Speed and position estimation for the speed sensorless control of induction motors using neural network

ABSTRACTMaster ThesisSPEED AND POSITION ESTIMATION FOR THE SPEED SENSORLESS CONTROLOF INDUCTION MOTORS USING NEURAL NETWORKAhmet ÖZMENFırat UniversityGraduate School of Natural and Applied SciencesDepartment of Electronics and Computer Education2006, Page : 69Speed and position information of induction motors are obtained from the mechanicalsensor mounted on the motor shaft, which is required for the speed control of the motor. Thesesensors have an important effect on the motor cost and, due to the mechanical strengths theyincrease the risk of the motor failure. Therefore, a speed observer, which estimates the motorspeed using the motor current and voltages, is required for the speed control of inductionmotors. Since the motor speed and position information have an important effect on the speedand vector control performance, a robust observer should be designed under the load andparameter variations.In this thesis, a neural network (NN) speed observer is presented for the sensorlessspeed control of induction motors. It is aimed that the NN observer estimates the motor currentsand fluxes by using the measurements from the actual motor currents and voltages and then itcalculates the motor speed using the currents and fluxes errors. The structure and initial weightsof the NN observer are determined from the dynamical model of the motor. For the parametervariations, training of the NN can be continued using the current errors. Speed sensorlesscontrol of induction motor including the trained NN observer is simulated with MATLAB-Simulink. Simulation results showing the performance of the control system are presented underthe load and parameter variations.Keywords: Induction motors, speed sensorless control, artificial neural networks, speedobserver.