Online identification for auto-tuning PID based on wavelet neural networks: An experimental validation on an AC motor

Based on adaptive wavelet neural networks, a PID discrete control scheme for induction motor drives is presented. An auto-tuning wavenet scheme is synthesized for the PID feedback gains by means of gradient-descendent algorithm through online plant identification. The latter uses input and output data for a radial basis neural network (RBNN) with different daughter wavelet activation functions. Infinite impulse response filter in cascade with the output of the RBNN is introduced to prune nodes that contribute poorly for an efficient identification even though gradient-descendent is used. A comparative experimental study under real noisy conditions and no load perturbation shows results for different wavelets (Morlet, RASP1, POLYWOG1 and Shannon); better performance of Morlet and RASP1 wavenet-based tuning are obtained due to its better identification properties.