Buildings owners can use machine learning to extract knowledge from data
While machine learning and artificial intelligence may sound like industry buzzwords rather than real cost-saving applications for building owners, these technologies are poised to play a significant role in reducing costs and increasing efficiency in building operations.
In the U.S. alone, the combined energy costs for nearly six million commercial buildings and industrial facilities is estimated at $400 billion. The average building wastes 30% of the energy it consumes due to built-in inefficiencies, and ongoing operating costs represent 50% of a building?s total lifecycle expenses over an estimated 40-year lifespan.
While IoT-driven management solutions provide real-time information about buildings using data from automation systems, fire safety, power systems, security systems, machine learning multiplies the value of data by turning it into knowledge that building owners can leverage to drive cost efficiencies.
What is machine learning?
Machine learning is nothing more than a means of extracting knowledge from data. It is a way of organizing data in a statistically significant way to predict future behavior. The primary benefit of machine learning is that it can manage and analyze mass amounts of data that humans can?t. In buildings, machine learning takes a static system and its data and makes it dynamic by learning from previously collected information from sensors, measuring devices and occupant behaviors.
There are two types of machine learning supervised and unsupervised.
Supervised machine learning or predictive modeling is the process of using data to make predictions. For example, an algorithm can be created based on temperature, sunlight, time of day, shades and the number of occupants, to determine precisely how much energy a building owner can save. Algorithms can also be created to predict when a replacement belt should be budgeted based on rising operating costs. This is supervised learning because it is used to determine a specific outcome.
Unsupervised machine learning is the process of extracting structure from data or learning how to represent data best. Unsupervised machine learning looks at raw data and spots patterns within it. For example, data set of the characteristics and purchasing behavior of occupants in commercial real estate building- the task may be to segment these occupants into enterprise customers and small business owners based on their actions and then use the information to provide solutions based on occupant needs.
Another example is analyzing, occupant traffic patterns from sensors or security camera data and correlating these patterns with energy consumption and cleaning requirements in specific areas of the building.
Big tech using machine learning?
Tech giants like Amazon, Microsoft, and Alphabet, are all developing machine learning engines in their cloud-based applications. Today the majority of IoT cloud-based platforms have some element of machine learning incorporated into their cloud-based analytics programs. IBM offers the Watson Internet of Things platform, and Microsoft Azure and Amazon also have machine learning services.
Alphabet, the parent company of Google was one of the first to invest in machine learning to reduce energy costs in its data centers. DeepMind, owned by Alphabet has successfully used a machine learning algorithm to reduce the company?s energy bills by nearly 40%.
Machine learning, IoT and energy savings
Many existing building systems are controlled by direct digital controls (DDC). These devices have static programming and are usually rarely adjusted or optimized after installation. These controllers are programmed to accomplish tasks such as the opening/closing a heating valve to maintain a 72-degree space temperature or turning on/off the lights based on a schedule. These devices use a limited number of sensors to adjust settings. Most building HVAC and lighting systems are most often on an off binary schedules: weekday and weekend or holidays.
However, the way we use these buildings is more complicated. Fridays may not require the same energy loads or number of working elevators compared with the day a company is hosting its annual investors meeting at headquarters.
By leveraging historical data, a machine learning algorithm can automatically react to real-life conditions and reduce consumption as needed.
Machine learning and building maintenance
Facilities managers are starting to use machine learning to develop more efficient maintenance plans. The schedules for critical equipment systems, such as lighting and HVAC, are often set in a relatively static way and are rarely analyzed or updated. This is because the data points involved in determining the degrees of occupancy is too vast and complicated for any human to compute.
Machine learning can take thousands of data points from equipment usage and various sensors to ?learn? the exact schedule of a building and provide building operators with insights about when and how to change equipment schedules to maximize efficiency and reduce costs.
It can also help to forecast long-term costs and improvements. If an owner knows the estimated knows that estimated occupancy rates are expected to increase or that a portion of the building is used more often by occupants, owners can use the data from machine learning to budget maintenance, repairs, security and other costs for high traffic areas in advance more precisely.
Alternatively, machine learning can help building owners to understand which areas are under-utilized such as conference rooms, common areas, and even bathrooms. If these areas aren?t being optimized then, owners and tenants can take actions to find a better use for those locations.
Leveraging occupant data for improved customer experience
The dramatic increase in the use of IoT devices and sensors is enabling building owners to leverage user-based data to deliver better outcomes for occupants through space utilization. Machine learning can analyze how occupants are navigating and using a building?s space to improve outcomes and cost savings for both tenants and building owners.
In the retail sector occupancy sensors are deployed to determine where and when shoppers are entering and exiting malls. While the data itself is useful, adding machine learning to it, can help retail space owners identify precisely how many dressing rooms, restrooms, displays are necessary during a particular time of year. Improving occupant experiences inside of large retail spaces can help to drive and anchor tenants in the long-run.
Further data collected and analyzed using predictive analytics can provide powerful insights to building owners about where and which tenants to place in specific locations.
For many building owners making occupancy data available to marketing teams has not only improved tenant retention but also become a massive differentiating factor for new tenants considering retail leasing space.
