PhD Studentship: Uncertainty-Aware Synchronisation for Sustainable Data Centre Digital Twins

system states. This growing uncertainty can further exacerbate the synchronisation challenge by making it difficult to determine when the Digital Twin has deviated sufficiently from the physical system to warrant an update.

Existing approaches based on fixed thresholds, periodic updates, or heuristic policies can lead either to excessive synchronisation and unnecessary resource consumption or to delayed updates that allow the Digital Twin to drift from the operational system.  Sustainable datacentres are one particular area where self-adaptive Digital Twins are increasingly being adopted to model and optimise renewable energy utilisation. In such environments, continuous synchronisation between the physical entities and their virtual counterparts remains a key challenge due to the involvement of multiple prediction modules, including renewable energy availability, datacentre and workload dynamics.  

Each of these modules introduces a degree of prediction uncertainty, where forecast errors can propagate through the Digital Twin, potentially increasing the discrepancy between the predicted and actual states of the physical system. Consequently, determining when synchronisation is necessary becomes particularly challenging, as the decision must account not only for the current state discrepancy but also for the uncertainty associated with future system states. This raises a central research question for this PhD project: How can a self-adaptive sustainable datacentre digital twin determine when synchronisation between an operational system and its Digital Twin is necessary, while accounting for physical drift, uncertainty, and stakeholder-defined operational and performance tolerances? This makes synchronisation fundamentally a decision-making problem rather than a simple update mechanism. Instead of asking whether the current drift exceeds a predefined threshold, the Digital Twin must decide whether the available evidence is sufficient to trigger synchronisation, whether the divergence remains acceptable, or whether additional information should be gathered before making a decision.     

Research Objectives  

1. Develop an uncertainty-aware synchronisation framework for Digital Twins: Design a self-adaptive framework that models physical drift, forecasting uncertainty, and stakeholder-defined operational tolerances to enable intelligent synchronisation decisions between physical and virtual systems.  
2. Develop and evaluate AI-based synchronisation strategies for sustainable data centres: Investigate AI techniques, such as reinforcement learning, probabilistic machine learning, and Bayesian inference, to learn adaptive synchronisation policies that improve Digital Twin fidelity while reducing unnecessary updates and supporting renewable energy optimisation.  

Expected Contributions  

• A novel self-adaptive decision-making framework for uncertainty-aware synchronisation in self-adaptive Digital Twins.  

• Adaptive AI-based synchronisation algorithms that combine predictive uncertainty with real-time operational context to support intelligent update decisions.  

• A multi-objective synchronisation strategy that balances Digital Twin fidelity, computational and communication overhead, renewable energy utilisation, and stakeholder performance requirements.  

• A comprehensive experimental evaluation and reproducible benchmark demonstrating the effectiveness of uncertainty-aware synchronisation for sustainable data centre Digital Twins. 

Funded by: EPSRC Engineering and Physical Sciences Research Council 

£21,805. The studentship will provide funding of fees and a stipend of £21,805 per annum for 2026/7.