Dual-stream BiLSTM framework with histogram-based shape features for household load forecasting

Authors

• CHANG XUFollow
• WONG JEE KEEN RAYMONDFollow
• HAZLEE AZIL ILLIASFollow
• HAZLIE MOKHLISFollow

Author ORCID Identifier

CHANG XU: 0000-0001-5093-3448

WONG JEE KEEN RAYMOND: 0000-0002-6656-826X

HAZLEE ILLIAS: 0000-0002-5061-1809

HAZLIE MOKHLIS: 0000-0002-1166-1934

Abstract

This study proposes a dual-stream BiLSTM framework for household load forecasting that integrates time-series dynamics with histogram-based daily shape features. Unlike existing models relying on weather or external data, the proposed method extracts intrinsic load-shape information directly from normalized daily curves. A multihead attention module fuses temporal and shape representations, enabling adaptive weighting of informative dimensions. Experiments on three real-world datasets show consistent improvements over the baseline BiLSTM, with up to 30.12%, 24.27%, and 19.03% reductions in MAE, RMSE, and SMAPE, respectively. The results highlight the framework’s robustness and efficiency for fine-grained load forecasting without external inputs.

DOI

10.55730/1300-0632.4193

Keywords

Attention mechanism, BiLSTM, load forecasting, pattern recognition, smart grid

First Page

604

Last Page

623

Publisher

The Scientific and Technological Research Council of Türkiye (TÜBİTAK)

Creative Commons License

This work is licensed under a Creative Commons Attribution 4.0 International License.

Recommended Citation

XU, C, RAYMOND, W, ILLIAS, H. A, & MOKHLIS, H (2026). Dual-stream BiLSTM framework with histogram-based shape features for household load forecasting. Turkish Journal of Electrical Engineering and Computer Sciences 34 (4): 604-623. https://doi.org/10.55730/1300-0632.4193