Sensorless Passive Control Algorithms for Medium to High Power Synchronous Motor Drives

This study is focused on the definition of sensorless algorithms for Surface-Mounted Permanent Magnet Synchronous Motors (SM-PMSM) and Electrically Excited Synchronous Motors (EESM). Even if these types of motors are rather different from a constructive point of view, they have some common issues regarding sensorless drives. Indeed, SM-PMSMs, which are usually used for low-medium power applications, have a low rotor anisotropy, therefore it is complicated to use sensorless active methods (which are based on high-frequency voltage injection), due to the low signal to noise ratio. On the other hand, active methods on high-power EESM have the drawback of high torque ripple. For these reasons, both for SM-PMSM and EESM, it is interesting to define and use sensorless passive algorithms (i.e., based on observers and estimators). The drawback of such algorithms is that their performance deteriorates significantly in the low-speed region. The aim of this thesis is to define a robust sensorless passive algorithm that could work in a wide speed region and that could start the motor from standstill even with a high load torque. The initial objective of the work is to find, among the various algorithms proposed in the technical literature, the most promising one. For this purpose, four different algorithms are selected. They are chosen considering the most recent articles presented in the technical literature on high reputable journals. Since many improvements are proposed in the literature for the different algorithms, the most recent ones are candidates for being the ones with higher performance. Even if the experimental tests of the four different algorithms are shown in the literature, it is difficult to evaluate a priori which offers the best performance. As a matter of facts, for each algorithm different tests are carried out (e.g., different speed and torque profiles). In addition to that, motor sizing and features are different. Moreover, the test bench characteristics can significantly affect sensorless performance. As an example, inverter features and non-linearities (e.g., switching frequency, dead times, parasitic capacitance) and current measures (e.g., noise, linearity, bias) play a key role in the estimation of rotor position. The added value of this thesis is to perform a fair comparison of the four algorithms, performing the same tests with the same test bench. Additional tests are performed on the most performing algorithm. Even if this sensorless technique is already proposed in the technical literature, a methodology for observer gain tuning is not shown, which is proposed, instead, in this thesis. Moreover, the algorithm is enhanced by adding a novel management of direct axis current, which ensures the stability during fast transient from medium-high speed to low speed. The algorithm is tested with different test benches in order to verify the control effectiveness in various operating conditions. As a matter of facts, it is tested at first in the University of Genoa PETRA Lab on two different test benches. The first test bench is composed of two coupled motors, in which the braking motor could realize different torque profiles (linear torque, quadratic torque and constant torque), whereas in the second test bench the motor is coupled with an air compressor, which is a demanding load since high and irregular torque is applied at standstill. After the test at the University of Genoa, the algorithm is implemented in Phase Motion Control and Physis drive and tested on a six-meter diameter fan. Regarding the EESMs, for these type of motor is necessary to estimate the stator flux amplitude and angle. Indeed, the stator angle is usually used to perform the Park transformations in the FOC scheme and the stator flux amplitude is used to control the excitation current. In this study, the RFO is adapted for estimating the stator flux of an EESM. Regarding the control for EESM, it is tested on a simulative model for high-power motors provided by NIDEC ASI and tested on a small-scale test bench at the University of Genoa