Non-Intrusive Load Monitoring for High Power Consuming Appliances using Neural Networks

The topic of Energy Conservation requires urgent attention worldwide to avoid the impending energy crisis and reduce the impact on the environment through emissions. A crucial step in energy conservation is to motivate individual consumers to reduce their consumption. Itemized energy consumption feedback on each appliance helps users to plan their consumption patterns in an optimum way. Non-intrusive load monitoring is a low-cost and low-maintenance method for identifying consumptions of individual devices from the aggregate data of the mains supply. However, high power-consuming devices with power patterns with varying states are generally difficult to identify, despite them making a huge impact on the overall consumption of a household. Research shows that machine learning techniques are a promising approach for this disaggregation process. This paper focuses on developing data preprocessing methods and neural network algorithms to accurately disaggregate four common household appliances including ones with multistate power patterns.