Abstract
The increasing replacement of synchronous generation by converter-interfaced photovoltaic (PV) plants introduces new challenges to the output-only modal identification in modern power systems. Besides reducing system inertia, converter dynamics modify the spectral characteristics of measured responses through fast control loops and control-induced interactions, redistributing modal energy and reducing the physical interpretability of conventional identification methods. This paper investigates the impact of converter-induced spectral contamination on two output-only modal identification techniques: classical Prony analysis and Total Least Squares–Estimation of Signal Parameters via Rotational Invariance Techniques (TLS-ESPRIT). Time-domain responses are obtained from the Kundur two-area benchmark system, where one synchronous generator is replaced by a utility-scale PV plant modeled using the standardized WECC dynamic models REGC_A, REEC_A and REPC_A. A fully synchronous reference case together with constant active-power and constant voltage-control operating modes is comparatively analyzed. To complement conventional reconstruction-based metrics, this work proposes the Physical Energy Relevance(PER) metric, which quantifies the fraction of identified modal energy associated with physically relevant dynamic phenomena through a frequency-dependent weighting function. Because PER does not require prior knowledge of the expected modal spectrum, it provides a flexible framework for evaluating the physical significance of output-only modal decompositions. The results show that, although Prony provides accurate reconstruction under synchronous conditions, converter-induced spectral contamination progressively shifts the modal energy toward high-frequency components with limited physical relevance. In contrast, TLS-ESPRIT consistently preserves the concentration of modal energy within the dominant electromechanical frequency bands. These results demonstrate that PER complements conventional reconstruction metrics by evaluating the physical relevance of the identified modal spectrum, highlighting the superior robustness of TLS-ESPRIT for converter-dominated low-inertia power systems.