Using Machine-Learning to Reduce Energy Consumption
February 08, 2020
Lineage handles a staggering 40 billion pounds of food annually — 120 pounds per person in the United States, to be exact. Storing such enormous amounts of food requires enormous amounts of electricity.
Starting in 2014, Lineage realized we needed to start thinking creatively to drive down the cost of our monumental electric bill and reduce our environmental impact.
Charged with finding a creative, technology-driven solution to this challenge, our Applied Sciences team went to work to rethink the way temperature-controlled warehouses consume energy.
Their breakthrough was an innovative process we designed to proactively manage our energy consumption called flywheeling. Through flywheeling, we changed our electricity from a flat consumption to one that absorbs excess production from renewable sources opportunistically. When the electricity grid lacks supply, we can eliminate our consumption, reducing reliance on fossil fuels, without affecting our customers’ product.
Flywheeling lets us take advantage of low-cost renewable sources during peak production and reduces our overall demand when prices are highest.
How it works.
Lineage first predicts when peak demand, and therefore peak price, will occur. We cool the warehouse to a lower-than-normal temperature in advance, avoiding peak energy charges.
This process effectively turns the facility into a battery, allowing greater proliferation of intermittent solar and wind production without matching fossil “peaker” capacity.
Fly-wheeling relies on advanced mathematical methods – from machine learning to predict the properties of the grid and the warehouse, to artificial intelligence that makes scheduling decisions based on those predictions.
The number of homes Lineage could power with its energy savings alone
The reduction in Lineage’s energy intensity from 2014-2017
Lineage's reduction of kilowatt hours year over year