Deep Neural Network Based Energy Disaggregation

In smart electricity grids, energy disaggregation significantly contributes to better demand side management, load forecasting and energy savings via estimating appliance level energy consumption from the aggregated smart meter data. This paper proposes a deep neural network based system by combining convolutional neural networks and variational auto-encoders for energy disaggregation. Domestic Appliance-Level Electricity dataset (UK-DALE) is used along with the standard error measures such as Mean Absolute Error (MAE) and Signal Aggregate Error (SAE) in order to evaluate the proposed system performance. Test results show that the proposed system improves the state-of-the-art performance by 44% and 19% based on SAE and MAE respectively.