VoltVision

Inspiration

High electricity consumption is a significant concern in U.S. households, with a vast portion being wasted daily. This wastage not only contributes to environmental harm but also poses the risk of potential energy crises. There are predictions of rolling blackouts by 2040 if the current consumption pattern goes unchecked.

What it does

VoltVision aims to map the energy footprint of the future using machine learning. It focuses on two main application areas:

Green Tech / Environment
Using machine learning regressions and data analytics.

The benefits of VoltVision include:

Cost Savings
Crisis Prevention
Preparations for Peak Usage

How we built it

Our initial implementation involved using regression, specifically time versus electricity usage. However, this approach didn't work due to fluctuations in the data.

To address fluctuations in the dataset, we then attempted to use a Simple Moving Average over 24-hour periods, which showed improvements but was still not ideal.

During our development process, we tried hyperparameter tuning and testing different models, but these approaches didn't yield the desired results either.

Ultimately, we experimented with using a 4D array as individual variables. The input (X) comprised the year, month, date, and hour, while the output (Y) represented the usage.

Challenges we ran into

Initial regression models faced issues due to data fluctuations.
Hyperparameter tuning and different model testing didn't provide the desired results.
Data processing and understanding the nuances of the dataset were challenging.

Accomplishments that we're proud of

We recognized the significance of the problem and took steps to address it using machine learning.
Experimentation with various models and approaches, ultimately leading to the 4D array solution.

What we learned

One of the most crucial takeaways from this project was understanding the importance of data processing. Proper data processing can significantly impact the effectiveness of machine learning models.