Skip to main content

Autonomous Anomaly Detection for MPC Forecasts of HVAC Systems in Residential Communities...

by Viswadeep Lebakula, Evgeniya Tsybina, Justin Hill, Jeffrey D Munk, Helia Zandi
Publication Type
Conference Paper
Book Title
2024 IEEE Power & Energy Society Innovative Smart Grid Technologies Conference (ISGT)
Publication Date
Page Numbers
1 to 5
Publisher Location
New Jersey, United States of America
Conference Name
The 2024 Conference on Innovative Smart Grid Technologies, North America (ISGT NA 2024)
Conference Location
Washington DC, Washington, United States of America
Conference Sponsor
Conference Date

The use of residential heating, ventilation, and air conditioning (HVAC) to shift peak demand or provide ancillary services is a potential solution in the presence of older grids and distributed renewables. However, to ensure the efficient use of devices, utilities need to accurately forecast the load and adopt error correction schemes when necessary. While significant theoretical research exists in the area of predictive control of HVAC, little experimental evidence exists. The lack of experimental data in turn causes researchers to be unprepared for unsystematic errors which emerge due to the higher complexity of the data generating process. This study offers an anomaly detection methodology that uses unsupervised machine learning algorithms to detect and isolate these errors with different forecast error ranges. The results of anomaly detection procedure can then be used for error correction and would eventually help develop better predictive controllers. The methodology is tested using real world data from a smart neighborhood that currently operates in Atlanta. GA.