It is appalling how much energy an average household consumes in the US. Research has shown that over 50% of the total energy consumed per month in California comes from household appliances like Air Conditioners and Heaters. Several attempts at decreasing the total energy consumption have failed ranging from increased electricity bills to setting limits for usage per household which is why we came up with the idea of building Savergy that aims to reduce energy consumption by deconstructing the energy chain system and targeting the users directly. By consistently using our app, users would be able to cut down energy usage from each of their homes and in unison, promote environmental sustainability by a reduction in CO2 emissions from less energy production.

What it does

Savergy uses machine learning through the Mage AI to predict the total energy cost of a household and from the information, recommends specific actions that the user can take to reduce the energy consumption in their house. One prominent function is how it estimates the energy cost of using ACs and Heaters and uses weather forecasting to recommend when to use and when not to use these appliances. By doing this, the users cut down on excessive and irrelevant power/energy usage which saves them both cost in dollars and energy and makes them partakers of environmental sustainability.

How we built it

For the back-end, We gathered and prepared a dataset with relevant variables then trained the data to maximize the accuracy of estimating the total energy cost of a household given the user input parameters by Machine Learning through Mage AI. We also deployed the model and hosted the API through Mage as well and parse it into our front end algorithm which we made using Python on Streamlit. We made a web app that a user could input a set of parameters and it would estimate the energy cost of their household. Then, it would forecast their weather conditions using the Weather API and recommend using the Twillo API whether they should put on their AC for that day or night or not based on the average temperature. If cold, No. If hot, No.

Challenges we ran into

  1. The time difference between user
  2. User Authentication was difficult to implement
  3. One of the team members lost electrical power to their home and had to be off-grid for a while.

Accomplishments that we're proud of

  1. We were able to successfully train the model with an R2 score of 0.92
  2. We also successfully, deployed the model, hosted the API, integrated the front and back ends and deployed the app.
  3. We made use of multiple APIs: Mage, Weather, and Twillo and made them all work

What we learned

  1. We learned that working together in groups is tasking and requires patience and understanding
  2. We also learned that it is better to have everyone know at least something about your app than for only one person to know everything.

What's next for Savergy

Savergy would be used on a large scale to monitor the energy usage of large corporations in large cities and help reduce energy consumption from their ends. We would also increase its generalizability by improving the models with a large database such that it can effectively predict the expected energy consumption for cities in the United States per month/season and help the electricity distribution companies efficiently distribute available energy for use per need in each city. To cap it up, Savergy could be integrated into the electricity distribution system and become the decentralized energy use tracking device for each user/ resident of a community where they can pay for their electricity expenses from the app and save cost as much as possible. This would benefit the earth by reducing Carbon emissions that energy production causes and fighting against climate change.

Built With

Share this project: