Inspiration
During the winters in buffalo the weather can become brutal with high winds and very low temperatures. In some instances when blizzards occur houses can go out of power and as a result the temperature indoors can get down to critical levels. If we could optimize a route that gets crews to more critical areas then we can help save more lives.
What it does
Our project uses a simulation based off previous big storms in Buffalo New York and using that simulation we are able to use an llm to optimize a route that targets the more critical areas based on winds, snow, and how fast a house can get to critical levels in temperature.
How we built it
We gathered data from the census to get an idea of the population within an area and we looked at data from big storms to get info on the wind conditions and snow levels. Then we put all of our data into tigerdata to have easy access to it. Afterwards we used an LLM that would predict the best route based on all of the data provided.
Challenges we ran into
Our simulated data estimate was far different from real baseline values. We trained our simulation data without accounting for actual weather conditions that could affect how realistic the baseline data was. We also had trouble getting our api call to treat every hour as a unique scenario and considering the same context data for every hour. Our overall cost for claude api credits were also growing faster than anticipated, so each rerun took not only a mental strain, but a growing financial worry too.
Accomplishments that we're proud of
We managed to create a full working pipeline in 24 hours by gathering data into our database and using our AI on a simulation using the database to make the best optimal routes to areas of critical conditions.
What we learned
We learned how to use an API for our simulations and how to get it to use the information we gave it to get the most optimal routes to restore power to critical areas. We also learned how to use a time-series database to make it easier to store all our replays and information to easily compare them.
What's next for Power Triage
As we grow in the future we want to expand and gather real indoor temperature signals from households to get a more accurate route. We also intend to expand the project to include more of western New York so that this can be expanded to more areas. .
