Learn the twin
NuGrid samples the sizing envelope, evaluates each point on the digital twin, persists a versioned dataset, and trains a surrogate that answers in milliseconds.
Size the microgrid.Prove the design.
NuGrid searches thousands of sizing candidates with a fast surrogate, then verifies the shortlist on a digital twin, before anyone commits capital to the design.
The problem
Everything you oversize is capital you burn.
Sizing a microgrid is one coupled decision, not eleven separate ones. Solar, wind, battery energy and power, hydrogen, heat pumps, thermal storage, EV chargers, and the ancillary reserve all trade off against each other and against the grid connection limit. Simulate every candidate on a faithful model and the search takes forever; guess with rules of thumb and you either overpay for storage or destabilise the connection.
The design space11
coupled sizing variables, from PV and wind capacity down to committed ancillary reserve.
A surrogate model makes searching this space tractable; the digital twin keeps the answers honest.
The objectives
Storage capex down, self-consumption up, EV charging in, stability intact.
NuGrid optimizes against the objectives grid operators actually carry: minimize storage capital cost, maximize renewable self-consumption, and integrate as much EV charging as the grid-stability constraint allows, with thermal demand served and ancillary revenue counted in the economics.
Method
Fast where speed helps. Faithful where it counts.
The surrogate exists to make the search tractable, not to be believed blindly. Every design that matters goes back to the twin for verification, and out-of-domain inputs are flagged rather than silently extrapolated.
Search the space
Differential evolution over the surrogate finds the weighted-score shortlist; non-dominated sorting exposes the Pareto trade-off front between capex, self-consumption, and EV integration.
Verify before you trust
Shortlisted designs run against the twin itself, and the lead candidate gets a forecast-driven dispatch plan. Results carry dataset and model versions, so every number is reproducible.
Honest about the twin
The current reference twin is a monthly energy-balance model.
It covers the yearly balance, battery, thermal, and hydrogen dispatch, EV demand, a grid-connection limit, and the component economics. It is a tractable reference, not an hourly high-fidelity twin; a provider's twin drops in behind the same interface when available.
Built on NuCore
One engine: surrogate optimization with verification, applied per domain.
NuGrid is the energy vertical of the NuCore engine: the same surrogate-accelerated optimization loop, uncertainty and out-of-domain layers, and versioned dataset and model registry, pointed at microgrid economics.
In practice
From scenario to dispatch plan.
Sizing and operation are one problem, so NuGrid answers both: what to build, and how to run it once built.
- Phase 01
Study
Describe the scenario: electrical and thermal demand, resource shapes, prices, the connection limit. NuGrid builds the dataset and trains the surrogate.
- Phase 02
Optimize
Get the weighted-score shortlist and the Pareto front, with every shortlisted design re-checked against the twin before it reaches you.
- Phase 03
Operate
The lead design ships with a forecast-driven dispatch plan: how the batteries, hydrogen, and thermal storage should actually run month by month.
Try it
Run a sizing study on live models.
Open the workstation, inspect studies and their Pareto fronts, and run live inference for a sizing point against a registered model version, with the verification results alongside.
Open NuGridStay in the loop
Sizing a real microgrid?
NuGrid is early-stage, built on the NuCore engine. Leave a work email to run a study on your own scenario data, or to follow releases as the twin interface matures.