Efficiency-optimized speed control for a separately-excited DC motor using adaptive neurocontroller

The artificial neural networks (ANN) direct inverse control and the direct adaptive control are presented in this paper to control the armature voltage and the field voltage of the separately-excited DC motor to yield maximum efficiency speed control. The optimal ratio K between the armature current and the field current giving minimum losses is analytically derived as a function of speed. Two identical ANN's have been utilized, one for armature control, the other for field control. An online training algorithm is presented for efficient and stable operation rather than fixed weights and biases for the ANN's. Experimental results are recorded for step change in the load torque and step change in the reference speed. Good agreement was found between the simulation results and the experimental results. A distinct improvement in the system efficiency is observed especially at light load conditions.