Event Matching Classification Method for Non-Intrusive Load Monitoring

Nowadays, energy management aims to propose different strategies utilize available resources, resulting in sustainability of systems and development smart sustainable cities. As an effective approach toward management, non-intrusive load monitoring (NILM), infer the power profiles appliances from aggregated signal via purely analytical methods. Existing NILM methods are susceptible various issues such as noise transient spikes signal, overshoots at mode transition times, close consumption values by appliances, unavailability a large training dataset. This paper proposes novel event-based classification algorithm mitigating these issues. The proposed (i) filters signals accurately detects all events; (ii) extracts specific features operation modes their respective intervals, dataset; (iii) labels with high accuracy each detected event appliance transition. is validated using REDD results showing its effectiveness disaggregate low-frequency measured data existing meters.