Resumo:
The purpose of
this work is the implementation of a method for
estimation of the dynamic torque to induction motor, using the three phases
voltages and currents samples, with a DSP (Digital Signal Processor) based
circuit tool.
Nowadays, the vector control is used in drives where high-performance
dynamic is required. The flux is measured directly in the gap, with Hall effect
sensors, or another measurement technique. The gap measured flux feedback
is made available to the drive system and the currents, which produce torque,
are synthesized. The direct flux measurement is expensive and the process
accuracy may not be enough.
currents and then, the stator
Another process consists in the measurement of the stator voltages and
flux is calculated by means of equations
processing. The process
strategy
is
the stator
flux
estimation by
the
programmable cascaded low-pass filter (PCLPF). This method is well known
because it permits the stator voltage integration, since extremely low frequency
to high frequency on the field-weakening range, without introducing any offset at
the output.
Besides the stator voltage and current measurement, the only parameter
under consideration to the stator flux estimation, using the programmable low-
pass filter (PCLPF) method, is the stator equivalent impedance, where the
significant component is the resistance. This stator flux estimation method
eliminates flux and speed sensors, decreasing cost and making the system
reliably better.
It is important to notice that the programmable cascaded low-pass filter
(PCLPF) was substituted by recurrent neural network (RNN), which permits the
fast and simple flux estimation. The recurrent neural network (RNN) response,
compared with another digital signal processing based PCLPF implementation,
shows equal behavior at steady state, but is superior at transient response. A
Kalman filter based algorithm is used to recurrent neural network (RNN)
training.