UK ENERGY LAW AND ELECTRICITY SYSTEM – FORECASTING AND ARTIFICIAL INTELLIGENCE REGULATION

Introduction

Forecasting and Artificial Intelligence regulation concerns the legal governance of AI, machine-learning and advanced analytical systems used to predict electricity demand, renewable generation, system congestion, wholesale prices, balancing requirements and network risks. In the United Kingdom, AI is increasingly relevant to electricity-system planning and real-time operation because a system containing large volumes of intermittent wind and solar generation requires increasingly sophisticated forecasting. Ofgem recognises that AI can improve planning, management and real-time operation of the energy system, while also creating risks requiring regulatory controls.

Legal and Regulatory Framework

The principal electricity legislation remains the Electricity Act 1989, supplemented by licence conditions, industry codes and Ofgem regulation. The Energy Act 2023 strengthens the institutional framework for a digitalised and decarbonised electricity system.

AI forecasting may additionally engage the UK GDPR and Data Protection Act 2018 where smart-meter, household, employee or other personal data is processed. Cybersecurity obligations may arise under the Network and Information Systems Regulations 2018, particularly where AI is connected to essential electricity infrastructure.

The UK’s general AI framework follows five principles: safety, security and robustness; appropriate transparency and explainability; fairness; accountability and governance; and contestability and redress. Rather than originally imposing a single comprehensive AI statute, the UK approach relies heavily on existing sector regulators applying these principles within their regulatory responsibilities.

Ofgem and AI Governance

Ofgem published dedicated energy-sector AI guidance in May 2025 and has subsequently developed an AI Regulatory Laboratory, allowing licensees and energy-sector participants to test proposed AI applications against regulatory requirements. Ofgem also decided to establish an AI technical sandbox pilot for controlled testing of AI technologies.

AI forecasting systems therefore require appropriate governance, testing, human oversight, auditability and risk management. An electricity company cannot simply rely upon an algorithm where inaccurate forecasting could adversely affect balancing, network security, customers or market outcomes.

Energy Data and Forecasting

Reliable AI forecasting depends upon high-quality energy data. Ofgem’s Data Best Practice Guidance, updated in November 2025, establishes expectations concerning effective management and sharing of energy-system data. Ofgem considers accessible and standardised data important for meeting demand, reducing costs and enabling decarbonisation.

This is particularly important where AI predicts distributed generation, electric-vehicle charging, heat-pump demand, flexibility services or network congestion. Poor-quality training data may produce inaccurate forecasts and potentially discriminatory or inefficient outcomes.

CASE LAW 1 – R (Eisai Ltd) v NICE [2008] EWCA Civ 438

Facts: NICE relied upon an economic computer model when deciding whether Alzheimer’s medicines were cost-effective. Eisai received only a restricted version and could not fully test assumptions and calculations.

Legal Issue: Whether refusing access to the fully executable model made the decision-making procedure unfair.

Judgment: The Court of Appeal held that adequate access to the model was necessary because the model played a central role in the regulatory decision.

Legal Principle/Ratio: Where computational modelling materially determines regulatory outcomes, procedural fairness can require sufficient transparency to permit meaningful scrutiny.

Significance: The principle is highly relevant to electricity forecasting. Where Ofgem, NESO or another public decision-maker materially relies on sophisticated forecasting algorithms, transparency and the ability to test important assumptions may become legally significant.

CASE LAW 2 – R (Bridges) v Chief Constable of South Wales Police [2020] EWCA Civ 1058

Facts: South Wales Police deployed automated facial-recognition technology that algorithmically processed biometric information.

Legal Issue: Whether automated technology was governed by sufficiently clear legal safeguards and complied with data-protection and equality obligations.

Judgment: The Court of Appeal found important aspects of the deployment unlawful, including excessive discretion within the governing framework and deficiencies relating to data-protection and equality duties.

Legal Principle/Ratio: Automated decision-making requires sufficiently clear rules, proportional safeguards, accountability and consideration of discriminatory impacts.

Significance: Energy AI used for forecasting, customer segmentation, flexibility markets or automated system decisions should similarly operate within defined governance structures rather than uncontrolled algorithmic discretion.

Conclusion

UK electricity forecasting is increasingly dependent on AI, but innovation operates within electricity licensing, data protection, cybersecurity, administrative-law and consumer-protection frameworks. Future regulation is likely to place increasing emphasis on algorithmic transparency, explainability, data quality, human oversight, model validation, cybersecurity and accountability. Effective regulation must allow AI forecasting to increase electricity-system efficiency while ensuring that critical energy decisions remain lawful, auditable, secure and capable of regulatory scrutiny.