The Center for Data Innovation spoke to Raphaël Vullierme, co-founder and chief executive officer of Paris-based startup Luko. Luko individually monitors the energy consumption of multiple household appliances, using AI and a device connected to an ordinary electricity meter. Vullierme talked about how AI can identify individual appliances just from the household electricity meter, and how customers could use this data to access third-party services.
This interview has been edited for clarity.
Nick Wallace: Luko can detect multiple household appliances and simultaneously monitor the electricity consumption of each without actually being attached to them. How is that possible?
Raphaël Vullierme: We develop a small piece of hardware that you stick on your electricity meter. And I literally mean you stick it on: there are no electrical contacts and it isn’t connected to the grid. Luko’s goal was to develop a small piece of hardware that anyone could install, even a ten-year-old kid, without any prior knowledge or tools, and with no safety issues.
Precisely how it monitors the electrical flows depends on your meter. The latest meters have an LED that lights up at a certain frequency. In the EU, it’s every watt-hour. This LED is normally used to calibrate the meter, but we use it to get the consumption data, by detecting how frequently it lights up. Older meters are mechanical—for example, many have a disc that spins at the required frequency—but they work on the same principle, and we can detect those movements to monitor consumption in the same way as with the LEDs. Some meters are in the basement and some people live in apartments a few floors up, so the device has a range of about 70 meters, which we hope to increase in future.
Once you get the overall consumption, it’s important to understand that every appliance consumes electricity in a slightly different way. A heater doesn’t consume electricity in the same way as a garage door, or a fridge, or a washing machine. This is where machine learning comes in. We analyze the total consumption data to identify the different energy footprints of your home appliances.
We are not the first company to do this—what’s different is that our technology can be used by anyone. The other players use electric clamps, which you have to attach to wires that may be in your fusebox, or hidden in the wall. That means opening the fuse box, cutting the current, and manipulating the wires. This creates a safety hazard, and the need to use tools. So many ordinary householders will never do this, or they’ll call out an electrician to do it for them.
But we wanted anybody to be able to have Luko at home and to be able to install it themselves. We wanted something very affordable, very accessible and easy to install, something that even the kids could use. The price we pay for that is the fact the data we collect is less precise than with electric clamps. We get data that is accurate to around one hertz—one measurement every second—but electric clamps can get up to eight or ten kilohertz measurements. That means we can detect major devices in the home, but not low energy devices.
Wallace: What does applying artificial intelligence to energy consumption data do for people?
Vullierme: We do two things for the customer. The first is we explain how energy is being consumed in the home. By just putting a number and a price on how people are consuming energy, intuitively people will be more attentive to how they consume energy, and then they will consume less. We also provide customized and targeted advice generated using our artificial intelligence, that says “if you change this specific behavior, you will save this much on your bill every year.” We can give customized recommendations for saving energy, to protect the environment and to cut energy bills.
The second thing we do is provide a “safety alert” and a “comfort alert.” If the customer leaves an electric oven on, or if the freezer has broken down, we send him a text message to let him know. Or if you’re always forgetting your laundry in the machine, we can send an alert when it finishes. Or if you go away every weekend, and we notice that there is no other electrical activity besides the heating, or the air conditioning, we’ll send you an alert the next weekend so you don’t do it again, to help you to be more energy efficient.
Wallace: Where did this idea come from? How did you get started?
Vullierme: Luko started with Benoît Bourdel, my co-founder. He was a researcher at École Polytechnique de Montréal in Canada, and was doing some research into the latest machine learning and deep learning technologies, as well as smart grids, and he wanted to give everyone a way to fight global warming using the latest deep learning technologies. He started working on connected hot water heaters to manage when the water heats up based on the price of the electricity, that was his first project, and then he moved on to other home appliances.
He always faced the same issue: the cost of installing smart energy solutions was too high because every time you had to send a technician. That’s not a good business model for this kind of technology, because it will result in limited market penetration and slow growth. Then he had the idea of having a really low-cost and affordable way of monitoring total consumption. He worked with small businesses to help them reduce their energy bills, and he got a contract with a lawyer’s office. He went to the basement to connect his electric clamps, but the wire was really thick, the clamps he had with him weren’t big enough so he couldn’t do it that day. That’s what gave him the idea of working with the calibration LED, and of developing the hardware and the app.
Wallace: Your website mentions a few appliances that are impossible to detect—why is that? What would it take to iron out those sorts of wrinkles?
Vullierme: I mentioned that we collect data at about one hertz, which isn’t enough to detect low-energy devices, because the footprint they leave on the grid is too small. I also mentioned that different devices use electricity in different ways. Devices with a constant, flat pattern of consumption, like lights or chargers, don’t leave a very distinctive energy footprint. A heater, by contrast, has a resistor that goes on and off at a particular frequency, and it consumes a certain amount of power, so that leaves a much more distinctive footprint.
It is possible to detect some of those appliances with the other solutions I mentioned. If you only have one data point per second, then you need a very distinctive electric pattern, whereas if you have 10,000 per second, you can detect subtler patterns that are left by your charger or your TV. We made a very radical choice to collect low-frequency data as a trade-off for ease of use. It costs ten times less to equip a home with our solution than it does to send out a technician to install an electric clamp.
However, we can detect the total consumption of your “always on” devices, and help you to reduce your consumption by turning them off—we just can’t identify them for you individually.
Wallace: Though different, Luko is clearly on the same spectrum of ideas as smart meters. How do you see the technology developing in the future? What do you think data and AI could achieve for household utilities?
Vullierme: The roll-out of smart meters in the European Union is mandatory, but we think the future of this kind of technology is to give people choice. The current generation of smart meters being rolled out collect around one data point every ten to thirty minutes, which isn’t enough to monitor individual devices. Our approach is to give people control of more data, and let them decide whether or not to share with third parties in exchange for services. If you imagine Facebook’s privacy controls, where you can choose what to share with apps—that’s exactly what we’re doing with Luko.
We connect all the appliances in your house, even if they aren’t smart. You’ll likely have more smart devices in future, but you’re not going to replace everything with a smart device. We know how much laundry you do, even if your washing machine isn’t smart. So we could, for example, integrate you with Amazon or your local grocer, and have them send you detergent at the right moment. We aren’t going to do it ourselves, but we will provide the third party integration so you can give access to the seller. We can also integrate with the smart home ecosystem, so if you want to play Beethoven when you come home and open your garage door, then we can integrate with your smart home provider to do that. These features aren’t live yet, but we are building the integration and are talking to some partners to do this.