Deep Neural Network Based Non-Intrusive Load Status Recognition

In this paper, we have targeted two appliances, microwave and fridge, from a single house and have built a neural network model to recognize their load status i.e, either ON or OFF. The data-set was obtained online from UK-DALE which contains appliance meter readings and site-meter readings of a few houses at a 6 second sampling rate. To train our network model, we use Intrusive Load Monitoring (ILM) by using the load status (0 or 1), recognized from the appliance meter readings, as the output and the site-meter readings as the input. The main objective of the paper is to successfully detect the change in the status of the appliance and the duration of this change, which denotes its ON/OFF status. We have used a novel approach of feature detection for a change in status using convolutional neural networks (CNN) followed by the use of Long-Short Term Memory (LSTM) recurrent neural networks (RNN). The results show the F1 score and the accuracy score of the implemented neural network on the test data and the actual vs predicted load status graph over a period of time for both the appliances is plotted.