Neural-net-based adaptive PID regulator with attenuating excitation signal

A NN adaptive PLD (NNAPID) regulator for complex plants with unknown models is proposed. To enhance the on-line self-learning ability and robustness, an attenuating excitation is introduced to excite all modes of the plants and produce the error needed for the self-learning process. To realize the self-learning and self-tuning, a function to evaluate the control effect is introduced to choose the learning samples from the on-line data.