• Citer

Modelling, Design, Implementation and Control of Grid-Tied Inverters in order to Reduce Unwanted Oscillations and Parasitic Capacitor Currents

par Zamani, Hasan
Président du jury Garone, Emanuele
Promoteur Gyselinck, Johan ;Abbaszadeh, Karim
Publication Non publié, 2023-12-19

Thèse de doctorat

• Accès en ligne
• Détails
• Contenu
• Statistiques

Résumé :

Global warming due to greenhouse gases is undeniable, and renewable energies have attracted attention as a solution to replace fossil fuels. The energy transmission system of renewable energies usually includes a DC-AC inverter that injects power into a power grid.To inject current with high quality and meet grid standards, an L filter or an LCL filter is usually integrated between the grid and the inverter. The advantages of the use of an LCL filter are the remarkable attenuation of switching harmonics, lower weight and size of the total hardware, and less cost-expensive with higher efficiency. But the LCL filter generates resonance that can make the closed-loop system unstable. Damping methods are generally divided into Passive Damping (PD) and Active Damping (AD). AD needs an extra sensor for measuring a state variable of the system, which increases the total cost and reduces reliability. PD adds a passive resistor in parallel or series with one element of the LCL filter and thus increases losses.Designing a high-performance control method that does not degrade the ability of the LCL filter to attenuate switching harmonics is the aim of this research. This control method removes the need for AD or PD and respects the requirements such as high performance and zero steady-state offsets in tracking the references, and robustness against the grid-inductance variations and mismatches in the LCL filter elements. A variety of control methods such as Sliding Mode Control (SMC), Deadbeat Predictive Control (DPC), and State-Feedback Control (SFC) are developed for grid-tied inverters. Basically, these control methods do not eliminate the need for AD or PD. Model Predictive Control (MPC) has received overwhelming attention in recent years in both industrial and academic societies.In this research, the concept behind CCS-MPC is developed for the LCL filter based grid-tied inverters. The proposed controller is designed in both stationary and synchronous frames. Regardless of the reference frame, the proposed controller shows inherent robustness against the LCL filter resonance. The optimization process of this control method is performed offline and just some simple numerical expressions are executed in real-time. Therefore, it does not need a powerful and expensive processor for implementation.Working with DC variables is the advantage of the design in the synchronous frame, and there is no offset in the response. However, the model order of the LCL filter in the synchronous frame is relatively high. In addition, there are cross-coupling terms between the dq axes. To deal with the latter, a Multi-Input Multi-Output (MIMO) CCS-MPC is designed, and to reduce the order, system identification is used to build the model.There is no coupling between the axes in the stationary frame. But the controller has to track sinusoidal references. It is shown that by parametrizing the control law's matrices while considering the reference trajectories, the offsets disappear and are close to zero.The parameters of the proposed controller are selected with sensitivity analysis in both frames. With this selection, the proposed controller shows enough robustness against grid inductance variations.The simulation and experimental results show the inherent robustness of CCS-MPC in both frames against the LCL filter resonance and the good performance in reference tracking.The other approach to reduce the number of sensors is to measure the capacitor voltage of the LCL filter, which is also used to synchronize with the grid. The derivative of the capacitor voltage results in the capacitor current that can be used for AD. But, traditional discretization methods like the backward Euler method, forward Euler method, and Tustin method do not preserve the phase and magnitude of the derivative operator around the resonance frequency. In this research, the construction of a discrete derivative function that is valid in a desired frequency range is formulated. The simulation results show the effectiveness of the proposed method for the digital implementation of the derivative operator.