How Accenture & Johnson Controls Are Curbing Climate Change with an Assist from Edge AI

By Taryn Engmark

Associate Editor

Embedded Computing Design

September 23, 2022


How Accenture & Johnson Controls Are Curbing Climate Change with an Assist from Edge AI

The International Energy Agency, or IEA, estimates that 10 percent of global electricity consumption comes from building HVAC systems and fans. So, what’s the impact of those systems on climate change?

“If you look at the total carbon footprint in the world, 40 percent of that is within the building environment,” says Vikrant Viniak, Senior Managing Director at IT services firm Accenture. “If you make an impact on that, you are making a big impact on sustainability.”

With initiatives like the United Nations Framework Convention on Climate Change (UNFCCC) setting aggressive goals to reduce global carbon emissions by 30 percent in the next eight years, and offset them completely by 2050, it’s clear that one of the biggest areas to address with sustainability efforts are our own homes, offices, retail outlets, and other structures. That sounds like a simple enough task for building management systems (BMS), which have been helping facility operators optimize and manage structures and their utilities since at least the early 2000s.

But a lot has changed since then. The rise of IoT means more in-building systems can now be connected to a BAS to monitor things like air quality in addition to resource consumption. And with more BAS now connected to the energy grid and the cloud, there’s exponentially more data to be analyzed.

When trying to compile all this data toward the goal of curbing climate change – especially within the timetables laid out by organizations like the U.N. – there’s no simple fix. Rather than go it alone, Accenture has partnered with building equipment leader Johnson Controls on OpenBlue, a suite of connected, AI-enabled smart building solutions designed to improve facility efficiency (Figure 1).


[Figure 1. The OpenBlue platform is the product of a partnership between Accenture and Johnson Controls to improve building efficiency and occupant safety as the world works towards sustainability goals.]

“The way that OpenBlue works is whatever devices are inside the building – whether they be a number of sensors or control systems that manage and measure different set points or chillers – these are always collecting different types of data. We have an OpenBlue edge gateway that we deploy as part of our solutions that ingest the data from all these different sources,” explains Vijay Sankaran, CTO at Johnson Controls.

“So, the whole idea is that you ingest all this information, you standardize it at the edge, you apply AI and ML models as applicable on the edge, and then some models get run up in cloud based upon the types of data that need to be actually integrated,” he continues. “So, we've developed a lot of different models like that, and then we also offer bidirectional control from our OpenBlue software directly into the building domain where we can adjust set points for HVAC units and things like air handling units that exist, in the space itself.”

Visualizing BAS Set Points in Real Time

While most commercial buildings today are equipped with sensors and BAS control systems that allow facility operators to monitor and manage things like HVAC units and thermostats, many of them are set-and-forget implementations. In other words, some set points are determined when the BAS is installed, then left alone, provided everything continues to operate as intended.

While that is automated, it’s not necessarily “smart.” And, for example, if a system like an air conditioner is running on a mild day the same way it runs on a hot day, a lot of energy is wasted resulting in far more carbon dioxide emissions than necessary.

That’s where OpenBlue’s edge AI comes into play.

“The way that a building has worked historically, both with Johnson Controls and its competitors, is you set up the control systems, you set the set points, and then nobody ever goes back and looks at it and the building operates fine,” Sankaran says. “But there’s no real-time adjustments of the buildings – set points, the speeds on the blower, or any inputs around how many people are actually utilizing the spaces.

“We recently acquired back in December a company called FogHorn Systems that's an expert in edge AI, where we're actually able to deploy models that are built in Python, R, and other languages directly on the edge so that we can perform a lot of those calculations in real time versus ingesting all of the data up into the cloud,” Johnson Controls’ chief technologist says. “We use everything from basic machine learning-type techniques to multi-stage neural network techniques in some cases, especially for video-type applications and text processing.

“Then we actually deploy that model into our OpenBlue AI-as-a-Service platform and integrate that into our OpenBlue Enterprise Manager, which actually presents customers with choices around optimizations based upon how many people are in a space at a time, and how we might improve ventilation in this space while balancing energy efficiencies,” he says.

Currently, OpenBlue provides operator-assisted control that precludes users from having to sift through mountains of BAS data but still gives them the final responsibility of issuing commands to BAS-controlled equipment on the ground. They are also working toward fully automated versions of the platform for customers looking for a hands-off approach.

“This can all be done in a completely closed-loop nature,” Sankaran adds.

Johnson Controls and MIT successfully demonstrated an end-to-end OpenBlue workflow in a case study that theorized increasing the ventilation of an enclosed space could reduce the spread of COVID-19 pathogens while also improving energy efficiency. The collaboration involved creating a physics-based multi-solve machine learning model in Python using a variety of test data sets and then deploying it with the OpenBlue platform to optimize different building environments.

“So, there's a lot of different techniques that we use as part of our AI team, but overall, it's the data and the connectivity that enables us to create a holistic set of offerings that ultimately add value using AI.”

AI’s in the House … and is the House

While impressive, that’s just one vector of AI and ML in the OpenBlue platform. All the data collected by the OpenBlue platform is stored in digital twin format and overlayed onto what’s called a building information map, or BIM. The BIM is an accurate, visual representation of a building complete with assets and alerting functionality that allows users to quickly locate, understand, and react to system failures when they occur.

Zooming back out, as these AI-enabled capabilities evolve they also continue to converge into a comprehensive strategy to address climate change. By combining real-time AI models and the predictive potential of simulating digital twins, building owners and operators will be able to adjust short-term equipment performance; ensure the efficient, green operation of their machinery in the mid-term; and, over the long term, gain insights that help inform the construction of new facilities.

“We’re constantly monitoring the space with real-time data and we’re taking in inputs like weather data – whether it’s going to be warm or cold outside – and ingesting all that data. We’re actually able to forecast and say, ‘Okay, if I increase the set point to this, because it’s going to be really cold, I can save this much energy.’ Or, ‘If I’m seeing that a chiller is running really heavy and I have a fault on one of my air handling units, I need to go dispatch a technician to fix that air handling unit or adjust that set point remotely so that my chiller isn’t operating as hard as it would be otherwise,’” Sankaran says.

“It’s really the data and the analysis that didn’t exist before that allows you to take action in real time within the building to adjust and really save energy through that process. That is just so powerful when you think about all the different devices that exist in the building and the ability to remove control via your building management system that wasn’t possible without a platform like OpenBlue,” he concludes.

Accenture and Johnson Controls have OpenBlue Innovation Centers across the globe to help tailor the platform and its AI models to customers' energy and emissions reduction objectives. For more information, go to