ESG Risk Assessment & Decarbonisation Tracking for Infrastructure Investments
An end-to-end ESG analytics platform built for infrastructure investment analysis — covering ESG scoring, SFDR PAI regulatory reporting, EU Taxonomy alignment, decarbonisation trajectory tracking, and ML-driven volatility prediction across 25 infrastructure companies spanning renewable energy, utilities, oil & gas, digital infrastructure, and social infrastructure.
Infrastructure investment is central to the energy transition and sustainable development. Institutional investors need robust ESG analytics to:
- Evaluate ESG risk during due diligence
- Monitor decarbonisation trajectories against Paris Agreement targets
- Produce SFDR-compliant PAI disclosures for regulators and LPs
- Quantify the relationship between ESG performance and financial risk
This project demonstrates these capabilities with real-world-calibrated data and production-ready tooling.
| Capability | Description |
|---|---|
| ESG Portfolio Scoring | E/S/G sub-scores, ESG profiles (Leader → Laggard) via K-means clustering, sector benchmarking |
| SFDR PAI Dashboard | All 14 mandatory Principal Adverse Impact indicators with traffic-light status and honest data-gap disclosure |
| EU Taxonomy Alignment | Portfolio-level taxonomy eligibility and alignment rates by sector |
| Decarbonisation Tracker | Scope 1+2 emissions trajectories (2020–2024), SBTi target coverage, net-zero timelines |
| ML Risk Model | Random Forest / Gradient Boosting volatility prediction from ESG features, leakage-free time-series cross-validation |
| Due Diligence Scorecard | Per-company ESG assessment with risk flags and investment readiness scoring |
| Excel Report Export | Multi-sheet formatted .xlsx workbook for investor/LP reporting |
| Interactive Dashboard | 6-tab Streamlit app with Plotly visualisations |
25 infrastructure companies across 6 sectors:
- Renewable Energy (8): NextEra Energy, First Solar, Enphase, Brookfield Renewable, Orsted, Vestas, SolarEdge, Plug Power
- Utilities (6): Duke Energy, Southern Company, Dominion Energy, National Grid, Exelon, AES
- Oil & Gas (4): TotalEnergies, Shell, BP, Equinor
- Infrastructure (2): Brookfield Infrastructure, Waste Management
- Digital Infrastructure (4): Equinix, Digital Realty, American Tower, Crown Castle
- Social Infrastructure (1): Welltower
# Install dependencies
pip install -r requirements.txt
# (Optional) Download stock prices and build panel dataset
python data_collection.py
# Generate Excel report
python export_excel.py
# Launch interactive dashboard
streamlit run app.pyThe dashboard works in two modes:
- Basic mode (always available): ESG scoring, SFDR PAI, decarbonisation, due diligence
- Enhanced mode (after
data_collection.py): + ML volatility model, financial correlations
├── app.py # Streamlit dashboard (6 tabs)
├── config.py # ESG universe data (25 companies, 15+ attributes each)
├── data_collection.py # Stock price download & panel dataset assembly
├── analysis.py # ML modelling, correlations, clustering
├── sfdr_pai.py # SFDR PAI indicators & EU Taxonomy computation
├── export_excel.py # Formatted Excel report generation
├── requirements.txt # Python dependencies
└── data/
├── panel_data.csv # Company × month panel dataset (generated)
├── esg_universe.csv # ESG attributes (generated)
├── stock_prices.csv # Daily prices (generated)
├── InfraESG_Report.xlsx # Excel report (generated)
└── charts/ # Static chart exports
Data: Company-level ESG attributes calibrated against publicly reported sustainability metrics (CDP, company ESG reports, proxy statements; FY2023–FY2024 representative values). Stock prices from Yahoo Finance. For production deployment, integrate with MSCI ESG, Sustainalytics, or Bloomberg ESG data feeds.
Financed emissions (PAI 1–2): Standard attribution approach — each company's emissions are attributed in proportion to the portfolio's ownership share (investment / company value), computed for an illustrative €1bn portfolio weighted by market cap. Indicators where issuer-level data is not collected (PAI 7–9) are disclosed as data gaps rather than silently zeroed — mirroring how real reporting teams handle first-year coverage under SFDR's best-efforts provisions.
ML Pipeline:
- Features: ESG score, E/S/G sub-scores, carbon intensity, renewable energy %, taxonomy alignment, board diversity, board independence
- Target: Annualised realised volatility (monthly, √252 scaled)
- Models: Linear Regression, Random Forest (200 trees), Gradient Boosting
- Validation: TimeSeriesSplit 5-fold cross-validation on a chronologically sorted panel — the model always trains on the past and tests on the future, so reported CV-RMSE is a genuine out-of-sample metric
- ~1,500+ company-month observations (25 companies × 60+ months)
Regulatory Framework (as of 2026):
- SFDR (EU 2019/2088, Delegated Regulation 2022/1288) — the 14 mandatory PAI indicators implemented here remain the applicable disclosure standard. The Commission's SFDR 2.0 proposal (November 2025) would streamline product-level PAIs and align them with CSRD data; this project's indicator engine is structured so thresholds and indicator sets can be swapped without touching the rest of the pipeline.
- EU Taxonomy (EU 2020/852) — 6 environmental objectives; eligibility vs alignment tracked per company.
- Omnibus / CSRD — the 2026 Omnibus package narrowed CSRD scope (1,000+ employees, €450m+ turnover) and simplified ESRS datapoints; most companies in this universe remain in scope as large issuers.
- SBTi — Science Based Targets initiative validation tracking.
Python, Pandas, Scikit-learn, Plotly, Streamlit, yfinance, openpyxl
Mridul Daga
- Email: mridul.daga@edhec.com