Aplicación de técnicas de identificación y control multivariable en convertidores en fuente de tensión conectados a la red eléctrica

Abstract

This thesis is focused on the design of the control of Voltage-Source Converters connected to the grid through LCL and L filters. Particularly it is focused on the improvement using multivariable techniques of the design and operation of the inner control loop of a grid converter: the current control. To that end, the thesis follows three main topics. The first main topic is identification of the VSC and filter model. The thesis proposes the application of techniques of multivariable identification to the Voltage-Source Converter connected to the grid by means of an LCL filter. The objective is to obtain a discrete model in the dq-frame that permits to directly synthesise a multivariable current controller. Three multivariable methods have been proposed: MOESP, N4SID and PEM. The thesis deals with determining the adequate identification method and the kind of inputs that should be applied to obtain a correct model estimation. Besides, an identification method based on linear regression and SISO models is proposed to estimate the values of the LCL filter parameters. The second main topic is the application of multivariable control techniques to the problem of the current control of the Voltage-Source Converter with LCL filter. The proposed MIMO controllers are based on servo controller structure and the optimal LQ technique. It could be combined with the use of full state vector estimators to reduce the number of necessary sensors to carry out the control. The thesis applies the combination of the LQ servo controller and Kalman Filter to the current control problem in noisy environments, resulting in the proposal of the LQG servo controller. The thesis studies a KF implementation, the so-called SSKF, which permits to reduce the computational complexity and facilitate his implementation on the control platform. The thesis contributes also with some design criteria and an algorithm based on Genetic Algorithms to auto-tune the parameters involved in the control design, helping in this way to the designer. The proposed algorithms have proven their versatility, being suitable to different current control problems. Besides, a structure based on the LQ servo controller and the use of resonators is proposed to mitigate the effects of grid harmonics in the current control. The formulation of the new structure still permits use the design tools applied in the basic structure. The third main topic is related with the adaptive control. The proposal deals with the improvement of a basic previous control approach for a Voltage-Source Converter with L filter. This previous approach is very simple by operation in the stationary reference frame as well as lack of AC line currents sensors but is very sensitive to variations of the line inductances. The modifications proposed in this thesis overcome this disadvantage using a Fuzzy Logic Controller (FLC) and grid current reconstruction to tune parameters and do more robust the basic control in view of line inductance parameters variations.