Smart meters and related technologies are becoming a noteworthy aspect of residential electrical power systems as proper energy feedback can result in significant energy savings. Smart meters also facilitate implementation of “time of use” (TOU) pricing for the residential sector. Results of Advanced Meter Infrastructure (AMI) pilot studies reveal that it is possible to save 12-15 % of energy saving and maximum demand shifting to low peak time when appliance specific disintegrated real-time energy feedback is provided to consumers or controllers. With current technology, appliance level detection requires costly hardware installation and is quite impossible to deploy in large scale AMIs. It would be much more cost effective to predict appliance level consumption using smart meters already installed in premises. However, smart meter has only one current transducer per phase, rendering appliance usage detection challenging. This paper proposes an algorithm which can be performed on the data collected by the smart meter and a real-world example of its application.