Sensorless vector control of induction motors at low speed

ABSTRACT PhD Thesis SENSORLESS VECTOR CONTROL OF INDUCTION MOTORS AT LOW SPEED Hayrettin CAN Firat University Graduate School of Natural and Applied Sciences Department of Electrical - Electronics Engineering 2003, Page:99 Vector control is known to have good transient and steady-state performance in induction motor drivers. It involves direct stator flux orientation methods that do not need speed information and are insensitive to parameter variations except stator resistance. However, the performance of these control strategies depends on accurate estimation of the stator flux. The voltage model is one of the methods used for estimating the stator flux. In this model, the stator flux can be computed by integrating the terminal voltage reduced by the ohmic losses. However, at low frequencies large errors occur due to the variation of the stator resistance, integrator drift, and noise. In this thesis, novel mathematical derivations and soft computing methods are developed to increase the performance of the drive, especially at tow speeds. First, existing integration algorithms used in the voltage model are discussed, and compared to each other. These integrators have magnitude and angle errors in the stator flux. Then, a new compensator is proposed to solve the problems associated with the integrator. The new compensation method is computationally fast due to underlying simple mathematical formulas. Additionally, a new fuzzy controller is developed for the feedback loop of the integrator. The most important advantage of the fuzzy controller is to provide a robust structure and simple design compared to the other methods. Computer simulations and experiment results of the algorithms mentioned above are also presented. Moreover, to reduce die voltage error between the reference and the actual stator voltages, an artificial neural network-based voltage compensator is designed and verified by empirical results from experiments. One of the main contributions of the newly developed system is that it accurately aligns the actual stator voltage with the corresponding reference voltage using a neural network-based compensator to correct for nonlinear voltage discrepancy due to the quantization error in the controllers, discrete-time processes, forward voltage drop of switching devices, and dead-time of the inverter. The nonlinear difference between the actual stator voltage and the corresponding reference voltage is a function of rotor speed and load, and is crucial especially at low speeds and light loads. XV