Evaluating Lightning-Caused Wildfire Risk to Alaska’s Power Grid Infrastructure

Accurate prediction of wildfire risk is crucial for power grid decision makers, allowing them to identify areas of high risk and optimize resource allocation through strategic planning. This is particularly critical in regions like Alaska, where the risk of wildfires poses a significant threat to remote and difficult-to-access power infrastructures. In this study, we aggregate multiple sources of regional climate, topographic, lightning, and fire data to analyze the risk to the electricity grid posed by lightning-caused wildfires in Alaska. A focus area is chosen based on notable changes in lightning activity and the occurrences of lightning-induced fires, along with its proximity to densely populated areas and electric utility infrastructure. To identify key factors, we perform a significance analysis on widely recognized contributors to lightning-ignited fires and subsequently eliminate redundant features. The risk of fire occurrence (target variable) is calculated as a weighted average of two factors: the frequency of fires in the region and the distance to the nearest fire. Machine learning models, including logistic regression and multilayer perceptron, are applied. Ultimately, a decision tree regressor is used to enhance the interpretability of the results in predicting fire occurrence risk. Assigning risk values to the power grid in the designated focus area offers electric utilities in wildfire-prone regions valuable insights, enabling them to improve wildfire planning and mitigation strategies.