HELIOFORGE / METHODOLOGY
Know what the model represents.
Hybrid-energy learning activities and research screens with explicit capability boundaries.
Hosted preview and local calculations
- Bundled synthetic examples are precomputed outputs; the public browser edition does not calculate new study-model results from edited inputs.
- The Terminal runs new deterministic, bounded 24-hour energy-balance cases in the browser; it uses a greedy policy and is not an optimizer.
- The separate local Python application runs new calculations for the advanced study models and optimization.
- Optional local external providers are separate, opt-in integrations; credentials and explicit consent are required where applicable. The public static preview makes no provider calls.
The local setup guide explains how to run the separate research application on your own machine. See the mode guide and versioned model registry for model-specific contracts.
Catalogue coverage
The source catalogue includes 36 architectures, 24 scenarios and 36 architecture-linked lessons. In the local application, 19 architectures have runnable electricity-balance screens and 17 advanced architectures are study-only. There are 20 numerical scenario presets and four dynamic/control study presets. A displayed architecture is not evidence that all of its specialized physics have been implemented. The study catalogue covers 59 named experiments across 12 workspaces.
Browse original system layouts, read assumptions and follow the lessons’ experiment prompts and self-check questions. The separate Terminal panel labels newly calculated reduced-model outcomes explicitly.
Energy-model boundaries
The hybrid hourly screen and the battery dispatch optimization are separate models. Their state-of-charge conditions, grid limits, units and energy-balance assumptions must be read in context. The hybrid methodology and numerical methodology are the detailed source references.
A demonstration day must not be multiplied into an annual business case without a reviewed annual model. A cheaper case that serves less demand is not automatically a better case. The interface’s geometry is explanatory; animated links are not solved electrical branch flows.
Finance, forecasts and evidence
Storage, project finance, PV, carbon and forecast examples are separate screening cases. No hidden automated revenue bridge connects them. Project and equity cash flows, nominal and real assumptions, and null or undefined metrics remain distinct. Synthetic examples are not calibrated forecasts, plant controls, investment recommendations or regulatory approvals.
AI and review roles
The public preview makes no external AI calls. The local application’s deterministic review mode is a checklist, not independently executed agents. Its optional external adapter requires explicit configuration and consent; it is not part of this hosted website.
Separate synthetic wind-market model
MKTrade methodology documents a separate synthetic wind-market model. Its illustrative market assumptions and results belong to a different project; they do not feed HelioForge’s bundled examples or models.
Verification and reproducibility
Read the quality report and release verification, and inspect CI runs. The website’s source revision and runtime SHA-256 are in app.json. Software checks do not establish independent scientific calibration or operational approval.