Non Intrusive Load Monitoring Based on Squeezenet-CBAM Model

Abstract

In today's non-invasive load monitoring research, household appliance recognition is an important sub research direction of non-invasive load monitoring. The appropriate feature representation method and appropriate model building method are still an unresolved problem. Recently, a load recognition method based on Adaptive Weighted Recursive Graph (AWRG) and Convolutional Neural Network (CNN) has been proposed, which generates an image like representation for a given period of current. However, the model is still not satisfactory due to resource constraints, resulting in poor performance. Therefore, we innovatively propose a SqueezeNet- CBAM model to address resource constraints. Experimental results on the PLAID dataset show that our model performs better under three different evaluation metrics.