As businesses look ahead to the easing of lockdown restrictions, energy managers are facing the challenge of how best to prepare their building portfolio for what will likely be a progressive return to normal operations.
Increasingly, facilities management professionals are turning to advanced digital technologies, including artificial intelligence (AI), to ‘map’ different energy consumption scenarios that they can implement quickly as occupier levels change or production ramps up.
Many are looking forward by looking back, using AI to interrogate years’ worth of historic half-hourly gas and electricity data to create multiple energy consumption models that reflect differing prevailing operating conditions. Analysing data on this scale manually would need an army of analysts – but with AI, it can be a matter of hours, even minutes to yield results.
For example, through its AMR DNA service, metering and data services specialists Energy Assets is enabling facilities managers in multiple sectors, including higher education, retail, local authorities and commerce, to use AI to develop accurate scenario planning linked to the status of the pandemic.
This means managers can adapt quickly to changes in COVID rules and be confident that whether it’s a full lockdown, localised tiered restrictions, a ‘COVID-secure’ normal requiring social distancing, or a return to pre-COVID normality, they are in a better place to maintain optimal energy efficiency.
“Through AI-informed scenario planning and reporting, energy managers can apply the correct measures to any given situation in order to minimise energy waste and optimise energy procurement,” said George Catto, Client Services Director at AMR DNA.
In scenario planning, Energy Assets uses kWIQly AI architecture to assimilate two years’ worth of half-hourly gas and electricity meter data into a performance model that can be measured against key criteria to identify waste. Energy waste can result from something as simple as leaving lighting on overnight or, more critically, failing to revise heating schedules when the clocks go back or forward.
Identifying and eradicating this waste can be a particularly pressing challenge in an environment of change, when building occupancy levels are fluctuating. In any portfolio, site-by-site performance varies with respect to its indicators, benchmarks and relative to prior performance. For example, by how much consumption falls overnight.
In the case of AMR DNA, AI finds and flags up areas for energy efficiency improvement because the system progressively ‘learns’ what best performance looks like.
From this benchmark, energy managers can use the AI-driven model to develop building profiles that fit multiple contingency planning scenarios and to sharpen their energy procurement strategies.
Says George Catto: “Managing energy efficiency boils down to doing only what is necessary at least cost. Since Covid-19, ‘what is necessary’ has changed, as have metrics of cost. Pre-pandemic it was enough to measure and predict kWh in terms of normal operations and interpret cost around carbon impact and money.
“Rarely was health considered a ‘cost factor’. However, requirements to keep a building healthy have changed, and not in one dimension. Today, COVID means that energy and facilities managers need to take account of ideal temperatures, air-change rates, humidity, fresh-air circulation, heat recovery strategy and even comfort. Why comfort? - because breathing through PPE changes ambient requirements.”
While AI cannot predict the economic well-being of an organisation, political sentiment or price of energy, what it can do is provide predictability using rock-solid data from historic automated meter readings married with variables such as weather, occupancy, operational policy etc..
This enables AI to remodel outcomes under any set of assumptions based on available data.
“What is quickly becoming apparent is that COVID is challenging our conventional understanding of energy management systems and how meter data can positively affect scenario planning,” says George Catto. “This ability to forecast what the future can look like, and adapt quickly to changing circumstances, is gaining a lot of interest because no organisation can afford to burn money unnecessarily in today’s trading climate.
“Being able to implement new scenario strategies quickly also gives managers the ability to revise their energy buying strategies with confidence, because they have a clearer understanding of their future requirements based on data-driven insights rather than guesswork.”
AI is just one example of how digitalisation through advanced metering and data analytics is helping businesses to take control of energy performance in buildings, bear down on cost and reduce their carbon footprint.
Covid Scenario Planning for Energy in Buildings
Some buildings have always been intended for continuous occupancy, so it is not surprising that not all buildings respond to being unoccupied in the same way. However, given any set of buildings (and an energy management team) some patterns or norms emerge when occupancy levels fluctuate.
This raises questions such as: What is achievable in terms of energy reduction, is the optimal state applicable and uniform across an entire estate, where can it be improved, how will the estate respond to increased ventilation under colder external conditions and how can energy purchasing decisions best reflect the different scenarios?
What is needed is a strategic approach. These graphs from AMR DNA illustrate how AI can help.
This chart (on the left) represents a set of buildings belonging to one owner where each dot is a gas meter. Some buildings have been switched off entirely (bottom line).
While it can be seen that one very large building (top-right) and a handful of others have increased consumption, the majority have seen a 20%-40% reduction over normal operating conditions, depending on the level of turndown. Having this knowledge means that managers can assess opportunities for additional energy saving measures.
This chart (on the right) shows on top a weather adjusted model by time of week and outside temperature (from which daily profile averages given forecast temperature can be calculated).
The section below shows the single daily profile of actual consumption (blue bars) compared with achievable (grey shaded). This immediately illustrates a problem; that although consumption inside occupancy (core hours) has improved, during evenings and before occupancy the consumption has risen.
Having this information to hand enables energy consumption under different scenarios across portfolios to be accurately tracked automatically, to identify not just the impact of turning down but also where buildings are not responding as anticipated. As well as enabling follow-up action, this data can greatly assist with budgetary and forecast scenario planning.
This article was first published in Energy in Buildings & Industry April 2021.