Verfahren zum Einsatz eines Beobachtermodells zur Drehmomentenschätzung und -vorhersage für eine Asynchronmaschine

摘要

A hybrid state observer is employed with observer errors and measurements taken from the motor as inputs. The resultant torque error of the observer is used to train a neural net such as a radial basis function net (RBF) consisting of three layers of neurons. During operation of the motor, the neural net delivers a correction to the current value of torque computed by the observer, to improve the accuracy of prediction. Such a structure can be used where the torque must be estimated as precisely as possible but is not directly measurable.