MCIR-CNN: Multichannel Imaging-Based Nonintrusive Load Identification With Renyi Entropy Window for Improved Sensor Data Processing

Nonintrusive load identification is an important aspect of load monitoring and can be accomplished using sensor data. In this article, we propose multichannel imaging-based methods with REW (MCIR)-convolutional neural network (CNN), a novel load identification method that leverages multichannel imaging of sensor time-series data and CNNs. We introduce a new method based on Renyi entropy for determining the sliding window size when segmenting sensor time series into subsequences, which we refer to as the Renyi entropy window (REW). Our method relies on sensor data to accurately classify loads, providing a nonintrusive solution for energy monitoring. We evaluate our method on publicly available datasets and show that MCIR-CNN outperforms the current state-of-the-art (SOTA) method, demonstrating the potential for sensors to improve household energy efficiency (EE) and reduce energy waste through load monitoring. By leveraging sensor data, our method has significant implications for the development of advanced sensor technologies in energy management.