The data science of domestic appliances: Energy companies have big plans for the smart home

Energy supplier EDF and smart home vendor Toon talk data-driven services and avoiding being seen as Big Brother

The electricity that flows into our homes has a unique story to tell. As well as our relative profligacy or frugality, analysis of the household power consumption using so-called disaggregation algorithms can tell which devices you are using and when. That's because each device has a unique energy usage profile.

Energy companies and smart home vendors are now taking this to the next stage, connecting electricity and gas meters to thermostats, smart plugs, boilers, sensors and appliances. Combine the current-drawing profile with additional data these sources and a very precise picture of activities in each household can be derived. It is now possible to ascertain the temperature of your wash, the efficiency of your fridge, the condition of your boiler and the effectiveness of your insulation and display this information on a smartphone app.

(Image courtesy of Toon)

Taken a little further, this data can be used to trigger an alert if someone turns on a light unexpectedly when the home owners are out, or to monitor the activities of an elderly person living on their own. This is all very useful but it's easy to see how it could cross a line and be seen as an invasion of privacy.

Smart home venders and energy companies need to move very carefully as they go down this track, suggests John Hutchins, head of connected home at EDF Energy. What can be done with data and what should be done can be very different things.

The trust equation

Energy companies rarely top the list of the most loved. Trust is a vital commodity because, with disruptive new entrants arriving in the market and customer loyalty on a downward trajectory, incumbents like EDF are vulnerable.

"We've spent 10 or 15 years trying to build up a level of trust - we're now second in Trust Index Survey among the major suppliers," Hutchins said, explaining that customers must ultimately own their data and decide for themselves how much privacy they are willing to trade for convenience.

Introduced with sensitivity, though, smart meters and connected home devices represent a way that energy companies can actually build trust, providing more visibility into charging and consumption and offering cost-saving advice to customers. At EDF Hutchin's team has a strategy of introducing new technology and services that's summarised by the mantra "show me, help me, do it for me".

"You have to take people on that journey before you can do the

'Big Brothery' stuff," Hutchins (pictured) told Computing at a recent dataIQ event.

"So, first 'help me' to visualise what's consuming energy. The next stage is 'show me' how much it costs each time I boil the kettle, then there's ‘do it for me', where we're working on algorithms to help you change behaviour".

It's all about getting the balance right, he went on, illustrating his point with a couple of jokey examples: "We will be able to tell you when your fridge is about to fail, but we'll also be able to say ‘you're watching too much TV,' or ‘you're cooking all your meals with microwave' or ‘you should get more exercise'."

The personal data conundrum

Some of the new smart-meter-based services being mooted, such as a carer receiving an alert when an elderly person's daily routine changes, require the storing of highly personal information for a certain length of time so that a picture of 'normal' can be built up. If that data were to be repurposed, say for marketing, or sold on to a third party, the damage to the brand could be substantial. Trust takes years to build and it can be lost overnight. The raft of new data protection and privacy legislation coming down the track just adds to the urgency for companies like EDF to ensure everything is above board.

[Turn to page 2]

The data science of domestic appliances: Energy companies have big plans for the smart home

Energy supplier EDF and smart home vendor Toon talk data-driven services and avoiding being seen as Big Brother

For both compliance and trust consent is crucial, and each service must be specifically signed up for. But machine learning algorithms require training on real-world data if they are to properly represent real life. So how do the data scientists working on connected home controllers get their hands on the data? At EDF they rely on a "living lab", opted-in volunteers whose information, augmented with dummy data, is used to train to feed their machine learning algorithms.

Dutch smart home vendor Toon has a similar system. "We invite our customers to join our "innovators group", Stephen Galsworthy, head of data science, told Computing at the recent Spark Summit. "Around one per cent of customers join up and they get to try out our new services early."

To avoid the possibility of a malicious actor piecing together various data sources they are all kept deliberately separate.

"We only keep what we need", said data engineer Telmo Oliveira, adding that the data is anonymised and in order to de-anonymise it they need to request a key from the customer. "For us they're just random sources of data, we cannot combine them unless the customer gives us the device key."

As well as consent, there is also the issue of ensuring the data is held securely. One high-profile breach could set the smart home business back years.

Both Toon and EDF use third-party cloud services and emphasised their use of anonymisation and encryption techniques. Hutchins spoke of EDF's pedigree: "We've got huge experience in data security because of our nuclear generation arm," he said.

Smart platforms

Connected homes generate a lot of data. A couple of years ago the team at Toon realised they needed to move to the cloud so their operations could scale up and out. Data gathered from smart meters and other sources is stored in Amazon S3. The data is then processed using the Databricks Spark-based Unified Analytics Platform.

"Apache Spark allows us to do machine learning on distributed data, and the Databricks platform allows us to manage our Spark clusters using Notebooks and a lot of the data engineering gets simplified," explained Galsworthy. "We can use Scala and Python to do R&D and create production algorithms using the same tools."

The Netherlands has one of the highest uptakes of smart thermostats in the world, and some experimental services have already been scaled up and rolled out. "We probably know more about how Dutch people wash their clothes than anyone else," Galsworthy joked.

(Image courtesy of Toon)

EDF's innovation centre 'Blue Lab' has built its own data platform to support data science activities. These include the development of a WiFi-enabled in-home display that can connect to other devices and algorithms to feed a mobile app. The company plans to start offering many of its data-driven services for free.

"You can start seeing real-time data away from home, so you're not just looking at your device," says Hutchins.

"You can look at what's going on in your house at any given time from your mobile app. You could easily already see if it's four o'clock and you're expecting your kids home, you can look on your mobile to see if your house is in its restive state or if someone's at home doing something."

Not so smart

The UK government's smart meter rollout has been roundly criticised for pushing outdated technology that was so insecure that In March 2016 GCHQ was forced to intervene. In March it was reported that EDF's own meters were failing to send data back to the company owing to a faulty firmware update. Every UK home is due to be fitted with a gas and electricity smart meter by the end of 2020 but it seems unlikely this target will be met. Hutchins acknowledged that the rollout has been problematic, but said that doing nothing was not an option.

"We recognise the next-gen smart meters will be better but in the interim, we're driving on because we have our regulatory targets to hit. The beauty of regulation in some ways is that it forces people to get on with things and roll some tech out that's going to make a difference," he said.

"As a company, we've been trying to do what's required today while leaving open the opportunity for the next generation".