CarbonOpt

Inspiration

Decarbonization shouldn't be guesswork and sustainability shouldn't bankrupt businesses. We were inspired to build CarbonOpt to help business owners to deal with environmental, social and governmental pressures with least profit damage and most importantly caring our environment.

What it does

CarbonOpt is an intelligent decision-support dashboard that helps corporates eliminate the guesswork from decarbonization. It answers a fundamental question: "What is the optimal combination of climate initiatives that hits our carbon reduction target without exceeding our capital budget or sacrificing our bottom line?" using baseline baseline forecasting, multi-action what-if simulator, NSGA-II, deterministic recommendation, risk-stress testing, peer benchmarking.

How we built it

There are four main work-streams:

1. ML and Forecasting
2. Action Engine and Optimization
3. Risk Analysis and Statistical Modeling
4. Interactive Dashboard ## Challenges we ran into
   • Integrating Individual Components into a Full Pipeline -Generating a Realistic and Simulative Synthetic Data
   • Multi-Objective Optimization with Hard Constraints
   • Strict Provenance and Honest Architecture ## Accomplishments that we're proud of We didn't build a toy questionnaire or a static dashboard. We built an actual multi-objective optimization pipeline that extracts true Pareto frontiers and handles realistic financial accounting. We treated this hackathon project like production software and worked as a team with high collaboration and communication, helping each other and fixing our teammates problems while trying to help business owners and the environment. ## What we learned Traditional business tools fail because spreadsheets are incapable of navigating Pareto-optimal trade-offs across non-linear dimensions because there are so many parameters and where a parameter can affect another one. ## What's next for CarbonOpt Next, we want to connect the platform directly to real accounting and supply-chain tools, expanding from a 12-month horizon into long-term transition roadmaps. More than anything, we want to prove that taking climate action doesn't have to be a scary financial sacrifice.

Built With

• conda
• github
• matplotlib
• montecarlo
• numpy
• pandas
• pareto
• plotly
• prophet
• pytest
• python
• scikit-learn
• scipy
• statsmodels
• streamlit
• xgboost