What do you want us to predict, power outage only? What are the assumptions?

Hi,

According to the challenge description: " A good model should not only capture whether and where an outage occurs, but also predict it sufficiently early to allow proactive measures. Moreover, modeling the severity and duration of outages is essential for assessing potential impacts on communities."

So I understand that we need to predict power outage only and not rare weather events, correct? It’s not so clear in the description: “Meteorological weather data is essential for forecasting extreme weather events and predicting power outages”.

Next, are external weather forecast models allowed in our prediction? So we can make the assumption that we know what the weather will be for the next P days?

For instance we’re at day T, we’ve weather history for T-1, T-2, T-N days and weather forecast for T+1, T+2, T+3, … T+P. Is such assumption acceptable? Then our model could predict power outage for T+1, T+2, T+3, … T+P, and even T+P+1, … T+P+M

Thanks for the clarifications.

Hi @MPWARE

You are correct in your understanding, the primary goal of the challenge is to predict power outages and how they correlate with extreme, rare weather events. While meteorological weather data is indeed essential for this task, the ultimate objective is to develop a model that can accurately forecast these power disruptions.

You should focus on building a model that predicts power outages, leveraging the relationship with historical and, as you asked, potentially future weather events.

Your assumption that you can use weather history for T-1 to T-N and weather forecasts for T+1 to T+P to predict power outages for T+1 to T+P+M seems reasonable within the context of the challenge, provided that the weather forecast data you use is from publicly available sources.

Remember to adhere to the submission guidelines regarding the inclusion of any external data sources and their licenses.

Good luck with the challenge!

ThinkOnward Team :o: