Purechain-Enabled Asymmetric Temporal Knowledge Transfer for Scalable Energy Forecasting
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
Energy consumption patterns display temporal asymmetry: peak hours (6-8 AM, 6-10 PM) feature dense, overlapping appliance signatures, whereas off-peak periods are cleaner but less diverse. Conventional Non-Intrusive Load Monitoring (NILM) methods treat these periods uniformly, leading to poor cross-temporal generalization and accuracy losses of up to 65% off-peak. We propose Asymmetric Temporal Knowledge Transfer (ATKT), a semi-supervised framework that leverages peak-to-off-peak knowledge transfer. ATKT combines multi-head attention for complex peak-hour disaggregation with LSTM-based modeling for off-peak sequences, unified under an asymmetric loss function that prioritizes peak periods. To ensure secure and distributed deployment, the framework integrates PureChain blockchain with a Proof-of-Authority-and-Association ($\text{PoA}^{2}$) consensus. Experiments on Energy Forecasting and Load Forecasting datasets show significant gains, with$\mathrm{R}^{2}$scores of 0.9324-0.9647 for load data and 0.997 for energy data, outperforming standard LSTM baselines. The blockchain layer delivers low latency (0.0317-0.0340s) and stable throughput (29.6-32.4 TPS), ensuring scalability. Overall, ATKT provides a comprehensive solution for intelligent energy management, enhancing grid stability and facilitating the integration of renewable energy sources.
Fields of Study & Subjects
Smart Grid Energy Management
Energy Load and Power Forecasting
Blockchain Technology Applications and Security
Publication details
- Journal / source
- International Conference on Artificial Intelligence in Information and Communication (ICAIIC)
- Pages
- 785–789
- Publication date
- 24 February 2026
- Language
- EN