An efficient energy controller for an induction motor drive with compensation of temperature using neural networks

A neural network (NN) control is used to implement optimal flux in a vector control variable speed drive. A model of the motor losses allows the NN to calculate the flux for a maximum efficiency in each working point of the induction motor (IM). Taking into consideration the fact that nominal flux is not necessary for load torques below the rated torque of the IM, losses can be transferred from iron to copper decreasing the real flux and reaching the minimum total losses point. A complex loss model of the motor, including magnetic and thermal deviations of its parameters, is used to estimate losses. This can be useful in many applications where induction motors work below the nominal torque most of the time. To account for iron and copper losses, some inductances and resistances of the IM must be computed. Estimators are implemented for this purpose. Changes in resistance values due to temperature and in inductances due to iron saturation curves are also computed. Algorithms for computing the optimum flux with NN and vector control with this optimum flux were implemented in a DS1102 controller board from DSpace. In the sub-system for flux computation the neural network is trained to estimate the optimum rotor flux. Inputs to the NN are torque, speed and estimated rotor resistance of the IM and the output is the rotor flux.