Answers are grounded in this paper and cite their sources.
This study developed a field-measurement framework for substation operating-state assessment and next-day prediction using 32 micro-phasor measurement units (μPMUs) at Yeonggwang Substation, Republic of Korea. Positive-sequence voltage deviation, voltage unbalance factor, frequency deviation, and normalized one-minute positive-sequence voltage range were used to calculate the Health Index (HI) and Running State Index (RSI). The daily RSI combined the full-day mean with the mean of the lowest 5% of valid minute scores. The 21-day record yielded 480 development and 160 held-out μPMU-day pairs. Four-fold expanding-window validation selected Ridge regression with F4 inputs (α = 100; MAE = 0.4672). On the independent test set, Historical Mean achieved lower μPMU-level and station-level MAEs (0.2993 and 0.1809 RSI points) than Ridge (0.3918 and 0.2105). In nine development-period sensitivity scenarios, changes in feature weights, scoring anchors, the voltage-range definition, and daily aggregation affected the resulting RSI; HI equaled 100 in 96.25% of valid minutes. All four models predicted scores above the cutoff for each of the 11 test cases with RSI at or below P10. The findings demonstrate an auditable measurement-to-assessment pipeline and the importance of simple baselines, while the short test period, upper-score saturation, and low-score misses limit operational deployment claims.
No takes yet. Share an insight, caveat, or question.
Li et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: