Adaptive neural network temperature control for thermoelectric refrigeration systems using online self-learning mechanism.

This article presents a new adaptive neural network control method designed for thermoelectric refrigeration systems. This approach leverages the capabilities of neural networks to address the uncertain nonlinear dynamics within the system. A self-learning mechanism is developed to enable the online training of the neural network’s weights, as well as the center points and widths of the basis functions, resulting in improved control performance. Additionally, an adaptive law based on a projection algorithm is introduced to prevent potential parameter drift and singularities of the basis functions. The stability of the closed-loop system is analyzed using Lyapunov stability theory. To demonstrate the effectiveness of this proposed method, both simulation and experimental results are presented. Compared with traditional neural network control method, our control method reduces the maximum error by 29.4% and the set time by 39.3% in the experiment.