SBIR-STTR Award

Commercial refrigeration systems are the largest energy consumers in food retail stores, constituting up to 62% of their electrical energy usage. Around 20% of this energy going to refrigeration is wasted due to the machines’ low energy efficiency. This results in over $4.4 billion lost annually by domestic food businesses. This low efficiency is coupled with low resilience in commercial refrigeration systems. Most food stores lack sufficient backup power or energy storage to continue running their entire refrigeration system in an outage, so they stand to lose $367,000 to over $900,000 in food inventory at each extended power outage. These problems will be addressed with non-invasive retrofit hardware that collects data and hosts the proposed artificial-intelligence-based software to run each building’s commercial refrigeration system in three modes. In fully-powered, normal operation (mode 1), the technology will measure refrigeration temperature and case traffic in real time. With this data the software will generate operational commands that manipulate temperature signals to the refrigeration thermostat, in this way shifting compressor cycles to reduce energy consumption. The technology will also be equipped to pre-cool units when sent a command that harsh weather and potential power outages are predicted (mode 2). For these situations, the proposed solution will preemptively adjust the temperature of certain refrigeration cases to keep food below the industry-critical temperature of 41?F during impending outages. Finally, during unexpected outages (mode 3), the technology will supply the highest-value assets with backup power (e.g., from battery storage or an onsite generator). Given the typically limited capacity of backup power, the technology will enable food retailers to optimally use backup power and maintain temperature in essential chillers (e.g., meat and dairy cases) while letting less sensitive cases (e.g., soda coolers) heat up gradually. Phase I will design and do initial testing on a proof of concept with machine learning software and hardware that hosts the control software locally. The company will develop software that cleans and analyzes operational data from commercial refrigeration systems and then derives optimized temperature signals from this data. Finally, Phase I will produce hardware that implements these signals by updating temperature settings in a commercial refrigeration unit in real time. To alleviate barriers to adoption and manufacturing challenges, Phase II will involve developing a prototype with a form factor similar to existing commercial refrigeration thermometers. In Phase II, the technology will also be tested and updated to receive wireless commands from a central server, a necessary step for commercialization. At the conclusion of Phase II, the technology will be available to sell as a Software as a Service package to food retail chains.

Perishable food items consistently account for over 50% of U.S. supermarket sales, but the refrigeration infrastructure vital for these items faces significant resiliency and efficiency issues. Due to low profit margins, the food retail sector routinely relies on legacy refrigeration equipment without investing in new, expensive alternatives. As a result, most food stores lack sufficient backup power or energy storage to continue running their refrigeration systems when a power outage strikes. In addition to lack of resilience, food stores face high operational costs due to the low efficiency of their commercial refrigeration systems. These systems are the largest energy consumers in food retail stores, constituting up to 62% of their energy usage. Through this SBIR project, Grid Fruit is developing Retrofittable Efficiency Technology for Resilient Refrigeration, an AIdriven controls technology that provides food businesses with resilience and efficiency benefits to save each supermarket an average of $36,000/year. Grid Fruit’s technology uses machine learning to put unused data to work for food businesses and power grids. The controls software improves compressor efficiency and minimizes peak demand by controlling commercial refrigeration and HVAC, reducing store greenhouse gas emissions by up to 15% while protecting food quality for our retail customers. In Phase I, Grid Fruit developed a machinelearningbased prediction platform with case temperatures as inputs to an optimization control algorithm to provide resilience and reduce aggregate demand of multiple refrigeration units. Their results show Grid Fruit’s optimal control reduces aggregate demand by 30% when no power outages are present. In an outage scenario with backup power, they demonstrate reductions in total electricity consumption of 28% across a twohour outage. They also reduce aggregate peak demand by 64% compared to typical thermostat controls when unlimited backup power is present. In Phase II, Grid will focus on the physical demonstration of the control schemes developed in Phase I and provide data visualization of these schemes in an online dashboard. The project will develop predictive analytics from the Phase I data and validate these analytics in commercial refrigeration systems. To alleviate barriers to adoption and manufacturing challenges, the project will develop an IoT hardware prototype with a form factor similar to existing commercial refrigeration thermometers. The refrigeration demo will test the IoT hardware and confirm it receives wireless commands from a central server, a necessary step for commercialization. At the end of Phase II, Grid Fruit will be ready to commercialize the technology as a Software as a Service package they will sell to food retail chains. Grid Fruit’s technology will enable commercial building refrigeration systems to be retrofitted for better energy resilience and energy efficiency. Based on pilot data, Grid Fruit estimates the energy resilience and efficiency gains will save food retailers 15% in operational and maintenance costs, a domestic savings of $4.4 billion annually.