Application of Self-supervised Learning in Non-intrusive Load Monitoring

Abstract With the proposal of smart grid, demand both source and load for fine monitoring control power is becoming increasingly prominent. Non-intrusive a technical means to better meet this demand. However, research at home abroad focuses on existing data sets labeled improve accuracy identification, while training method model under massive unlabeled in actual scene still relatively blank stage. Aiming problem how make full use training, non-intrusive-load based self-supervised learning proposed paper. This designs task, so that can eliminating step manually labelling data; Based encoder-decoder structure, deep established, identified through characteristic vector output by encoder, has generalization performance. In paper, AMPds2 set used verify method, test examples effectiveness method.