LINEA²D: Scheduled renovation of overhead power distribution infrastructure

Enedis conducts highly detailed inspections of its power lines to identify degraded equipment and prioritise replacement operations.

The current limitations of manual analysis by human experts (slow analysis, limited number of technicians capable of performing the analysis, cost of analysis) prevent a significant increase in the number of kilometres inspected, thereby limiting its ability to prevent incidents before they occur.

How can reliable and fast diagnosis be automated through the analysis of aerial images taken by a drone?

Our SaaS and edge solution

The proposed solution incorporates a Cloud services platform and Boxes hosting these services locally. This combination provides a distributed decision support system based on Artificial Intelligence and Computer Vision.

Human-Machine Interfaces facilitate interactions between Business Experts and AI, both in the learning phase and during recognition.

Client benefits

Fast, efficient and precise, with reduced production costs

Reduced costs of integration with Business Information Systems, thanks to a modular and secure architecture.

Centralisation of data uploading to a single platform, allowing sharing with stakeholders.

Harmonisation of presentation formats used by Business Experts

In summary

The developed application allows:

Acquisition of images of overhead infrastructure in real-time via a drone in flight,
Automatic real-time analysis of videos to detect power cables and various components connecting the cables to the pylon,
Segmentation of images of those elements, in interaction with a specialised diagnostic operator,
Analysis of those components to diagnose the degree of wear.

Enedis operates the largest European power network, with 1.4 million kilometres of lines in France. These include 230,000km of overhead medium-voltage lines (attached to poles). Due to exposure to weather events, this part of the network accounts for 70% of incidents affecting service quality for customers. These lines are therefore subject to regular advanced inspections performed by an expert, either on foot, by helicopter or, more recently, by drone. The significant aspect of drone acquisition is that the expert has to analyse the video after the inspection, sometimes image by image, to detect anomalies (damaged pole, broken insulator, worn bracket, etc.) – which is a very time-consuming process. Our breakthrough was to use artificial intelligence as a tool to assist in detecting these anomalies from drone videos. We turned to A2D to make this idea a reality. A2D’s approach and expertise allowed us to offer experts a bespoke solution, significantly reducing their analysis time, without changing their working practices, since the AI is hidden behind the app they normally use. Finally, by proposing a solution based on edge-computing, the processing is carried out locally, without depending on computer network speeds, while benefiting from and contributing to the AI’s learning matrix. Throughout this development, Pierre and his team have been very informative, regularly updating us on their progress and consistently putting themselves in the user’s shoes. The solution resulting from this A2D-Enedis collaboration addresses the challenges of maintaining the power network at a controlled cost.