Modernizing the Cold Chain: UCSB Researchers Team Up with Industry Partners to Bring AI-based Control to Industrial Refrigeration
Project PIs: Jason Marden and Mahnoosh Alizadeh, Profs. ECE, CCSP
Perishable products such as food, beverages, and chemicals are stored in large refrigerated warehouses before reaching consumers. These cold storage facilities operate continuously to maintain strict temperature requirements that preserve product safety and quality. As a result, the industrial refrigeration sector alone accounts for nearly 10% of U.S. energy usage, yet much of it still relies on legacy control technology. This reliance presents a significant opportunity for improved efficiency via modern AI techniques. However, implementing these modern control strategies is challenging due to the scale and complexity of industrial refrigeration facilities. A typical medium-sized facility spanning 100,000 square feet consumes around 2.5 million kWh annually. To address this challenge and modernize how these massive systems operate, UCSB ECE Professors Jason Marden and Mahnoosh Alizadeh are leading a unique research project.
At the heart of this project is a deceptively simple question: how do you teach an algorithm to manage a system that never behaves the same way twice?
Industrial refrigeration facilities are constantly contending with shifting variables, i.e., product loads, ambient weather conditions, equipment wear, and fluctuating electricity prices, making static, one-size-fits-all control strategies fundamentally inadequate. Critically, any solution must also be scalable: algorithms developed for one facility need to translate across the diverse range of equipment configurations and operational profiles found throughout the industry. The research team is tackling this through a combination of digital twin simulation, system identification from real operational data, and learning-based control methods designed with built-in safety guarantees. That's where industry partners CrossnoKaye and Lineage Logistics prove indispensable, providing vast operational datasets and the physical infrastructure needed to move solutions out of the simulation environment and into widespread practice. Early results are already encouraging as learning-based control approaches have shown the potential to cut operating costs by up to 22% without compromising storage safety.
This project thrives on the complementary knowledge of the diverse team. Professor Jason Marden is an expert in optimal control and has served as a technical consultant for CrossnoKaye, while Professor Mahnoosh Alizadeh contributes crucial expertise in electricity markets and safe reinforcement learning. Dr. Jesse Crossno and Dr. Alex Wolf represent CrossnoKaye and Lineage Logistics respectively and provide firsthand experience in industrial refrigeration. As graduate student Yohan John notes, "The industrial refrigeration project is a rare opportunity in academic research for developing AI with a positive real-world impact". Fellow team members echo this enthusiasm for practical applications with Arghavan Zibaie adding, "Industry collaboration in this project makes it possible to move beyond theory and face the real challenges of developing learning-based control solutions".
