Condition Monitoring of Pole Transformers (PMT Scan)

Funding mechanism	Network Innovation Allowance (NIA)
Duration	Jul 2026 - Mar 2027
Estimated expenditure	£234,234
Research area	Data and digitalisation. Optimised assets and practices. Supporting customers in vulnerable situations.

The PMT Scan project aims to develop and test a scalable approach for assessing the external condition of pole‑mounted transformers using high‑resolution aerial imagery and automated visual analysis. By combining routine helicopter and drone image capture with computer‑vision techniques, the project will identify and classify visible condition indicators such as corrosion, oil leakage, structural condition, and bushing health, and convert these into standardised, auditable asset condition data aligned with Common Network Asset Indices Methodology (CNAIM). The project will also compare the cost, efficiency, and suitability of helicopter versus drone‑based capture for this use case. The outcome will be evidence on how pole‑mounted transformer conditions can be monitored more consistently and efficiently, supporting earlier intervention, improved network resilience, and readiness for ED3 regulatory reporting.

Problem(s)

Pole-mounted transformers are inspected through existing foot patrol and aerial activities, but the outputs are not structured in a way that provides a reliable, asset-level view of condition at scale. The current process is largely manual, time-consuming, and dependent on subjective judgement. Assessments can vary between inspectors, records can be incomplete, and condition information has not always been linked directly to the transformer asset itself. As a result, it is difficult to track deterioration trends, compare assets consistently, identify which transformers are at greatest risk, and make confident decisions on prioritisation, replacement, and longer-term investment.

Aerial inspection using helicopters can provide broader coverage but does not always capture the parts of the transformer that matter most, such as the underside, early-stage oil leakage, or defects that are obscured by viewing angle, mounting position, or surrounding infrastructure. Ground-based inspection faces similar limitations from the topside, where achieving consistent condition grading across a large and geographically spread fleet is difficult in practice. Together, these constraints mean that important deterioration can go unidentified, under-recorded, or inconsistently classified.

The problem is more pronounced in coastal areas, where corrosion can progress more quickly and assets may deteriorate earlier than expected. In these locations, limited condition visibility increases the risk of missing developing deterioration before it leads to failure, oil leakage, environmental contamination, or safety concerns.

This is now a growing regulatory concern too. PMTs are becoming more important within CNAIM and Network Asset Risk Metric (NARM)-related processes, and condition data needs to be consistent, traceable, and auditable to support those requirements. Where it is not, there is a real risk of applying the wrong condition assumptions to assets, allocating incorrect risk points, and producing Health Index outputs that cannot be properly defended. That creates compliance exposure and weakens the quality of risk-based decision-making at a time when regulatory expectations around structured PMT condition evidence are increasing ahead of ED3.

Method(s)

The PMT Scan project will develop and test an AI-enabled aerial inspection and condition assessment process for pole-mounted transformers, covering the full pathway from image capture through to structured condition outputs that can support asset health and risk processes.

Stage one

The first stage will establish the image capture standard needed for reliable assessment and collect a representative dataset of PMT images across coastal, inland, and rural environments. This will cover image resolution, viewing angles, coverage, lighting, and occlusion tolerances, with quality assurance checks applied to confirm that images are suitable for analysis and correctly linked to asset records. The project will use both helicopter and drone-based capture methods and compare their relative image quality, practicality, time efficiency, and cost to inform the most suitable approach for future use.

Stage two

The second stage will develop the condition assessment approach using computer vision models trained on the labelled image dataset. The models will be trained to identify the transformer, separate it from the background, and detect visible condition indicators including corrosion severity, oil leakage or staining, tank body defects, bushing defects where visible, and structural deterioration of mounting components. Outputs will be produced as condition classes and confidence scores, with the process also testing whether parts of the assessment can be automated or semi-automated to reduce manual effort at scale.

Stage three

The third stage will validate the method against manual inspections and engineering judgement on a selected cohort of assets across different environments, including coastal locations where corrosion risk is higher. This will be used to test performance, refine classification thresholds, and improve accuracy, particularly for higher-risk defects such as severe corrosion and oil leakage.

Stage four

The fourth stage will translate the condition outputs into CNAIM-aligned condition categories and test how they can be integrated into NGED's Health Index and risk processes. This will include developing mapping logic, maintaining full traceability from the original image through to the mapped condition output, and demonstrating an auditable pathway from captured image to structured condition data. The project will also test how outputs could flow into existing data architecture, including NGED's internal asset data and NARMs reporting systems, to confirm alignment with current reporting pathways.

Stage five

The final stage will assess how the method could be scaled across the wider PMT fleet. This will cover rollout planning, operational requirements, cost-benefit considerations, and the findings from the capture method comparison, to define the most practical and cost-effective approach for future deployment.

Measurement Quality Statement

Measurement quality will be controlled through defined image capture standards, quality assurance checks, and structured validation of model outputs. The project will confirm whether images meet the required quality threshold for analysis and will measure model performance using recognised metrics such as precision, recall, and F1 score, supported by comparison with engineering judgement and inspection evidence.

Data Quality Statement

Data quality will be managed through asset-linked image records, structured labelling, and validation checks during dataset creation, model training, and output generation. The project will apply clear metadata standards, consistent labelling rules, and traceable condition outputs so that results can be linked to the correct transformer asset and used reliably in later analysis and reporting.

---

• Scope

  The PMT Scan project will develop and test a practical method for assessing the external condition of pole-mounted transformers using aerial imagery and structured condition analysis. The scope covers visible external deterioration that affects asset health and risk, including corrosion, oil leakage or staining, and damage to components such as tanks, radiators, brackets, and mounting arrangements.

  The project will include defining image capture requirements, creating and labelling a representative image dataset, developing and validating a condition assessment process, and converting outputs into structured condition information that can support asset health and risk processes. It will also assess how condition information can be linked more effectively to the transformer asset itself and used within existing asset management systems and reporting frameworks. The project will test whether parts of the assessment process can be automated or semi-automated, which could reduce manual effort and make large-scale condition assessment more practical going forward.

  The work will be carried out through a pilot across a representative sample of transformers in different operating environments, including higher-exposure locations where deterioration may progress more quickly. The scope also includes comparing helicopter and drone-based capture methods to understand their relative image quality, practicality, time efficiency, and cost.

  This phase is intended to prove whether the method is workable, accurate enough, and practical for wider use. It is not intended to deliver full fleet rollout, replace existing inspection activity, or carry out network-wide replacement works.

  Net Benefits for Consumers

  Pole-mounted transformers are widely used in rural and village networks, and failures in these areas can be more disruptive and harder to recover from quickly, particularly for customers who are more vulnerable to the effects of a loss of supply. The project will help NGED identify deteriorating transformers earlier and act before issues develop into more serious failures, reducing the risk of unplanned interruptions in these communities.

  There are also safety and environmental benefits. Earlier identification of deteriorated assets will help reduce the risk of failures that could lead to falling equipment or oil-related environmental incidents. This is especially relevant in coastal and higher-exposure areas where environmental conditions may accelerate deterioration and increase the likelihood of asset failure.

  Consumers also benefit when investment decisions are based on better evidence. More consistent condition data will help direct resources toward the transformers that present the greatest risk, improving the efficiency of maintenance and replacement planning and helping to avoid poorly targeted interventions.

  Direct Financial Benefits to National Grid 

  The most direct financial benefit is a reduction in the manual effort currently needed to assess PMT condition at scale. A more structured image-based approach will lower the time and cost involved in turning inspection activity into usable condition data, and if parts of the process can be automated or semi-automated, that saving could be more significant across a large fleet.

  Better condition visibility will also support earlier and more targeted intervention, reducing exposure to the higher costs that come with reactive maintenance, emergency replacement, and unplanned response. Identifying deterioration earlier, particularly in coastal areas where corrosion progresses quickly, will help convert reactive failures into planned works, which are considerably cheaper and less disruptive to manage.

  Improved condition data will also strengthen the quality of asset health and risk inputs used for investment planning. More accurate and consistent condition scoring will support better prioritisation, reduce the risk of misinformed investment decisions, and improve the defensibility of Health Index and NARM outputs ahead of ED3. That has financial value in its own right, as poorly evidenced condition data can lead to misallocated spend and increased regulatory scrutiny.

  Finally, testing the method at pilot scale before any wider commitment reduces the financial risk of investing in an approach that may not perform as expected. The project is designed to generate the evidence needed to make that decision with confidence.

• Objectives

  The PMT Scan project aims to prove whether aerial imagery and structured condition analysis can provide a more reliable, consistent, and scalable way of managing pole-mounted transformer condition. To do that, the project will:

  • Develop a practical method for assessing the external condition of pole-mounted transformers from aerial imagery, covering visible deterioration such as corrosion, oil leakage, tank body defects, bushing defects where visible, and structural deterioration of mounting components.
  • Define the image capture requirements needed for reliable condition assessment, including resolution, viewing angles, coverage, and quality standards suitable for use across coastal, inland, and rural environments.
  • Create a labelled PMT image dataset and condition grading approach that can support repeatable and consistent defect classification across a representative range of asset conditions.
  • Test whether parts of the condition assessment process can be automated or semi-automated, reducing manual effort and making large-scale assessment more practical and consistent.
  • Validate the method against manual inspections and engineering judgement, including assets in coastal environments where corrosion risk is higher, to confirm whether outputs are accurate and reliable enough for operational use.
  • Develop the logic needed to translate condition outputs into CNAIM-aligned condition categories and test how those outputs can be integrated into NGED's Health Index and NARM-related processes ahead of ED3.
  • Compare helicopter and drone-based capture methods to understand their relative image quality, practicality, time efficiency, and cost for future deployment.
  • Define what would be needed to scale the approach across the wider PMT fleet, including operational, data, and integration requirements.
  • Generate practical learning that supports future rollout within NGED and provides transferable insight for other network licensees facing similar challenges.
• Success Criteria

  NGED will consider PMT Scan successful if it demonstrates that a structured image-based approach can produce PMT condition information that is accurate, consistent, and practical enough to support future asset management use.

  Success will be evaluated against the following: 

  • A representative pilot dataset is created: A set of PMT images is captured across coastal, inland, and rural environments with sufficient quality, coverage, and asset linkage to support model training, testing, and validation. The coastal cohort should include locations where corrosion risk is known to be higher, such as Cornwall.
  • Image quality standards are met: Captured images meet the agreed requirements for resolution, viewing angle, and coverage, with a high proportion passing quality assurance checks and confirmed as suitable for condition assessment.
  • The condition assessment method produces credible results: The method identifies and classifies visible condition indicators, including corrosion, oil leakage, tank body defects, bushing defects where visible, and structural deterioration of mounting components, with performance levels that are considered reliable for asset management use. Performance will be measured using precision, recall, and F1 score, together with review of false positives and false negatives.
  • Outputs are validated against engineering judgement and inspection evidence: Results are shown to be broadly consistent with subject matter expert review and ground-truth inspection evidence across both coastal and inland environments, demonstrating that the approach is repeatable and suitable for practical use.
  • The potential for automation or semi-automation is assessed: The project provides a clear view of which parts of the condition assessment process could realistically be automated or semi-automated, and what that would mean for manual effort, consistency, and scalability.
  • Condition outputs can be linked to the transformer asset and used in risk processes: The project shows that outputs can be correctly linked to PMT asset records and translated into a format that supports CNAIM, Health Index, and NARM-related processes ahead of ED3.
  • The approach produces more consistent condition information than current methods: The project demonstrates that the method can provide more consistent, asset-specific, and auditable condition information than the current largely manual approach.
  • The capture method comparison produces usable findings: The comparison of helicopter and drone-based imagery provides enough evidence to understand their relative image quality, practicality, time efficiency, and cost, and to inform which approach or combination of approaches is most suitable for future use.
  • A credible deployment roadmap is produced: The project delivers a clear roadmap setting out the technical, operational, data, and commercial requirements for scaling the method across the wider PMT fleet.
• Potential for New Learning

  PMT Scan will generate practical learning on how pole-mounted transformer condition can be assessed more consistently using aerial imagery and structured condition analysis. That includes what image quality and coverage is needed to assess visible deterioration reliably, how well corrosion, oil leakage, bushing defects, and external component damage can be identified from captured imagery, and how outputs can be validated against engineering judgement and inspection evidence across different environments, including coastal locations where corrosion risk is higher.

  The project will also produce learning on how condition information can be properly linked to the transformer asset and translated into a form that is useful for asset health, risk, and regulatory processes. This covers data quality, asset linkage, condition classification, and how image-based condition evidence could support CNAIM, Health Index, and NARM-related inputs ahead of ED3. Alongside this, the project will test whether parts of the assessment process can be automated or semi-automated, and what that could mean for scalability, manual effort, and the practicality of keeping condition data current across a large fleet.

  The comparison of helicopter and drone capture methods will also generate useful learning on their relative image quality, efficiency, practicality, and cost, helping inform the most suitable approach for future deployment.

  Learning will be shared through a final project report, deployment roadmap, and supporting technical documentation covering the method, findings, and limitations. It will also be disseminated through internal stakeholder sessions, lessons learned workshops, and industry knowledge-sharing routes including the ENA Smarter Networks Portal. This will help ensure the findings are accessible to other network licensees considering similar approaches to PMT condition assessment.