pybamm — Claude Code skill: battery models and setup
Runs PyBaMM lithium-ion battery experiments, compares predictions with measurements, and checks solver and mesh sensitivity.
MIT 467 installs ★ 58,036 updated 2026-10-19
npx skills add K-Dense-AI/scientific-agent-skills --skill pybamm -a claude-codeWhat pybamm does
pybamm runs single-cell charge, discharge, and rest experiments with PyBaMM. It supports SPM and DFN models, C-rate protocols, voltage cutoffs, parameter provenance, and CSV comparison. The helper script scripts/simulate_battery.py writes curves, exported parameters, a report, and measurement residuals.
The workflow separates solver error from mesh discretization error by comparing tighter tolerances and a doubled mesh. Current signs matter: positive current discharges and negative current charges. Measurement files need time_s, voltage_V, and current_A columns. The described models are isothermal and exclude aging, mechanics, lithium plating, and pack control.
Run it with Python 3.12, pybamm==26.9.0.0, and pybammsolvers==0.10.0 using IDAKLU. Without that environment, the simulation will not start. Network access is used for installation and optional upstream dataset retrieval; bundled runs and CSV comparisons work locally without credentials.
What it helps with
- simulate a cell charge, discharge, and rest protocol
- compare SPM or DFN predictions with cycling data
- measure solver and mesh effects on voltage curves
- calculate prediction residuals against a CSV file
Example prompts for the agent
- Run the protocol in
assets/chen2020-protocol.jsonand save the results tobattery-reference. - Compare
protocol.jsonwithmeasured.csv, use 30 mesh points, and write the output tobattery-comparison. - Check how solver tolerances and mesh size change the voltage curve for this protocol.
How to install pybamm
1Skills CLI (Vercel)
npx skills add K-Dense-AI/scientific-agent-skills --skill pybamm -a claude-code2Manually: put the files in place
~/.claude/skills/pybamm/SKILL.mdCopy the whole skill folder; for one project use .claude/skills/pybamm/
1Skills CLI (Vercel)
npx skills add K-Dense-AI/scientific-agent-skills --skill pybamm -a codex2Manually: put the files in place
.agents/skills/pybamm/SKILL.mdFor all projects use ~/.agents/skills/
1Skills CLI (Vercel)
npx skills add K-Dense-AI/scientific-agent-skills --skill pybamm -a cursor2Manually: put the files in place
.cursor/skills/pybamm/SKILL.mdCursor also reads .agents/skills/ and .claude/skills/
1GitHub CLI
gh skill install K-Dense-AI/scientific-agent-skills skills/pybamm2Manually: put the files in place
.github/skills/pybamm/SKILL.md1Skills CLI (Vercel)
npx skills add K-Dense-AI/scientific-agent-skills --skill pybamm -a gemini-cli2Platform CLI
gemini skills install https://github.com/K-Dense-AI/scientific-agent-skills --path skills/pybammManually: put the files in place
.devin/skills/pybamm/SKILL.mdThe legacy .windsurf/skills/ path still works
Skills CLI (Vercel)
npx skills add K-Dense-AI/scientific-agent-skills --skill pybamm -a opencodeClaude.ai
Zip the folder and upload it in Settings → Capabilities → Skills
Step by step, with commands for each platform: How to install skills in Claude Code: five ways
Questions about pybamm
How do I install pybamm in Claude Code?
Run npx skills add K-Dense-AI/scientific-agent-skills --skill pybamm -a claude-code, then check ~/.claude/skills/pybamm/SKILL.md.
What does pybamm do?
It runs PyBaMM charge, discharge, and rest experiments, compares SPM or DFN predictions with measurements, and checks solver and mesh sensitivity.
What is required to run pybamm?
Use Python 3.12 with pybamm==26.9.0.0 and pybammsolvers==0.10.0 through IDAKLU. Network access is needed for installation and optional upstream data retrieval.
Source text
SKILL.md · MIT
--- name: pybamm description: Simulates lithium-ion battery charge, discharge and rest experiments with PyBaMM, records parameter-set provenance, checks mesh and solver sensitivity, and compares predicted voltage curves with measured cycling data. Use for SPM or DFN electrochemical battery modeling, C-rate protocols, voltage cutoffs, parameter studies and numerical validation of battery simulations. license: MIT compatibility: Requires Python 3.12 with PyBaMM 26.9.0.0 and pybammsolvers 0.10.0 (IDAKLU). NumPy and CasADi are supplied by PyBaMM. Network is needed for installation and optional upstream dataset retrieval; bundled simulations and CSV comparisons run locally without credentials. metadata: version: "1.1" skill-author: K-Dense Inc. upstream-version: "26.9.0.0" last-reviewed: "2026-10-01" --- # PyBaMM battery experiments ## When to use Use this skill to model single-cell constant-current charge/discharge and rest, examine voltage and charge trajectories, or compare SPM/DFN predictions to cycling measurements. The helper runs real PyBaMM experiments and three numerical resolutions; it does not control a battery cycler or establish an operating envelope for hardware. ## Runtime and tested case ```bash uv venv --python 3.12 battery-env uv pip install --python battery-env/bin/python pybamm==26.9.0.0 pybammsolvers==0.10.0 ``` This release requires `pybammsolvers>=0.10.0`, NumPy 2 or newer, and CasADi 3.8.1. The tested environment used Python 3.12.10, NumPy 2.5.3 and SciPy 1.18.1. IDAKLU is the recommended solver; `CasadiSolver` and `ScipySolver` are deprecated in this release. Refer to the **26.9.0.0** manual below, since `latest` can describe unreleased APIs. The included [assets/chen2020-protocol.json](assets/chen2020-protocol.json) is a synthetic isothermal 298.15-K SPM case: 80% initial SOC, discharge at 0.5C for 600 s, rest for 120 s, charge at 0.5C for 600 s. Chen2020 supplies an LG M50 parameterization with 5-Ah nominal capacity; here 0.5C means 2.5 A. This is an executable reference example, not a claim that an arbitrary user's cell has those parameters. The helper disables PyBaMM usage telemetry unless the caller has already explicitly configured that variable. ## Workflow 1. Establish the cell chemistry, geometry, nominal capacity, initial state, temperature and current-sign convention. Use an appropriate parameter set and explain its source. Distinguish a paper's fitted parameters from measurements of this particular cell. Do not transplant degradation parameters without checking their meaning and applicable conditions. 2. Convert the requested protocol to the JSON contract in [references/protocol-and-comparison.md](references/protocol-and-comparison.md). Positive simulation current discharges; negative current charges. Every step has a finite duration. A specified voltage cutoff can end it earlier; the report records actual termination times. C-rates use the selected set's nominal capacity. A change in that capacity changes current. 3. Choose SPM when its reduced transport assumptions are adequate; use DFN when resolving electrolyte/electrode transport matters. The helper's tested models are isothermal and exclude aging, mechanics, plating and pack control. Increasing rate can invalidate SPM predictions even if numerical convergence is excellent. 4. Run the helper. It validates protocol fields, rejects unknown or overridden-by-protocol parameter inputs, uses IDAKLU, and snapshots the base parameters **after SOC initialization**. Keep the protocol with that snapshot: experiment steps supply their own currents. Infeasible or skipped steps are errors, rather than silently presenting a partial protocol as complete. 5. Read the two numerical comparisons separately: baseline versus tighter tolerances isolates solver error; tight tolerances on the original versus doubled mesh isolates discretization. Compare voltage differences and event-time differences against the accuracy the question needs. Refine again when these are too large; one doubling does not prove convergence. 6. If measurements are available, check current, time origin, temperature, SOC and capacity before interpreting residuals. Supply matching seconds, volts and amps. The helper reports voltage RMSE/MAE/bias and current RMSE, preserving residuals. A small voltage error under a mismatched input current does not validate the model. This workflow compares curves; it does not claim to identify unique kinetic parameters from voltage alone. ## Run and inspect From the skill directory, point `battery-env/bin/python` at the environment created above: ```bash battery-env/bin/python scripts/simulate_battery.py assets/chen2020-protocol.json \ --output battery-reference # measured.csv is user data with time_s,voltage_V,current_A columns. battery-env/bin/python scripts/simulate_battery.py protocol.json \ --measured measured.csv --mesh-points 30 --output battery-comparison ``` The first command was executed as written with an external output location. The second uses illustrative user filenames; the measurement path was exercised against a frozen synthetic reference curve in the tests. Output directories must be new. | Artifact | Interpretation | | --- | --- | | `curve.csv` | Baseline time, step, voltage, current and **net** discharge capacity | | `tight-tolerance.csv` | Same mesh, tighter solver | | `refined-mesh.csv` | Doubled mesh with tighter solver | | `parameters.json` | Base parameters after SOC initialization; step currents remain in the protocol | | `report.json` | Protocol/checksum, package versions, parameter source, numerical comparisons and terminations | | `measurement-residuals.csv` | Prediction minus measurement and current mismatch, when measurements were supplied | The reference case conserved integrated charge: 600 s at 2.5 A yielded 0.4166667 Ah, then equal charge returned net discharge capacity to zero. Voltage stayed within the Chen2020 limits in this case. Tightening tolerances changed voltage by about 1 microvolt; doubling mesh from 20 to 40 points changed it by about **2.17 mV**, so claiming sub-millivolt mesh accuracy would be unjustified. A separate real DFN test stopped at the requested 3.9-V event and verified its charge integral. Native tests also exercise charge cutoff, infeasible discharge, capacity-to-current conversion, and replay of the exported SOC-adjusted parameters without reinitializing SOC. The frozen reference is numerical regression evidence, not measured-cell validation. ## Primary references - [PyBaMM experiment API examples](https://docs.pybamm.org/en/pybamm-v26.9.0.0/source/examples/notebooks/getting_started/tutorial-5-run-experiments.html) - [Mesh refinement workflow](https://docs.pybamm.org/en/pybamm-v26.9.0.0/source/examples/notebooks/getting_started/tutorial-9-changing-the-mesh.html) - [IDAKLU solver options](https://docs.pybamm.org/en/pybamm-v26.9.0.0/source/api/solvers/idaklu_solver.html) - [Release source and changes](https://github.com/pybamm-team/PyBaMM/blob/pybamm-v26.9.0.0/CHANGELOG.md) The linked release manuals, bundled helper, parameter serialization and optional DataLoader recipe in the reference were verified against PyBaMM 26.9.0.0.
© K-Dense-AI · MIT · K-Dense-AI/scientific-agent-skills