NILM in high frequency domain: A critical review on recent trends and practical challenges
Petros G. Papageorgiou, Georgios C. Christoforidis, Aggelos S. Bouhouras
Abstract
The benefits of monitoring aggregated electric load consumption and distinguishing it into its components are numerous and are mainly related to more efficient household energy management. Non-intrusive load monitoring (NILM) as a key method to accomplish this task has attracted the interest of the research community. The basis of this method is the acquisition process of the aggregated electrical signal, which can be performed at different sampling frequencies depending on the specifications of the disaggregation model. Nevertheless, the use of higher sampling frequencies can improve the performance of the disaggregation process. The development of different disaggregation models should be based on their applicability under practical conditions. This means that during their development, these models should include all parameters that will be met in real application. In this way, they will have a higher chance of effectively applying the load disaggregation in practice. Although there are several published review papers about NILM, there is a lack of a systematic review focusing on both the high-frequency domain and the practical aspects that need to be considered during model development. To this end, this study systematically reviews 40 recent papers in the high-frequency domain published between 2019 and 2022. Then, the practical issues derived from them are discussed and referred to the different steps of NILM development. Finally, 30 of them are evaluated in terms of their disaggregation performance and whether their development is focused on practical applications, according to several criteria. • Recent high sampling rate NILM developments are examined. • The practical challenges regarding the real-world application of NILM are presented. • A critical comparison of recent high sampling rate NILM frameworks is presented. • Directions regarding the effective real-world application of NILM are given.