An error driven hybrid neuro-fuzzy torque/speed controller for electrical vehicle induction motor drive

The paper presents a novel elastic neuro-fuzzy speed/torque controller for electric vehicle induction motor drives and is based on artificial technologies (fuzzy logic and neural network). The salient feature of this technique is the hybrid control action and online output scaling between the fuzzy logic(FLC) and neural network(ANN) controllers, based on the induction motor absolute value of the normalized speed error and using an adaptive weighting factor(B). The hybrid action tolerates any inaccuracies in the fuzzy logic assignment rules or in the neural network stationary weights and online tuned weights. The neural network based controller is not required to be fully trained and the neural network weights need not be exact as they are tuned online using the error driven back-propagation algorithm. The reinforcement signal used in online training is the actual motor normalized speed error. The outputs of the two separate controllers are combined and scaled to adjust the inverter switching frequency and the output voltage of DC/DC chopper of the induction motor drive system.