Model Order Reduction of High Gain DC-DC Converters using Proper Orthogonal Decomposition

This study aims to address the challenges of excessive computational burden and control complexity in high-gain, voltage-multiplier DC-DC converters by seeking compact yet accurate surrogate models. We implement the Proper Orthogonal Decomposition (POD) algorithm in MATLAB on a fifth‑order converter, generating snapshot ensembles from impulse responses and constructing Reduced-Order Models (ROMs) of orders 1 through 4 via projection onto dominant singular modes. A comparative evaluation of the time‑domain Mean‑Square Error (MSE) and the semilogarithmic Bode magnitude and phase error metrics demonstrates a clear monotonic decline in error with increasing order. The fourth‑order ROM achieves amplitude deviations below 0.05 dB and phase errors under 1°, whereas the third‑order surrogate delivers a 40% reduction in state dimension with under 0.1 dB amplitude loss. These results confirm that moderate‑order POD surrogates (r ≥ 3) strike an effective balance between dimensionality reduction and dynamic fidelity, enabling faster simulation and real‑time control without compromising essential behavior. The results also point to future work on adaptive snapshot selection and integration with predictive control strategies.