`dc_ohms_from_percent` fails on single-Array ModelChain with list input

[mokazemi]

Describe the bug

ModelChain._run_from_effective_irrad (and the dc_ohmic_model step it calls, dc_ohms_from_percent) determines whether self.results.dc should be treated as a tuple based solely on whether the input data passed to run_model_from_poa() / run_model_from_effective_irradiance() was a list/tuple, rather than on PVSystem.num_arrays.

Meanwhile, PVSystem.dc_ohms_from_percent() is decorated with the _unwrap_single_value helper, which returns a scalar float (not a tuple) whenever the PVSystem has exactly one Array, regardless of how the input data was shaped.

When a user passes a length-1 list/tuple of POA (or effective irradiance) DataFrames into a ModelChain built from a PVSystem with a single Array, this mismatch causes:

• self.results.dc to be stored internally as a 1-element tuple (because the input was a list), while
• Rw = self.system.dc_ohms_from_percent() to be returned as a plain float (because num_arrays == 1).

dc_ohms_from_percent() in modelchain.py then does:

if isinstance(self.results.dc, tuple):
    self.results.dc_ohmic_losses = tuple(
        pvsystem.dc_ohmic_losses(Rw, df['i_mp'])
        for Rw, df in zip(Rw, self.results.dc)
    )

Since Rw is a float, zip(Rw, self.results.dc) fails with:

TypeError: 'float' object is not iterable

This only reproduces for single-Array PVSystems whose POA/effective irradiance data happens to be passed as a list/tuple rather than a bare DataFrame; with multiple Arrays it works fine because both code paths agree on "tuple," and with a bare DataFrame input on a single-Array system it also works fine because both code paths agree on "scalar."

To Reproduce

Steps to reproduce the behavior:

1. Build a PVSystem with a single Array using a Sandia module from pvlib.pvsystem.retrieve_sam, and set array_losses_parameters={"dc_ohmic_percent": <some value>} on that Array.

2. Build a ModelChain for this system with dc_ohmic_model="dc_ohms_from_percent".

3. Prepare a single POA DataFrame poa_df, then wrap it in a one-element list: poa_input = [poa_df].

4. Call mc.run_model_from_poa(poa_input).

5. See error:

   TypeError: 'float' object is not iterable

   raised from ModelChain.dc_ohms_from_percent at the zip(Rw, self.results.dc)
   line in modelchain.py.

Minimal, runnable reproduction:

    import pandas as pd
    from pvlib.location import Location
    from pvlib.modelchain import ModelChain
    from pvlib.pvsystem import PVSystem, Array, FixedMount, retrieve_sam
    from pvlib.temperature import TEMPERATURE_MODEL_PARAMETERS

    sandia_modules = retrieve_sam("SandiaMod")
    module = sandia_modules["Canadian_Solar_CS5P_220M___2009_"]
    cec_inverters = retrieve_sam("cecinverter")
    inverter = cec_inverters["ABB__MICRO_0_25_I_OUTD_US_208__208V_"]
    temp_params = TEMPERATURE_MODEL_PARAMETERS["sapm"]["open_rack_glass_glass"]

    array = Array(
        mount=FixedMount(surface_tilt=30, surface_azimuth=180),
        module_parameters=module,
        temperature_model_parameters=temp_params,
        modules_per_string=2,
        strings=3,
        array_losses_parameters={"dc_ohmic_percent": 1.5},
    )
    system = PVSystem(arrays=[array], inverter_parameters=inverter)
    location = Location(35.6889, 51.3897, tz="Asia/Tehran", altitude=1040)
    mc = ModelChain(system, location, aoi_model="physical",
                     dc_ohmic_model="dc_ohms_from_percent")

    index = pd.date_range("2025-06-01", periods=3, freq="1h", tz="Asia/Tehran")
    poa_df = pd.DataFrame({
        "poa_global": [500.0, 600.0, 550.0],
        "poa_direct": [400.0, 480.0, 440.0],
        "poa_diffuse": [100.0, 120.0, 110.0],
    }, index=index)

    mc.run_model_from_poa([poa_df])  # fails with TypeError
    # mc.run_model_from_poa(poa_df)   # succeeds

Expected behavior

ModelChain.run_model_from_poa (and run_model_from_effective_irradiance) should produce consistent, correct results regardless of whether the caller wraps a single Array's data in a one-element list/tuple or passes it as a bare DataFrame. Either:
(a) ModelChain should normalize/unwrap single-element list input to match num_arrays == 1 before running the rest of the chain, or
(b) dc_ohms_from_percent's tuple/scalar decision should be based on self.results.dc's actual type consistently with how Rw is computed, e.g. by wrapping Rw in a tuple whenever self.results.dc is a tuple (and vice versa), so that no TypeError occurs and ohmic losses are computed correctly either way.

Screenshots

N/A (traceback only):

pv_farm_sim/simulate.py:59: in run_system_model
    mc.run_model_from_poa(poa_list)
.venv/lib/python3.14/site-packages/pvlib/modelchain.py:1807: in run_model_from_poa
    self._run_from_effective_irrad(data)
.venv/lib/python3.14/site-packages/pvlib/modelchain.py:1834: in _run_from_effective_irrad
    self.dc_ohmic_model()
.venv/lib/python3.14/site-packages/pvlib/modelchain.py: in dc_ohms_from_percent
    for Rw, df in zip(Rw, self.results.dc)
TypeError: 'float' object is not iterable

Versions:

• pvlib.__version__: 0.15.2
• pandas.__version__: 3.0.3
• python: 3.14

Additional context

• Workaround: only pass a list/tuple to run_model_from_poa/ run_model_from_effective_irradiance when len(pv_system.arrays) > 1; pass the bare DataFrame otherwise. This avoids the mismatch entirely.
• The bug is specific to the combination of (single Array) + (list/tuple input) + (dc_ohmic_model="dc_ohms_from_percent"). Other DC/loss models in the chain don't perform this tuple-vs-scalar cross-check, so they don't
  surface the inconsistency — which is why removing dc_ohmic_model="dc_ohms_from_percent" "fixes" the symptom without fixing
  the underlying shape mismatch.
• Related but distinct multi-array issues: Multiple arrays in a PV system #1067 (multi-array support), AttributeError when using ModelChain with multiple Arrays #1759 (AttributeError with multiple Arrays) — this report concerns a narrower single-Array-with-list-input edge case that doesn't appear to be tracked elsewhere.