Introducing the rank-weight method into the

basic ant colony optimization (ACO), we use the modified

ACO to optimize the weights and thresholds value of neural

networks (NN). And when the BPNN is being trained, this

method can solve the disadvantages of running into the

minimum easily, and enhance the convergence speed. So we get

a heuristic method, which is good at time efficiency and

derivation efficiency that is ACOrw-BP. Then this method was

used to identify the speed of the motor in direct toque control

(DTC).The results of the simulation showed that: the modified

ACOrw-BP neural network not only has the ability of mapping

widely, but also enhancing the operation efficiency obviously.

The speed of the motor can be identified accurately by this

method, and the result is good. So, it implements the direct

toque control of speed sensorless.