In our day-to-day work as design engineers, this term is typically used in design discussions with an owner to imply some level of building data collection and analysis. In practice, the energy dashboard is usually realized as a neat display at the entrance of a new building, neither understood by the building occupants nor well-utilized by the building operators. While intended to help owners and occupants make better decisions about the use and operations of the building, the execution of the dashboard is usually less inspired.
The discussion surrounding the energy dashboard is, nonetheless, an indicator of the foothold that data collection has gained both in our industry and everyday lives alike. The key difference between data collection and Big Data is the resulting analysis of the data points. The built environment represents a prime candidate for this type of computational analytics, marked by an understanding of, and action associated with, the analysis of these massive data sets.
Concurrently, market research estimates the compound annual growth rate (CAGR) of the building automation sector as 11% over the next five years. One driver behind this growth is the desire for tools to improve the financial decision-making capabilities of building owners, namely refining maintenance procedures and opportunities for decreasing energy consumption.
The convergence of wide-scale data collection (and analytics) and improved building automation system (BAS) functionality offers an opportunity for building owners to move beyond the tired concept of an energy dashboard and toward a new paradigm, where evidence-based decision-making is made possible through a strategically designed data management plan.
Improving Data Collection
In traditional BAS systems, coded (usually proprietary) controller logic, multiple signal pathways, and preset response patterns all require considerable computing power to read data from end devices. However, the BAS, as we currently know it, is experiencing a significant shift as sensors, actuators, and equipment capable of communicating over BACnet/IP become more readily available. Freed from the limits of controllers and proprietary logic, considerable financial and computing resources can be reallocated to smarter building solutions.
In the April 2019 issue of Engineered Systems, SmithGroup published an article discussing the elimination of proprietary controllers and control sequences in favor of a multifunctional reinforcement learning system agent. Responsible for the monitoring and control of all sensors and actuators in the building, the agent would most importantly be independent. By cutting ties to any single communications protocol or BAS software, building owners gain full control of those devices and the data they can provide.
However, to ensure functionality beyond the aforementioned energy dashboard, the agent must empower building owners to analyze the vast amounts of data now available to them. By deploying bots and artificial intelligence (AI) as part of the package, the agent can help building owners transform the way they maintain their buildings and manage their data.
At their core, bots (short for “web robots”) are software applications that run tasks over the internet. First developed to execute repetitive tasks more effectively than their human creators, bots now vary as widely as the types of tasks they perform. Perhaps the most ubiquitous type that exists today is the chat bot, which is able to retrieve weather forecasts and sports scores, convert currency, and even order pizza. More recently, though, chat bots have been deployed to help building occupants report maintenance issues.
Restricted to identifying an issue and where it took place (and possibly creating a work order), current facility bots are limited in their ability to help building operators efficiently address maintenance issues. These bots have, however, laid the groundwork for better bots to work in coordination with a BAS. As part of the larger platform, the agent’s bot would act as an intelligent interface, clearly communicating the nature of the issue between the building occupant and the building operator. With the filters in the agent’s bot (illustrated in the next section), building occupants could provide data vital to addressing the issue to the building operator without knowing anything about the building’s HVAC system.
Consider the following scenario: Parissa works in an office building served by a multi-zone VAV rooftop unit and hydronic reheat VAV boxes. She is seated in a corner office that is an independent HVAC zone that has its own box and thermostat. On a December afternoon, Parissa opens the agent’s app on her phone after feeling cold all day, follows the prompts, and answers the bot’s questions (see Figure 1).
Instead of simply creating a work order for a building operator, the bot’s filters use Parissa’s responses to identify the source VAV box and its operating characteristics, such as damper position, reheat valve position, discharge air temperature, current space temperature, and any other helpful information. With this data provided as part of the work order, the building operator can evaluate and address the issue more efficiently.
Taken a step further, as part of the agent platform, AI could further evaluate and potentially diagnose issues. Returning to the comfort issue Parissa was experiencing, the bot’s filters have now provided the relevant data points (see Figure 2) to the building operator. Based on this data, the AI could even suggest possible sources of the issue to the building operator as well as an order of likelihood (see Figure 3). In this modified scenario of a typical hot/cold call, it’s evident that AI can not only assist in identifying possible maintenance issues but also drastically simplify the response process.
Shifting the Maintenance Paradigm
While AI will never fully eliminate the necessity for reactive maintenance, it can be leveraged to help transform the implementation of predictive maintenance. The current practice of
maintenance is less predictive and more preventive in approach. Both dated and inefficient, the preventive method relies (often solely) on timetables and schedules based on a manufacturer’s suggestions and daily visual observations. Ignored is the vital contextual information that defines equipment or system operations, including such data as design operating conditions and recent performance trends.
To illustrate this proactive approach, consider an 800-ton, water-cooled, centrifugal chilled water plant with a chiller capable of operating at 0.46 kW/ton given 70% loading and 75°F condenser water (factory selection data points). With the load and condenser water conditions as stated, the chiller is noted to be operating at 0.53 kW/ton. This represents a 15% degradation in chiller performance from the baseline performance capabilities of the machine.
The AI would compare power consumption trend data against the factory selection points, evaluate scheduled maintenance cycles, and identify that the condenser tubes are due for cleaning in less than a month. Evaluating this scenario against other possible chiller maintenance issues, the AI would suggest and prioritize the most significant change agent in performance for consideration by the building operator.
Much in the same way a car can tell drivers to “Service Engine Soon,” AI can take operational data from smart sensors/equipment and combine it with established maintenance schedules to inform building operators when maintenance may be required ahead of its scheduled time. The concept of predictive maintenance is already driving advances in operational efficiency in the railway and manufacturing industries. Why shouldn’t the BAS be next?
Embracing Energy Management
Given the global demand for reduced energy consumption, it’s no surprise the built environment has seen a continual evolution of the policy and standard of care associated with design and operation of buildings and the infrastructure serving them. While the straightforward aspects of energy code compliance are generally well understood and documented during the design phase of a project, the realization of energy savings is too often the burden of the building owner. Even in the instance where an owner has a skilled facilities maintenance team, the strategic and tactical decision-making and operational practices associated with energy management are not traditionally the focus of these teams.
SmithGroup was recently engaged by an owner following the retrofit of several of their existing buildings with a combined total of more than 1,000 metering points. Trend data from this network of sensors and meters, collected in 15-minute intervals, amounted to almost 35 million data points annually. Lacking a means to organize and understand this data (for several years), the owner tasked SmithGroup to act as its energy manager to organize and evaluate the data. By post-processing the data each month, SmithGroup created an evidence-based approach to providing ongoing recommendations for control sequence modifications, identification of possible equipment failures, or areas for behavioral (facilities and staff) changes. Figures 4 and 5 represent a sampling of the summaries delivered to the owner at monthly energy manager meetings.
However, even this model of post-processing is rapidly becoming outdated given advancements in AI and machine learning. Similar to the model of predictive maintenance, the increase in building data capture can help move an owner toward a state of near-real-time energy management. While options for open-sourced AI offer flexibility to a capable building owner or staff energy manager, companies that offer an analytical suite of tools existing in partnership with the BAS provide another solution to an owner otherwise untrained in the fields of data analytics and energy management. This service makes the advantages of near-real-time energy management approachable to almost any building owner looking to make an immediate financial impact to their operations.
In the previous example of the chiller plant operating below the design efficiency, the maintenance cycle was the primary driver to understanding the performance issue. With the additional layer of energy management, factors such as weather conditions, building occupancy, and past performance become tools in the analysis. In this example, consider a weekly trend in chiller consumption. The energy management system (or software) can evaluate the previous two weeks of chiller operation as well as those exact two weeks in the previous year. Further, any chiller data from a period recognized as having similar weather and building occupancy conditions can be evaluated. The results indicate the chiller is 10% outside the normal range of energy consumption given the associated conditions. The building owner can now use the BAS to search for possible causes, such as set points that may be locally overridden or other equipment that may be in alarm (outside air damper stuck open, etc.) to help curb additional unnecessary chiller plant energy consumption.
As sensors and equipment continue to evolve and communicate more freely, engineers and manufacturers have a tremendous opportunity to collaborate with building operators to change the way data is used. Armed with open, intuitive platforms to organize and visualize system data points, and using AI to help make sense of that data, building operators can spend less time triaging issues as they arise and focus their efforts on running their buildings as comfortably and efficiently as possible.
Putting the Big Data analysis into the hands of AI or a software company should not relieve the engineering design team of their partnership with the owner in the successful design and operation of the building. By taking an active role in design phase conversations about data management, the engineering design team can more adequately ensure the correct information is collected and that the data serves a functional role in operational decision-making. This includes recommending specific control points for trending, creating relevant groupings of control points to understand the comprehensive impact of a system or subsystem, and representing this information graphically for effective consumption by the owner/operator. Critical to this phase is the engineering team’s ability to bring together the input of the owner, facilities team, commissioning agent, and any other project stakeholders.
While near-real-time building management helps to ensure typical (maintenance) and efficient (energy) operation, strategic decision-making at a macro level should include an ongoing support role for the design team. Perspective on new control sequences and improvement of system design based on feedback and data collection are prime examples of this level of owner support. Where the traditional project life cycle tends to end at occupancy, recognizing the engineer as a trusted advisor through the life of the building is more aligned with the goal of providing holistic support of the owner and his or her data management plan.
Every new design project is defined by a unique set of site, program, infrastructure, and financing complexities that make a single approach to advancing a data management plan impossible. Understanding each of these priorities and limitations early in the design process is critical to ensuring the approach is tailored to the project and, more importantly, the needs and goals of the owner. The convergence of data analytics and BAS improvements with more involved engineering design support is a critical step toward helping owners make evidence-based decisions about the operation of their buildings.