Two Similarity Measure Approaches to Whole Building Fault Diagnosis

Abstract

This paper introduces two methods based on cosine similarity and Euclidean distance similarity respectively to diagnose possible causes of abnormal whole building energy consumption. The concepts of cosine similarity and Euclidean distance similarity are defined and the methodology for implementing the proposed whole building fault diagnosis approaches is presented. Cosine similarity and Euclidean distance similarity are applied to two field observed fault test cases, and both the cosine similarity and Euclidean distance similarity methods indicated that the most probable fault was the fault observed in the field survey.