04SuryaChain
Planned · not yet deployedConsumer energy runs on two questions: how much will this array make, and who gets paid for it. SuryaChain answers the first at the edge and the second on a ledger, without a single meter reading leaving the site that produced it.
01The problem
Three constraints on a distributed grid
Generation is localSite-specific
Irradiance, soiling, shading and inverter behaviour are properties of one site. A model trained on somebody else’s array is a worse model.Meter data is commercialCannot pool
Half-hourly generation and consumption reveal how an operator runs its plant, and how much it earns. Pooling it is not a technical problem.Settlement needs proofNeeds a ledger
Payment against a reading only works if every party can verify the reading was not changed after the fact.02The flow
MeterForecastCommitReconcileSettle
The controller on the array reads its own meters and runs its own forecast. What leaves is a signed commitment to the reading and a gradient for the shared model. Never the reading itself.
01PLANNED
Forecast at the edgeDay-ahead and intra-day generation and load, computed on the site controller.02PLANNED
Federated modelEach array improves the shared forecaster without shipping its own readings.03PLANNED
Commit & attestEvery reading is committed as a signed event with a verifiable timestamp.04PLANNED
SettleContribution and export are reconciled against the ledger, not against a spreadsheet.03Telemetry into the chain
Stays on site
Raw meter readings
Inverter telemetry
Consumer load profiles
Tariff agreements
Goes on chain
Forecast model weights
Signed reading commitments
Settlement records
04The ledger
Readings committed
Readings committed5,840
Last hour244
Settlements run18
Disputes open0
01peak 49614
05Contribution
Federated forecasting · sites
SiteCyclesForecast accΔ gridContributionShare
ARRAY-0130/3096.2%+1.400.2828%
ARRAY-0230/3095.1%+0.900.2222%
ARRAY-0329/3093.8%+0.300.1919%
ARRAY-0430/3092.4%-0.400.1414%
ARRAY-0528/3090.7%-1.100.1010%
ROOFTOP-A26/3088.9%-1.800.077%
Forecast quality · scored in place
Day-ahead MAPE4.1%Sample
Intra-day MAPE2.3%Sample
Peak-hour error5.8%Sample
Curtailment avoided11%Sample
Meter data exported0