This paper presents an Ensemble Extreme Learning Machine (ELM) approach for predicting railway delays, part of an ongoing PhD project with Network Rail and Loughborough University. With only 68.3% of station stops on time in early 2024, improving prediction accuracy is vital. The Ensemble ELM model, combining multiple ELMs, addresses key challenges in Train Delay Prediction (TDP) such as data quality and model generalization. Initial results show significant improvements in accuracy and efficiency over traditional methods. The model will be validated using UK railway data, contributing to more reliable and scalable delay prediction systems.