Introduction
The transition towards Net Zero requires power systems that can accommodate vast amounts of distributed renewable energy, while maintaining security, reliability, and cost-effectiveness. In this respect, Artificial Intelligence (AI) offers significant potential to optimise power system operations
However, unlike traditional power system operation which rely on centralized control, future energy systems will increasingly be operated by crowds of decentralised AI agents, each controlled by different stakeholders, that can be potentially vulnerable to cyber-attacks.
The core research question is: how can we guarantee safety in power systems with decentralised AI agents operating across market and physical layers, ensuring resilience against malicious interference? While individually “safe” AI agents may comply with safeguards, their collective behaviour can still lead to unsafe system-wide outcomes, as highlighted in control theory and multi-agent systems research. This problem is largely unaddressed in the current literature, where most safety frameworks assume centralised oversight.
This PhD project proposes to develop and validate an integrated workflow for detecting and mitigating unsafe emergent behaviours within crowds of AI agents in power system operations.
The research will draw on methods from optimisation, control theory, game theory, and cybersecurity to ensure resilience not only under normal operating conditions but also in the presence of adversarial threats. These approaches will be tested in scenarios including electricity markets, distributed resource coordination, and renewable energy integration. By systematically benchmarking these methods, the project aims to deliver the first scalable approach to guaranteed safe and resilient decentralised AI for power systems.
Join us to make AI-driven power systems safer by reducing the multi-billion-pound costs of balancing the UK grid, enabling higher renewable generation, and lowering the risk of large-scale blackouts. With this PhD project you will work on advancing the field of safeguarded AI and tackle one of its most critical application areas: the energy transition.