As demonstrated in the ADViCE project, artificial intelligence (AI) will be a key tool in enabling and managing the transition to Net Zero. With increased monitoring of controllable technologies, AI will be able to optimise the deployment of smart technologies, facilitate new products and services, and help to plan and operate the networks. Barriers to entry are being lowered and the technology is more readily available to an increasing array of practitioners, accelerating its adoption into the energy system and creating new opportunities for decarbonisation and energy efficiency. 

However, with automation and AI comes new risks to consumers and the energy system. Automated systems will likely interact in new ways that are detrimental to the whole network. Algorithms applied in different parts of the energy network may have conflicting objectives that lead to undesirable oscillations, creating suboptimal behaviour and increasing costs for consumers. More concerningly, errors in one algorithm could trigger a reinforcing feedback loop that could lead to a wide scale system failure. 

As part of this Innovate UK funded project, this report presents our investigations from stakeholder interviews and desktop research into some of the potential risks and impacts on energy networks, including possible ways to mitigate and manage these risks. The report identifies eight core areas where there could be significant challenges and emerging risks in adopting AI: 

1. AI strategy and visibility 
2. Collaborative AI approaches 
3. Organisational culture 
4. National security and AI assurance 
5. Data and modelling 
6. Skills for AI and related disciplines 
7. Consumer protection 
8. Regulation and governance 

Our research and analysis have outlined the scale of the challenge in designing and delivering an AI-enabled future energy sector, as we move towards a more decentralised and automated energy network. The challenges identified in our report represent an opportunity for the sector to take a proactive and collaborative approach to AI risk management, creating a regulatory market that nurtures the innovations that will support rapid decarbonisation, efficiency improvements and new market entrants. 

The following recommendations and proposals for future areas of work, expanded upon in the report, will require engagement and validation from the wider stakeholder community to ensure buy-in.  

Recommendation 1: Initiate a cross-sector AI special interest group. Understanding the risks and best ways to mitigate them will require cross-industry collaboration. The sector could harness the strengths of its own safety culture to better understand what AI is being applied and share learnings of risks and near-misses. Skills should also be a focus, as the increasing penetration of AI will necessitate upskilling in several areas, including risk management, and AI tool assessments.  

Recommendation 2: Develop a sandbox testing environment. Appropriate whole-system risk management strategies will require realistic simulations and testing of operational implementations of AI. This will only be possible through the development of a clear AI roadmap and strategy for identifying future uses cases, emerging system-level risks and innovation funding requirements. Further, the sharing of data will be essential to ensure any outcomes from the sandbox testing is replicable, trustworthy, and scalable.  

Recommendation 3: Create best practice AI risk guidance. Regulation and governance will play a key role in setting best practice AI risk guidance for the energy sector. It will be important to understand the regulatory gaps and priority areas, and the appropriate levels of oversight and enforcement for AI applications and risks.