Short-Term Prediction of Electricity Outages Caused by Convective Storms

Abstract

Prediction of power outages caused by convective storms which are highly localised in space and time is of crucial importance to power grid operators. We propose a new machine learning approach to predict the damage caused by storms. This approach hinges identifying and tracking of storm cells using weather radar images on the application of machine learning techniques. Overall prediction process consists of identifying storm cells from CAPPI weather radar images by contouring them with a solid 35 dBZ threshold, predicting a track of storm cells and classifying them based on their damage potential to power grid operators. Tracked storm cells are then classified by combining data obtained from weather radar, ground weather observations and lightning detectors. We compare random forest classifiers and deep neural networks as alternative methods to classify storm cells. The main challenge is that the training data are heavily imbalanced as extreme weather events are rare.

Predicting power outages caused by extratropical storms

“…Trees in the ensemble are constructed with four steps: 1) use bootstrapping to generate a random sample of the data 2) randomly selected subset of features at each node 3) determine the best split at the node using loss function 4) grow the full tree (Breiman, 2001). RFC is also found to provide adequate performance with imbalanced data (Tervo et al, 2019;Brown and Mues, 2012). We use RFC with the Gini impurity loss function.…”

“…Rossi (2015) developed a method to detect and track convective storms. The method was later developed to predict power outages (Tervo et al, 2019).…”

“…We present a novel method to identify, track, and classify extratropical storm objects based on how much power outages they are expected to induce. We adapt storm object detection (Rossi (2015), Tervo et al (2019), Cintineo et al (2014)) to 2 https://doi.org/10.5194/nhess-2020-205 Preprint. Discussion started: 3 August 2020 c Author(s) 2020.…”

Predicting power outages caused by extratropical storms

“…Rossi (2015) developed a method to detect and track convective storms. The method was further developed to predict power outages (Tervo et al, 2019).…”

“…We present a novel method to identify, track, and classify extratropical storm objects based on how many power outages they are expected to induce. We adapt convective storm object detection (Rossi, 2015;Tervo et al, 2019;Cintineo et al, 2014) to find potentially harmful areas from extratropical storms by contouring objects from pressure and wind gust fields. Instead of highly localized convective storms, we aim at larger but still regional geospatial accuracy so that, for example, damage in western and eastern Finland can be distinguished.…”