The Guarino Infrastructure Dependency Metric, Paper IV. Winter Storm Elliott: Does the Buffer Group Travel?

Papers I to III of this series built the buffer group ΠB = Ta/Tx for the edge between two infrastructure nodes, tested it plant by plant on the February 2021 Texas chain, and priced its hours with a cost group ΠC = ca/(vCd), all on one event. This paper reads a second. Winter Storm Elliott, 22 to 26 December 2022, ran the same chain, wellhead to pipeline to generator to grid to tap, through the Eastern Interconnection with exposures of two to six hours on the grid edge against Texas's 70.5, and the hypotheses are stated with those exposures as inputs and the edge-level structure as the prediction. Dry gas production fell 16 percent, gas suppliers and pipelines failed firm shippers for about as much generation as they interrupted interruptible ones, instrumentation froze at coal and gas units from the Carolinas to Kentucky, and 90,500 MW was unavailable at the peak; six entities shed over 5,400 MW, TVA for the first time in ninety years. The same edges bound that bound in Texas, the weather edge at the wellhead and at the unit and the gas fuel edge behind a frozen wellhead; the coal pile was slack at all eleven plants, 10 to 46 days deep, and the oil on site at the turbine stations, 14 to 295 hours of it, held. The grid-to-water edge, which bound for 44 percent of Texans on community systems, held for every Eastern system with a day of storage, at buffer groups of 4 to 12, and bound at one town of 1,400 whose pumps did not restart; the notices that reached cities, Memphis's 440,000 customers for 4.8 days, came through the water node's own weather edge, mains and plants, and the tail again belonged to recovery. Priced from the public record, TVA's own winterisation spend of 35 to 123 million dollars over its 21,245 MW fleet gives an hour of autonomy on the weather edge at 34 to 121 dollars per MW-hour against a 48-hour design and ΠC of 0.004 to 0.013 at the Texas value of lost load, 0.025 to 0.086 at a residential floor of 1,400 dollars per MWh and 0.001 to 0.003 at the 2024 ERCOT average of 35,685; the national estimates give the Duke edge 0.0001 to 0.1 at the Texas value, with one corner above one, the top of the press range over a two-day design at the floor, that the paper reports. Firm gas transportation, below one in Texas, prices at 0.6 to 1.5 against Duke's windows at the Texas value and at 3 to 8 at the floor, because a year's premium charged to a short event is dear per hour prevented, and the record shows the buffer did not hold in any case; oil on site, which Texas could not test, is the cheap hour on the Eastern fuel edge at ΠC of order 0.001. The ranking travelled: the first dollar goes to the weather edge, the coal pile gets none, and the water plant gets a generator and not a tank, which is the order the operators bought in. What did not travel is the water edge's realised group, because it was never a property of the systems, and the paper reads that as the closure doing its work: a ratio with the exposure in its denominator predicted that shortening Tx from days to hours would move whole classes of edges across the boundary at once, and it did. This record carries the typeset paper and the complete working deposit in seven zip parts: the six Python scripts that compute every table and ratio in the paper, every results table as deposited, the three figures with their pgfplots sources, and every source the paper cites as downloaded (the FERC, NERC and Regional Entity inquiry; TVA's after-action report and IEEE presentation; the North Carolina Utilities Commission's order in Docket M-100 Sub 163; the South Carolina Office of Regulatory Staff examination; PJM's report; the 2022 EIA-860 and EIA-923 files; an extract of EIA-930 hourly balancing-authority data; ASOS hourly weather observations; EPA's SDWIS community-water-system inventory for four states; Transco's and Tennessee Gas Pipeline's tariff rate sheets and FERC eTariff records; the value-of-lost-load studies; and the cited web pages saved to PDF). Every part carries a README and a SHA-256 manifest of all files; a clean run of the scripts on the unpacked deposit re-creates every results table exactly. Only the journal articles cited by DOI and one EIA data download are not redistributed.