FeederLens LLC is the legal home of one independent research project: measuring stress on the North American power grid from public telemetry, and compressing it into derived indices that can be discussed openly — without redistributing anyone's raw data.
View Research Methodology ↓The core of the current work is a one-way abstraction. Raw grid constraints pass through a 30-day rolling Z-score, then a sigmoid squash onto a 0–100 scale. The proprietary baseline never leaves the pipeline, so the published number cannot be inverted back to the source series — that is the whole point.
// Research output sample — 2026-07-17 (semantic buckets only; node-level identifiers are stripped by design) { "model_iteration": "v1.0-research", "region": "FRCC", "fgsi": 82, "band": "AMBER", "structural_observation": "Elevated local scarcity detected during solar ramp-down." }
Dated notes on model iterations and data-treatment methods. All work uses public or synthetic data (EIA-930, ISO public filings, LBNL, NREL/Pecan Street). No proprietary or third-party operational data.
A heterogeneous, deterministic-first pipeline: public telemetry in, derived structural heuristics out. Every figure is traceable to a public source.
Interconnection-queue workbooks, EIA-930 hourly telemetry and public LMP archives, parsed deterministically. Regions are isolated: one dark feed never poisons the continental picture.
Merit-order dispatch models in Python. Fuel cost plus carbon shadow price reorders the stack; a weighted multiplier ties regional slack back to fundamental priors.
Every change runs the full 1,300+ case regression suite before it lands — under 15 seconds. If a test fails, nothing ships.
Dormant watchdogs — hurricane demand shield, nuclear-outage monitor, carbon registry — stay armed through long quiet stretches, and degradation is always regional: the run finishes even when a source does not.