Cooling Load Forecasting for Multi-user Scenarios: A Unified iTransformer Framework with Personalized Knowledge Fine-Tuning

Mingxuan Zou

doi:10.1016/j.enbuild.2026.118066

Cooling Load Forecasting for Multi-user Scenarios: A Unified iTransformer Framework with Personalized Knowledge Fine-Tuning

2026-08-06

0

PRE

AI

M

Mingxuan Zou

W

Wenjun Huang *

J

Jianxiang Jin

Y

Yaochun Hou

Y

Ying Wang

DOI：10.1016/j.enbuild.2026.118066

Original

Original request for help

Save

Abstract

En 中文

• Introduced a novel multi-user cooling load forecasting framework balancing general and personalized knowledge. • Developed multimodal user profiles integrating sector semantics and statistical features for precise user characterization. • Employed parameter-efficient fine-tuning to personalize modeling across diverse sectors, significantly improving forecasting accuracy. • Validated the superiority of UiT-PKFT over state-of-the-art baseline models in large-scale real-world tasks. • Achieved Pareto optimality between forecasting accuracy and computational overhead for 1,000 users.

Keywords:

Cooling load forecasting

Multi-user heterogeneity

Multimodal feature processing

iTransformer model

Parameter-efficient fine-tuning

Cooling Load Forecasting for Multi-user Scenarios: A Unified iTransformer Framework with Personalized Knowledge Fine-Tuning

Abstract

Journal

Organization

Cited Papers

Citing Papers