The Mechanism
Natural gas doesn't move itself. Pipeline compressor stations push gas from wellhead to power plant, and most of those compressor stations run on grid electricity. When the grid sheds load, compressor stations lose power. When compressor stations lose power, gas stops flowing to the plants that were still running.
The amplification is severe: losing 1 MW of compressor power doesn't take 1 MW of gas offline. Each compressor station serves an entire pipeline segment, so 1 MW of compressor loss takes roughly 20 MW of gas supply offline. That's the pipeline amplifier — and it turns a linear problem into an exponential one.
The question is: at what coupling strength does this feedback loop become self-sustaining? When does a gas shock stop at "bad" and tip into "total collapse"?
Phase Diagram: Shock vs. Collapse
Each bar shows the final gas offline percentage after the cascade converges. The initial shock is on the x-axis; different coupling strengths are shown as separate series. When a bar hits 95%, the system has collapsed — the feedback loop became self-sustaining.
Coupling Sensitivity: Final Gas Offline % at 25% Initial Shock
| Coupling Coefficient | Gas Offline (25% shock) | Interpretation |
|---|---|---|
| 0.5% | 29.3% | Shock contained — partial failure |
| 1% (ERCOT est.) | 34.7% | ~35% offline — significant but recoverable |
| 2% | 51.8% | Approaching cascade threshold |
| 3% | 95.0% | Total collapse in most scenarios |
| 5% | 95.0% | Rapid total collapse |
The coupling coefficient — how much one node's failure increases failure probability in adjacent nodes — is the single most influential parameter in this model. Most grid planners do not publish this value.
50% isn't the worst case — it's the starting point. Q3 treated gas outages as fixed: 50% offline for 72 hours, done. But compressor stations run on grid power. Load shedding kills compressors, which takes more gas offline, which causes more load shedding. With that feedback loop in place, 50% is where the cascade begins, not where it ends. The question goes from "how do we weatherize plants" to "how do we break the dependency."
Model: iterative coupled solve. Exogenous gas shock → hourly dispatch → load shedding → compressor failure (coupling × load_shed × pipeline_amplifier × demand/gas_ratio) → more gas offline → repeat until convergence or collapse (>90% offline). Pipeline amplifier = 20 (1 MW compressor → 20 MW gas supply). Uri crisis window (Feb 14–19, 2021). 70% RE scenario.