AI anomaly detection on power lines Data Challenge

Last year we took part in a data challenge initiative led by Elewit and the Department of Line Maintenance of Red Eléctrica de España to develop the image processing platform required by the DALIA (Detecting Anomalies in Lines Inspection Autonomously) project.

The project was looking to develop a solution to optimise the analysis of data of images captured during the high voltage power lines inspection using Computer Vision and Artificial Intelligence. In addition, the solution being sought was looking to contribute to the homogenisation and standardisation of the line diagnosis activity and, maybe most importantly, reinforce the safety of the technical engineers and specialists.

The challenge

For the challenge 26 vendors were given two different image datasets to demonstrate their AI capabilities with the following information:

  • A training dataset of 1,234 real images – a mix of various transmission towers.
  • Six types of failures that must be detected: rusted insulators, broken insulator glass, polluted insulator, flashover insulator, rusted tower structure and bent tower bars.
  • A second dataset of 404 images to be used to evaluate and validate the AI accuracy model, in a maximum of 24 hours.
  • The overall process had to be completed and the results delivered in less than 2 weeks.

All proposals were evaluated and judged on the following criteria:

  • Platform architecture
  • Staff capabilities
  • Project planning for the implementation of the final product
  • Client user-case references
  • Metrics associated with the anomaly detection on the validation dataset

The results

In less than 2 weeks, the team at Inspection² was able to prepare, train and analyse datasets and detect multiple defects in the infrastructure. We expected that the AI detection results were not at an ideal production level, given the limited dataset and nature of some of the failures. There was simply not enough training data. However, we demonstrated our software solution’s capability and how it can hit the ground running in such a short space of time. With more time and data in the next operational phases, we expect the AI accuracy to hit production-level performance.

“The quality of the 26 proposals received from a wide range of potential technology partners was very high. The Inspection² proposal was among the challenge’s top five performers. The architecture and technical features of the Inspection² software cover a high percentage of Red Eléctrica’s criteria for this project. The strong staff profiles also covered the needs during the implementation project.  We have seen plenty of clear evidence from their clients to trust in their project execution skills as a provider for a client like Red Eléctrica.”

Iago Veiras Lens

Artificial Intelligence Partner, Elewit Ventures

Whilst this challenge focussed on AI, it’s true to say that this is not the only solution. AI can be very powerful, but it can’t do everything. If the data you use to feed an AI algorithm is not adequate – it’s bad quality or there’s not enough of it – AI alone cannot be the sole solution. There are other scenarios where different software techniques are required. At Inspection², we approach every project on a case-by-case basis as every customer has different requirements. The Inspection² mature software platform offers a wide range of technologies to solve component or anomaly identification problems such as thermal analysis, geo-positioning, optimal character recognition (OCR), and more.

Read more on the winner announcement here (in Spanish)

About Red Electrica

Red Electrica de Espana maintains and operates Spain’s electricity transmission grid. The Company extends the high-voltage grid and coordinates the production and transmission systems. Founded in 1985, the company employs more than 1,800 people and its network stretches across 44,000 kilometres (27,000 miles).

Elewit, the technology platform of the Red Eléctrica Group offers innovative solutions to the new challenges of the electricity and telecommunications sectors, with one goal: to promote energy transition and connectivity for a sustainable future. In order to drive innovation in these areas, Elewit focuses on six technologies: Internet of Things, Industry X.0, Platforms of the Future, Applied Intelligence (AI and Advanced Analysis), Satellites and New Telecommunication Technologies and Cybersecurity.

Pin It on Pinterest

Share This