Speak tweet to power

I am on a plane returning from New Jersey where I was attending a workshop hosted by the Kostas Institute . It was supposed to be a workshop to synthesize lessons on transportation resilience in light of Hurricane Sandy’s impact in New York and New Jersey.

Many, if not most, conversations would actually turn to power systems–how transportation resilience is tied to the resilience of power systems.

People talked about the need to prevent power outages and restore power faster, but spent a lot of time talking about how social media can and should be used to inform power customers when and where power will (or has been) restored.

I have studied several power restoration events and have found that practices of communicating restoration via social media–Twitter–varies considerably. The best I have come across so far was by San Diego Gas and Electric after the county-wide September 8, 2011 power outage. The results of this study are published in the .

For that study, SDG&E wouldn’t be interviewed or provide data, but we were able to assemble a decent spatio-temporal picture of the restoration using (relatively) publicly available data– dynamic load profiles , WebEOC logs from City of San Diego, and, of course, SDG&E’s tweets during the outage and restoration. At the time, I had help gathering tweets from the awesome folks at HEROIC . Since, I have gotten the hang of Twitter’s API–a good subject for another post some time.

The most important thing that SDG&E did and what allowed us to create a useful spatio-temporal picture was to post to Twitter each time a set of circuits within a particular neighborhood were restored. And they did this sequentially. Obviously, it was not real-time information, since the information would have to be sent to their PIO (or unpaid intern) before posting. But it was close, as you can tell in the restoration curve below. The tweets, the load profiles, and the WebEOC updates (number of customers with power) line up quite nicely to give a fuller picture of what happened.

That’s the image at the top of this post. Each neighborhood name represents the last tweet for that respective neighborhood (i.e., multiple tweets might have occurred if circuits in a neighborhood were not restored simultaneously).
With the geographic information included in SDG&Es tweets, we were also able to make a crude choropleth map. Unfortunately, the geographic names they used were not standard. And for some reason they never made announcements for some areas, even though the entire county lost power. Maybe those neighborhoods are still without power today.

If you’re curious what SDG&E’s tweets were like during the restoration, I embedded the ones that had geographic information below. . Feel free to play around with it.

[field name=iframe]