# data/ -- CEC certified PV module database (Sandia / CEC 5-parameter single-diode)

## Source

- **Primary registry**: California Energy Commission "Solar Equipment Lists
  -- Photovoltaic Modules", https://solarequipment.energy.ca.gov/Home/PVModuleList
  (CEC maintains the list; updated monthly as new modules are certified).
- **Snapshot used**: the database bundled with
  [`pvlib-python`](https://pvlib-python.readthedocs.io/), accessed via
  `pvlib.pvsystem.retrieve_sam("CECMod")`. This is the SAM (NREL System
  Advisor Model) snapshot of the CEC list that ships inside
  `pvlib/data/sam-library-cec-modules-*.csv`. The snapshot in this
  benchmark corresponds to pvlib version `>= 0.10`.
- **Primary technical references**:
  - De Soto, Klein, Beckman 2006, *Improvement and validation of a model
    for photovoltaic array performance*, Solar Energy 80(1):78-88,
    doi:10.1016/j.solener.2005.06.010 -- defines the 5-parameter
    single-diode model whose parameters each CEC row carries. Full
    derivation is in De Soto's 2004 MS thesis at UW-Madison
    (`../reference/desoto2004_thesis.pdf`).
  - King, Boyson, Kratochvil 2004, *Photovoltaic Array Performance
    Model*, SAND2004-3535, https://www.osti.gov/biblio/919131 -- defines
    the PTC test conditions and the empirical temperature-response
    model used by the CEC certification process
    (`../reference/king2004_sapm.pdf`).
- **Retrieved on**: 2026-04-22.
- **License**: California state government open data (public). The CEC
  list is a public registry and the pvlib snapshot is redistributed
  under the pvlib BSD-3-Clause license with attribution to CEC / NREL.

Each row is a PV module certified by an accredited test laboratory
(TUV, SGS, ETL, Intertek) against the Sandia / CEC 5-parameter
single-diode model, with parameters `(a_ref, I_L_ref, I_o_ref, R_s,
R_sh_ref)` fitted by SAM to match the manufacturer datasheet key points
`(I_sc_ref, V_oc_ref, I_mp_ref, V_mp_ref, P_mp_stc)` at STC, plus
temperature coefficients measured under NOCT.

## How to rebuild `data.csv`

```bash
bash download.sh
```

The script installs `pvlib` (via `pip install pvlib`) if it is not
already present, calls `pvlib.pvsystem.retrieve_sam("CECMod")`,
transposes the module-as-column layout into one-module-per-row, enforces
a stable column order, and writes the result to `data.csv`. The whole
process is offline once `pvlib` is installed; the CEC database is
bundled inside the `pvlib` wheel.

## Columns

21 535 modules x 26 columns. Column 0 is the primary output;
columns 1..25 are inputs / metadata kept verbatim from upstream so that
alternative ground-truth formulas can be added later without re-fetching
the data.

| index | name | units | type | description |
|---|---|---|---|---|
| 0  | `P_mp_stc`  | W     | output (float)   | **Output.** Module max-power-point power at STC (1000 W/m^2, 25 deg C, AM1.5). Stored as `STC` upstream; renamed for clarity. By CEC convention `P_mp_stc == I_mp_ref * V_mp_ref` exactly. |
| 1  | `I_sc_ref`  | A     | input (float)    | Short-circuit current at STC. |
| 2  | `V_oc_ref`  | V     | input (float)    | Open-circuit voltage at STC. |
| 3  | `I_mp_ref`  | A     | input (float)    | Current at the max-power operating point at STC. |
| 4  | `V_mp_ref`  | V     | input (float)    | Voltage at the max-power operating point at STC. |
| 5  | `alpha_sc`  | A/K   | input (float)    | `I_sc` temperature coefficient (SAPM convention). |
| 6  | `beta_oc`   | V/K   | input (float)    | `V_oc` temperature coefficient. |
| 7  | `gamma_r`   | %/K   | input (float)    | Relative `P_mp` temperature coefficient. |
| 8  | `T_NOCT`    | deg C | input (float)    | Nominal Operating Cell Temperature (test conditions: 800 W/m^2, 20 deg C air, 1 m/s wind). |
| 9  | `a_ref`     | V     | input (float)    | Modified ideality factor `a = n * N_s * k * T_cell / q` at STC. One of the 5 De Soto parameters. |
| 10 | `I_L_ref`   | A     | input (float)    | Light-generated (photo) current at STC. De Soto parameter. |
| 11 | `I_o_ref`   | A     | input (float)    | Diode reverse-saturation current at STC. De Soto parameter. |
| 12 | `R_s`       | Ohm   | input (float)    | Series resistance (lumped). De Soto parameter. |
| 13 | `R_sh_ref`  | Ohm   | input (float)    | Shunt resistance at STC. De Soto parameter. |
| 14 | `Adjust`    | %     | input (float)    | SAM empirical `I_sc` vs. `alpha_sc` adjustment factor. |
| 15 | `N_s`       | count | input (int)      | Number of cells connected in series in the module. |
| 16 | `A_c`       | m^2   | input (float)    | Module aperture area. |
| 17 | `Length`    | m     | input (float)    | Module length. |
| 18 | `Width`     | m     | input (float)    | Module width. |
| 19 | `PTC`       | W     | input (float)    | Power at PVUSA Test Conditions (1000 W/m^2, 20 deg C air, 1 m/s wind). |
| 20 | `Technology`| --    | input (str)      | `Mono-c-Si` / `Multi-c-Si` / `Thin Film` / `CdTe` / `CIGS`. |
| 21 | `Bifacial`  | bool  | input (0/1)      | Bifacial construction flag. |
| 22 | `BIPV`      | --    | input (Y/N)      | Building-integrated PV flag. |
| 23 | `module_id` | --    | identifier (str) | Manufacturer + model string (unique). |
| 24 | `Version`   | --    | provenance (str) | SAM DB version that generated the row. |
| 25 | `Date`      | --    | provenance (str) | SAM DB entry date. |

### Columns consumed by the shipped ground-truth formulas

- `single_diode_pmp`:  `I_L_ref, I_o_ref, R_s, R_sh_ref, a_ref` (the 5
  De Soto parameters).
- `power_from_imp_vmp`: `I_mp_ref, V_mp_ref`.
- `efficiency_area_scaling`: `A_c`.

All other columns are retained for future ground-truth additions
(e.g. SAPM PTC temperature correction, per-technology efficiency fits,
`V_oc` from simplified diode, ...) -- see `../description.md`.

## Shape

21 535 rows x 26 columns.

## Notes on the upstream table

- The five single-diode parameters `(a_ref, I_L_ref, I_o_ref, R_s,
  R_sh_ref)` are not directly measured: CEC's accredited labs measure the
  I-V curve under STC and NOCT, and SAM fits the 5 parameters so that
  the resulting De Soto I-V curve matches the measured
  `(I_sc, V_oc, I_mp, V_mp, alpha_sc, beta_oc)`. Consequently the
  single-diode formula `single_diode_pmp.f(X)` reproduces `P_mp_stc` to
  numerical precision -- this is by construction and is the intended
  upper bound / physics check for the benchmark.
- A handful of older entries have `I_o_ref` so small (`< 1e-18` A) that
  the single-diode solver hits its lower numeric bound. The bundled
  scorer treats these rows as valid and they do not materially affect
  aggregate metrics.
- Technology counts in this snapshot: 11 221 Multi-c-Si, 9 725 Mono-c-Si,
  561 Thin Film, 20 CdTe, 8 CIGS. 133 modules are flagged bifacial.