New journal article - Learning from other cities: Transfer learning based multimodal residential energy prediction for cities with limited existing data

A groundbreaking study by lead author Yulan Sheng, Research Fellow in Urban Heat Systems has unveiled a powerful new method for predicting residential energy consumption, combining multimodal neural networks with transfer learning to overcome data limitations. By integrating both tabular and visual data, the model significantly outperformed traditional approaches in case studies across Barnsley, Doncaster, and Merthyr Tydfil.

The research showed a dramatic improvement in prediction accuracy, with the model reducing errors by up to 63.6%.

To ensure transparency, the team employed explainable AI techniques, confirming that features like floor and wall insulation were key drivers of energy use. The study offers vital insights for policymakers, paving the way for smarter, data-informed energy efficiency strategies across the UK’s diverse housing stock.

Energy & Buildings, Volume 338, July 2025

Learning from other cities: Transfer learning based multimodal residential energy prediction for cities with limited existing data