Measurement schemes with reduced number of sensors for modular multilevel converter

Abstract

During the last four decades there has been considerable development in voltage source converters (VSCs), which are widely contributed in multilevel converter topologies. Since then, multilevel VSC topologies have been used for applications with different power rating owing to the improvement of the output waveforms quality and minimising filtering requirements. In comparison with the conventional multilevel converters, modular multilevel converter (MMC) is considered as the most attractive topology for high and medium-power applications mainly due to the series connection of a high number of submodules (SMs). The challenges associated with the implementation of a high number of SMs includes: voltage-balancing of the distributed SM, cost, reliability and the increased complexity in the circuit configuration. Furthermore, achieving efficient and fast closed-loop control of the MMC requires the accurate knowledge of the voltage and current measurements, which means a considerable number of sensors are usually required to operate the MMCs. The main objective of this research is to propose several novel strategies for the converter to achieve voltage-balancing with fewer number of sensors to produce comparable performance to the sensor-based method. Four different sensorless schemes have been investigated, where two are current sensorless-based techniques and two are voltage sensorless-based techniques. The proposed current sensorless schemes are based on developed sorting algorithm, and the proposed voltage sensorless schemes employ two novel different recursive algorithms with the standard sorting algorithm. In regards to the voltage sensorless schemes, the first proposed method uses an exponentially weighted recursive least square (ERLS) algorithm, while the second proposed method employs a Kalman filter (KF) to estimate the SM capacitor voltages. Capacitance uncertainty has been investigated for the proposed voltage sensorless schemes. The proposed methods have been implemented via simulation but also on a scaled-down laboratory prototype. II The thesis also deals with capacitor diagnosis where a new scheme has been proposed which may be used for health monitoring technique, a comparison with an existing technique has been evaluated. Detailed simulations and experimental tests are carried out to investigate the performance of the proposed sensorless schemes, and results are compared with the sensor-based approach. These various schemes have been implemented and tested in real-time using a commercial floating point microcontroller where a 4-level single-phase MMC was employed. The results achieved for these novel schemes show an important improvement in the performance of the MMC under different operation conditions while fewer sensors were used.