We are delighted to announce that Keen AI, in collaboration with National Grid Electricity Transmission, has been awarded funding by Ofgem’s Strategic Innovation Fund (SIF) to develop a Foundational Shared Model (FoSMo) for asset management across the UK’s electricity network.
Ofgem has highlighted the ambition of this project, noting that:
FoSMo is “considered innovative, novel, and risky. It is the first attempt to develop a shared ML model across all UK networks, introducing technical, regulatory, and commercial challenges. The risks include the complexity of data integration, the need for cross-industry collaboration, and the development of governance frameworks for shared model use.”
This recognition underscores the pioneering nature of FoSMo and its potential to transform how electricity networks deploy AI for asset management.
Tackling Inefficiency in Computer Vision for Energy Networks
Today, network operators across the UK are independently developing computer vision models to detect the same types of defects and assets across similar infrastructure. This siloed approach results in:
- Duplicated development effort across different organisations
- Inconsistent performance due to fragmented datasets
- Increased costs that ultimately impact bill payers
FoSMo directly addresses this inefficiency by creating a shared, industry-wide foundational computer vision model, continuously updated with shared datasets.
The model will be permissively licensed and openly available, allowing operators to focus on value-adding activities rather than duplicating effort.
What is FoSM0?
FoSMo (Foundational Shared Model) is designed to act as a base model for computer vision tasks across electricity transmission and distribution networks. Instead of each operator starting from scratch, FoSMo provides:
- A pre-trained, high-quality vision model trained on a large and diverse dataset of network assets and defects
- Transfer learning capabilities, enabling operators to fine-tune the model for specific use cases (e.g. insulator defects, vegetation management, corrosion monitoring)
- Improved accuracy and robustness through pooled data contributions across the sector
- Permissive licensing, ensuring accessibility without costly licensing fees
By adopting FoSMo, the UK electricity industry can benefit by:
- Reducing duplicated effort across organisations.
- Accelerating AI deployment, cutting time-to-value for inspection and monitoring solutions
- Lowering operational costs for both operators and consumers
- Enhancing the resilience and accuracy of AI models through wider data sharing and collaboration
- Supporting innovation, enabling SMEs and new entrants to build solutions on top of a common foundation
- Enabling more efficient asset management, delivering tangible benefits to consumers through fewer service interruptions and improved network reliability
FoSMo also introduces a new collaborative framework for Transmission Operators (TOs) and Distribution Network Operators (DNOs), allowing industry-wide optimisation that currently does not exist.
Jon Hunte, Innovation Engineer at National Grid Electricity Transmission, one of our key collaborators on FoSMo, remarks that:
“FoSMo has the potential to address a fundamental gap among the networks – employing a unified ML model for asset condition assessment, which allows the model to learn from the collective datasets across networks. This cross-collaborative approach can lead to a higher quality model and potentially lead to reduced costs for networks and consumers.”
What next?
Over the coming months, Keen AI will begin work on developing FoSMo, including:
- Data integration and curation from across networks.
- Model architecture design and initial training.
- Establishing governance frameworks for shared model use.
Keen AI are actively seeking engagement with transmission and distribution operators who are interested in shaping this foundational capability.
If your team wants to shape and benefit from the UK’s first shared vision model, we would be delighted to discuss potential collaboration!
