How to calculate measure of fuel cost per mwh generated

[bendnorman]

This question was submitted to [email protected].

I'm trying to find a specific piece of data, and I was hoping that someone on your team has already thought about this. I'm looking for a measure of "fuel cost per mwh generated". I'm thinking total generation and fuel consumption from out_eia923__boiler_fuel combined with out_eia923__fuel_receipts_costs. I noticed that some tables have fuel_cost_per_mwh but this variable seems to be missing for a substantial portion of gas and coal observations. I was hoping you might have a suggestion for what to use. Ultimately I think I will end up using these data for analysis as the utility-year level if that makes any sort of difference.

[bendnorman]

Answer from @zaneselvans

Hey there,

We do have our best estimate of this number. It's in the out_eia__monthly_generators table. This is a wide table with a ton of information about generators reported or calculated on a monthly basis. The column you want is probably fuel_cost_per_mwh.

• PUDL data dictionary entry
• Browse this table in Datasette
• Download this table as a Parquet file
  Query the Parquet table directly (e.g. using pandas.read_parquet() in Python) from: s3://pudl.catalyst.coop/nightly/out_eia__monthly_generators.parquet

It turns out that this seemingly simple number is actually pretty hard to calculate for a couple of reasons. First, fuel consumption and net generation aren't reported on the same basis -- generation is associated with generators, and fuel consumption is associated with boilers... except that also in some cases it's reported to EIA by plant and prime mover, which makes it ambiguous what generator or boiler the generation / fuel consumption came from and it has to be allocated as best as possible. And to know how much fuel was used to generate how much electricity you need to calculate the heat rate, which depends on knowing what collection of boilers and generators are actually connected to each other, which is only partially reported by EIA. Second (Third?) the fuel costs are reported on a per-delivery basis at the plant level, so there's a slightly tenuous connection between the cost of fuel delivered in a given month, and the cost of fuel burned in that month. For simplicity we assume the fuel is burned in the same month it's delivered.

Also note that there are some big holes in the fuel price reporting, particularly for natural gas in the Northeastern US -- EIA allows utilities to redact those prices, and we do our best to fill them in with the aggregate fuel prices from EIA's bulk API data for the fuel, month, and state in question, but that doesn't catch them all.

If you care about this value at the utility-year level, you'll probably want to do a custom aggregation starting with the monthly generator-level data, since it is more complete than the yearly data since any NA value in the annual aggregation will cause the whole year to become NA -- you'll need to figure out a reasonable way to work around the NA values.

Feel free to sign up for office hours if you want to talk through this in more detail. @cmgosnell is intimately familiar with these calculations.

If you care about non-fuel operating costs or capital investments, there's data in the FERC Form 1 large steam table, and we also publish a linkage between the FERC & EIA plants & utilities, though the coverage and linkage are not 100% complete or perfect. @cmgosnell and @katie-lamb are familiar with those tables.