DON Systems LLC builds software that measures grid entropy from physical conditions and routes power through lowest-risk paths. Our approach is deterministic, auditable, and grounded in mathematical physics — not machine learning.
California utilities spend over $6 billion annually on wildfire mitigation. Public Safety Power Shutoffs (PSPS) de-energize entire regions to prevent fires, affecting millions of customers. Current monitoring systems rely on weather models or machine learning trained on historical fire data — approaches that struggle with novel conditions and produce opaque decisions that regulators cannot audit.
Firebreak measures the physical state of every transmission segment in real time. When grid entropy — a mathematically derived measure of segment instability — exceeds the bifurcation boundary, the system routes power through safer paths automatically. The routing decision is deterministic: same physical inputs on the same grid topology produce the same result every time. There is no training data, no neural network, and no stochastic variation. Every decision traces back to specific physical measurements at specific times on specific grid segments.
DON Theory and its grid applications are protected by a patent portfolio covering the mathematical framework, entropy calculation methods, and routing algorithms.
Detailed technical briefings on the DON Theory mathematical framework are available under NDA to qualified utility partners, grid operators, and regulatory bodies.
Every number on this website comes from either a backtest output or a cited public source. The backtest methodology is designed for independent reproduction.
Each fire was backtested independently on the CEC transmission topology for the fire's county. ERA5 weather was fetched at every node's geographic coordinates for the fire date. The system ran 24 hourly cycles with no prior knowledge of fire occurrence or location. Detection is defined as any hour where network entropy exceeds the bifurcation boundary (θc = 0.42).
False positive analysis used 100 randomly selected non-fire dates from fire season (June–October, 2015 and 2017) across all three primary topologies, yielding 300 topology-day analyses. With the unified detection gate (DON maturity + sustained hours ≥5), 0 of 300 analyses triggered false alerts (0.0%).
All financial data is sourced from public regulatory filings. PG&E wildfire claim settlements ($25.5B) from SEC filings during 2019 bankruptcy proceedings. Annual mitigation spending ($6.17B) from PG&E's 2024 Wildfire Mitigation Plan filing. Consumer impact estimates are derived from these filings with methodology available on request.
The system detected dangerous grid conditions in 13 of 14 historical wildfires — covering 140 of 140 deaths — with zero false positives across 300 non-fire days. The decisions are deterministic and fully auditable. The approach requires no new hardware.
The system does not predict fires. It measures grid entropy and routes power through safer paths. The unified gate achieved zero false positives across 300 non-fire topology-days. The single miss — Jerusalem Fire — caused zero deaths and zero structure loss. The system is an intelligence layer, not an autonomous controller.
We provide detailed technical briefings under NDA for utility engineering teams, grid operators, and regulatory staff. The briefing covers the mathematical framework, validation methodology, and deployment architecture.
Contact Us