Non-invasive load Decomposition Method of 3DU-net Based on Multi-task and BiLSTM

Aiming at the problem that the current non-invasive load monitoring methods are not accurate enough to identify low-power appliances and multi-state appliances, a load recognition method based on multi-task DUnet model is proposed. In this method, the DUnet model is optimized and the task of decomposing the switching state of the appliance is added to assist the bus power decomposition. Secondly, the bidirectional long and short term neural network is added to the model to solve the problems of gradient disappearance, gradient explosion and long distance information dependence. Finally, the algorithm is trained and tested by using UK-DALE dataset. Through comparison experiments with different models, it is analyzed and verified that the method has a good decomposition effect.