What Happens When You Ask an LLM to Read 20+ Years of AI Research?

From early studies on distributed renewable control in the early 2000s, through to today’s work on deep learning, digital twins and even large language models, the papers tell a much longer story about AI in energy than you might expect.

Just to give a flavour of that range:
• Early work like, involving Phil Taylor, explored intelligent, decentralised control in off grid renewable systems
• Later research such as Network State-Based Algorithm Selection for Power Flow Management Using Machine Learning showed how machine learning could support real time grid decisions
• More recent papers from Furong Li and Alessandra Parisio demonstrate how advanced AI techniques are embedded in forecasting and planning
• And newer contributions, co authored by Chenghong Gu, explore the role of large language models in energy systems


“While AI may feel like it appeared overnight, the research tells a 20 year story of steady evolution.”

________________________________________

One of the most striking things that comes through is just how long this community has been working in the space — long before it was commonly labelled as “AI”.

Back in the early 2000s, researchers such as Phil Taylor were already developing distributed control approaches for renewable energy systems. At the time, these were framed as engineering and control challenges — but many of the ideas map directly onto what we now describe as intelligent, autonomous systems.

As the field developed, the work expanded. Researchers like Jianzhong Wu developed data driven approaches to state estimation and system monitoring, improving visibility in complex networks. Furong Li led work on data driven load profiling and forecasting, helping to make sense of increasingly distributed demand.

Alongside this, contributions from Victor Levi advanced anomaly detection and robustness in system estimation helping ensure reliability as energy systems became more complex and data rich.


“AI in energy isn’t new — what’s new is how visible it has become.”

________________________________________

As the years progressed, the research didn’t suddenly pivot into AI — it evolved into it.

Machine learning became central to managing uncertainty. Work by Robin Preece explored neural network based wind forecasting, while Furong Li and collaborators advanced deep learning approaches for electricity price forecasting.

In parallel, Phil Taylor’s work on intelligent control systems demonstrated how optimisation and AI could support real time operational decision making in grids.

At the same time, the idea of distributed and collaborative systems continued to develop. Early decentralised control approaches evolved into today’s smart, flexible energy systems. Research into resilience also kept pace, with work exploring how systems respond to disruption, including cyber attacks Using Self-Organizing Architectures to Mitigate the Impacts of Denial-of-Service Attacks on Voltage Control Schemes.


“The tools have changed, but the challenge hasn’t: managing complex, uncertain, and interconnected energy systems.”

________________________________________

Looking at more recent work, the same themes continue — but with new tools and scale.

Deep learning models developed by Alessandra Parisio and collaborators now predict electricity, heat and gas demand across multi energy systems. Reinforcement learning approaches, explored by Chenghong Gu for example are enabling optimisation of community level energy systems and flexibility markets.

Advances in system visibility continue through work by Jianzhong Wu, including digital twin based battery state estimation, and by Victor Levi, whose research improves state estimation under uncertainty.

There is also a strong emphasis on real world application. David Greenwood’s work demonstrates how AI forecasting models can be deployed in operational environments, bridging research and implementation.

At a system scale, Sheridan Few’s research shows how AI enabled modelling can support planning for low carbon technologies across thousands of local electricity networks.

And now, the research is coming full circle — with work exploring how AI itself can help manage system complexity. Contributions such as, involving Chenghong Gu, examine how large language models might be integrated into smart grid systems.


“This wasn’t a step change into AI — it was a natural progression of work already underway.”

________________________________________

A pattern emerged, across the work, there is a clear progression: Distributed intelligence → machine learning and optimisation → modern AI systems. Or, more simply:

“We’ve been building ‘AI for energy’ long before it became a buzzword.”