A fusion framework using integrated neural network model for non-intrusive load monitoring

Smart grid has been developed greatly over the years and the users are paying more attention to detailed energy consumption information. Non-Intrusive Load Monitoring (NILM) technique enables the users to get the power of each appliance by analyzing aggregated power. In this paper, we propose a fusion framework and use an integrated neural network model to estimate the power of each appliance. First, we detect the working state change events in aggregated power by using cumulative sum method. To correlate each event with the corresponding appliance, we then calculate the possibilities of working state for each appliance based on different event features. Then we use the fusion framework to decide which appliance has caused this event. Next, we generate a reference power curve for each appliance and refine this curve by an integrated neural network model using aggregated power. The experimental results show that the proposed method can achieve good performance on a variety of indicators even on low sampling-rate data.