Non-intrusive Load Monitoring Using Empirical Mode Decomposition and AutoML

Load monitoring plays a pivotal role in intelligent power consumption. This paper proposes a non-intrusive load monitoring (NILM) approach that utilizes automatic machine learning (AutoML) and empirical mode decomposition (EMD). To begin with, the high-frequency power signal is decomposed using EMD technology. Subsequently, AutoML is employed for load disaggregation, thereby alleviating the burdens associated with model creation and hyperparameter tuning. Experimental evaluation conducted on the UK-DALE real-world dataset demonstrates the efficacy of the proposed method, particularly in terms of achieving high disaggregation accuracy for five specific appliances, namely, fridge, washing machine, dishwasher, microwave, and kettle. Additionally, the utilization of EMD enhances the performance of load disaggregation as compared to the corresponding baseline method.