The ongoing heating and cooling needs of the world’s buildings account for 28% of energy related carbon emissions, according to the World Green Building Council. It’s no secret that creating a more sustainable environment depends upon our ability to innovate and quickly adopt smart technologies that can greatly reduce our carbon footprint. Among the companies leading the charge on this front is Montreal-based startup BrainBox AI.
BrainBox AI provides an innovative artificial intelligence solution that learns a commercial building’s thermal behaviors, and combined with other input data like weather and occupancy, controls the building’s HVAC system to reduce energy consumption without compromising occupant comfort. To date, BrainBox AI is deployed worldwide in over 100 million square feet of commercial building space.
As one might expect, the BrainBox AI algorithms rely heavily on data, most of which comes from HVAC equipment channeled through the building automation system (BAS). The quality of the data is of paramount importance because erroneous data points may affect BrainBox AI’s ability to optimize control strategies. To mitigate this possibility, BrainBox AI turned to Visual BACnet, Optigo Networks’ software for troubleshooting and monitoring building automation networks.
BrainBox AI’s earliest use case for Visual BACnet begins before any control strategies can be implemented and answers a critical question: can the building’s BAS network support the amount of network traffic generated by BrainBox AI’s data polling requirements? After all, the average age of a commercial building in the United States is over 50 years old, and the operational technology (OT) infrastructure needed to facilitate modern day technologies like BrainBox AI can often be lacking.
Making smart decisions about how to control a commercial building’s HVAC systems requires thousands of temperature, occupancy, and other data points to be polled throughout the network every hour, which can negatively affect network performance and potentially cause controllers to go offline, in turn creating broadcast storms. In a more typical scenario, the network traffic can cause dropped packets, missed alarms, incomplete data in trend logs, and other unintended consequences. By deploying Visual BACnet early-on, BrainBox AI can determine if the network is robust enough to support additional traffic. If it isn’t, BrainBox AI can work with systems integrators and building owners to determine appropriate next steps.
The second, and more important use case for BrainBox AI is to ensure the integrity of data used to continuously improve its AI-driven building control strategies.
To illustrate this use case, imagine a hypothetical example of two Variable Frequency drives (VFD) controlling the fans of an air handling unit (AHU). These fans are typically controlled by physically wiring to the control points, not through the BAS network. However, the status points are normally communicated to the BAS and are often used by applications like BrainBox AI to optimize control strategies.
If both of these VFDs had the same BACnet ID, then the data being collected could originate from either VFD at any given moment, causing the values to be wrong at least half of the time. While this mistake wouldn’t directly affect the AHUs being controlled, the outcome would be erroneous data being trended, alarms not triggering due to incorrect input values, and applications like BrainBox AI interpreting erroneous data – potentially leading to suboptimal control strategies and false positives.
A simple mistake like this can go unnoticed for quite some time absent a tool like Visual BACnet which continuously scans the network for duplicate devices and other BACnet anomalies. Once detected, remediation is a simple matter of changing the BACnet ID of one of the duplicate devices to a unique instance.
This type of network issue can be particularly pernicious as increasingly stringent building codes and local laws like LL97 in New York City and Title 24 in California force commercial building owners to improve their building control strategies to reduce energy consumption while ensuring healthy levels of fresh air. Failure to do so may result in heavy fines for non-compliance as early as 2025.
As building owners develop plans to update and retrofit their buildings with more efficient HVAC equipment to meet these tougher regulations, they are deploying solutions such as BrainBox AI to their existing systems to control them. Yet, even that process can yield its own set of problems.
Take for example a building with brand new boilers and chillers connected to a small MSTP network with a dozen controllers or less. A well meaning but green technician arrives on site to install an additional device on that network. Unfortunately, he forgets to ground the new controller causing the entire MSTP trunk to lose communication.
Like other technicians who haven’t yet discovered Visual BACnet, he/she proceeds to bifurcate the network – an arduous task since the other controllers on the network are located in hard-to-reach areas of the building. Meanwhile, due to the communication loss, the boilers and chillers have reverted back to their failsafe settings, causing them to operate at full capacity until the problem is resolved – potentially a days long affair.
Here again, a communication loss on a small MSTP network can lead to a number of negative consequences, namely:
- Increased energy consumption to run the mechanical equipment in failsafe mode while technicians troubleshoot the problem, leading to increased energy costs, and reduced equipment lifespan.
- Finger-pointing between vendors, facility managers, and building owners.
- Complaints from building occupants, causing friction between tenants and building owners.
By connecting a laptop installed with the Visual BACnet Capture Tool to any device on this MSTP network, our technician would have quickly seen a failed Checksum Errors diagnostic, indicating that the communication loss was due to a physical wiring issue. That would have turned a potentially days-long affair into an hour-long task.
As evidenced by the above example, the prevalence of tools like BrainBox AI and other machine learning (ML) technologies has not (and will not) completely remove human error from the equation, at least not anytime soon.
One thing is certain, it has become fundamentally clear that a healthy BAS network is table stakes. Fortunately, verifying network health has never been easier with Visual BACnet, which allows all building owners and system integrators to quickly ascertain their network health baseline, for free.