Examenspresentation av Anna Ferm och Erik Gustavsson

Conducted at Aros electronics AB supervised by Daniel Chädström and Filip Karlsson

Opponent: Daniel Quach

Abstract
This thesis describes how to construct different control schemes and evaluate them in regards to controlling a current in a system using a grid connected converter and an LCL filter. The control schemes in question are a PI controller that controls the current in the rotating reference frame as well as a PR, an LQR and an LQG
controller using the standard reference frame. The controlled current will be analysed in how well the controllers are able to mitigate peak errors when drastic changes occur as well as the ability to follow a steady state when the system is noisy. Additionally the controllers are tested with extra harmonic mitigation tactics such as additional integrator terms in the frequency of the harmonic component, changing the placement of the current sensor and the controllers are also tested with a Smith predictor that looks one and two steps ahead.

The conclusion is that for the base system the LQG is the absolute best controller because it follows the reference closely and reacts quickly to great changes while still having very low amounts of oscillation. The LQR is the second strongest controller but has oscillation problems. However the LQG underperforms if the grid impedance is changed and the model is not updated. Since it is a model based controller that also uses a model based Kalman filter, it overreacts a lot if the grid impedance is weak. The result is an unacceptable amount of oscillation which leads to the conclusion that a PI might be better when the grid is unknown or that the grid needs to be estimated before using the LQG. The problems discussed for the LQG are the same for the Smith predictor, it gives an overall stronger performance to the system but causes oscillation when it poorly estimates the true grid impedance.

Welcome!

Anna, Erik and Torbjörn