A study on transient response of digital PID control for DC-DC converter with weighted predictions of neural network

Abstract

In this paper, a reference modification method for the PID control is studied to improve the transient response of the output voltage of the power converter using weighted prediction information with a neural network predictor. Several predictions of the output voltage with the neural network are used to obtain the modified reference value in the PID control by minimizing the objective function. The objective function uses weight factors which can represent how the predictions are treated. In this study, a buck type dc-dc converter is used as a test study to evaluate the presented method. The evaluation results show the presented method has a superior performance compared to the conventional PID control. It is confirmed that the presented method provides a function of prediction based approach, which is widely required in the field of renewable energy since the input-output relation of power conversion varies widely and dynamically.