Sensorless speed control of induction motor using sliding mode and neuro-fuzzy observers

ABSTRACTPhD ThesisSENSORLESS SPEED CONTROL OF INDUCTION MOTOR USINGSLIDING MODE AND NEURO-FUZZY OBSERVERSCafer BALFırat UniversityGraduate School of Natural and Applied SciencesDepartment of Electrical and Electronics Engineering2007, Page:120Vector control technique is desired for high performance control of induction motordrives. A shaft encoder is used as a speed and position sensor in order to measure the motorspeed required for the control of motor drive systems. The encoder mechanically coupled to themotor shaft reduces the operation reliability and increases the cost of the drive system.Additionally, in some applications such as high speed drives, the encoders cannot be coupled tothe motor shaft. Therefore, sensorless speed control of an induction motor is needed. In the mostof sensorless speed control methods used in literature, the motor speed is determined from theestimated motor currents and/or fluxes using the dynamic equations of the motor. However,robustness and stability of the speed estimation cannot be achieved with the designed observersusing the dynamic model of the motor due to the unmodelled dynamics and parametervariations which change with the operating condition of the motor.In this thesis, a fuzzy neural network current observer is proposed for the sensorlessspeed control of induction motor. The proposed fuzzy neural current observer offers a differentapproach than the current methods since it is used as a direct controller instead of the flux andthe speed estimation. Sliding mode current controller is designed for the current control in thesystem. Thus, the proposed sensorless control method is not affected from the parametervariations. Fuzzy neural network current observer is trained online with the experimental data.Performance of the control system including the trained fuzzy neural network current observerXIIis tested for various operating conditions of the motor. Training algorithm of fuzzy neuralnetwork and the control algorithm are prepared at the MATLAB/Simulink environment and it isimplemented using the digital signal processor DSPACE-DS1104 card. Experimental results areacquired from the developed interface using Control Desk Developer software which allowsonline access to the variables. The performance of the fuzzy neural current observer iscompared with the current observer designed using the dynamic model of the motor. Theperformance of proposed sensorless control method is verified with the simulation andexperimental results under different speed and load conditions.Keywords: Induction motor, fuzzy neural network, sensorless control.XIII