An improved flux and torque estimation strategy of speed sensorless induction motor

The direct torque control has emerged recently as a popular control technique for high performance induction motor drives by means of machine parameters minimization, algorithm complexity reduction and linearization feedback. The induction motor torque effectively relies on constructing the stator flux which is closely related to the voltage drop on the stator resistance. A new method of stator resistance identification using wavelet network is brought forward to improve the low-speed dynamic performance of induction motor. Considering advantages of wavelet transform in time-frequency domain, the combined information can be obtained from both magnitudes and arguments of complex coefficients. Then the time-varying signal characteristic will be extracted and increases the control precision for direct torque control. The improved orthogonal least squares algorithm is used to complete wavelet network training and parameter initialization, and then the accurate stator flux vector and electromagnetic torque are acquired, optimizing the inverter control strategy. During training phase, the wavelet network not only learns adequate decision functions defined by the weight coefficients, but also looks for parameter space which fits for reliable input signals. The simulations as well as experimental results are shown to illustrate the improved performance of the proposed method, indicating torque ripple reduction and system stability.