Speed estimation in rotating reference frame

An angular velocity neuroestimator synthesized in the rotating reference frame is demonstrated.

This model illustrates the possibility to use a feedforward neural network (static neural network) to estimate (more precisely to approximate) mechanical speed of the asynchronous motor. The main challenge in such a task is to construct an effective approximation base. In http://www.mathworks.com/matlabcentral/fileexchange/48012-speed-sensorless-induction-motor-drive 6 heuristic signals have been proposed. Here the very basic 4 signals, i.e. stator voltages and currents, span the approximation space. These 4 signals are observed in the flux-oriented rotating reference frame and thus one of the key objectives to keep them constant at a steady state is fulfilled.