Research On The Control Methods Of The Speed Sensorless Field Oriented Control Of Induction Motor Drives

The rotor field oriented control has brought essential advances in AC variable speed drive system. Speed sensorless induction motor drive promotes the simplicity and robustness further, and two problems must be solved in the system: the speed estimation and rotor flux observation. Based on the former researches, the speed estimation and rotor flux observation methods are studied using the theory of Model Reference Adaptive System. Considering the time variability of the stator and rotor parameters of induction motors, the methods make real-time identification of the parameters to retain the dynamic performance of the drive system. The simulation results show the MRAS-based field oriented control system has good static and dynamic performance.To great extent the performance of field oriented control depends on the precise measurements of motor parameters. The initial values of motor parameters can be measured by the no-load and locked rotor tests. But some parameters will vary with the temperature rise and magnetic saturation, and they must be real time identified. Just using the stator currents, stator voltages, and velocity, the method proposed in the paper makes real time identification of induction motor parameters based on the least squares identification algorithm. The method doesn't use the rotor flux signal, avoiding the coupling between the rotor flux observation and the parameter identification. After identifying the parameters, because of the multiplication terms of state variables, the induction motor model is still the non-linear state equations. To estimate the state variables of motor model and gain the rotor flux and velocity signals, the paper proposes a method to estimate them using extended kalman filter. The simulation of the method gets satisfied results.Artificial neural networks have mighty learning ability. After being properly trained, the multiplayer networks are capable of approximating any nonlinear functions with any precision. Therefore it has become a powerful tool in nonlinear system identification field. The paper makes research into the multiplayer feedforward networks and dynamic recursive networks, and proposes a method to estimate the speed and rotor flux of induction motors using the dynamic recursive networks. To the used dynamic recursive network model, the off-line dynamic BP algorithm has been reasoned out so as to observe induction motor state variables. The paper makes simulation studies into the field oriented control systems based on multiplayer feedforward networks and dynamic recursive networks respectively, and compares the convergence speed and approximate extent between the two learning algorithms.Experiment is based on the DSP design system for digital motor control. Software programs carry out extended kalman filter algorithm to estimate the rotor speed and fluxes, and space vector pulse width modulation algorithm. The satisfied experimental results prove that extended kalman filter algorithm can real time estimate rotor speed and flux very accurately, and based on which the speed sensorless drive system has good static and dynamic performance.