Development of an Edge AI Based Embedded System for Appliance Level Energy Monitoring and management for Smart City- Homes

With the high increase in the development of the latest tech in modern world, the use of electronic devices and
smart appliances has gone way up in the daily life. Consequently, the patterns of energy consumption continuously
change, with highly dynamic behaviour over time. Precise monitoring of the real-time loads variation is critical
in the grid management process and for the improvement of energy efficiency. Energy disaggregation, which uses
the total aggregated load data to estimate the power consumption of individual appliances, is a highly promising
and economical method to monitor electricity usage, and in real time. It offers useful information to consumers,
utility providers, researchers and policy makers by enabling informed decision making and efficient grid
operations implementation strategies. Non- Intrusive Load Monitoring (NILM) is a data-driven method to
ascertain the power consumption of individual appliances, based on measurements taken from a single point of
measurement (usually a main energy meter). This approach eliminates the need for multiple sensors on each
appliance and thus makes it cost-effective and appropriate for smart homes. This thesis is on the design and
implementation of an efficient NILM framework based on energy disaggregation methods for residential smart
home applications. The proposed research is divided into four major phases. The first phase is a detailed review
and comparative analysis of current NILM techniques used with a variety of load characteristics and a focus on
their applicability to residential energy monitoring. The development of such techniques will allow for the proper
disaggregation of individual appliance loads from aggregated consumption data, which will increase the
effectiveness of NILM in the energy disaggregation process. In the second phase, different energy disaggregation
algorithms which are suitable for smart home environments are analyzed and the most suitable method for
residential load monitoring applications is determined.