BIoT Canada
Feature

Q&A with BrainBox AI co-founder Jean-Simon Venne

We wanted to know where the AI would deliver the most value, and so far we’re consistently seeing 25% to 35% energy savings and reducing greenhouse gas emissions by a huge percentage, about 40%.


June 28, 2019  


Print this page
(this article first appeared in the Summer 2019 edition of BIoT Canada)

By Doug Picklyk

artificial

Jean-Simon Venne, CTO BrainBox AI

In early May, Montreal-based BrainBox AI announced the launch of its unique technology that combines deep learning, cloud-based computing and algorithms to autonomously-control a building’s environment.

It’s the first artificial intelligence (AI) technology designed specifically for heating, ventilation and air conditioning (HVAC) systems in commercial buildings. The company’s vision is to redefine building automation and help lead the green building revolution. BIoT Canada spoke with the company’s Chief Technology Officer and co-founder, Jean-Simon Venne to learn more.

Please explain what BrainBox AI is bringing to market.

We have an edge computing device that we deploy on top of a building’s existing master control system for the HVAC system. Communicating with the cloud, our technology uses AI to study a building’s behaviour, sending more efficient operating instructions and modulating the building in real time to resolve issues related to energy consumption, efficiency and occupant comfort.

We want it to be as simple as receiving the box, plug it in, connect it and the rest is automatic.

What inspired this project?

When I saw the autonomous car I was absolutely fascinated. The car is moving, and there are other cars, pedestrians, bicycles, cats and dogs, all of these things are moving and you have milliseconds to react to prevent a catastrophe.

So we thought, if they’re able to develop that solution with existing computing technology, we should be able to take the same chips and do that for a building, which is not moving.

Thermodynamics moves relatively slow. To change one degree in a conference room may require 15 to 20 minutes, so you have a lot of time to react, plan and calculate.
So we said, there’s no reason it should not work. That was the initial thinking to launch the development of BrainBox AI.

How long has this been in development?

We started to develop the technology in 2017 and began testing in a beta program about 1-1/2 years go. We’re still in the pre-commercialization program, and we will probably launch at the end of the year.

How many buildings are you in now?

We have around 20 buildings in the Montreal area and as far as Quebec City where we’ve deployed the technology at different stages. We don’t want to be too far from the head office to begin.

Over the summer we should be adding another 20, reaching across the country and into the U.S. to check the technology out in different climates.

What types of buildings are you in?

For initial testing we selected different building types with different HVAC equipment. For example, we have a typical office building in the downtown core of Montreal; a flat warehouse in the suburbs; a retail space; hotels; and even a college campus that’s over 100 years old with a museum-like collection of HVAC equipment.

And we’ll keep adding more building types. We’re adding several buildings in the Toronto area, and also in Alberta, which is quite different again in terms of climate.

What has early testing revealed to you?

We wanted to know where the AI would deliver the most value, and so far we’re consistently seeing 25% to 35% energy savings and reducing greenhouse gas emissions by a huge percentage, about 40%.

You connect to the HVAC systems in a building. What about lighting systems?

For us the lighting controls are too easy. The lights are on, off or dim, you don’t really need AI for that. And with the adoption of LED lighting, the energy footprint of the lights is becoming so small, that there are a lot less savings to be found.

The real money savings from an energy perspective is on the HVAC side. It’s probably 50 to 60% of a building’s total energy bill.

And HVAC is much more complex because it requires mastering all of the thermodynamic situations—modulating the internal environment versus the outside weather, getting the heat and cooling to the right locations depending on the sun’s location, and more.

It’s a very complex calculation, that can be done by hand, but requires dedicated people doing that all day, which is not financially viable. So let’s teach the AI to do it. It runs 24/7, and it never gets tired.

What is your business model?

The intention, when we reach commercialization, is to have a software as a service (SaaS) fee, which is much less than the savings the building will receive.

And there will be no real up-front cost, no capital expense hit for the building operator/owner. It’s a low-cost edge device that we install, and we don’t want to charge an installation fee, if we do it will be minimal. And the user doesn’t need to do much to deploy the technology.

Will it only work with new HVAC equipment?

It needs to be digital of course, so if you have a pneumatic system and there is no interface to it, that’s going to be a problem.

So far, the challenge is developing an edge device that is capable of connecting to several different types of controllers. There are about 700 different HVAC control protocols around the world, and some of them are very closed (proprietary). We want to tackle the major ones (about 15 to 20) that cover about 80% of the market.

Does the type of HVAC system the building uses make a difference?

So far we have tested a retail building where they have a minimal set up, only a rooftop and a thermostat, all the way to a building with an entire system with a water tower and different loops, boiler, hot water system, chilled water system, a VAV (variable air volume) box. So far, if the systems are connected to a control system, we usually can bring value to the building.

So your solution will actually run the HVAC system?

Yes, it’s a completely autonomous system, no human intervention. The building operator/owner will have access to a dashboard so they can see what’s happening, but they’re not really doing anything.

And the AI is continuously learning so it keeps evolving, the more data it’s exposed to and the more experience is has the better it gets over time.

What’s the priority for the building owners—energy savings or occupant comfort?

That’s a fascinating question. We have a team that is working to optimize the quantity of energy a building is consuming, and then we have another team and their job is to address all of the comfort issues.

Most of the time there is no conflict, but it does happen, so then we have a coaching algorithm which provides a ruling, in terms of how we balance comfort versus savings.

To make that judgement, we needed to inform the algorithm about the values of that building’s operator/owner. For example, we may have a hotel that wants to make sure that the comfort is always respected, no matter the cost—so that’s their value.

What role are your three research partners playing in the product development?

IVADO [Institute for Data Valorization] is the Montreal-based organization pushing research on AI and how to extract value from data. This is a large organization headed by Yoshua Bengio, one of the Canadian fathers of AI and deep learning. It’s a collaborative, and they give us access to PhDs to do research on certain aspects, and then we can take that development and incorporate it into our solution.

The U.S. Department of Energy’s National Renewable Energy Laboratory (NREL) in Denver, Colorado, is doing a lot of AI work in energy production, and we are working with them to develop an autobot, an AI engine that we unleash into a new building that will search and do an entire mapping of the building, in terms of all of the data points on the HVAC, which is something we do manually now when onboarding a new building.

The ETS [Ecole de technologie supérieure] is a university in Montreal doing AI research and helping us to push along some of the deep learning predictive algorithms.

You also promote a preventive maintenance aspect to the technology, please explain.

We didn’t plan for this, it kind of happened by itself. We discovered the AI is bothered by unconventional behaviour.

For example, the system reported that a room was able to reach it’s desired set point temperature in the morning in 19 minutes, and then it was taking 21 minutes, and the following week it was taking 24 minutes.
The AI gets aggravated by that and reports back information like: “It would seem that my heating capacity in that zone is increasing over time with similar conditions in outside weather, so there must be something wrong with the system.”

We called in a service contractor to take a look at that unit, a regular call not an emergency call, and discovered that yes, probably within two months that unit would have broken down.

Do you have any competitors in this market yet?

We’ve been talking to different product vendors and our offering is unique to the market. There are a lot of analytics products out there, but a product that is really driving control in real time, and giving a real autonomous building without any human intervention, this is a first.

Canada is a leader in AI, but unfortunately a lot of big AI development is happening outside of the country. So for us we’re proud that this is a Canada-first, world first, development.

Jean-Simon Venne, P.Eng., is CTO and co-founder of Montreal-based BrainBox AI.


Print this page

Related


Leave a Reply

Your email address will not be published. Required fields are marked *

*