Smart homes are part of a fundamental smart grid concept for smart cities. Home energy management system (HEMS) became an essential tool for smart homes to monitor electricity consumption and prevent energy wastages. In this paper, a Power Alert System (PAS) is proposed for HEMS to indicate the power consumption level of smart homes. In the proposed system, K-means algorithm is applied to classify power consumption into three levels, namely high, average and low. To evaluate the developed system, real-time unsupervised power consumption data has been collected from a Singapore smart home. In this practical residential environment, the developed system has performed a reliable clustering and created energy awareness for the residents. This classification of the energy consumption was compared with a traditional classification.