Feature Extraction and Classification of Household Refrigerator Operation Cycles: A Study of Undisturbed Refrigerator Scenario

Household refrigerators are cyclic appliances whose operations comprise of disturbed and undisturbed cycles. While disturbed cycles are due to the interference of human activities, undisturbed cycles often occur late at nighttime and until early morning when there is less or no interference from building occupants. Better understanding of the operation cycles of refrigerators will not only help in improving the energy use of the appliance, but also develop the necessary constructs that will enable efficient operation via Smart Grid technologies. This paper uses an unsupervised learning, the k-means clustering method, to classify household refrigerator operation cycles for undisturbed periods. A typical household refrigerator was monitored for 11 weeks. Two parameters namely, maximum electricity consumption density and total electricity use for each refrigerator operation cycle were extracted from the data collected and were used for the classification of operation cycles. In order to analyze the actual working status (i.e., without human interference) of the refrigerator, data obtained between 1:00 a.m. to 6:59 a.m. were used for this study. Results show that the undisturbed refrigerator operation cycles can be distinctly classified into three groups namely, “general working cycle,” “auto-defrost working cycle,” and “auto-defrost affected working cycle.”