Inspiration
The inspiration behind the "Explosive Shells" project was given by this year's Engie Challenge. The need to better understand and optimize solar panel performance, combined with the availability of historical energy data, provided fertile ground for a data-driven solution to analyze solar panel efficiency and contribute to sustainability efforts.
What it does
The Explosive Shells project provides a comprehensive solution designed to investigate and analyze the performance of solar installations at the University of Iowa. It calculates the conversion efficiency of two solar arrays - the Electric Vehicle Charging Array (EVCA) and the CAMBUS Array (CA) - over time and under varying environmental conditions. The project also includes data visualization tools and predictive modeling to provide insights into solar panel performance and the route to the most explosive energy production.
Train 90% and Test 10% Prediction Flow Tag Table
| Electrical Vehicle | Bus Car |
|---|---|
| Features | Features |
| *Ridge Regression: r=78.26 score=0.27 | *Ridge Regression: r=38.03 score=0.34 |
| *Bayes Regression: r=78.31 score 0.27 | *Bayes Regression: r=38.08 score=0.34 |
| *Random Forest: r=24.55 score=0.77 | *Random Forest: r=18.59 score=0.68 |
| *Decision Tree: r=48.83 score=0.55 | *Decision Tree: r=35.43 score=0.38 |
| Removed Features | Removed Features |
| *Ridge Regression: r=81.98 score=0.24 | *Ridge Regression: r=40.61 score=0.29 |
| *Bayes Regression: r=81.94 score=0.24 | *Bayes Regression: r=40.58 score=0.29 |
| *Random Forest: r=52.55 score=0.51 | *Random Forest: r=30.31 score=0.47 |
| *Decision Tree: r=84.47 score=0.22 | *Decision Tree: r=56.21 score=0.02 |
How we built it
-We built the project primarily with Python and its various libraries for data analysis and modeling.
-Data collection was achieved through API integration, extracting historical solar panel generation data from the University of Iowa's energy system.
-Data analysis and visualization components were developed using Python libraries NumPy, Pandas and Matplotlib
-Predictive modeling was implemented using Scikit-Learn, a machine learning framework for regression tasks.
-GitHub and Google Colab for team collaboration and version control.
Challenges we ran into
-Data Integration: Accessing and integrating historical solar panel generation data from the University of Iowa's energy system API presented challenges related to data format, completeness, and potential discrepancies.
-Granularity: Achieving fine-grained granularity in conversion efficiency calculations, especially at hourly intervals, required complex data processing.
-Predictive Modeling: Developing accurate predictive models for energy production based on weather forecasts was challenging due to the dynamic nature of environmental patterns.
And most importantly,
-Artistic Vision: eking out that perfect AI text-to-image prompt to generate the optimally coolest 'exploding shell' images
Accomplishments that we're proud of
-Successfully collected and processed years of historical energy data from the University of Iowa's energy system API.
-Developed algorithms to calculate conversion efficiency at various granularities, contributing to a deeper understanding of solar panel performance.
-Created data visualizations and predictive models that provide actionable financial insights.
-Collaborated effectively as a team, overcoming technical challenges and delivering a comprehensive solution.
-Braved a 36-hour coding marathon, thwarting the relentless advances of Mr. Sandman.
What we learned
-Improved skills in data analysis, including data cleaning, transformation, and visualization.
-Gained experience in predictive modeling and forecasting techniques.
-Enhanced proficiency in using Python for scientific computing and data manipulation.
-Developed a deeper understanding of renewable energy systems and solar panel performance factors.
-Engaged in our inaugural (and second) Hackathons.
What's next for Explosive Shells
-Pushing the limits of discomforting exploding shell images fueled by sleep deprivation and Dall-e
Ben Burnham, Ian Olmstead, Colin Sampey, Michael Van
Log in or sign up for Devpost to join the conversation.