Reporting Conditions

An ASTM E2848 capacity test compares the measured capacity to the modeled capacity at a single, agreed-upon set of reporting conditions (RCs) — the representative irradiance, temperature, and wind speed the plant is rated at. Because both regressions are evaluated at the same point, pvcaptest models the reporting conditions as one value owned by the test: rc.

This page explains how the single test RC is established, overridden, tracked, used in filtering and results, and persisted across a yaml round-trip. It assumes a CapTest instance ct has been created and set up as described in CapTest Workflow.

The single test reporting conditions

rc is a one-row pandas.DataFrame (or None before any have been established) holding one value per regression variable:

>>> ct.rc
     poa  t_amb  w_vel
0  805.1   24.7    2.1

Its provenance is tracked by rc_source, which is one of 'meas', 'sim', or 'manual' — recording whether the conditions were computed from the measured data, computed from the modeled data, or supplied directly.

Note

A standalone CapData used without a CapTest keeps its own cd.rc and is unaffected by the test-level ownership described here. Inside a CapTest the test RC is authoritative and ct.meas / ct.sim resolve reporting values from it.

Computing reporting conditions

rep_cond() computes the reporting conditions from one dataset using the selected test setup’s default aggregations. For the standard E2848 setup, POA uses the 60th percentile of the filtered POA while ambient temperature and wind speed use the mean.

ct.rep_cond()              # compute from measured data (rc_source -> 'meas')
ct.rep_cond(which='sim')   # compute from modeled data  (rc_source -> 'sim')

Whichever side it is computed on becomes the single test RC: the call updates rc and sets rc_source accordingly. Calling ct.meas.rep_cond() (or ct.sim.rep_cond()) directly has the same effect — any reporting-conditions calculation on a test member flows up to ct.rc (last writer wins). With no which argument, ct.rep_cond() defaults to the current rc_source, so re-running it does not silently switch sides.

When a computation changes the source — for example computing from sim after the RC was previously taken from meas, or overwriting a computed RC with a manual one — a UserWarning is emitted so an unintended switch is visible. Re-computing on the same side is silent.

Anchoring irradiance filters on the reporting irradiance

After the reporting conditions are computed it is common to apply a second, narrower irradiance filter around the reporting irradiance. rep_irr_filter_low and rep_irr_filter_high provide the fractional bounds (0.8 and 1.2 with the default rep_irr_filter=0.2), so the same window is applied consistently to the measured and modeled data:

ct.meas.filter_irr(
    ct.rep_irr_filter_low,
    ct.rep_irr_filter_high,
    ref_val='rep_irr',
)

ct.sim.filter_irr(
    ct.rep_irr_filter_low,
    ct.rep_irr_filter_high,
    ref_val='rep_irr',
)

Passing ref_val='rep_irr' resolves the reference irradiance from the single test RC: within a CapTest both ct.meas and ct.sim read rep_irr from ct.rc. This means a modeled filter can anchor on the test’s reporting irradiance even when the conditions were computed from the measured data — without passing the value by hand. If no test RC has been established, ref_val='rep_irr' raises a ValueError directing you to compute or set the reporting conditions first.

Setting reporting conditions manually

Sometimes the reporting conditions should be supplied directly rather than computed — for a sensitivity study, or to reproduce a reviewing party’s stated values. Assigning to rc is the single public way to do this; it records rc_source='manual':

# a one-row DataFrame, a Series, or a dict of regression variable -> value
ct.rc = {'poa': 800.0, 't_amb': 25.0, 'w_vel': 2.0}

The value must provide a number for every right-hand-side variable of the (shared measured/modeled) regression formula; interaction terms such as I(poa * t_amb) are unwrapped to their component variables. Extra columns are preserved. The assignment validates the input and raises:

RuntimeError if setup() has not run (the regression formula is unknown);
ValueError if the measured and modeled formulas differ, if the value resolves to more than one row, or if a required regression variable is missing (the message names the missing variables);
TypeError if the value is not a DataFrame, Series, or dict.

A manual RC is treated as the authoritative value: it is not overwritten by the filter-replay on load, and a later rep_cond call that would change the source back to a computed value emits the source-change warning.

Reporting conditions and results

captest_results() (and captest_results_check_pvalues()) predict both the measured and modeled regressions at the single test RC ct.rc and return the capacity ratio. ct.rc_source is reported for provenance. If no reporting conditions have been established, captest_results raises a ValueError — call ct.rep_cond(...) or assign ct.rc first.

Saving and restoring reporting conditions

The single test RC round-trips through to_yaml() / from_yaml() (see Saving and Reproducing Tests):

Computed reporting conditions are not value-serialized. They are recomputed on load by replaying the RepCond step in the rc_source side’s filter pipeline, so the restored RC reflects the same filtered data.
Manual reporting conditions carry their values in a reporting_conditions_values block. On load they are re-validated and seeded before the filter pipelines replay, so a self-anchoring ref_val='rep_irr' filter resolves correctly.

On load the configured rc_source side’s pipeline is replayed first, so a cross-side ref_val='rep_irr' filter (for example a measured filter anchored on a modeled reporting irradiance) resolves against an already-established ct.rc.

Adjusting the default aggregation

The reporting-condition recipe — how each regression variable is aggregated — comes from the selected test setup and can be customized per call (e.g. ct.rep_cond(func={'poa': perc_wrap(55)})) or in the config file. See CapTest Workflow for the rep_conditions configuration and percentile helpers.