Sliding Window Approach for Online Energy Disaggregation Using Artificial Neural Networks

Energy disaggregation is the process of extracting the power consumptions of multiple appliances from the total consumption signal of a building. Artificial Neural Networks (ANN) have been very popular for this task in the last decade. In this paper we propose two recurrent network architectures that use sliding window for real-time energy disaggregation. We compare this approach to existing techniques using six metrics and find that it scores better for multi-state devices. Finally, we compare ANNs that use Gated Recurrent Unit neurons against those using Long Short-Term Memory neurons and find that they perform equally.