A Digital Twin Framework for Visual Perception in Electrical Substations Under Dynamic Environmental Conditions.

Area of Science:

• Robotics and Computer Vision
• Electrical Engineering
• Digital Twin Technology

Background:

• Electrical substations are complex, safety-critical environments requiring robust inspection methods.
• Variable lighting and restricted access pose significant challenges for traditional visual inspection.
• Developing and validating vision-based inspection systems requires controlled and repeatable testing environments.

Purpose of the Study:

• To present a novel sensor-centric digital twin (DT) framework for electrical substations.
• To enable the development and validation of vision-based inspection algorithms.
• To generate realistic synthetic datasets for benchmarking robotic perception.

Main Methods:

• Developed a DT framework integrating accurate 3D substation geometry with physically based lighting dynamics.
• Implemented a Render-In-The-Loop (RITL) pipeline for synthetic data generation.
• Configured sensor models, variable lighting, and time-dependent material responses, including dynamic object properties.

Main Results:

• The framework successfully reproduces typical perceptual challenges in substation inspection.
• Generated synthetic datasets with realistic diurnal variations, sun-pose control, and optical imperfections.
• Demonstrated strong potential for supporting the development and testing of robotic perception algorithms.

Conclusions:

• The sensor-centric DT framework offers a controlled and repeatable context for vision-based inspection development.
• This approach facilitates the generation of reproducible datasets for benchmarking and pre-deployment testing.
• The research benefits utility operators, asset management teams, and robotics researchers by enhancing inspection system validation.