Validation of gf_eia923

This notebook runs sanity checks on the Generation Fuel data that are reported in EIA Form 923. These are the same tests which are run by the gf_eia923 validation tests by PyTest. The notebook and visualizations are meant to be used as a diagnostic tool, to help understand what's wrong when the PyTest based data validations fail for some reason.



In [ ]:

    
%load_ext autoreload
%autoreload 2



In [ ]:

    
import sys
import pandas as pd
import sqlalchemy as sa
import pudl



In [ ]:

    
import warnings
import logging
logger = logging.getLogger()
logger.setLevel(logging.INFO)
handler = logging.StreamHandler(stream=sys.stdout)
formatter = logging.Formatter('%(message)s')
handler.setFormatter(formatter)
logger.handlers = [handler]



In [ ]:

    
import matplotlib.pyplot as plt
import matplotlib as mpl
%matplotlib inline



In [ ]:

    
plt.style.use('ggplot')
mpl.rcParams['figure.figsize'] = (10,4)
mpl.rcParams['figure.dpi'] = 150
pd.options.display.max_columns = 56



In [ ]:

    
pudl_settings = pudl.workspace.setup.get_defaults()
ferc1_engine = sa.create_engine(pudl_settings['ferc1_db'])
pudl_engine = sa.create_engine(pudl_settings['pudl_db'])
pudl_settings

Get the original EIA 923 data

First we pull the original (post-ETL) EIA 923 data out of the database. We will use the values in this dataset as a baseline for checking that latter aggregated data and derived values remain valid. We will also eyeball these values here to make sure they are within the expected range. This may take a minute or two depending on the speed of your machine.



In [ ]:

    
pudl_out_orig = pudl.output.pudltabl.PudlTabl(pudl_engine, freq=None)
gf_eia923_orig = pudl_out_orig.gf_eia923()

Validation Against Fixed Bounds

Some of the variables reported in this table have a fixed range of reasonable values, like the heat content per unit of a given fuel type. These varaibles can be tested for validity against external standards directly. In general we have two kinds of tests in this section:

Tails: are the exteme values too extreme? Typically, this is at the 5% and 95% level, but depending on the distribution, sometimes other thresholds are used.
Middle: Is the central value of the distribution where it should be?

Fields that need checking:

These are all contained in the frc_eia923 table data validations, and those should just be re-used if possible. Ugh, names not all the same though. Annoying.

fuel_mmbtu_per_unit (BIT, SUB, LIG, coal, DFO, oil, gas)



In [ ]:

    
gf_eia923_orig.sample(10)

Coal Heat Content



In [ ]:

    
pudl.validate.plot_vs_bounds(gf_eia923_orig, pudl.validate.gf_eia923_coal_heat_content)

Oil Heat Content



In [ ]:

    
pudl.validate.plot_vs_bounds(gf_eia923_orig, pudl.validate.gf_eia923_oil_heat_content)

Gas Heat Content



In [ ]:

    
pudl.validate.plot_vs_bounds(gf_eia923_orig, pudl.validate.gf_eia923_gas_heat_content)

Validate Monthly Aggregation

It's possible that the distribution will change as a function of aggregation, or we might make an error in the aggregation process. These tests check that a collection of quantiles for the original and the data aggregated by month have internally consistent values.



In [ ]:

    
pudl_out_month = pudl.output.pudltabl.PudlTabl(pudl_engine, freq="MS")
gf_eia923_month = pudl_out_month.gf_eia923()



In [ ]:

    
pudl.validate.plot_vs_agg(gf_eia923_orig, gf_eia923_month, pudl.validate.gf_eia923_agg)

Validate Annual Aggregation



In [ ]:

    
pudl_out_year = pudl.output.pudltabl.PudlTabl(pudl_engine, freq="AS")
gf_eia923_year = pudl_out_year.gf_eia923()



In [ ]:

    
pudl.validate.plot_vs_agg(gf_eia923_orig, gf_eia923_year, pudl.validate.gf_eia923_agg)