Intelligent Control of Refrigerators and Freezers

Abstract

As the number of consumers has been growing and global warming continues to affect the environment, the refrigeration sector is becoming an increasingly important part of the global economy. Refrigeration products have recently received significant attention primarily due to their energy consumption and food safety concerns. Today, conventional controllers are used for adjusting the compressor speed to regulate the temperature in most refrigerators and freezers. This thesis aims to develop a fuzzy logic control concept that incorporates machine learning-based forecasts of the door opening events using the MATLAB 2020b Toolboxes. The performance of the smart controller was compared to conventional controllers in terms of keeping the desired minimum cabinet temperatures at optimum energy efficiency. In the first step, Bayesian neural networks, logistic regression, and decision tree machine learning techniques are investigated to predict user behavior based on the data collected from 18 households that use domestic refrigerators. The simulation study has demonstrated that logistic regression is the best method for predicting door opening events on an hourly basis after only one week of training, with more than 80% prediction accuracy. As a second step, fuzzy logic controllers are designed to regulate the configuration parameters of the main refrigerator controller based on the predicted door opening. These configuration parameters are: the maximum speed of the compressor, the air temperature setpoint, and defrost interval of the fresh food compartment for a refrigerator. In the third step, the performance of the smart controller is compared with the traditional controllers with simulation studies on a domestic refrigerator model developed in MATLAB Simscape. The model utilized for simulations has been developed based on an existing refrigerator currently in use and validated with actual laboratory tests. For sample day profiles of an inactive and active user, daily simulations of a refrigerator were performed for ambient temperatures of 16°C, 25°C, and 32°C. According to simulation results, the intelligent controller can maintain the required cabinet temperatures while achieving 2.5% and 4.5% energy gains for active user and passive user profiles, respectively. Lastly, the designed controller was integrated into a commercial freezer. The selected freezer was tested with a fixed and variable speed compressor to investigate the impact of the intelligent controller on different compressor types. Similar to the methodology applied in the simulation studies, the chosen commercial freezer was tested at 25°C ambient temperature based on the door opening profiles of active and passive users. The test results have proven that an intelligent controller can reduce the maximum package temperature by approximately 0.5-1°C and energy consumption by 5-7% compared to standard controllers for freezers with fixed and variable speed compressors.