list(
tar_target(
name = load_one_year,
command = load_eaglei_year(2021)
),
tar_target(
name = state_month_hour,
command = summarise_mo_hr(load_one_year,
c("state", "month", "hr"))
),
tar_target(
name = county_month_hour,
command = summarise_mo_hr(load_one_year,
c("county", "state", "month", "hr"))
)#,
## IGNORE below for now
# tar_target(
# name = ny_ecdf,
# command = state_county_ecdf(county_month_hour,
# c("New York"),
# c("Kings", "Erie"))
# ),
# tar_target(
# name = eia_ecdf,
# command = state_ecdf(state_month_hour,
# c("Maine",
# "Texas",
# "West Virginia",
# "Mississippi",
# "Louisiana",
# "Michigan",
# "Kentucky",
# "Oregon",
# # Least
# "District of Columbia",
# "Delaware",
# "Florida",
# "North Dakota",
# "Nevada"))
# )
)
Pilot episode (updated 8/29)
This post is the first in a series where I will explore the DOE’s Eagle-I dataset on energy grid outages.
It’s a quick one, focusing on mise en place. I’ll share some snippets of code for setting up a pipeline with the targets package to save compute time further downstream in the analysis, given the size of the data (available via figshare below). On my local machine, I will definitely save meaningful time by not loading the same csv multiple times. And especially during more exploratory phases of analysis, I enjoy fewer breaks for computation because I’ll stay in a flow-ier state.
For setup preceding the steps I’ll show, I recommend the targets walkthrough which runs through a minimally viable project.
Before showing snippets for tasks within the Eagle-I workflow, here’s the _targets.R script that connects the tasks together. On good days, and ideally always, I’ll sketch out this script before coming up with code for the component tasks. In short, I want to load one year of Eagle-I data, then create monthly summaries of outage time & affected customers in every state and county. I’ve chosen 2021 for reasons we’ll get into on a future post.
For the first few targets, I’ll be using the here and tidytable packages to get my work done. The load_eaglei_year function will do little more than load the right subset of outage data, and the summarise_mo_hr function will do the heavy lifting for the two other targets we’ll look at in the next installment.
Shoutout to Mark Fairbanks and the tidytable package!
Next steps
In future posts, I’ll cover individual targets in use here, then compare Eagle-I data with what the Energy Information Administration (EIA) visualized in this article. For now I’ve copied the charts here for reference. The first shows how long the average customer deal with power interruptions from 2013 through 2021, split by whether a “major event” triggered the interruption.
The second disaggregates the data further into points for each state, plotted by total interruption time (horizontal axis) and number of discrete interruptions (vertical).
There be dragons…
With yet more installments, I’m debating whether to create a golem-based Shiny application or build a package for easier access to EAGLE-I data within the R universe.
