Development of a fuzzy sales forecasting system for vending machines

Vending machines operate 24 h a day, allowing consumers to obtain products anytime of the day and night for added convenience. However, they consume large amounts of electricity, equaling the total power output of two nuclear reactors in Japan. As the global environment deteriorates, energy conservation is becoming increasingly important. There is therefore an urgent need to reduce the power consumption of vending machines. To respond to this problem, we conducted research to develop a system that would reduce the energy used for cooling by canned beverage vending machines. In our research, we forecasted the number of cans dispensed daily so that electricity would be used to cool only the required number of cans. We also used fuzzy logic and a multiple regressive model to correct the prior forecast value on the day of forecast for improved forecast accuracy. We conducted simulation experiments using this method and confirmed that the cooling energy could be reduced to approximately 1/10.