Sub-Nyquist Harmonic Current Component Extraction using Band Pass Filters for NILM

Abstract As environmental issues emerge, interest in monitoring power consumption is increasing, and Non-Intrusive Load Monitoring (NILM) has gained attention as a method for this. Recent advancements in artificial intelligence (AI) have improved the performance of NILM systems in terms of accuracy and inference time. However, due to insufficient research on data acquisition methods, the practicality of implementing these systems remains limited. This study focuses on the development of an IoT-based edge device acquiring current harmonic component with sub-Nyquist sampling using a Band Pass Filter (BPF) approach. The primary objective is to extract high-frequency harmonic components from AC power signals at sampling rates lower than traditionally required rate, addressed by the Nyquist-Shannon sampling theorem. To validate the effectiveness of the collected current harmonic components, an SVM algorithm was embedded into the device to perform a simple classification task. The device achieved 100% accuracy with an inference time of 584 ms. This innovative approach enables simultaneous multi-channel monitoring using a single low-performance NILM device. This contributes to reducing the costs associated with energy monitoring while enhancing the scalability of energy management points.