Energy crisis and climate change have caused global concern and motivated e orts to take steps to reduce energy consumption. Studies have shown that providing appliance-level consumption information can help users conserve a signi cant amount of energy. Non-Intrusive Load Monitoring (NILM) is an attractive method for energy disaggregation, as it can enable disaggregation and classification of individual electronic device operation based on one aggregated measurement data from one electrical sensing point in the house or building. It has been considered as a potential means of providing energy management tools where energy consumption patterns, usage habits, and other energy relevant information are available to end users. The NILM system monitors the aggregated electric signal from the main circuit panel of the house (or building) and infers power ON/OFF, other operational states, and the corresponding power consumption of house appliances. In addition, NILM can potentially provide an early detection of device malfunction. With NILM technology, the end user can receive detailed energy usage reporting. Such information is critical for home energy management and other smart home management tools.
Although numerous studies have been devoted to developing e ective models for NILM from high-sample-rate data with higher costs, limited progress has been made in low frequency energy disaggregation by exploring discriminant features of low-sample-rate data for different appliances. Most of existing methods focus on learning signal signatures. However, unlike high-sample-rate data, which is generally discriminant for different appliances, many devices share similar signatures due to similar behaviors in low frequency mode. There is a lack of research work considering inherent characteristics of the data. Due to these special characteristics, how to maximally exploit them to improve energy disaggregation is the key challenge in NILM. The goal of this project is to develop ML (Machine Learning) based algorithm solution(s) to disaggregate several major appliances operations (HVAC, refrigerator, dryer, dishwasher, cooktop, cloth dryers, microwave, water heater, lights) and properly detect/classify each device, and estimate its power consumption with the performance confidence from the low-sample-rate power measurements. In order to enhance the overall algorithm performance, other contextual information such as user behavior and environmental information will be considered as complements. In addition, non-real time/post processing methodology after the completion of the device operation can be considered. Overall, this success of this project will significantly improve the state-of-the-art in Energy Disaggregation and NILM, and inevitably broaden this research area by opening up and addressing many new research themes.
PI: Junzhou Huang (Ph. D.)
