Power Outage Modeling
Welcome to the power outage modeling section! In this section, we dive deep into power outage and weather data and link them to find grid reliability matrix to see how reliability changes both temporally and spatially over Florida. We go even deeper, and based on the power outage data, weather data, grid information, and social vulnerability, we build power outage models to predict different types of power outages (no/minor, moderate, and major) for each county in Florida.
Prof. Thomas Wahl
Mohammad Siddiqur Rahman
Meghana Nagaraj
Alejandra Rodriguez Enriquez
Power outage models rely on historical weather information, power outage records, and environmental factors. To ensure model reliability, we require training data with long temporal coverage and high spatial resolution. We retrieve power outage data from various sources, such as governmental agencies, private providers, or utility companies. We assess data quality and select the appropriate dataset for model development. In cases where multiple datasets are available, merging is done to enhance the spatial and temporal granularity of the power outage information.
We developed weather-related outage risk maps for Florida using historical weather and outage data. The Environment for the Analysis of Geo Located Energy Information (EAGLE-I) power outage data with 15-minute temporal resolution is linked to weather data to build the first version of the risk maps. Maps can be derived for various metrics, at the moment we derive commonly used reliability metrics such as system average interruption duration index (SAIDI) and system average interruption frequency index (SAIFI). The framework used to derive the metrics and maps is described in more detail in the following paragraph.
In the first step, we want to filter EAGLE-I power outage data to remove minors that were unlikely to be weather-related and could be due to daily routine systems operations. Those events are short in duration and only affect a small number of customers. The challenge is to identify an appropriate threshold to remove all those minor events initially. After testing different values, we decided to use the 85th percentile value derived from the raw outage time series and only retain events that exceeded that threshold in terms of customers affected. Next, using a delustering window of 60 minutes (where no outages occur) between events, we identify individual power outage events and calculate duration, maximum number of customers out, and customer hours of interruption (duration multiplied by maximum number of customers). Typically, reliability metrics like SAIDI and SAIFI are derived by using all power outage events. However, our goal is to create risk maps for weather-related power outage events. To identify weather-related outages, we link the outage data to various weather variables that can potentially trigger or contribute to outage events. We found that precipitation and wind speed were the most dominant drivers causing outages in our study region. Hence, we identify weather-related power outages as events when either precipitation or wind speed exceeds their respective 95th percentile values or both when both exceed their 85th percentiles simultaneously. All other outage events are removed (i.e., values set to zero).
SAIDI indicates the total duration (in minutes) of interruption for the average customer in a given time period (e.g., over a year).
SAIFI Calculation:
The System Average Interruption Frequency Index (SAIFI) is calculated by dividing the total number of customer interruptions by the total number of customers served.
Power outage maps were generated for Florida considering weather-related outage events. The figures below show, as an example, SAIDI and SAIFI metrics for each year where power outage data is available.
Fig. 1: SAIDI calculated from weather-related power outage events show spatial variability over the state of Florida.
Fig. 2: MED SAIDI calculated from weather-related power outage events when SAIDI values exceed MED threshold
Non-MED SAIDI is great for mapping minor power outage events that do not exceed the MED threshold, which means it is not significantly influenced by the big events but rather a combination of all the minor events.
Fig. 3: shows Non-MED SAIDI from weather-related events
We aim to develop a robust data-driven weather-outage model using the outage data from power outages and weather data during outage events. This is done by evaluating various machine learning algorithms, such as artificial neural networks or random forests, and selecting the most suitable ones based on their performances. The model output will encompass key metrics such as whether a day is an outage day and a multiclass classification of outages based on duration and/or number of customers impacted. Finally, our goal is to identify vulnerable hotspot areas prone to weather-related outages and assess options to improve the resilience of power systems in those regions.
Based on the duration of the power outage for each day. Events were divided into three different classes.
Minor/no outage: When outage duration is <=1 for a day
Moderate outage: When outage duration is 1< and >=4, it is considered as minor outage day.
Major outage: Daily outage duration is >4 for a major outage day.
Finally, machine learning models were fit to each county to predict the frequency of different types of outages. We utilize this model to predict past and future outages when Eagle-I data is not available.
Fig 4: Power outage modeling shows for year 2021, observed and predicted power outages for each county in Florida.
