Article Top Ad

Why this cold storage warehouse operator warmed up to artificial intelligence

The energy team supercools its facilities to help offset energy costs. Intelligent software helps figure out when they should be switched back on. Read More

(Updated on July 24, 2024)
Lineage Logistics
The Lineage Logistics facility at the port of Oakland

This article is adapted from GreenBiz’s newsletter, VERGE Weekly, running Wednesdays. Subscribe here.

If you think your organization has a challenging energy consumption profile, try running a network of cold storage facilities — with most sites in not-so-cold locations, such as California and Georgia.

One of the more intriguing examples of smart power management I’ve come across in recent months is an initiative under way at Lineage Logistics, which runs more than 200 warehouses across North America, Europe and Asia. It handles food for more than 2,500 customers, which are all businesses themselves, such as grocery stories and food services organizations.

Each of the Lineage buildings — an average of 250,000 square feet in size — basically stores about the same amount of food that you’d find in 770,000 home freezers. There’s a range of equipment involved including refrigeration units and blast freezers that get the temperature inside down to minus 20 degrees. In aggregate, Lineage’s annual energy consumption is on par with a mid-sized U.S. city. It’s the company’s second highest cost after labor, and there’s a corporate goal (established back in 2014) to improve energy intensity at least 5 percent annually. 

The approach that the Lineage operations team has taken to dramatically reduce that load uses some pretty sophisticated technology, including artificial intelligence, wireless sensors and cloud software services. 

Yet at the heart lies a profoundly simple idea: Rather than keeping the facilities super-cooled all the time, the team figured out a way that Lineage could use the presence of all that frozen food — and all the cold thermal energy it emits — to its advantage.

The solution: overcool the buildings overnight — or when power costs are lower — to about five degrees below the safe temperature for storing frozen food (that would be zero), taking advantage of lower electricity costs. The freezers are then turned off until the temperatures rise to a pre-indicated level. “You don’t want to go above zero, but you can actually go colder,” observes Alex Woolf, principal data scientist for Lineage, who was instrumental in designing this system.

Depending on the size of the facility, the specific sorts of food being stored, the region, and outside weather and ambient air conditions, the freezers might sit idle for anywhere from five to 10 hours before they become more active again. For safety reasons, Lineage invested significant hours on safety checks before embracing this approach, Woolf says, which it’s using in close to 80 facilities. Sensors from industrial.io keep watch over conditions constantly, communicating the data with an AT&T wireless hookup to an analytics dashboard that lives in the cloud.  

Since Lineage embraced this approach, it has reduced annual electricity consumption by 33 million kilowatt-hours, saving about $4 million along the way. Some facilities actually have managed to cut their power costs in half, Woolf says.

So how do you figure out the precise moment to make sure the equipment is re-engaged without having a person watch the thermostat and some complex dashboard on a 24/7 basis? That’s where the aforementioned machine learning technology (PDF) comes in — it works in collaboration with the internet of things technology that Lineage has been putting into place across its network. “I don’t want to sit in front of a computer all the time. That’s what the AI does, and it scales,” Woolf says. 

Without getting super geeky, the approach that Woolf’s data team developed — it calls the process “flywheeling” — builds on the investment that company made in installing more than 1,000 sensors through its warehouses over the past few years.  

The AI software correlates the data the sensors are collecting with other information, such as the price of electricity, the suggested cooling schedule for an individual building and other more unpredictable variables such as weather conditions. Then, it takes action based on certain predefined conditions. For example, if temperature dips more quickly than expected or, conversely, whether the power can be kept off longer than anticipated because of colder outside temperatures.

It also can serve as a demand response mechanism, allowing Lineage to shift its load based on peak pricing changes from its utilities. That’s a big deal, especially in markets where the intraday price per kilowatt-hour can swing from 2 cents to 80 cents based on electric grid demand.

“We can modulate our load based on what we need,” Woolf says. “The computer gives you the solution.” Of course, there’s always an override button in case of an unforeseen emergency that requires human intervention.

The flywheeling application was piloted in Mira Loma, California, where the company was able to achieve a normalized cost of 38 percent in refrigeration costs per pallet. Now, it’s being rolled out to about a half-dozen other facilities. Woolf says now that the AI program has been written, it only takes a few weeks and “minor tweaks” to get the capability up and running at a location. He has received multiple patents related to the technology, and Lineage was recognized this summer with a Better Practices award for energy efficiency by the U.S. Department of Energy’s Better Plants Program.

If you’re coming to VERGE 19 Oct. 22 -24, you’ll have a chance to hear Alex Woolf talk more about Lineage’s profoundly practical approach during a workshop I’m moderating right after lunch Oct. 23: “The Future is Now: Energy Blockchain and AI in Action.”

Trellis Briefing

Subscribe to Trellis Briefing

Get real case studies, expert action steps and the latest sustainability trends in a concise morning email.
Article Sidebar 1 Ad
Article Sidebar 2 Ad