Working Status Mining Enhanced Sequence-to-sequence Network for Non-intrusive Load Monitoring on Industrial Power Data

Non-intrusive appliance load monitoring (NILM) is the process of decomposing the total energy consumption of the combined electric system into its contributing appliances. In this study, we proposed an approach by which the working status mining and sequence translation model were applied to NILM. First, we determined the working power of each appliance in different working statuses, and the combined encoding was generated to represent the working status of all electrical appliances. Subsequently, the total energy consumption signal and combined status code were trained on the sequence-to- sequence model, which considered the time correlation during operation. The constructed model integrated the time scale information and signal amplitude characteristics of electrical working status, and translated the energy consumption into a status code for load decomposition. Finally, we evaluated our model using the electrical consumption data of the gas station (the overall accuracy of load decomposition reached 87.6%), where we successfully monitored the power consumption of 12 appliances. Moreover, our solution provided a new perspective for the application of NILM in the field of industry and commerce and has the potential to be an effective energy management tool.