Pico-Grid Smart Home Energy Management System Using Mel Frequency Cepstral Coefficients

Residences consume a significant amount of electrical power. Unfortunately, some of it goes to waste as homeowners often unintentionally consume excessive amounts of power. Home energy management systems aid homeowners in their power consumption decisions by measuring their power consumption and presenting that information in a user-friendly format. In this paper, we propose the enhancement of the Pico-Grid Smart Home Energy Management System by using Mel Frequency Cepstral Coefficients. It is a consumer-grade energy management system capable of identifying loads by analyzing current draw and is distinguished from other home energy management systems by its incorporation of a novel hardware design created to ensure reliable sampling and a classifier consisting of several artificial neural networks (ANN). Each neural network corresponds to a classified device which allows the system to be trained to recognize additional devices without having to re-train the entire existing network. The classifier converts the current signal to the Mel-spaced Cepstrum domain in order to extract relevant features achieving an overall classification accuracy of 97.8% with average positive and negative predictive values of 93.2% and 98.7% respectively. This enables the system to potentially have a lower cost but higher accuracy than other similar designs that have been suggested thus far. This system aids homeowners in making energy-conscious decisions such as monitoring the energy costs of their home appliances and scheduling heavy loads during periods of low demand.