System identification of PWM dc-dc converters during abrupt load changes

Abstract

System identification has an important role in control system design, particularly in auto-tuning and adaptive control applications. Recently, it has emerged as a significant research topic in power converter design. For this reason, this paper proposes a new on-line approach to identify the parameters of a dc-dc converter. The proposed method is capable of rapidly detecting and accounting for abrupt load changes during transient periods. In most situations the controller is blind to this change. However, the estimation algorithm is able to update the model parameters before the output reaches the peak overshoot value. As a result, the controller can minimise the effect of any parameter change. The proposed method depends on prediction error variation during a rapid change of load and is designed around a simple fuzzy logic structure. An adaptive forgetting factor is used to optimise the identification process, and this varies very quickly and smoothly with the step load change. Simulation results show that the identification model matches the plant during the transient period. Importantly, the convergence rate and parameters error, the two factors use to validate the algorithm, are very good during system startup and after any abrupt load change.