Our inspiration for this project came from our desire for impact. The Engie challenge served as the perfect catalyst to work on real-world issues, namely sustainability and climate change.

What it does

We built a web app to help Engie understand their carbon footprint with historical, real-time, and predicted data.

There are three parts to our project. The first section correlates daily historical campus energy load with daily historical CO2 emissions for the last year. The second is a real-time dashboard for visualizing current CO2 output. Additionally, it breaks down grid energy production by sustainability. Finally, the third part of our project predicts the energy requirements for the next 24 hours.

How we built it

We utilized python and pandas for data analysis and Streamlit for visualization. Tensor flow and Keras were used to train our machine learning model (DNN with one hidden layer). We pull relevant data the PI API for campus energy usage, and the EIA MISO API for grid information.

Challenges we ran into

The main challenge we had to overcome was retrieving the correct data from relevant APIs and getting it into a workable form. Additionally, while Streamlit provided an easy environment to deploy our app quickly, there were complications that arose with API authentication. Also, we there are difficulties in finding an appropriate model architecture for a time-series based data-set.

Accomplishments that we're proud of

There are two accomplishments that we're most proud of. First, is the simple and easy-to-use web GUI. And second is our predictive model, which is a deep neural network implemented in Tensorflow and trained on 10 years of historical electricity load data.

What we learned

We learned both technical skills and domain knowledge about our current energy systems.

What's next

We'd like to fine tune our model for more accurate predictions. A UIOWA account is needed to access the API, and so while we have basic login functionality, further support is needed.

Built With

Share this project: