Non-Intrusive Load Monitoring (NILM) using a LSTM with socio-economic parameters

Abstract

Nonintrusive load monitoring (NILM) deconstructs aggregated electrical us age data into individual appliances. The dissemination of disaggregated data to customers raises consumer awareness and encourages them to save power, lowering 𝐶𝑂2 emissions to the environment. The performance of NILM sys tems has increased dramatically thanks to recent disaggregation methods. However, the capacity of these algorithms to generalize to various dwellings as well as the disaggregation of multi-state appliances remain significant ob stacles. In this paper we propose an energy disaggregation approach by using socio-economic parameters with the aggregated data. The suggested approach helps in creating more accurate load profiles, which improves the accuracy and helps in better detection of the appliances. The proposed model outperforms state-of-the-art NILM techniques on the PRECON dataset. The mean absolute error reduces by percentage 5%- 10% on average across all ap pliances compared to the state-of-the-art. Thus, improving the detection of target appliance in the aggregate measurement.