Electrical Appliance Identification through Signal Processing of Electrical Wave Signal: A CNN-Based Approach

The escalating global energy demand, driven by population growth and technological advancements, underscores the need for effective energy management strategies, particularly in residential settings where electrical appliances dominate power consumption. Current methods, relying on kWh meters, often fail to provide nuanced insights into individual appliance usage, prompting the exploration of innovative approaches for appliance identification. This research investigates the application of Convolutional Neural Networks (CNNs) for household appliance identification based on electrical signatures. The study begins with a comprehensive literature review, exploring appliance recognition algorithms, convolutional neural networks, device identification methods, visual signatures, spectrogram analysis, and current sensors. The methodology chapter outlines a systematic approach, including the conceptual framework, system process flow, experimental setup, program development, data collection, CNN model training, and performance evaluation. Spectrograms are generated from wave signals of the five appliances namely: electric kettle, Air-Conditioning Unit, Rice Cooker, Electric Fan, and Light. CNN program is utilized to train and test appliances based on spectrogram. The final testing phase assesses the program's ability to accurately identify appliances using appliances that were not on the initial list. The result shows that the three-appliance identification was more accurate than the four and five appliance identification, with a model loss of 0.000481 and 1.0, a model loss of 0.1422 and an accuracy of 0.9297, and a model loss of 1.4150 and 0.5409 respectively.