Overview

What it is

OffshoreLH2 treats an offshore hydrogen project as one connected system rather than a set of separate spreadsheets. In every simulated hour, electricity, water, gaseous hydrogen, liquid hydrogen and vessel cargo move through physical states with explicit limits. Every run ends with conservation checks and an independent audit.

The same engine drives single simulations, sensitivity and uncertainty studies, multi-year comparisons and design optimization. You can use it through a desktop application, a command-line tool or a Python API.

Motivation

Why I built it

Offshore hydrogen systems are tightly coupled. Wind sets production, storage absorbs the mismatch, and waves decide when cargo can move. Annual averages hide those interactions, and paper-specific scripts are hard to reuse or check.

I wanted each new study to add to one tested model instead of starting another script, and every result to be traceable to its data, settings and software version. OffshoreLH2 began as the Python models behind my offshore-hydrogen studies and has grown into a general research platform.

Physics core: C++20, deterministic hourly state evolution and conservation ledgers
Interfaces: pybind11 → Python SDK, command-line tool, PySide6 desktop app
Weather input: ERA5 reanalysis (automatic) or hourly CSV
Project files: .olh2 projects and .olh2study recipes (JSON)
Outputs: JSON summary, hourly CSV, HTML report, provenance manifest
Platforms: Windows (primary) and Linux; Python 3.12–3.13

Capabilities

What the current release can do

Status labels follow the software's own capability register. Verified means internally verified through unit, analytical, conservation and regression tests. It does not mean external experimental validation.

Verified internally verified, integrated physics
Implemented working data, analysis or workflow feature
Research stage implemented but not enabled for normal use
Planned no implementation yet

Implemented

Site and metocean data

Site selection on an offline world map or by coordinates
Automatic ERA5 wind and wave download for complete years, with caching, stop and resume, and source hashes
Quality control: full calendar years, range checks, short gaps filled (6 h by default), long gaps rejected, hub-height wind correction, design-year selection
Read-only context from GEBCO bathymetry and the NGA World Port Index
Ocean Networks Canada observations compared with ERA5 (bias, MAE, RMSE, correlation) without replacing the simulation weather

Verified

Wind power

Offshore wind farm based on the IEA 15-MW reference turbine power curve
Hub-height wind correction and output capped at nameplate rating
Wave and tidal generation exist only as research-stage compatibility modules

Verified

Water and hydrogen production

PEM electrolysis with part-load operation
Seawater reverse-osmosis desalination with water and energy ledgers
Optional chronological stack degradation and automatic stack replacement

Verified

Gaseous hydrogen buffer

Simple inventory model, or real-gas dynamic storage that tracks pressure and temperature
Hydrogen properties from the Leachman (NIST) equation of state
Staged compression with power and flow limits

Verified

Liquefaction and LH₂ storage

Dynamic liquefier with capacity, turndown and energy use
LH₂ tank with simple boil-off, or an equilibrium-thermal model: heat ingress, self-pressurization, venting and relief-pressure checks

Verified

Electrical storage and platform loads

Battery storage
Underwater compressed-air energy storage (UCAES), as an energy store or a thermodynamic model with thermal storage
Platform loads and a capped backup generator, with unserved energy reported explicitly

Verified

Marine logistics

Terminal loading only inside wind–wave weather windows
Vessel loading, departure, voyage, customer unloading and return, tracked call by call
Cargo boil-off and tank pressure on board; optional reliquefaction, which must be energy-accounted

Verified

Pipeline export pathway

Alternative to shipping: compressed hydrogen sent through a single export pipeline
Real-gas compression work, Darcy friction losses, leakage and customer back-pressure
Quasi-steady model: no line-pack or transient flow

Implemented

Economics

CAPEX, OPEX and discounted replacements
Delivered cost of hydrogen (DCOH) at the transfer point, including shipping, and at the customer
Default costs are declared study assumptions, kept separate from the physics

Verified

Research Lab

One-at-a-time sensitivity and Latin-hypercube Monte Carlo
Interannual comparison of complete, quality-checked weather years
NSGA-II multi-objective sizing; every proposed design is re-checked with a full-year physical replay
Storage-portfolio comparison (none, battery, UCAES, combined)

Implemented

Studies

Versioned study recipes that reference the platform physics and never copy it
Published-study reproduction, templates, cloning and per-experiment readiness
Capability register that keeps research and planned features from running by accident

Implemented

Reproducibility and usability

Each run saved with its configuration and weather-file hashes and the software version
Run history, run comparison and a generated methods statement
Run-readiness checks, settings search, activity log, and cancellation at safe checkpoints

Model fidelity

Simple where possible, detailed where it matters

Most subsystems have a fast reduced-order model for screening and a higher-fidelity model that resolves more physical state. Together the higher-fidelity options form the "P8 integrated advanced chain", the frozen and internally verified physics baseline. A more detailed model is not the same as a validated one, and the software labels the two separately.

Fidelity options in v0.4.0-alpha.4
Subsystem	Reduced-order model	Higher-fidelity option
GH₂ storage	Mass inventory with prescribed compression energy	Real-gas dynamic: pressure, temperature, compressibility and staged compression
Stationary LH₂ tank	Prescribed boil-off	Equilibrium thermal: heat ingress, liquid–vapour equilibrium, self-pressurization, venting and pressure limits
UCAES	Energy store with charge and discharge efficiencies	Thermodynamic isobaric: depth pressure, stored air mass, compression and expansion, thermal storage
Vessel cargo	Prescribed voyage boil-off	Equilibrium thermal cargo state through loading, voyage, unloading and return
Electrolyzer	Constant performance	Chronological degradation with operating-hour history and stack replacement

Architecture

Physics in one place

Each layer has one job. The C++ core owns state evolution, dispatch, inventories, marine-transfer state and the hourly conservation ledgers. Python owns data, economics, analysis, studies and reporting. The desktop application never duplicates model equations.

User interfacesPySide6 · CLI · Python
Desktop application with eleven workflow pages; offshorelh2 command-line tool for reproducible runs and ERA5 download; OffshoreH2Model Python API for scripted studies.
Studies and analysisPython
Study recipes, capability register and runner; sensitivity, Monte Carlo, interannual and NSGA-II workflows; economics and delivered-cost boundaries; reports and provenance manifests.
Platform servicesPython SDK
Project schema and validation; metocean ingestion and quality control; site-data providers (ERA5 via Copernicus CDS, GEBCO, NGA World Port Index, Ocean Networks Canada); cyclic steady state, spin-up and checkpoints.
Typed bridgepybind11
Bindings only, with no hidden formulas or fallbacks. Long simulations release the Python interpreter lock so the interface stays responsive.
Physical coreC++20
Wind, electrolysis and water, real-gas hydrogen thermophysics, liquefier, LH₂ storage, battery and UCAES, terminal and vessel logistics, pipeline export, hub dispatch and hourly mass and energy ledgers.

The integrated chain, hour by hour

MetoceanHourly wind and waves
Wind powerTurbine curve, hub height
PEM + ROElectrolysis and desalination
Real-gas GH₂Buffer and compression
LiquefactionTurndown, energy
LH₂ storagePressure, boil-off
TerminalWeather windows
VesselVoyage, cargo state
CustomerDelivered hydrogen
EconomicsCAPEX, OPEX, DCOH

Workflow

From a map point to a citable result

The desktop application follows the order of a study. Each page owns one part of the project, and a run-readiness check lists anything that would block a run or affect how its results should be read.

DashboardCurrent study status, run readiness and shortcuts to each step.
StudiesPublished studies, templates and your own study recipes.
Project SetupName the study and choose the export pathway.
Site & MetoceanPick a site, prepare or import weather data, run quality control.
System DesignChoose technologies and fix, optimize or disable each component.
Operations & ShippingShipping procedure, vessel and terminal sizing, route and weather limits.
EconomicsFinancial and cost assumptions, separate from the physics.
SimulationRun a finite chronological or cyclic steady-state year with hard checks.
ResultsKey results, hourly charts, run history, comparison and export.
Research LabSensitivity, uncertainty, interannual, optimization and storage studies.
Validation & EvidenceWhat has been verified, screened against literature or still needs validation.

Read the step-by-step guide

Interface tour

The desktop application

Screenshots from the current build (v0.4.0-alpha.4) running the bundled 2020 North Atlantic demonstration case. Select an image to open it at full resolution.

Site and Metocean page with an offline world map, a selected point at 45 degrees north and 60 degrees west, coordinate fields, a complete-years selector and a Prepare Site Data button. — **Site & Metocean.** Choose a point on the offline map or type coordinates, choose complete years, and prepare ERA5 wind and wave data with recorded provenance. Weather quality control runs before any simulation.

System Design page listing the wind farm, PEM electrolyzer and gaseous hydrogen storage, each with a technology choice, a Fixed or Optimize sizing mode, minimum and maximum bounds and a value source. — **System Design.** Every component is Fixed, Optimize (with bounds) or Disable, and the source of each value is shown.

Operations and Shipping page with shipping procedure, vessel cargo capacity and terminal transfer settings. — **Operations & Shipping.** Shipping procedure, vessel and terminal sizing, and route and weather limits, with expert overrides kept separate.

Simulation page with run controls, a run-readiness list in which all items pass, and a run report listing core hard checks PASS, independent audit PASS, closure errors of order 1e-14 and pressure safety checks PASS. — **Simulation.** Blockers in the readiness list must be resolved before a run. Afterwards the run report lists hard checks, the independent audit, mass and energy closure and pressure-safety results.

Results page with key performance cards including delivered liquid hydrogen, delivered cost, service level, operability, CAPEX, water use, marine calls and an independent audit PASS. — **Results.** Key results for a reduced-order, cyclic steady-state run. Default costs are study assumptions, not a project estimate.

Results chart of liquid-hydrogen tank pressures over the year for the offshore tank and the vessel cargo tank, with relief and maximum allowable working pressure lines. — **LH₂ pressures.** With the higher-fidelity chain, offshore-tank and vessel cargo pressures are tracked against relief and maximum-pressure limits.

Results chart of hydrogen inventories over the year: gaseous hydrogen buffer, offshore liquid-hydrogen tank and liquid hydrogen aboard the vessel, with each loading event marked. — **Hydrogen inventories.** The offshore LH₂ tank fills between vessel calls and empties at each loading, while the gas buffer absorbs hour-to-hour variation.

Run history and comparison tab listing saved runs with key results, checks and configuration hashes, and a methods statement generated from recorded provenance. — **Run history and comparison.** Every run is saved as an immutable bundle. Two runs can be compared by results, inputs and provenance, and a methods statement is generated from the record.

Research Lab with buttons for sensitivity, Monte Carlo, interannual, NSGA-II sizing and storage portfolio studies, study controls, a read-only design-variable table and a sensitivity results table. — **Research Lab.** Sensitivity, Monte Carlo, interannual, NSGA-II sizing and storage-portfolio studies run through the same physics core. Each analysis is saved with its seed and inputs.

Studies page showing the platform capability register, a table of capabilities with their category, status such as Verified, availability and the next evidence gate. — **Studies.** The capability register shows what each study relies on. Research-stage capabilities need an explicit opt-in, and planned ones cannot run.

Validation and Evidence page with a table of subsystems, the evidence implemented for each, its current status and the next scientific gate. — **Validation & Evidence.** Internal verification, literature screening and external validation are kept apart for every subsystem, each with its next evidence gate.

Research applications

Questions the platform is built to answer

Because the whole chain is simulated together, a constraint in one stage shows up in all the others. That makes it possible to ask questions which isolated component models cannot answer.

Resource and siting

How do wind and wave climate, and year-to-year variability, change production and marine operability at a candidate site?

System sizing

Which combination of wind, electrolysis, liquefaction and storage capacity performs best within declared bounds and service constraints?

Storage

When do batteries, compressed-air storage or larger hydrogen inventories reduce delivered cost or improve service, and when do they not?

Hydrogen chain

How do gas pressure limits, LH₂ tank thermodynamics, electrolyzer degradation and water demand constrain operation?

Marine logistics

How do weather windows, vessel availability, boil-off and delivery schedules affect the hydrogen that actually reaches the customer?

Economics and uncertainty

How sensitive is delivered cost to technical and economic assumptions, and how wide is the spread under declared uncertainty?

Model fidelity as a research question. OffshoreLH2 can run the same hardware with simplified and with detailed physics, so a change caused by the design can be told apart from a change caused by the model. An earlier LH₂-hub study has been rebuilt as a study recipe. It reproduces the original result under the original assumptions, and the higher-fidelity chain then shows how far the conclusions move once tank thermodynamics, vessel cargo state and degradation are resolved.

My 2026 IJHE article on operability thresholds studied the same class of system with a dedicated hourly production, storage and vessel model. Turning studies of that kind into OffshoreLH2 recipes is ongoing work.

Reproducible studies

Papers become recipes, not new code

The platform separates three things. The physical models are the ingredients and the analysis methods are the equipment. A study is a recipe that says which of each to use, in what order and with which data. A recipe never brings its own physics.

A study recipe (.olh2study, versioned JSON) declares:

sites and datasets, with SHA-256 hashes and the weather years that passed quality control;
the export pathway, technologies and fidelity levels;
which components are fixed and which are optimized, with their bounds;
cost boundaries, optimization settings and seeds;
experiments, requested outputs and provenance, including what the recipe was cloned from.

Each experiment is shown as Ready, Declared (valid, but not yet connected to a runner) or Blocked (with the reason). Only Ready experiments run, and each run writes to a new folder with a manifest of its inputs and hashes. Generic runners currently execute baseline, sensitivity, uncertainty, interannual and NSGA-II experiments for single-site studies.

What every run leaves behind

File	Contents
`summary.json`	Key results, economics and check outcomes
`timeseries.csv`	Hourly simulation output
`report.html`	Portable, human-readable report
`project_snapshot.json`	The exact configuration used
`manifest.json`	Software and core versions, build commit, configuration and weather hashes
`analysis_context.json`	Seed, sample counts and inputs for each Research Lab analysis

A cancelled run never writes a result bundle. It is marked CANCELLED / INCOMPLETE so that partial work cannot be mistaken for a finished result.

Verification and evidence

What has been checked, and what has not

The physics baseline (called P8) was frozen and qualified on Windows at a single source commit in September 2026. This establishes internal software verification and reproducibility of that baseline. It is not independent experimental validation of every subsystem.

18 / 18native C++ tests passed at the frozen baseline commit

8,784 hfull leap-year integrated acceptance run with an independent audit

7 yearsconsecutive state-continuity stress test passed

~10⁻¹⁶maximum hourly relative error of the electrical ledger in the canonical case

Also passed at the frozen baseline

Checkpoint and restart give results identical to an uninterrupted run
Water balance and ship-energy closure
Optimizer proposals replayed with full physics; infeasible designs stay infeasible
Hydrogen-property check against the NIST/Leachman equation of state
A deliberately unsafe LH₂ tank design is rejected

Before the v0.4.0-alpha.3 release, the integrated suite reported 432 Python tests passed and 1 skipped.

Still open

LH₂ self-pressurization against NASA K-site tank data (comparison harness prepared)
Matched experimental comparisons for UCAES, PEM part-load operation and vessel boil-off
Liquefier plant comparison and project-specific marine operability evidence
Project-specific cost sources before economics are treated as forecasts

Scope. OffshoreLH2 is research software. It is not a certified engineering-design, class-approval, structural, CFD, process-safety or marine-assurance tool, and its default costs are study assumptions rather than vendor quotes.

Status and roadmap

Where development stands

OffshoreLH2 is in active development as a research preview. New capabilities are added one at a time, and each must pass its own evidence gates before it can be used in normal studies.

v0.4.0-alpha.4

Available now

Liquid-hydrogen pathway, end to end
Gaseous-hydrogen pipeline pathway (quasi-steady)
Higher-fidelity thermophysics and electrolyzer degradation
ERA5 site preparation and Ocean Networks Canada comparison
Research Lab, Studies, run history and provenance

Not enabled

Research stage

Compressed-hydrogen shipping pathway
Ammonia carrier pathway
Wave and tidal generation (compatibility modules)
Fidelity comparison and physics attribution, currently available only through the published-study workflow

No implementation yet

Planned

MILP and MPC dispatch back-ends
Multi-site study execution and cross-fidelity design replay
Offshore solar; alkaline and solid-oxide electrolysis
Hybrid offshore energy hubs and offshore CO₂ capture, transport and storage
Lifecycle-emissions objective

Recent releases

3 Oct 2026 · alpha.4
Ocean Networks Canada observational validation
Live Oceans 3.0 observations, targeted station discovery and a Campbell River wind comparison workflow; matched-sample bias, MAE, RMSE and correlation against ERA5; QA/QC flags and provenance kept; API credentials redacted from errors.
2 Oct 2026 · alpha.3
Integrated research-platform preview
Gaseous pipeline pathway integrated alongside LH₂; study recipes and published-study reproduction; run readiness, run history, run comparison and methods statements; cooperative cancellation; pathway-aware Results Explorer.
25 Sep 2026 · alpha.2
Research workbench
Dashboard, offline world map, ERA5 acquisition with provenance, GEBCO and port context, Research Lab and the Validation & Evidence workspace, built on the frozen P8 physics baseline.

Access and citation

The source repository is private while the first public release is prepared. The source carries an Apache-2.0 licence and citation metadata, and an archived release with a DOI is planned. Researchers interested in evaluation or collaboration are welcome to ask for access.

Request access by email Read the guide