Induction motor rotor time constant inverse estimation using neural network
Boukhemis Chetate
This paper proposes a new estimator for induction motor rotor time constant basing on a simple hidden layer neural network. An online algorithm is developed for the training of the suggested neural network estimator parameters. This algorithm use the model reference principle by comparing measured and estimated stator currents. So, beside the measured currants which present the reference model (ideal model), it suggested to reconstitute these currants via the motor model which present the actual model (real model). thereafter the errors between measured and estimated currants are exploited for the neural network parameters adaptation. The obtained results show that beside its simplicity, the proposed estimator provides satisfactory performances even at very low speeds.