This is a project which is currently making use of HPC facilities at Newcastle University. It is active.
For further information about this project, please contact:
This project develops and evaluates deep learning models for energy forecasting and performance modelling of air source heat pumps (ASHPs) in domestic smart buildings. Using a multi-system monitoring dataset of several hundred UK heat pump installations, the work trains and benchmarks sequence models (LSTM and related architectures) alongside gradient-boosted baselines to estimate electrical consumption from concurrent heat demand and environmental conditions. The research emphasises transferability across heterogeneous buildings and post-hoc explainability of model predictions. HPC resources support GPU-accelerated model training, hyperparameter search, cross-validated evaluation, and explainability analysis.
GPU-accelerated deep learning for time-series energy modelling of air source heat pumps. Models are developed in Python using PyTorch (LSTM/sequence architectures) and XGBoost, with scikit-learn for preprocessing and cross-validation, and pandas/NumPy for data handling. Post-hoc explainability is applied via SHAP-based attribution methods (e.g. TimeSHAP/WindowSHAP) to interpret model predictions. Workloads cover model training, hyperparameter search, batch inference, and explainability analysis over multi-system smart-building datasets, run on GPU nodes for the deep learning and attribution stages and CPU nodes for data preparation.