Use of IoT for Refrigeration Equipment in a Selected Supermarket Chain

This thesis is part of a solution to an optimization problem for a Czech supermarket chain that uses IoT devices to monitor in-store activity. The business problem is quite extensive, therefore it is divided between two authors. In the first part of the solution, author Gorylová created models predicting energy consumption and also identified key factors impacting the energy consumption of a supermarket. The second part of the solution which is provided by this author focuses on the optimization problem. Based on SARIMAX models and Monte Carlo simulations, we estimate expected costs associated with the possibility of goods spoiling. With the use of the findings of author Gorylová, we predict expected energy consumption costs via XGBoost model. Total expected cost is the sum of the two aforementioned expected costs and the supermarket's optimal temperature setting is the one that minimizes total expected costs. In other words with the help of machine learning methods, we develop an algorithm that estimates the total expected costs of a supermarket associated with refrigeration and freezing based on specific input parameters. The parameters include for example the setting of the temperature of refrigeration and freezing, inside temperature, outside temperature, or the value of goods.