Audit-ready microgrid optimization

Built for the real world, not for academia

Deterministic PV + battery + diesel techno-economics with hard constraints, hourly dispatch, and a reproducible run package on every result. The bankable alternative to heuristic, single-seat tools.

No sign-up · runs entirely in your browser · 20 reference projects included

Live model · Island resort, 300 kWfeasible
Average-day dispatch, computed in-browser by the engine
01252503755000006121823kW
LCOE
$0.516
Renewable
64%
vs diesel
$6.67M

Built to ingest primary public data

The demo runs on a transparent, synthetic climate model; the platform is designed to ingest these measured public sources (ingestion on the roadmap).

NSRDBPVGISNASA POWERERA5OpenEI URDBNREL ATBEIANREL EULP

The incumbent’s blind spots

Legacy sizing tools earned their position when rough off-grid sizing was enough to secure funding. Modern developers need more.

Heuristic “optima”

A derivative-free black-box search can settle on local minima and report designs with 0.04–0.06% unmet load even when you asked for zero.

Single-bus fallacy

Energy-balance modeling ignores network topology, so an economically “optimal” design can violate voltage and thermal limits on real wires.

No engineering guardrails

Physically impossible configurations run silently. No real-time compatibility checks, no DC/AC ratio validation, no feasibility alerts.

No reproducibility

Desktop, single-seat, no run artifacts. Results are hard to defend to a project financier or re-run identically months later.

What we do differently

Make validation and reproducibility the product — not an afterthought.

Deterministic optimization

An exhaustive, constraint-enforced sizing sweep returns the provably lowest-NPC feasible design in the search grid — never an infeasible one.

Constraints are hard

Zero unmet load means zero. Minimum renewable fraction, excess caps, and reserve limits are enforced, not approximated.

Reproducible run package

Every run emits a hashed bundle: inputs, dataset versions, engine version, and outputs. Identical inputs always reproduce identical results.

Portfolio & batch

Compare scenarios side-by-side, sweep sensitivities, and run whole pipelines — not one desktop study at a time.

Public-data foundation

Weather, tariffs, load shapes, and cost baselines from primary public sources, normalized and version-pinned.

API-first & cloud-native

Built for automated intake and integration from day one. The interactive Modeler is just one client of the engine.

CapabilityHeuristic desktop toolsMicrogridModeler
OptimizationDerivative-free black boxDeterministic constraint-enforced sweep
Constraint handlingApproximate (residual unmet load)Strict (0% means 0%)
ReproducibilityBest-effort, single seatHashed run package, every run
CollaborationDesktop licenseCloud workspace, batch & API
Baseline comparisonManualAutomatic diesel-only NPC / payback / IRR

Frequently asked questions

What MicrogridModeler is, how it works, and how it differs from heuristic sizing tools.

What is MicrogridModeler?

MicrogridModeler is an audit-ready, cloud-native platform for techno-economic optimization of off-grid microgrids combining solar PV, battery storage, and diesel generators. It runs a deterministic, physics-based engine that sizes systems, simulates hourly dispatch over the full project life, and reports lifecycle economics — entirely in your browser.

How is it different from heuristic sizing tools like HOMER?

Legacy desktop tools use derivative-free black-box search that can settle on local optima and report designs with residual unmet load even when you asked for zero. MicrogridModeler runs an exhaustive, constraint-enforced sizing sweep that returns the lowest-net-present-cost design that is provably feasible in the search grid — hard constraints are enforced, not approximated.

Is the model deterministic and reproducible?

Yes. The engine uses a seeded pseudo-random generator, so identical inputs always reproduce identical results. Every run emits a content-fingerprinted package — inputs, dataset versions, engine version, and outputs — so any result is re-runnable and defensible to a lender or auditor.

What does the engine actually model?

A physics-based solar resource (solar geometry, Hottel clear-sky envelope, Erbs diffuse split, POA transposition, NOCT cell-temperature PV model), an 8,760-hour-per-year multi-year dispatch with battery state-of-charge propagation and generator minimum-loading, and lifecycle economics including net present cost, levelized cost of energy, a dispatch-modeled diesel-only baseline, simple payback, and incremental IRR.

Which data sources does it use?

Cost, weather, tariff, and load baselines are drawn from primary public sources: NSRDB, PVGIS, NASA POWER, ERA5, OpenEI URDB, NREL ATB, NREL EULP, and EIA. The product is built clean-room from public methods and data — no proprietary code or scraped results.

Do I need to install anything or sign up?

No. The interactive modeler and all 20 reference benchmark projects run entirely client-side in your browser, with no sign-up and no backend to stand up. You can also bring your own 8,760-hour load CSV.

Model your first system in under a minute

Pick a benchmark, tune the design, and watch the economics update live. Then let the optimizer find the least-cost feasible system.

Launch the Modeler →