Since there are no currently active contests, we have switched Climate CoLab to read-only mode.
Learn more at
Skip navigation
Share via:


Real-time big mobility data (location, keywords, ...) can be used to predict short-term demand for energy, greatly improving efficiency.



Use case 1:  It has just snowed 3 centimeters early this Monday morning.  Power companies are uncertain whether electricity usage will follow normal patterns. 

Shortly before the morning rush hour, using randomly-sampled mobility patterns and keyword searches from Twitter and other related sources, perhaps including a consumer application for the smartphone, statistical models predict that 20% fewer consumers will travel to work, and that instead they will stay at home. 

The power company finds with 90% confidence that they do not need to purchase power or turn on a plant that is normally required to supply downtown businesses, thereby saving a large amount of energy, CO2 emissions, and money.  These savings can then be passed back to the consumer. 

Also, this increase in consumption predictability allows use of more sustainable energy sources, some of which are also less predictable.


Modern mobility patterns and other real-time information can be very useful to reduce uncertainty for infrastructure providers.  That is, most sustainability problems are the result of large numbers of people using resources supplied over an infrastructure.  These infrastructures have inefficiencies and waste, as they must distribute a large number of resources through a huge network, and there is a lot of uncertainty about how individual usage will aggregate back up to the source.

(For example, one or two people turning on a lightbulb at the same time may seem insignificant, but millions doing it at the same time may cause problems!)

Here, we propose that real-time 'big data' offers a way to reduce uncertainty about what is happening at the individual level.  Even a small amount of randomly-sampled information can allow short-term forecasting about what will happen within a few hours, and this kind of prediction could be especially useful to providers where time-scales are relatively short, such as power supply, internet use, but perhaps also mobility demand or even water consumption.

Category of the action

Reducing emissions from electric power sector.

What actions do you propose?

- Working together with power companies to compare real-time big-data statistics to power meters, consumption patterns, and supply and thereby build consumption models using randomly-sampled (anonymous) big data sources.

- Development of a power-consumption mobile app including anonymized location and other data.

- Understanding the connections to control theory of networks.

Who will take these actions?

Where will these actions be taken?

Berlin, Germany and perhaps the USA.

How much will emissions be reduced or sequestered vs. business as usual levels?

Currently, according to discussions with a power company representative, there is a lot of uncertainty about power demand, occasionally resulting in power plants having to be switched on, producing large amounts of wasted excess energy and costs to customers.

This implies that there is a lot of inefficiency in the system, and therefore excess production, energy consumption, and CO2 emissions.

The 'binary' (on or off) nature of some large conventional power plants means that reducing uncertainty about the need to turn them on is very valuable.

Building models of consumer load and supply will help us understand the amount of waste.

What are other key benefits?

An increase in predictability of load demand implies that less predictable supply means (wind, solar power) can begin to be used, and gradually our power supply system can more tightly track daily load demand, and that the whole system can shift toward sustainability.

Electric power companies and other large infrastructure providers have in some sense, due to lack of predictability, been chained to "gross" solutions to uncertainty - turn on or purchase enough power to handle peak loads, thereby sometimes having large amounts of wasted resources.  

Beginning to track and predict large-scale user demand is a bit an extension of what we are seeing with the world-wide-web - the tailoring and adaptation of supply to details of real usage, rather than only the coursest trends.

Similar to biological organisms in nature itself, the large linear "economy of scale" infrastructure systems we have relied on since the industrial revolution can become much more resilient, adaptive, and efficient.

What are the proposal’s costs?

Time line

Related proposals


Hannak, Aniko, et al. "Tweetin'in the Rain: Exploring Societal-Scale Effects of Weather on Mood." ICWSM. 2012.