Appliance classification using energy disaggregation in smart homes

In this work we have addressed the problem of appliance classification and power consumption anomaly detection using energy disaggregation and machine learning techniques. The active power consumption data, received from a smart-meter, has been used as the only parameter for solving our problem. We have implemented a decision tree algorithm to classify appliances based on thresholds of their power consumption. Additionally, we have also proposed and implemented an algorithm for unusual fluctuation detection based on average magnitude of such fluctuations and an appliance quality recommender based on power-factor of the appliance. Initial results are promising as the classifier works correctly for 74% of instances, while the anomaly detector works correctly for 80% anomalies.