A Wasserstein Metric Based Distributionally Robust Synthesis Sparse Coding for NILM

Non-Intrusive Load Monitoring (NILM) is a pivotal technology in energy management and conservation. It involves disaggregating the total electrical consumption of a household into individual appliances without the need for dedicated sensors on each appliance. Sparse coding, a powerful technique in machine learning and signal processing, has gained attention in recent years for its potential to enhance the accuracy and robustness of NILM algorithms. This paper constructs an ambiguity set by using Wasserstein distance, and furtherly proposes a distributionally robust Synthesis Sparse Coding for the NILM problem. An approximation, which is a semi-infinite optimization problem, of the proposed model is given and solved by using constraints generating method. The dataset UK-DALE is used in our experiments to verify the effectiveness and robustness of our model. The simulation results show that the proposed model is effective and suitable for applying in NILM.