A new predictive model for more accurate electrical grid mapping | Hacker News Reader

A new predictive model for more accurate electrical grid mapping

code.fb.com

57 pointsManu19877 comments

Their model uses night-time lighting data collected by the SUOMI-NPP satellite to infer medium-voltage grid connections. Grid paths are based on Dijkstra’s shortest path, with accommodation to follow roads and avoid water. The authors state an accuracy of 70 % of their model predicting grid paths within 1 km of existing power lines. Prior attempts to model grid paths using satellite imagery failed due to the diversity of shape and orientation of poles, and false positives attributed to vegetation, shadows and similar infrastructure.

The rationale for creating this model is the lack of up-to-date electric grid location data, especially in developing countries. While I find the approach fascinating and well documented, I think the error in this model’s results is much too high for further use. I would not want to rely on any electricity system planning that utilizes these model results, because the model may omit entire branches of the grid.

Can anyone think of any useful applications of these model results?

> The rationale for creating this model is the lack of up-to-date electric grid location data

Certain details about the electrical grid are regarded as a matter of national security in Norway and I would think most other developed countries share this view.

For example, if the exact location of all transmission lines in a country are known to the public then an enemy state or a group of terrorists could use that knowledge to completely cut the power to a city as part of their attack.

The willingness of some to attack the power grid of someone else was demonstrated for example in the December 2015 Ukraine power grid cyberattack.

https://en.wikipedia.org/wiki/December_2015_Ukraine_power_gr...

What prevents those details from being mapped by volunteers though? Consider openinframap, for most western nations that looks to be pretty spot on?

Nothing, but some of the details are not available to volunteers. Even if you know where all the lines are, it may not be clear how to attack those lines. There are redundancies and branch switches and feeder cross connections. If you knew where all the switches and cross connections were, though, then you could really take advantage of any weaknesses in coverage.

Makes sense, I've seen mostly only high-voltage maps, not MV. Cool examples [warning, big PDFs!] include:

* Continental europe: https://docstore.entsoe.eu/Documents/Publications/maps/2018/...

* Spain: https://www.ree.es/sites/default/files/01_ACTIVIDADES/Docume...

This is correct. I work in this field and the grid model is always treated as highly confidential. You can find some models online for some power systems, but often these are simplified or slightly altered to make it no longer representative actual system.

At most with the approach used in the article, you would be able to identify some radially connected regions, or poorly connected areas, but nothing that is a national security risk. To be able to maliciously attack such a system efficiently you'd need an idea of instantaneous flow, component limits, substation topologies and any available mitigating actions.

I don't mean to create the illusion that power systems are impervious to attack. They aren't, and can actually be quite fragile if attacked at the right place and right time. Thankfully the 'arms race' is cyber and not physical. It is easy for grid companies to share best practice on cyber defense, but physical defense and mitigation is very system specific. It is also hard/expensive to defend infrastructure that spans your country or even beyond your border. The Metcalf Sniper Attack is a good example of such a 'difficult to defend against' problem. To perform such an attack you'd need more than a connectivity model.

Financial modeling of electricity pricing . It's only useful in grid regions governed by Independent System Operators with available market data, but I can see a lot of use here, not because it's good at. predicting where the lines, are, but because it gives a model of what portions of the grid are actively utilized and which are not.

Reply on news.ycombinator.com