Privacy Constrained Load Profiling using Smart Meter Data

Smart meters have transformed energy management by enabling precise measurements, real-time data collection, and enhanced communication capabilities. The availability of detailed smart meter data presents significant opportunities for load profiling, a vital component of electricity consumption analysis. However, privacy concerns arise due to the sensitive nature of individual energy consumption patterns. Moreover, challenges related to bottleneck, data heterogeneity, and communication overhead occur because of the need to aggregate and analyze data from diverse sources, transmit large amounts of data efficiently, and preserve privacy by minimizing data transmission and ensuring secure processing. To address these challenges, this research paper proposes an unsupervised federated learning approach that incorporates collaborative model training and autoencoder. The proposed approach aims to overcome the privacy concerns by encoding data locally and reducing communication overhead through the transmission of compressed representations. Furthermore, it addresses challenges related to bottleneck and data heterogeneity by leveraging collaborative model training with Irish (C.E.R) dataset which is particitioned into weekdays and weekend. Numerical attributes are used for evaluation, comparing the proposed approach with model averaging and gradient sharing methods. The results demonstrate the effectiveness of the proposed approach in extracting meaningful features and patterns from unlabeled smart meter data while ensuring privacy, optimizing communication efficiency, and improving load profiling accuracy.