It’s hard to believe that we are really only five months into a pandemic that has upended almost every aspect of our lives. I want to share with you some of the ways Gridium has adjusted its analytics in response to changing patterns of energy use in buildings. This is going to be a long message with a lot of details, so I’m going to jump right in.
Gridium builds a statistical model of energy use in your building based on its historical behavior. We use this model in all sorts of ways:
- To detect possible faults or controls issues when energy use diverges from expectation.
- To break down bill variance by source, such as weather or rate changes.
- To forecast demand peaks in the upcoming week.
- To calculate long-term energy budgets.
In all cases, the statistical model works by assuming that building behavior in the future is going to be similar to the past. Covid challenges that assumption.
On March 4, the date that most of the country went into lockdown, we also put our model into lockdown. That is, we locked the historical baseline for the model to the pre-Covid period of March 2019 – February 2020. Since early March, Gridium’s analytics have treated your building’s pre-Covid energy use as its expected energy use moving forward.
Why do this? Locking the baseline served a useful, short-term purpose: it kept our energy use predictions stable, revealing exactly how Covid is affecting your building on a week-by-week basis compared to its pre-Covid, “normal” behavior.
During this period, the chart of your energy use vs. expectation probably looked like the one above, showing a massive decline in energy use due to reduced occupancy.
Introducing: budget accrual
At the same time, we launched a new budget accrual feature that took the opposite approach to the locked baseline. To help people forecast their monthly energy costs under highly volatile conditions, the accrual feature adjusts its estimates based on your most recent week of energy use. This focus on recency allows more accurate cost estimates to be provided to the accounting department.
Next week: ending lockdown
We can’t keep the model baseline locked forever. Eventually the world is going to settle into a new normal. Although we’re probably not quite there yet, we do have five months of post-Covid data, which is enough for the model to start reflecting your building’s new pattern of energy use with reasonable accuracy.
Starting on August 3, we are unlocking the baseline period used for our statistical model. The model looks a year back, so the new baseline will incorporate seven months of pre-Covid data and five months of post-Covid data. As time passes, more and more of the model’s predictions will be based on the post-Covid time period. Expect to see the model’s predictions continue to adapt over time.
You should notice some immediate changes:
- The actual vs. expected figures for your energy use will settle down to less extreme values, as the expected values will no longer be based on your building’s pre-Covid behavior.
- For the same reason, peak forecasts will become much more reliable. Demand management hasn’t been a pressing issue for the past few months, because buildings have been mostly empty. As occupants return, you can once again rely on Gridium’s peak forecasts to help manage demand costs.
Budget forecasts will remain locked
One model that we’re not unlocking is the budget forecast. We don’t want next year’s forecast to mirror the enormous and hopefully one-time drop that occurred starting in March of this year, so we will continue to base budget forecasts on the pre-Covid baseline.
You may be expecting changes to occupancy next year based on the ongoing pandemic, in which case you can use the existing tools built into the budget forecast to make adjustments.
We added an information box at the top of your weekly emails that lets you know how your energy use over the past week compares to your pre-Covid baseline. We’re going to keep that feature for the foreseeable future, so you can continue to get an easy read on how the pandemic is affecting your operations.
The accrual feature will likewise continue to work as it has been, hopefully growing even more accurate over time as your building operations stabilize.
Summing it all up
These are a lot of details, but the big picture is that we will continue to refine our analytics on an ongoing basis so that they deliver the most useful information possible. Covid has injected a great deal of uncertainty into, well, everything. Unfortunately, there’s no mathematical trick that will make that uncertainty disappear, but by tuning our analytics we can make sure we continue to deliver operational insights that help you run your building as efficiently as possible during these very strange times.