Project II : Deep Learning for Household Energy Management
As we navigate the complexities of modern energy consumption, understanding how households utilize electricity is paramount. The household sector significantly contributes to overall energy demand, and optimizing its usage is crucial for both economic and environmental sustainability.
This project focuses on energy disaggregation, a technique that breaks down a household’s total electricity consumption into individual appliance-level usage. By doing so, we gain deeper insights into energy consumption patterns, which enables better management and efficiency improvements. Among various methods, sequence-to-point (seq2point) learning has emerged as a powerful framework for non-intrusive load monitoring (NILM).
Seq2point learning learns a mapping from the mains electricity signal to the appliance-level consumption, allowing us to derive a probabilistic distribution of appliance usage. This approach is flexible and can utilize architectures such as CNNs, RNNs, and more, making it adaptable for various applications.
