VAREN

VAREN (Vegetation and Asset Recognition for Electrical Networks) provides a Python library that automates the interpretation of mobile LiDAR point clouds collected along overhead distribution networks. It converts raw laser scan data into structured asset inventories and corrected network geometry, and it produces prioritized vegetation work orders.

Vegetation accounts for roughly 45% of outages on the Hydro-Québec distribution network. VAREN was built to deal with that problem at scale. It replaces largely manual point cloud interpretation with an automated pipeline that gives distribution utilities actionable outputs they can act on.

Project Lifecycle Stage
Sandbox

Project Special Interest Group
Artificial Intelligence SIG

The Challenge

Mobile LiDAR (laser scanning from vehicle-mounted survey units) gives utilities detailed, current views of their overhead distribution networks and many collect it routinely. The bottleneck is interpreting the data once it has been collected.

Turning raw point clouds into usable network information is slow, manual work. Vegetation can hide conductors. Assets are difficult to distinguish from objects around them, and sections of the network may be missing or incomplete in the scan data. Without automation, there is still a wide gap between collecting a survey and acting on it.

The consequences are practical:

• Asset inventories fall out of date, reducing confidence in GIS and network models
• Vegetation encroachment on conductors goes undetected until it causes an outage
• Pruning and maintenance programs rely on field inspection rather than data-driven prioritization
• Each utility or research group that wants to automate LiDAR interpretation must build the processing pipeline from scratch

VAREN addresses all four by providing a shared, open pipeline that utilities and researchers can adopt, adapt, and improve without duplicating foundational engineering work.

Key Features and Use Cases

Point Cloud Classifier

A neural network labels each point in a mobile LiDAR scan as a distribution asset or environmental object: poles, wires, transformers, vegetation, buildings, and other features. The classifier distinguishes medium-voltage, low-voltage, and telecom lines from one another, providing the labeled data on which all downstream analysis depends.

Overhead Line Geometry Reconstruction

Obstruction or low point density sometimes leave gaps in the scan. In those places,VAREN uses mathematical reconstruction algorithm to model the power line geometry across gaps. The pipeline then gives a continuous representation of conductor geometry even where the survey data is sparse or occluded. The resulting asset geometry data can be prepared for distribution network digital twins, power flow models and capacity planning tools.

Pole and Span Extraction

A network topology algorithm extracts poles and spans from the labeled and reconstructed scan. It rebuilds the physical layout of the distribution network without relying on existing GIS records. The result is a network model based directly on what was observed in the field. Utilities can use it to validate, correct, or replace existing GIS data.

Vegetation Encroachment Detection

VAREN finds vegetation encroaching on conductor spans, producing risk-zone classifications, and generating pruning work orders. The outputs are structured to go directly into the utility vegetation management application and enterprise asset management systems. Vegetation programs that are currently reactive and inspection-driven can then be scheduled from survey data that is already being collected.

Key Contributors

• Hydro-Québec

Collaboration Opportunities

VAREN welcomes participation from distribution utilities, LiDAR survey providers, GIS and asset management platform developers, and researchers working on:

• Mobile LiDAR processing and point cloud analysis for utility infrastructure
• AI and machine learning applied to distribution asset classification
• Vegetation management automation and risk assessment
• GIS data quality and network model maintenance
• Digital twin development for distribution networks

Utilities that collect mobile LiDAR data and want to automate interpretation, or that want to contribute additional scan data and use cases to the project, are encouraged to engage. Researchers developing point cloud AI methods for infrastructure applications will find an applied, real-world dataset and pipeline to build on.

VAREN provides the shared LiDAR processing foundation that distribution utilities need to act on survey data at scale, without each organization building the same interpretation pipeline independently.

Upcoming Meetings

View the meeting calendar

View the calendar of all LF Energy events