The electricity industry spends a gargantuan amount of time and money trying to improve system reliability. There are myriad conferences, studies and reports that delve into the uptime of the U.S. electricity system, slicing data in a hundred different ways. Entire careers are spent trying to decrease power outage frequencies from 0.04% to 0.03%.
But recent hurricanes put things in perspective. Puerto Ricans are likely to be without power for months, according to San Juan Mayor Carmen Yulín Cruz.
— Michael Murray (@murraytelegraph) September 6, 2017
I’ve seen the techno-babble of reliability metrics before, such as the System Average Interruption Duration Index (SAIDI) and System Average Interruption Frequency Index (SAIFI). There is also the CAIDI, MAIFI and ASAI (the latter of which reminds me of the Brazilian açai berry more than anything else). I’m not a reliability expert by any means, so I wondered, how much is a prolonged Puerto Rican outage going to affect overall U.S. numbers? Given that we spend billions of dollars in the U.S. mainland every year to improve electric reliability, what is the impact of this year’s hurricane season going to be on the metrics that are the subject of constant debate at DOE, FERC and state public utility commissions across the nation?
First, it was necessary to decide on which metric to evaluate. I learned that SAIDI is an index of the duration of power outages, whereas SAIFI is an index of the frequency of outages, regardless of how long those outages last. With SAIFI, a brief dimming of the lights is counted the same as a devastating hurricane. So I went with SAIDI because duration is what I’m interested in. SAIDI is measured in minutes of outage per average customer per year.
Second, I needed to find some historic U.S. SAIDI numbers. I downloaded the EIA-861 data from 2015. Only 62% of utilities reported data to EIA, but I calculated a U.S. mainland SAIDI of 203.4 minutes. To verify that I’m doing my math correctly, I found this report from LBNL:
The chart shows median SAIDI in the U.S. is ~200 minutes per customer per year, versus the 203.4 minutes I calculated. So I’m in the ballpark.
Now we can plug in numbers for a hypothetical four- to six-month outage in Puerto Rico and see how badly it would increase the U.S. average SAIDI. While Puerto Rico is home to 3.5 million people, its beleaguered utility, PREPA, has 1.57 million customers, or about 2.2 people per household. Graphing U.S.-wide SAIDI against number of months of a Puerto Rican outage, it looks like this:
I used 95% because 95% of Puerto Ricans are currently without power. Before the hurricanes, Puerto Rico had a low SAIDI — only 55 minutes. But prolonged outages for 1.57 million customers rapidly blows up the U.S. SAIDI as the weeks go by, putting the U.S. grid on par with Turkey’s or Kazakhstan’s. A 50% island-wide outage is hypothetical, and shows that even if power is partially restored, the bellwether metric of grid reliability goes off the chart.
Here is where the parsing of numbers will begin. Utilities love to exclude “major events” such as earthquakes and hurricanes from their SAIDI metrics, even though major events can cause 80% of outages. Excluding “acts of God” can be useful for benchmarking against utilities in different parts of the country, but it’s of little consolation to the millions of American citizens without power in Puerto Rico.
Then there is the issue of Puerto Rico being a colony, not a state. EIA tracks Puerto Rico separately in “U.S. territories,” along with Guam and American Samoa. Policymakers in D.C. are unlikely to include Puerto Rico in their U.S. statistics — conveniently saving face at the next G20 summit. Next year, FERC, DOE and EIA will probably issue reports that highlight the Texas and Florida outages, because those are states that affect our national numbers. But not Puerto Rico. Here’s how our reliability compares with other wealthy nations:
Reliability is a big deal. Billions of dollars gets invested in the grid every year to eek out a few more minutes of power here or there. But let’s not forget the painful disparity between small, incremental gains in reliability and metric-wrecking natural disasters like the ones we experienced this summer. §