Precise Temperature Control for Sensitive Medicines

As we have all experienced during the COVID pandemic, the storage and transportation of vaccines require a certain temperature control. The tighter the temperature control, the more predictable the time-wise viability of the vaccine can be guaranteed. This paper aims to provide an Adaptive PID (Proportional-Integral-Derivative) control based on reinforcement learning and supported by user usage habits that gives more precise temperature control for sensitive medicines, vaccines and samples. Data collected from IoT fridges will aid acquiring usage habits and ambient conditions from users and devices deployed in the field. Experimental results show that 45%-75% decrease in cabinet temperature oscillation with same energy level for energy tests, 4%-12% decrease in energy consumption, and 10%-24% decrease in cabinet temperature oscillation in user tests when compared with classical cooling method can be obtained.