Machine Learning is the process by which a computer acquires knowledge from patterns. How does this relate to HVAC? Firstly, control of the air deals with how people occupy rooms in a building. We are creatures of habit, and from our habits arise patterns. A machine learning algorithm, recognizing these patterns of behavior, will develop a more precise HVAC system. The ability to recognize patterns in human behavior opens new pathways for HVAC efficiency and effectivity.
No Shortage of Approaches
Some organizations pay a lot of money for a third party to perform analytics on their buildings. It’s no secret that HVAC is one of the largest energy consuming activities in buildings. It accounts for over 40% of the energy use in large commercial buildings, yet many people still aren’t comfortable. Clearly there’s room for improvement. In order to make their HVAC use more efficient, organizations need to know how people move through the building. These analytics companies gather data on energy use and how people occupy the building. The client organization will then take this knowledge and adjust their practices to increase efficiency.
People try to make HVAC more efficient and more effective in different ways. Solutions centered around analytics aim to improve functioning by being more knowledgeable. Sensor based solutions detect occupancy in real time, adapting to changing conditions. Some try to make HVAC better through automating building operations. A variety of solutions tackle the issues with HVAC by trying to make it more efficient, better at providing comfort, or both.
The Answer is Multi-Faceted
A truly complete HVAC solution needs all of the following:
- An awareness of how many people are in the building at a given time
- In order to deliver precise air conditioning and ventilation to a room, it is necessary to know how many people are in that room. A maximally occupied room has vastly different HVAC needs than a room occupied by a single person.
- Knowledge of users’ level of comfort
- If the goal of HVAC is mainly or partly to keep occupants comfortable, then it is important to know the quality of their experience. Otherwise, you are merely guessing for what users like. And everyone has different preferences, so it’s important to know their level of comfort down to the individual level for optimum comfort delivery.
- Data about the first two points over time for optimization
- Knowing occupancy and user comfort preferences over time gives you evidence that can inform future decisions. With enough reliable data over time, you can make accurate assumptions about future conditions. This is crucial for optimization.
- The ability to automatically make control decisions
- You need a computer that can take in the data and take action on the HVAC system in order to have optimum output of efficiency. Human input is slow and laggy. The optimum solution involves computer control.
Machine learning is the glue that holds these points together. It’s what synthesizes the data provided by occupancy and comfort detection, analyzes it over time, and picks out the best course of action for the HVAC system. Without machine learning, you just have a mass of data. No optimized course of action, no meaningful insights gained.
The Role of Machine Learning
The rise of Big Data, together with increasingly greater processing capabilities, has moved machine learning out of academia and into mainstream tech. Big data is food for machine learning algorithms. Collecting data on occupancy and users’ comfort is important in delivering more effective and efficient HVAC in buildings. This information won’t solve anything by itself. It’s also impractical to expect humans to manually sift through the data. We need the right machine learning algorithms to read through and make conclusions about optimal HVAC functioning.
Machine learning algorithms filter through the occupancy and comfort data and make sense of it. It synthesizes the data to deliver comfort and efficiency. The most efficient HVAC is one that is turned off. So efficiency isn’t the only goal. The true challenge is in delivering comfort to building occupants in the most efficient manner. The best machine learning algorithms take in the raw data, and compute the most optimally balanced decisions based on preferences. It will look at how occupant activity fluctuates for a given room, and it will look at user comfort preferences for that space, and based off these readings, it will determine an optimal action plan to efficiently condition the space for just that amount of people, with their preferences kept in mind. Some organizations will prioritize comfort over efficiency. Some the opposite. The right algorithm will allow space for these organization-specific preferences while keeping comfort and efficiency as priorities.
In time, these machine learning algorithms will have the power to predict the future. Humans are more predictable than we think. Analyzing long periods of occupant activity, they can make accurate guesses about what future activity will be. This is where the magic enters. They adjust the system in preparation for accurately predicted occupancy activity. It’s a new door opened for efficiency. HVAC’s blindness to time is addressed. A history of user comfort preferences and occupancy within rooms will inform the system in anticipation of real activity. The result is a seamless experience where the air is magically heated or cooled to your preference before you enter the room and does it just enough as to be as efficient as possible.
Machine Learning makes peoples’ jobs both more powerful and easier. For building, facilities, and energy managers, having a built in machine learning capability into their system of control grants them powerful analytical capabilities. Instead of contracting out to some company to get analytics about their building to improve service, with a technology like this, analytics can be done in real time. It gives them the tools for a full range of efficiency and effectivity. With the knowledge and capabilities provided by machine learning, managers can choose to orient the system more around efficiency or comfort. It’s not just that machine learning makes things more efficient, it’s that it puts more tools in the hands of those in control. It makes their job both easier and more robust. Without machine learning, you would have the old way: humans looking through a big stack of data, manually figuring out the best course of action. It’s simply a necessary part of the next generation of smart and green buildings.