Citation

Bushra, B. N. and Rajesh, K. and You, Ah Heng and Wong, Wai Kit and Ng, Poh Kiat (2026) Dual-Horizon Forecasting of Residential Load and Solar Generation for Smart Residential Energy Management using a Hybrid TCN-Transformer Model. In: 2026 8th International Conference on Inventive Material Science and Applications (ICIMA), 13-15 May 2026, Namakkal, India.

Text 4.pdf - Published Version Restricted to Repository staff only Download (1MB)

Abstract

Advanced residential energy management platforms facilitate the real-time supervision and control of domestic energy consumption and on-site renewable resources, enabling efficient operation, lower energy expenditure, and improved interaction with the utility grid. Accurate short-term and day-ahead forecasting of load demand and renewable generation is crucial for real-time control, demand response scheduling, and energy trading in residential environments. However, traditional statistical and machine learning (ML) approaches often exhibit limited ability in modelling the nonlinear, non-stationary, and highly stochastic characteristics of residential energy consumption and weather-driven PV output, resulting in degraded precision under dynamic operating conditions. This study offers a dual-horizon deep learning (DL) forecasting framework based on the integration of a Temporal Convolutional Network (TCN) and a Transformer encoder to jointly predict residential load demand and photovoltaic (PV) generation at (t + 15 min) and (t + 1 day) horizons using a multivariate home energy management system (HEMS) dataset. The architecture uses a shared latent feature representation with parallel decoder branches to capture both long-term dependencies and short-term variations. It achieves strong performance, with RMSE of 0.1184 and 0.0548 for short-term and day-ahead load forecasting, and 0.0876 and 0.0894 for PV forecasting, along with R2 scores of 0.8893 and 0.9312. The approach supports reliable integration into advanced residential energy management systems, enhancing efficiency, resilience, and sustainability.

Item Type:	Conference or Workshop Item (Paper)
Uncontrolled Keywords:	Smart residential energy management systems, Dual-horizon forecasting, Temporal convolutional network, Transformer encoder, Deep learning
Subjects:	T Technology > TK Electrical engineering. Electronics Nuclear engineering > TK9001-9401 Nuclear engineering. Atomic power
Divisions:	Faculty of Engineering and Technology (FET)
Depositing User:	Ms Suzilawati Abu Samah
Date Deposited:	04 Aug 2026 04:30
Last Modified:	04 Aug 2026 04:30
URII:	http://shdl.mmu.edu.my/id/eprint/16486

Downloads

Edit (login required)