Analysis of energy disaggregation techniques in non-intrusive appliance load monitoring

Carbon dioxide emission reduction goals have intensified interest in researching new methods to improve our efficient use of electricity. It has been proven that providing consumers with appliance usage patterns can have significant energy savings. Non-intrusive appliance load monitoring (NIALM) research aims to facilitate the large scale installation of mechanisms that provide such usage information. NIALM is the process of using the whole home electricity signal to determine the energy consumption information of appliances in the home without direct measurement. In this paper, we propose a fast and efficient non-parametric technique for disaggregating the whole home energy signal to determine individual appliance power consumption with high precision. We evaluate our proposed technique with the REDD dataset and show that it performs better than existing approaches in practice. We also propose modifications to known sparse coding techniques for energy disaggregation. Lastly, we evaluate the feasibility of employing Gaussian Process Regression for the purpose of NIALM.