Abstract:
Permanent magnet synchronous motors (PMSMs) are replacing induction motors because of
their better performance in industrial applications. Due to the availability of fast DSP processors
now-a-days complex control algorithms can be implemented efficiently. In conventional-drive
circuits the six-switch inverter ruled for a long time because of the ease of its control; however,
a four-switch inverter provides the cost effective efficient replacement. The dynamic model of a
machine provides an insight into the dynamics of its operation and control.
A PMSM has been modeled in Matlab Simulink in different frames of references and the model
has been validated under different operating conditions. The space-vector algorithm of a six-
switch three-phase inverter has been modified for the control of a four-switch three-phase
inverter (FSTPI) by establishing the symmetry between the voltage space-vector planes of both
inverters. A FSTPI exhibits the current-unbalancing problem because of the unbalanced voltage
distribution across the capacitor. The current-unbalancing issue has been resolved by
introducing a ‘duty waveform drift’ method. Artificial neural networks (ANNs) have been used
to optimize the performance of a four-switch inverter. The detailed modeling of the space-vector
technique for both the conventional and four-switch inverters has been presented.
A PWM rectifier has been implemented using sinusoidal pulse width modulation and space
vector pulse modulation to ensure that the drive maintains unity input power factor. A modified
optimized space-vector algorithm, used for the inverter operation, has been introduced in the
control of PWM rectifiers to ensure the best computational performance.
There exist a number of control techniques for a PMSM: direct-torque control (DTC) and field-
oriented control (FOC) have been investigated in detail for the proposed four-switch inverter.
The flux, torque, speed and current controller designs for both the DTC and FOC are presented
in detail and have been implemented on Texas Instruments’ digital signal processors using
Matlab Simulink’ code generation facility. Because of the ANN based modules the designed
digital system presents precise and accurate speed, current, flux and torque tracking with four-
switch space-vector modulator up to 30-kHz sampling frequency. The ANN has reduced the
implementation' complexity of the speed and position estimator. The initial rotor-position
detection has also been incorporated while collecting the data for the training of ANNs. The
results have been presented for evaluating the performance of a controller with the proposed
four-switch inverter and ANN based estimators.