An Enhanced Transformer Model with Multivariate Data Fusion for Steel User Adjustment Willingness Prediction and Optimization

Posted on 18 Dec 2025 — The copyright holder is the author/funder. All rights reserved. No reuse without permission. — https://doi.org/10.22541/au.176601796.63156965/v1 — This is a preprint and has not been peer-reviewed. Data may be preliminary. An Enhanced Transformer Model with Multivariate Data Fusion for Steel User Adjustment Willingness Prediction and Optimization Yumeng Zhang1, Bin Li 2, Chenle Yi 3, Ziqian Zhang 3, and Zixuan Meng 3 1North China Electric Power University - Baoding Campus 2North China Electric Power University School of Electrical and Electronic Engineering 3North China Electric Power University December 18, 2025 Abstract Industrial users, as major urban electricity consumers, play a pivotal role in ensuring power grid stability by adjusting their consumption patterns. However, conventional load forecasting methods often fall short in addressing the need for real-time flexibility and dynamic demand response. To bridge this gap, we propose a hybrid predictive framework that integrates Long Short-Term Memory (LSTM) networks with Transformer architectures. This model harnesses LSTM’s capability for capturing temporal dependencies alongside the Transformer’s strength in global attention and parallel computation, enabling precise identification of regulatory patterns in steel industry operations. Using real world operational data from a steel plant in Tianjin including equipment flexibility indices, maintenance scheduling, and historical response behavior the proposed model demonstrates a 30% reduction in prediction error compared to four benchmark algorithms. These results underscore the model’s effectiveness in accurately quantifying industrial users’ demand side adjustment potential, paving the way for more intelligent grid and user coordination and enhanced energy system resilience. Hosted file manuscript_word1.docx available at https://authorea.com/users/1011156/articles/1371264- an-enhanced-transformer-model-with-multivariate-data-fusion-for-steel-user-adjustment- willingness-prediction-and-optimization 1