Actual work.
Actual limits.
Software, numerical experiments and source-audited energy research—with a link to the evidence behind each claim.
Open Energy Methods is a personally maintained research workbench, not an incorporated organization. Only EnerLint is currently hosted within its GitHub organization. ThermoTwin-F, OpenSteamOpt and Kerala2040 are separate, publicly accessible personal projects included here as related work by the same creator.
EnerLint
A second look at exported mass and energy balances.
What exists
Independent mass-flow and steady first-law closure for a declared non-reactive control volume; energy accumulation for one storage interval; strict SI JSON contracts; residual thresholds; a command-line interface, Python API and machine-readable evidence reports with an input fingerprint.
The implementation computes residuals separately from the solver's convergence report. It still relies on the exported values, chosen system boundary and declared modelling assumptions.
Results recorded in the repository
The 27 October benchmark report records 42 locally passed tests and links one successful CI workflow. These are developer-selected software checks, not fault-detection accuracy on real models.
ThermoTwin-F
Gas-turbine and combined-cycle modelling, with the approximations exposed.
What exists
A Fortran educational simulation framework for Brayton cycles and optional heat-recovery/Rankine extensions, design and degradation analysis, transient response, inverse diagnostics, measurement-uncertainty studies, and grid-related scenario experiments. The native Windows GUI is experimental; it is not a verified distributed executable for the current beta.
Published software record
Released 27 October 2026. The immutable GitHub prerelease corresponds to commit ee04616. DOI is an archival identifier, not scientific validation.
OpenSteamOpt
Steam and power scheduling, solved and independently accounted for.
What exists
A two-boiler, two-header Modelica plant with a carbon-aware, hourly Pyomo/HiGHS mixed-integer scheduling layer, an FMU co-simulation route using OpenModelica/FMPy, a Python shadow twin and a Streamlit advisory interface. All operational, price and plant parameters are synthetic.
Committed 24-hour reference scenarios
| Scenario | Comparator | Optimized | Difference |
|---|---|---|---|
| Base | €81,472.30 | €78,749.13 | 3.34% lower |
| Price spike | €83,880.69 | €80,287.06 | 4.28% lower |
| Higher carbon price | €102,928.01 | €93,719.71 | 8.95% lower |
The committed manifest hashes reference files. A local FMU receipt records OpenModelica 1.27.0 and 1,441 samples across 24 hours; neither record proves real-plant accuracy.
Kerala2040
When the observations have limits, the conclusions must have limits too.
What the public research record supports
Audited Kerala electricity-consumption statistics, installed generation-capacity records, source-day availability of operational reports, and a retrospective weather–daily-demand analysis using observed SLDC daily targets and ERA5 reanalysis.
Different denominators, different meanings: the 354 SLDC days sum to 30,666.2569 MU under the operational reporting boundary; this does not replace the 29,311.76 MU annual consumer-side statistical series. Do not directly compare their totals. Read the definitions ↗
A reproducible result is not automatically a validated one.
These descriptions reflect the projects' public records checked on 27 October 2026. Results come from different systems, models, case definitions and evidence tiers. They are not directly comparable, and they should not be combined into a claim about OEM's independent scientific validation.
Each linked repository controls its own license, release policy, authorship and scientific status. If a source changes, the published note should be rechecked rather than silently rewritten. Third-party and government datasets retain their own usage terms.