in low phase lead compensation resolution in RC, which leads to control inaccuracy; third, conventional RC does not have frequency adaptability to reference frequency fluctuation, and even a small reference frequency fluctuation can lead to severe performance degradation.

To overcome the conventional RC limitations, two advanced design methods are proposed in the thesis: fractional order delay and virtual variable sampling. The method of fractional order delay approximates the non-integer delay part by building a finite impulse response filter. This improved method is not only able to be applied on a period delay unit but also on phase-lead compensation. The accurate period delay and phase lead compensation show a noticeable improvement in RC performance. Although fractional order delay can meet the requirement on most of the applications, it also has a minimal adjustable range on the reference frequency. To achieve an essential solution to this problem, the virtual variable sampling (VVS) method is developed. The VVS approximates a variable sampling unit instead of the fixed system unit for RC and its filters, in which RC is able to be frequency adaptive. Comparing with the method of fractional order delay, the VVS method can provide a much more extensive adjustable range on the reference frequency.

Based on the system performance under the conventional controller, power converter always has uneven distortion distribution. To further improve the stability and eliminate harmonic distortions efficiently, two selective harmonic RC schemes are introduced - nk ± m order harmonic RC and DFT-based selective harmonic RC. However, these selective RC schemes also suffer from the particular requirement of system sampling frequency and low reference frequency adaptability. Applying VVS methods on these two schemes can effectively present an improvement on their frequency adaptability. To verify the proposed methods’ effectiveness, a complete series of power electronics applications are carried out. These applications include single-phase and three-phase DC/AC power converter, single-phase AC/DC power converter, and single-phase grid-connected power converter. The detailed system modeling and the proposed RC schemes are presented for each power electronics application.


© 2019, Zhichao Liu