In a lively panel discussion at last week’s IEEE Industrial Electronics Society meeting in Montreal, two questions related to the smart grid (the prospective electrical distribution system that will set prices dynamically and let consumers sell electricity to other users easily) arose that I think we’ll hear much more about in coming years:
Who will own the data? One important feature of the smart grid will be integration with layers of software at the level of individual machines attached to it — everything from industrial furnaces to home clothes dryers. The idea is that these devices will constantly send data about their usage into a variety of optimization schemes that seek to balance energy usage by adjusting prices and advising power sources on expected demand.
If this data is valuable — and the smart grid’s proponents suggest it is — then someone will find value in capturing it. Who will claim it? Manufacturers might require licenses to decode data from their devices, and data clearinghouses might require that manufacturers license their standards in order to participate. Squabbles over data ownership could delay adoption and hurt systemwide gains.
Industrial users have presumably addressed this question in various ways. Readers who can put their hands on an industrial data usage agreement or two are welcome to send them my way.
Will users be overloaded by decision making? The smart grid promises to balance demand and let flexible users save money through dynamic pricing. Large electricity users already enjoy discounts for electricity at off-peak hours and adjust their work schedules accordingly, but this kind of pricing will soon be available to consumers, and at highly dynamic levels — imagine a display in your laundry room that tells you what it will cost to wash your clothes now and predicts the cost of washing them overnight instead. If the laundry isn’t urgent, the overnight cycle might be an easy choice, but consumers could be besieged by trade offs to which they’re nearly indifferent.
“The smart grid means lots of new decisions,” remarked Mo El-Hawary, an engineering professor at Dalhousie University. “Do I turn on that toaster now, or do I wait 10 minutes and save a few cents?”
Another engineering professor, Siddarth Suryanarayanan of Colorado State University, suggested that this will become an application for machine learning: train a system by letting it observe your preferences for a while and then let it make small decisions for you. “Look at Netflix,” he said, “watch 10 movies, and the next one it suggests is bang-on.” Appliance usage is often highly routinized: if, during a training period of 10 days, your revealed preference suggests you’ll wait an extra 10 minutes for coffee for savings of 15 cents, that’s likely to hold over longer periods.
Since dynamic electricity prices are themselves set by machine-learning and prediction algorithms, this would be a fascinating example of decision-making software that’s needed to act on the results of other decision-making software: a gigantic pyramid of artificial intelligence.
Join this discussion on Quora. My colleague Jim Stogdill asks: “What are some interesting real world examples of Internet of Things in industrial settings?”
This is a post in our industrial Internet series, an ongoing exploration of big machines and big data. The series is produced as part of a collaboration between O’Reilly and GE.