Identification of Photovoltaic and Electric Vehicle Profiles in Distribution Networks Using Long Short-Term Memory Network

The widespread implementation of global initiatives focused on achieving net-zero carbon emissions and the electrification of transportation has resulted in the extensive deployment of distributed energy resources (DERs) within the low-voltage distribution network. The rapid integration of DERs has introduced technical challenges, altering the electrical characteristics of conventional distribution networks. This challenge is exacerbated by the absence of monitoring infrastructure on the low-voltage side. Non-intrusive load monitoring (NILM) methods offers a chance to enhance the traditional electric measurements and boost the visibility of distribution network. The present work proposes a long short-term memory based NILM framework for the disaggregation of photovoltaic and electric vehicle profiles from the aggregated measurements in the distribution network. The comparative analysis has also been carried out with other machine learning classifiers Random Forest and k-Nearest Neighbors for the same dataset. The proposed approach has been rigorously validated for dataset with different input time frames to ensure robustness and reliability and found to achieve average F-scores in excess of 99.52% and 92.29% for identification of PV and EV profiles respectively.