Demand Response Based on Anomaly Detection and Non-Intrusive Load Monitoring

Non-Intrusive Load Monitoring (NILM) can give a breakdown of energy consumption per appliance downstream of the energy meter. It is planned in India that the Time of Use tariff be implemented in the residential sector with tariff during solar hours being less by 10-20% and increased by 10-20% during peak periods. To help the user make informed decisions about his energy usage, appropriate alerts and recommendations should be provided especially in case of peak load conditions in the grid and the ability to control or schedule loads. NILM can be used to disaggregate loads which can then be used to check anomalous behaviour of loads. An experimental setup was developed to enable NILM and scheduling of appliances and anomaly detection. The challenge in detecting anomalous energy consumption is choosing the right feature for detection and the pass/fail criteria. Anomaly detection using cumulative sum (CUSUM) of energy consumption and a model of average consumption is used to detect unusually high energy consumption with z score as the margin criteria.