D-14: VC Imputation — Distribution Analysis

Variable cost distributions for dispatchable technologies where D-14 imputation applies. Only observed (non-NaN, non-zero) raw VC values shown.

MADD §5.9 Table 35: "Lower Bound Application: assign Minimum Variable Cost of its technology class."
Current implementation uses non-zero median. This analysis shows the full distribution to evaluate alternatives.

Statistics Summary (non-zero VC only)

MIN(>0) — global minimum
E[minh](>0) — mean of hourly min
P10
P25
MEDIAN(>0) — current impl.
MEAN(>0)
TechnologyRecordsNaNVC=0% Imputed MIN(>0)E[minh](>0)P05P10P25 MEDIANMEANP75P90
Hidroeléctrica de Embalse 117,850 56,369 5,890 52.8% 0.0 46.1 26.3 45.6 59.8 76.7 99.3 103.7 129.5
Termoeléctrica CC 131,745 17,566 1,200 14.2% 63.4 81.2 64.6 167.0 184.4 209.4 220.3 263.4 329.4
Termoeléctrica TV 307,405 131,745 0 42.9% 41.9 46.7 51.2 54.4 59.2 65.6 77.1 75.5 139.5
Termoeléctrica TG 342,537 35,132 0 10.3% 180.0 191.1 219.6 240.2 270.0 299.8 313.6 347.5 424.7
Motor de Combustión Interna 1,192,280 859,750 0 72.1% 15.0 15.0 24.3 219.8 240.9 269.8 277.5 332.7 362.9

% Imputed = records with raw VC=NaN or VC=0 (reclassified by D-13) that receive the fill value.
Red = MIN(>0), the MADD-literal value. Orange = E[minh](>0), mean of hourly minimums — a less extreme "lower bound" interpretation. Green = MEDIAN(>0), the current implementation.

Histograms (observed non-zero VC, capped at P99)

Interpretation

E[minh](>0) ("mean of hourly minimums excluding zeros") is arguably the most faithful interpretation of "Minimum Variable Cost of its technology class": it answers "what does the cheapest unit of this technology typically cost?" without being distorted by a single extreme observation across the entire year. It is more conservative than MEDIAN (lower fill value → lower DIF → higher chance of being marginal) but avoids the pathological MIN=0 for embalse or MIN=15 for MCI.