Non-Invasive Load Monitoring Method Based on Compressed Sampling Signal

To solve the problem of massive data and high-frequency sampling in Non-Invasive Load Monitoring (NILM), this paper introduces Compression Sensing (CS) in NILM. Compared to the way with load identification based on reconstructed signal, the load identification of NILM based on compressed sampling signals in CS is more practical. Therefore, this paper focuses on the preliminary exploration of NILM based on compressed sampled signal (NILM-CSS). Firstly, the basic model of NILM-CSS was established and the specific process of load identification in the model was proposed. Then, for the two key elements of the model, a binary sparse matrix and back propagation artificial neural network were selected as the measurement matrix and machine learning algorithm respectively. The case analysis shows that the proposed NILM-CSS in this paper has similar load identification accuracy to the traditional NILM when NILM was first proposed, which also indicates the feasibility and further research value of NILM -CSS. Meanwhile, the key elements selected in this NILM-CSS are reasonable.