Artificial Intelligence is not just a subject of science fiction anymore. It is very much a reality today. Just as the internet disrupted the way we share, transact, communicate, and consume information, AI’s potential to transform businesses in terms of product design, sales, supply chain, and customer service is undeniable. In a short span of time, AI has managed to entwine itself in our lives in various forms such as Amazon’s virtual assistant Alexa or Tesla’s self-driving cars.
How AI can help
A traditional automated refrigeration control system uses a thermostat that receives temperature information from sensors located at various points inside the warehouse to monitor cooling. The cooling system automatically switches ‘on’ or ‘off’ based upon present conditions and pre-set temperature constraints. The amount of electricity needed to cool the space vary every day because of a range of factors such as weather patterns in the region, time of the day, activity level in the warehouse, thermal properties of both the enclosure and inventory, thermal characteristics of the incoming inventory, number of material handling equipment currently in use, etc. The unit cost of power also varies in a day as some utility providers use variable pricing to charge customers more during periods of high demand, and less during times of low demand. As a result, even though the length of the power consumption cycles is roughly equal for the most part, the associated energy costs vary drastically depending on what time of the day the power was consumed.
But what if we could take advantage of low off-peak pricing?
Similar to how batteries are charged during periods of power supply and discharged to power loads when there is a power outage, cold storage facilities can be over-chilled during periods of off-peak pricing typically in the night and allowed to slowly rise back to a normal cooled state during the day, reducing the need to consume power during peak hours. I have seen many households in rural India intuitively apply this concept to their domestic refrigeration needs at times of energy crisis to preserve their food. This method of over-chilling is also effective in offsetting the impact of variable pricing on industrial cooling costs. For example, let’s say a cold storage facility has an agreement with its customer to maintain inventory at -20°C. In standard automation, the freezer unit would switch on and off periodically to keep the goods at -20°C, regardless of the cost of power. Instead, with the help of AI, we could over-chill the contents to -25°C in the night when prices are low. When the peak-pricing time arrives the next day, a portion of the power consumption needs for that period can be reduced by allowing the warehouse to discharge that extra cooling till the temperature reaches close to -20°C. It is possible to build an AI model to arrive at an efficient operating plan for the refrigeration system that optimizes both cost and power consumption at any point in time.
What data is needed?
For the AI model to identify the most optimal refrigeration cycle, it requires a constant stream of key information. For one, the thermal interaction between various elements in the ecosystem should be studied to identify the rate at which the facility warms up or cools down at any given point during the day. This relationship can be modeled using data such as heat capacity of the materials used for construction of the warehouse, the thermal capacity of the objects in inventory, the type of packaging used, the weather condition in the area where the warehouse is located, nature of the incoming inventory, activity level in the warehouse (more activity usually means more heat from increased number of workers, forklifts and constant movement of personnel through the doors). Secondly, historical price charts or projected cost of power can be obtained from the utility provider to calculate operating costs based on anticipated power consumption. Thirdly, IoT sensors located in the various refrigeration equipment can provide real-time data on the performance of the cooling system, enabling the engineers to identify the most optimal working condition for these machines. Using these three inputs, the AI model can be designed to identify the optimal cooling strategy that avoids running the freezers during periods of peak pricing, at times when it inherently takes a long time to cool because of high atmospheric temperature or when the conditions are not ideal for optimal equipment performance. The new set of instructions is then relayed from the AI module to the control module that manages the overall refrigeration system.
What the future state could be
In the future, there is an opportunity to enhance this automation further by integrating the control module, the AI module and a new blockchain module to provide a seamless flow of information such as temperature, packing materials and content of the incoming inventory across these systems. This integration addresses another pain point of traditional cold chains - the lack of accountability. As the cold inventory changes multiple hands, there is no clear transfer of responsibility. With no mechanism of consensus at the time of hand-offs, it is very difficult to identify under whose custody temperature deviation/spoilage occurred. Using Blockchain, the quality control measurements can be collected using IoT devices and recorded immutably on a decentralized ledger shared across all supply chain partners. This not only increases accessibility and visibility to real-time information but the various supply chain actors can transfer ownership of the consignment through a clear chain of custody. When a warehouse receives a consignment, its blockchain record can be verified for temperature anomaly and the quality control team can take action accordingly. But as soon as they accept it, they sign their responsibility, making it possible to know “who is responsible for what” throughout the cold chain and enable quick claim settlements.
Artificial Intelligence along with digital innovations such as IoT, Cloud, and Blockchain has the potential to transform the way cold chains operated traditionally. It allows for a greater degree of efficiency by enabling managers to take decisions proactively. As the cost of prediction continues to drop, Artificial Intelligence is soon becoming a source of competitive advantage that is increasingly shaping business strategies of large organizations such as Amazon. As companies articulate their AI strategy, it is equally important to focus on their data strategy, AI skill development and change management for a successful AI implementation.