Recurent neural network based IM state estimation

The field oriented control also known as the vector control is a useful high-performance technique to control an induction motor (IM). With high-performance control of IM are used field oriented controlled drives where there are needed state variables as rotor fluxes, stator fluxes and rotor currents to be known. In particular for speed sensorless IM control, estimation of the rotor fluxes that can not be measured directly is very important. For high-performance IM control, estimation of unmeasurable state variables as well as estimation of changing parameters or the parameter adaptation is also of great importance.In this thesis study, state variables of state space mathematical models of the induction motor based on d-q axis system has been organized primarily. After, asymtotic observers, Kalman Filter (KF) and Extended Kalman Filter (EKF) algorithms and Artificial Neural Network (ANN) algorithms based on the state estimation has been investigated for different operating conditions for the high performance field compatible IM control. To estimate the rotor flux components especially for indirect field oriented control there has been proposed two new estimation algorithms based on Elman Artificial Neural Network (ENN) and PI-ENN. Proposed algorithms and EKF algorithm has been tested separately with on-line and off-line simulational and experimental IM measurements based on under different working conditions with different waveformed supply voltages. For estimation and actual results obtained by the devoloped algorithms and EKF are compared with each other with making the necessary examinations.