Robust adaptive event detection in non-intrusive load monitoring for energy aware smart facilities

This paper presents a robust adaptive goodness-of-fit (GOF) χ² test event detector for non-intrusive load monitoring applications. We derive a closed form for the decision threshold and a guideline for choosing the size of the detection data window. Using real-world power data collected in residential buildings, the proposed GOF detector shows superior performance compared to the conventional generalized likelihood ratio detector.