Handle plant startup consistently

[gailin-p]

Fossil fuel power plants have a startup period between between a period of inactivity and the start of normal operations. During these startup periods, there are differences in plant operations and data handling that impact our estimated plant-level emission rate data. For very small BAs (eg DEAA), this can also impact BA-level emission rate estimates.

Differences in plant operations during startup include:

• Oil and coal plants are required to use low-sulfur fuels (eg, natural gas) during startup prior to the initiation of electrical generation. For coal plants, this pre-generation part of start up can last 24+ hours, with a mode of 6 hours
• During the first few hours of electrical generation, but after a coal/oil plant has switched away from its startup fuel (still considered start-up), some emissions control devices are not yet operational, leading to higher sulfur and NOx emissions.

Differences in data during startup include:

• Some plants report default gross generation values to CEMS during the pre-generation portion of start-up. These default values may be up to 2.0 MWh. Not all plants report a default value (some report zero)

Other data describing start-up includes:

(Identified by @grgmiller )

• EIA actually reports separately the "operational" energy source codes and "startup" energy source codes. We account for both operational and startup energy source codes in our EIA allocation process, but we don't explicitly separate them out. In the future, we might consider whether we could somehow use the startup energy source codes to improve our startup emission rate estimates.
• We could think up some methods for identifying startup and shutdown hours, but the hourly CEMS data actually reports (kind of) which hours are startup and shutdown hours. Previously, the CEMS data reported through AMPD and the FTP site didn't explicitly include this data column (which I understand to be a flag column of whether the hour was a startup or shutdown hour), but the version of the data that exists in their FACT API does include this. When EPA was developing their new CAMPD portal (which replaced AMPD) I suggested that they include this data column in the new data, but I haven't followed up to check if it made it into CAMPD.

Action items related to start-up

• [ ] Use information about startup fuel and startup hours to estimate missing emissions data for CEMS plants: startup hours should use a different fuel code
• [ ] Use information about startup fuel type to refine emission rate estimates for non-CEMS plants (eg, provide both startup and normal operation emission rate estimates), since the average over both startup and non-startup times may not be an accurate representation of the actual hourly emission rate of a plant
• [ ] Identify startup hours for CEMS plants as a potential new data output. Some of this data could come from CEMS; other could come from inference based on EPA's 2014 rule changes on startup

Examples

CO2 emissions vs. gross generation for two plants: 51 (coal plant) and 55282 (peak-y natural gas plant). Periods where emissions rise before generation represent plant start-up. Note that start-up is longer for the coal plant, but that during startup the coal plant is burning a relatively clean fuel.

References

EPA rule changes relating to start-up, 2014: https://www.federalregister.gov/documents/2014/11/19/2014-27125/reconsideration-of-certain-startupshutdown-issues-national-emission-standards-for-hazardous-air#citation-3-p68779 EPA research on start-up related to above rule change: https://www3.epa.gov/airtoxics/utility/matsssfinalruletsd110414.pdf EPA Rule 75 (describes all CEMS rules): https://www.ecfr.gov/current/title-40/chapter-I/subchapter-C/part-75

[grgmiller]

the CEMS data reported through AMPD and the FTP site didn't explicitly include this data column (which I understand to be a flag column of whether the hour was a startup or shutdown hour), but the version of the data that exists in their FACT API does include this

Specifically, according to the FACT API documentation, we'd be interested in the following endpoint:

"MATS_FLAG_CODE": {  
        "D": "Shutdown hour",  
        "U": "Startup hour"  
      },

The easiest way to integrate FACT API data might be through PUDL. See https://github.com/catalyst-cooperative/pudl/issues/617