The development of algorithms that operate with minimal human intervention within the energy system will be an increasingly common occurrence and arguably a prerequisite of any future energy system, particularly as flexibility markets develop.

The Energy Digitalisation Taskforce recommended that the energy sector proactively manages new digital risks, such as the increased prevalence of algorithmic decision making. 

Large numbers of behind the meter devices will be incorporated into smart systems, utilising automation, machine learning and other techniques from the data produced by these devices to optimise for consumers, networks, and whole system outcomes.

The complexity of consumer choice alone means that automated decision-making based on user inputs, forecasts and models will be necessary for the operation of the energy system. This creates ample opportunities to create a more engaged, dynamic energy system which accounts for the considerable variety in energy use patterns through the transition to a Net Zero economy. This could manifest in more bespoke consumer offerings and incentives, or in new types of market for organisations seeking to deploy energy generation or storage at scale.  

While downside risks do exist for the automation of processes that algorithms enable, effective use of algorithms can enable several positive outcomes. Benefits include:  

• Enabling high frequency decisions to be made without manual intervention to assist balancing the energy system
• Allowing for consumer choice and confidence that their preferences will be catered for
• Creating new markets for energy products and services
• Network managers can transition towards more anticipatory outcomes, rather than reactive. 

However, the deployment of algorithms, without oversight at a sectoral level could cause issues that will need to be understood such as: 

• Cascade impacts across the energy system derived from the interactions between algorithms
• Bias or discrimination against individuals or groups
• Distortion or manipulation of markets.  

What is Algorithm Governance?

The utilisation of these algorithms does necessitate monitoring by a variety of market participants, principally the regulator and system operator(s), to help map and manage the interdependencies between the various algorithms in use by the sector and reduce systemic risks from developing undetected.

In the finance sector, the Office of the Comptroller of the Currency (OCC) provides guidance on a similar approach for effective management of risks that arise when using quantitative models which are used in bank decision making. Guidance from OCC highlights the need for proper governance of banking models, with residual risks from the models to be managed through other processes. By taking a similar, albeit whole system approach to monitoring algorithms, the increasingly decentralised energy system would be able to assess the risks associated with algorithms on their own services and systems. The regulator, and system operator(s) by comparison, could take a market wide view and intervene as appropriate.

Key points

To meet the challenges of a growing number of automated decisions being made about the energy system, and to reduce the risk of cascade impacts, this paper proposes the following solutions should be adopted to ensure all relevant organisations can access information to do their own risk analysis and mitigation:

1. Algorithm Metadata Standard
2. Register of Algorithms.  

Combined, these two recommendations would put in place the foundational information required, as well as a means to access it – to enable stronger governance of algorithms to be built over time. 

In registering and cataloguing algorithms, the regulator and system operator(s) gain a mechanism through which they can ascertain information about the technical operation of the sector, conduct risk analyses and develop robust governance to mitigate any issues.