Targeted Adaptive Non-Intrusive Load Monitoring

NILM (Non-Intrusive Load Monitoring) technology is a cost-effective load monitoring technique that infers the operating status and working conditions of each appliance in a system by disaggregating the system's overall electricity load recorded by the electric meter. Recently, deep learning has significantly enhanced the accuracy and computational efficiency of NILM techniques. However, the generalization ability of most deep learning algorithms is limited due to the data scarcity of the training dataset. To enhance the generalization ability of current deep-learning-based NILM, we propose a new method named TAN-NILM (Targeted Adaptive Network for Non-Intrusive Load Monitoring), which contains a main disaggregation network and an auxiliary network and is trained with the multi-task learning strategy. The experimental results demonstrate that TAN-NILM exhibits superior disaggregation accuracy and generalization ability compared to state-of-the-art NILM algorithms.