DOI: 10.35377/saucis...2015758 ISSN: 2636-8129

Discrete-Time Modeling and Observer Design for a MISO Power Converter: A Comparative Study of Discretization Methods for Prediction and Current Estimators

Onur Demirel
Show PDF Cite
This study provides a comprehensive analysis of estimation and discretization structures for discrete-time state observers applied to a continuous-time multiple-input single-output (MISO) DC–DC buck converter. The analysis focuses on model accuracy, estimation performance, noise sensitivity, and computational cost. The continuous-time state-space model of the power converter is established, and discrete-time models are derived using zero-order hold (ZOH), first-order hold (FOH), and the Tustin method. Discrete-time observers are formulated as prediction-estimator (PE) and current-estimator (CE) structures, reflecting the sequence in which measurement information is incorporated into the state estimation process. The ability of each discretization method to represent the continuous-time system dynamics is evaluated through frequency-response analysis, and the correspondence between continuous- and discrete-time models is assessed. The dynamic performance of the observer structures is compared based on estimation error and sensitivity to measurement noise. Furthermore, the arithmetic computational load per sampling period, execution times, and embedded hardware resource requirements of the discrete-time observer algorithms are analyzed. The results indicate that discrete-time observer performance should be evaluated not only by model accuracy, estimation error, and noise robustness, but also by real-time implementability and computational efficiency. Among the structures examined, the ZOH–CE variant is particularly suitable for embedded-system applications, achieving adequate model accuracy, good estimation accuracy, and noise robustness while minimizing arithmetic operations, execution time, and hardware resource usage. This study clarifies the trade-offs among model accuracy, estimation performance, noise robustness, and implementation cost in discrete-time observer design, and provides a comprehensive evaluation framework for selecting appropriate observer structures and discretization methods for real-time embedded power-electronics applications.

More from our Archive