Energy consumption and food safety issues of household refrigerators have attracted considerable attention in recent years. Today, most household refrigerators use conventional controllers for adjusting the compressor speed to regulate the air temperatures inside cabinets. This research aims to design an intelligent control concept that integrates machine learning-based forecast of door opening events with fuzzy logic controllers. Firstly, bayesian neural network, logistic regression, and decision tree techniques are investigated to predict user behavior with data obtained from 18 real users of domestic refrigerators. Results show that logistic regression has the best performance in hourly predicting the door opening events after one week of training with more than 80% accuracy. Secondly, fuzzy logic controllers are designed to use the door opening predictions to regulate configuration parameters of the main refrigerator controller: maximum compressor speed, air temperature setpoint of fresh food compartment, and time offset to control the time of defrosting events. Finally, simulation studies are performed on a domestic refrigerator model developed at MATLAB Simscape. The developed model is based on an existing product in the market, and the accuracy of the model is verified by actual lab tests. Daily simulations of the household refrigerator for sample day profiles of an inactive and active user are performed for ambient temperatures of 16 °C, 25 °C, and 32 °C. Results show that the designed smart controller can achieve up to 2.5% and 4.5% of energy gain for active and passive user-profiles respectively, while maintaining the desired cabinet temperatures.
Bibliographical notePublisher Copyright:
© 2022 Elsevier Ltd and IIR