GRIDRUSH

one policy, four operators, no per-region tuning — and what real power flow does to it

2025 hourly replay, same agent policy

CALCAISO
4.30×
net value vs best honest plan
85.4MW at month 12
2491sellable MW·months
0.0MWh grid violations
32of 40 sites energized
open dashboard →
TEXERCOT
3.41×
net value vs best honest plan
84.1MW at month 12
2492sellable MW·months
0.0MWh grid violations
27of 40 sites energized
open dashboard →
MIDAPJM
4.07×
net value vs best honest plan
82.4MW at month 12
2416sellable MW·months
0.0MWh grid violations
31of 40 sites energized
open dashboard →
MIDWMISO
3.88×
net value vs best honest plan
81.3MW at month 12
2390sellable MW·months
0.0MWh grid violations
30of 40 sites energized
open dashboard →
<

reality check: measured AC power flow inside the loop

AC PFphysics-in-the-loop (CAL) measured admission limit 6.53 MW / substation
2.39×
net value vs best honest plan — with the real AC power-flow limit inside the loop
44.7MW at month 12
1313sellable MW·months
29sites, all AC-PF clean
6.50max MW on one real feeder
Profile model (ICA-style headroom curve) gives 4.30×; real physics cuts it by 44% because a real 20 kV MV feeder of this class takes ~6.5 MW of new load, not the 20–60 MW the synthetic capacities assumed. Both numbers are reported; the delta is the value of real utility ICA data.
open physics dashboard →

reality check: real PG&E substations instead of synthetic ones

REALPG&E GRIP substations (CAL) 8 real substations · 1,620 MW rating · 1,081 MW 2026 load
1.22×
net value vs best honest plan — the same agent on real utility data
397.7MW at month 12
11,533sellable MW·months
40of 40 sites energized
0.0MWh grid violations
Real: substation identity, coordinates, facility rating and 2026 loading (PG&E GRIP Data Portal, public). The premium collapses to 1.22× because at the substation aggregate level these substations are large and not the bottleneck — which is exactly where PG&E's own Grid Needs Assessment points: in a sample of 2,000 circuits, 124 are already short for 2026 (up to 12.4 MW). The constraint is per-circuit, not per-substation — that is the next data we need.
open real-data dashboard →