Classifying smart buildings: a guide to four levels of smart
Workplace technology is in a constant state of beta, so how can a successful smart building strategy be deployed? Luke Bailey of UnWork explores the four stages of smart building implementation in the first instalment of a trilogy on strategy
Due to the nature of self-learning technology and the ability to retrofit different facility systems as and when new innovations are released, smart buildings tend not to be deployed as the finished article. Instead, there are four broad stages that a building progresses through to become truly smart; the four levels of smart can be seen as a transformational process at the core of a smart building strategy or as a method of evaluation, a benchmarking tool that helps understand just how intelligent a building really is.
Four Levels of Smart
The very best smart buildings harvest vast amounts of data from as many different sources as possible; building systems and sensors are capable of providing huge volumes of operational data regarding status, usage levels, maintenance needs and more. Data capture provides a foundational layer for the smart building and one that should be entirely interconnected, providing a comprehensive ecosystem of trackers and sensors, each producing real-time data to be stored in a data lake for aggregation, contextualisation and analysis further down the line.
For the smart office there is no more powerful data stream than that of real-time occupancy and people counting. Understanding which workstations are currently being used, how many people are flowing through each space and the historical patterns in utilisation unlock a plethora of compounding applications, ranging from demand-driven HVAC and lighting to improved agility and success in activity-based working.
While the data gathering stage is not necessarily ‘smart’ in its own right, it is absolutely pivotal to the intelligence of buildings. As with all IoT solutions, data is the lifeblood of the smart structure.
A fully deployed smart building contains hundreds of thousands of different data points, all feeding live information to the building automation system (BAS). While the most advanced sensors are capable of processing information at the edge to provide purpose and context, much of the information gathered is raw. For storage, the two structure classifications of data – raw and processed – are kept in different locations.
Pre-processed data can be sent straight to a data warehouse, a repository specifically designed for structured, filtered data that has been predefined for a specific use/uses. Most data, however, flows directly into the data lake; larger in storage capacity and more accommodating to malleable, unprocessed data, the data lake is essentially a giant bank of untouched data. Data lakes are highly effective in aggregating data from a variety of different disparate systems – analysts can use the lake to pull out data and produce bespoke dashboard and reports without setbacks through the common issues of privacy and ownership.
This also applies to smart building analytics software – facility managers can use the aggregated data stored in these repositories for a number of pain-free use cases. Preconfigured live reports from the data warehouse could be used on a global dashboard for property owners to view performance of the whole portfolio of assets in an accessible and stakeholder-friendly way; workplace analysts can access raw occupancy data to run advanced analytics that their solution provider doesn’t offer; and building operations teams can view the status of all building systems on one unified platform, known as the single pane of glass.
While the data gathering and processing stages are entirely necessary phases of smart building development, neither of which showcase any fundamentally ‘smart’ capabilities themselves – instead they provide the platform for the introduction of artificial intelligence and machine learning (a type of artificial intelligence) in the third level of smart: the autodidactic stage.
The primary purpose of machine learning’s use in the built environment is to create analytical models capable of continuously learning from available data, unassisted by humans. Machine learning algorithms can work through huge quantities of raw or processed historical data to turn the numbers into powerful and actionable intelligence; the process, known as mining, helps to inform better operational decisions by spotting patterns and trends in the data and contextualising them with other building systems or external data sources to understand exactly why certain things happen.
‘Machines make more accurate calculations than humans…’
The process can then be applied to make better informed decisions for a number of different smart building applications. Predictive maintenance identifies specific factors that commonly lead to breakdowns of specific hardware or systems, flagging to the FM team when a fault is expected so that the issue can be resolved with minimal downtime; AI-enabled energy management systems can be entirely demand-led, recommending sections of the building to shut down based on predictions of future occupancy levels; and ambient intelligence, a pervasive computing element of AI, can respond to the presence of the identified occupant, adjusting the immediate environment and workstation to meet their previously-monitored preferences.
Though it is unlikely any building will reach this stage for two to three years, a cognitive building is one that harnesses AI and machine learning such that its processes become largely automated. The autonomous building uses historical building data and external data with digital twin technology – a dynamic virtual representation of the physical asset depicted as an interactive 3D model, created using computer imaging and incorporating the vast quantity of data smart buildings provide – to assess the probabilities of outcomes of various scenarios by running multiple simulations on the model.
At this stage, the vast quantity of data combined with the accuracy of the experienced machine learning algorithms mean that any prediction or calculation made by the building is far more accurate than a human would be capable of. In the automation stage, rather than just making predictions to the FM team, the building responds to current demand and anticipates predicted future events to automate building processes without need for human input. The technology constantly learns from any errors and over time the building is able to maximise operational efficiency with very limited human interaction.
The four layers of smart denote the technological transformation even new-build structures experience to become genuinely intelligent. The process focuses on the evolution in sophistication of the technology that facilitates building operations, how the benefits increase as the amount of available data grows and the way in which systems optimise by analysing historical and predicted trends.
This smart building framework is a useful guide to evaluating how intelligent a building actually is, but it can also be seen as a roadmap for those looking to develop a smart building strategy.