Online training of neural network control for electric motor drives

Abstrad This paper proposed a simple on4ine trained neural network direct adaptive control for dc 'drives with compensation for learning saturation The neural network direct adaptive algorithm presented doesn't need any offline training or computing of the Jacobian m a t h of the plant controlled A new compensation approach for learning saturntion is developed and applied in order to overcome the poor convergent speed this may cause Simulations are carried out wi?h variations in armature resistance and measuring channel noise The simulation res& have proved that the proposad scheme has good robustness and convergent speed for real time controL