From the perspective of deep learning and edge computing, electric power IoT compatible non-intrusive load monitoring (NILM) system is studied in this paper. Firstly, in view of the NILM system deployment solution in IoT scenarios, a novel edge computing framework is proposed, of which the component roles are discussed. Furthermore, aiming at online extraction of load activations, a dispersion evaluation and load behavior regularity analysis-based activation detection strategy is proposed; for load feature of low-frequency sample, a CNN architecture that can automatically extract activation signatures and recognize the load type is proposed. Moreover, by the analysis of activation background power, power value fluctuation, etc., three generic features are defined as a supplement to the CNN extracted features. Finally, a verification experiment is carried out on a domestic dataset, of which the result proves the improvement of the proposed algorithm on generalization and computing efficiency.
