Risk Assessment of Distribution Networks Considering Climate Change and Vegetation Management Impacts

TL;DR

This paper tackles vegetation-related outages in power distribution networks under climate influences by building an integrated risk assessment framework that combines outage history, meteorology, and vegetation metrics. It compares logistic regression, XGBoost, and LSTM, selecting logistic regression with SMOTEENN balancing as the best performer on imbalanced data, validated on an 618-sample test set. The model identifies wind speed, vegetation density (via EVI), and wet snow as key outage drivers, with notable interactions such as vegetation density moderating high-wind effects. The approach yields an 80% match to real data within an error tolerance of ±0.05, offering practical guidance for targeted vegetation management and resilience planning, with future work including LiDAR-derived vegetation height data to capture non-linearities.

Abstract

This paper presents a comprehensive risk assessment model for power distribution networks with a focus on the influence of climate conditions and vegetation management on outage risks. Using a dataset comprising outage records, meteorological indicators, and vegetation metrics, this paper develops a logistic regression model that outperformed several alternatives, effectively identifying risk factors in highly imbalanced data. Key features impacting outages include wind speed, vegetation density, quantified as the enhanced vegetation index (EVI), and snow type, with wet snow and autumn conditions exhibiting the strongest positive effects. The analysis also shows complex interactions, such as the combined effect of wind speed and EVI, suggesting that vegetation density can moderate the impact of high winds on outages. Simulation case studies, based on a test dataset of 618 samples, demonstrated that the model achieved an 80\% match rate with real-world data within an error tolerance of \(\pm 0.05\), showcasing the effectiveness and robustness of the proposed model while highlighting its potential to inform preventive strategies for mitigating outage risks in power distribution networks under high-risk environmental conditions. Future work will integrate vegetation height data from Lidar and explore alternative modeling approaches to capture potential non-linear relationships.

Figures (6)

• Figure 1: Simulation framework.
• Figure 2: EVI trends from Jan 2016 through Jul 2024.
• Figure 3: Wind speed distributions illustrating the comparison between outage and non-outage events.
• Figure 4: Outage rates by wind speed bins.
• Figure 5: Outage rates by snow types.