Can machine learning methods be used to predict likelihood of tree failure? This is the research question which has guided the AI for Tree Risk Prediction project. The motivation behind this work lies in the importance of vegatation and trees in particular in infrastructure management. Large portion of power outages are caused by trees which results in high direct cost to utility companies. A study by Graziano et al. 1 estimated that $8.3 billion in revenue was lost in CT alone between 2005 and 2015.

Traditionally, likelihood of tree failure assessments have been conducted through drive-by windshield surveys and walk-by surveys. In our study we examine the potential for AI to process and predict tree failure near roadside electrical distribution lines. We are designing a framework which automates tree risk classification using a convolutional neural network. Ultimately, our model strives to promote efficiency, reduce costs and increase reliability in modern infrastructure management.

  1. Graziano, M., Gunther, P., Gallaher, A., Carstensen, F. V., and Becker, B. (2020). “The wider regional benefits of power grids improved resilience through tree-trimming operations evidences from Connecticut, USA.” Energy Policy, 138, 111293 ↩