Using optimized transient excitation technique to stabilize model-based sensorless control of induction machine

High dynamic control if induction machine is only achievable by some sort of field oriented control. Realizing this type of control without using a shaft sensor has the advantage of increased reliability and decreased costs. Although used in the higher frequency range, methods considering only fundamental wave quantities of the machine deteriorates in their performance at low or zero stator frequency due to the lack of signal and an increasing influence of noise and parameter uncertainties. The only way to establish a stable and controlled operation at and around zero stator frequency so far is to utilize non-fundamental effects of the machine. These effects can be exploited by determining the high frequency or transient electrical parameters of the machine. Parasitic effects like spatial saturation, slotting, or anisotropy of the rotor influence these transient electrical parameters. Any movement of one of these saliencies thus leads to a modulation of the signal obtained enabling thus the detection of the flux and/or rotor position. These estimated saliency position signals can be noisy what may cause problems in the control loop. Using conventional filtering methods leads to a reduction of the dynamic control performance. In this paper an excitation of the machine with voltage pulses is applied and methods to reduce disturbing harmonics in the control signals as well as ways to realize effective filtering are described. The reduction of disturbing harmonics is based on a compensation of non-ideal inverter/sensor properties and can be realized by optimization of the excitation pulses and/or placing of the sample instants. The effective filtering without loosing dynamic performance is based on a combination of the transient excitation method and a fundamental wave model of the machine. The signal obtained from the transient excitation is used to stabilize the model what results in a smoothed output much more suitable for the machine control. The advantage of this - tructure is, that the dynamic performance is not reduced, as is the case with conventional filtering methods