Predicting power outages caused by extratropical storms

Abstract. Strong winds induced by extratropical storms cause a large number of power outages, especially in highly forested countries such as Finland. Thus, predicting the impact is one key challenges for grid operators. This article introduces novel method to predict storm severity employing ERA5 reanalysis data combined with forest inventory. We start identifying objects from wind gust and pressure fields using contour lines 15 m s?1 1000 hPa, respectively. The are then tracked characterized features derived surface weather parameters vegetation information. Finally, classified supervised machine-learning based on how much damage they expected cause. Random classifiers, support vector naïve Bayes processes, Gaussian multilayer perceptrons were evaluated classification task, classifiers providing best results.