Quantifying the Risk of Wildfire Ignition by Power Lines under Extreme Weather Conditions

TL;DR

This paper introduces WRAP, a machine learning-based method to identify power line segments at high wildfire risk during extreme weather, improving safety and reliability over traditional PSPS strategies.

Contribution

The paper presents a novel risk quantification model using 3D vibration data and machine learning to optimize power shut-offs for wildfire prevention.

Findings

WRAP outperforms naive PSPS in wildfire risk mitigation.

The surrogate model accurately predicts ignition risk under various weather conditions.

Sensitivity analysis highlights key parameters influencing wildfire risk.

Abstract

Utilities in California conduct Public Safety Power Shut-offs (PSPSs) to eliminate the elevated chances of wildfire ignitions caused by power lines during extreme weather conditions. We propose Wildfire Risk Aware operation planning Problem (WRAP), which enables system operators to pinpoint the segments of the network that should be de-energized. Sustained wind and wind gust can lead to conductor clashing, which could ignite surrounding vegetation. The 3D non-linear vibration equations of power lines are employed to generate a dataset that considers physical, structural, and meteorological parameters. With the help of machine learning techniques, a surrogate model is obtained which quantifies the risk of wildfire ignition by individual power lines under extreme weather conditions. The cases illustrate the superior performance of WRAP under extreme weather conditions in mitigating wildfire risk and serving customers compared to the naive PSPS approach and another method in the literature. Cases are also designated to sensitivity analysis of WRAP to critical load-serving control parameters in different weather conditions. Finally, a discussion is provided to explore our wildfire risk monetization approach and its implications for WRAP decisions.