Improvement of Transient Response of PI Controller with Reference Modification for Digitally Controlled DC-DC Converter

In the switching converter control, a digital feedback control is widely used to regulate the output voltage of power supply. However, its feedback gain selection becomes more sensitive and difficult since the switching frequency tends to be faster in recent converters to satisfy the requirements of miniaturization and power efficiency. Therefore, a novel control method is needed to improve the transient response of output voltage effectively. Recently, the reference modification method has been proposed to improve the transient response. It is expected to have a superior performance since it can be adopted without sensitive gain selection. This paper investigates improved characteristics by employing the reference modification method based on a neural network predictor for a digital PI controller of dc -- dc converters. The reference modification is a method which is added to the conventional feedback controls without gain selection such as auto tuning method during the transient state to improve the transient response. Therefore, it is expected that the unstable phenomena from gain selection can be avoided. Additionally, the reference modification is realized by a neural network predictor, which learns the dynamical behavior of the output voltage of converter. The designing of the feedback gain parameter becomes a easier task since the reference modification can adjust the command value to be more suitable one without changing the gain parameter. The effect of improvement is shown in simulated results and it is confirmed that the reference modification can be improve the transient characteristics compared to the conventional digital PI control.