Predicting the wind speed in Miami.

Inspiration

One of the biggest concerns in the field of renewable energy is power generation depending on natural resources that are uncontrollable by humans. -AZO CLEANTECH

One of the most important things for wind turbine efficiency is, of course, wind speed. If we could provide a better prediction for future wind speed, we would be able to maximize the efficiency and reliability of wind turbines. We can also better estimate how many wind turbines are needed to get specific energy outputs in certain areas and times of year.

What it does

Using the city and date given by the user, the ML model predicts the future wind speed on that date and in that location using our modern + minimalistic application.

How we built it

The UI is built using HTML, CSS, and JavaScript, using jQuery to handle user input and displaying results. The ML model uses a KNN regression algorithm to take advantage of the distances of the data features. We cleaned and extracted relevant weather data for our model. Then, we optimize the model by training it on various distance functions, k-neighbors parameters, and 10 k-fold cross validation to pick the best performing parameters for our model. This is done in Python.

Challenges we ran into

We struggled to find weather dataset with ZIP Codes, so we ended up using cities instead of zip codes for locations. We also had some bugs while setting up the backend server, but this was thankfully resolved.

Accomplishments that we're proud of

The average difference between the wind speed we predicted and the actual wind speed is about 0.5 mph! With more weather data incorporated in the future, we can push the accuracy to be even better.

What we learned

Everyone helped each other out, so people who were weaker in one area was able to gain more knowledge and experience with those technologies/concepts.

What's next for Windner

Currently, Windner is limited to the populous cities for which we found a data set, so expanding the breadth of the locations would be more useful for wind turbine location. If the backend is expanded to all of North America, a better UI would show a heatmap where future wind speeds can be looked at concurrently. The updating of new weather data in the future will also help our model to perform even better.

Built With

Submitted to

Knight Hacks

Created by

I helped creating the front-end part of the site, using HTML and CSS, along with helping out with the logic of the solution to the problem.

Sriharsha Aitharaju
I cleaned + extracted the data, trained, created the model, performed statistical analysis, and created the backend server.

Anh Nguyen
I helped created front end, Help to organize data and Trained the data

Syawadhilah Pradipta
I created the front-end, making a simple UI using HTML, CSS, and JavaScript. This was my first hackathon so it was interesting to create a project in such a short period of time.

Private user