European Cross-Commodity Risk Monitor

Gas + Carbon → Power Curve Implications

An automated daily workflow that converts public energy market fundamentals into a desk-ready trading narrative for European power (Day-Ahead to curve).

---

Repository Structure

european-risk-pack/
├── risk_monitor.py          # Main automated pipeline
├── requirements.txt
├── .gitignore
├── README.md
├── desk_note/
│   └── desk_note.tex        # LaTeX source for the 1–3 page desk note PDF
└── sample_outputs/          # Pre-generated artifacts (charts, brief, CSV)
    ├── chart_1_macro_drivers.png
    ├── chart_2_power_spreads.png
    ├── chart_3_spread_history.png
    ├── metrics_snapshot.csv
    ├── prompt_log.txt
    └── daily_desk_brief.md

---

Quick Start

1. Install dependencies

python -m venv .venv
source .venv/bin/activate      # Windows: .venv\Scripts\activate
pip install -r requirements.txt

2. Set your Gemini API key

export GEMINI_API_KEY="AIza..."        # macOS / Linux
$env:GEMINI_API_KEY = "AIza..."        # Windows PowerShell

Get a free key at Google AI Studio. The script runs without a key — it skips only the AI narrative step.

3. Run the pipeline

python risk_monitor.py

All outputs are written to ./output_artifacts/.

---

Daily Monitor Metrics (7 core)

#	Metric	Unit	Relevance
1	TTF Gas Front-Month	€/MWh	Primary gas prompt risk benchmark
2	EUA Carbon Front-Dec	€/t CO₂	Carbon compliance cost signal
3	German Baseload Front-Month	€/MWh	European power curve proxy
4	EU Gas Storage Fill	%	Fundamental supply buffer; drives winter curve risk
5	Clean Spark Spread (CSS)	€/MWh	CCGT margin → gas-fired gen profitability
6	Clean Dark Spread (CDS)	€/MWh	Coal-plant margin → coal gen profitability
7	Coal-to-Gas Switching Price	€/tCO₂	EUA threshold above which gas beats coal at the margin

Bonus computed metrics (internal)

• Spark–Dark Differential (fuel-switching signal)
• Rolling 30-day Gas-on-Power beta

---

Metric Formulae

CSS = P_power  − P_gas/η_gas   − E_gas  × P_carbon
CDS = P_power  − P_coal/η_coal − E_coal × P_carbon

Coal-to-Gas Switching Price =
    (P_gas/η_gas − P_coal/η_coal)
    ─────────────────────────────────────────────────
    (E_coal/η_coal − E_gas/η_gas)

Where:

• η_gas = 0.50 (CCGT efficiency)
• η_coal = 0.35 (hard coal plant efficiency)
• E_gas = 0.20 tCO₂/MWh_thermal
• E_coal = 0.34 tCO₂/MWh_thermal

---

Pipeline Steps

fetch_market_data()          # Pulls / simulates OHLC for 5 commodities
    ↓
calculate_metrics()          # Computes all 7 metrics + rolling beta
    ↓
build_metrics_snapshot()     # Today vs prior day table → CSV
    ↓
generate_charts()            # 3 production-quality charts
    ↓
generate_ai_narrative()      # Gemini API → logs prompt + response → MD brief

---

AI / LLM Integration

The generate_ai_narrative() function:

1. Structures the prompt with all 7 live metrics and a strict 2-paragraph template.
2. Calls Gemini 2.5 Flash via the google-genai SDK.
3. Logs the full prompt and raw model response to output_artifacts/prompt_log.txt.
4. Writes a formatted Markdown desk brief to output_artifacts/daily_desk_brief.md.

This reduces the daily analyst write-up from ~20 minutes to under 30 seconds.

---

Production Extension Points

Current (mock)	Production swap-in
np.random simulation	blpapi / refinitiv-data / Montel API
Single-day run	Cron job or Airflow DAG (daily 07:00 CET)
Gemini Flash	Any OpenAI / Anthropic / local LLM via same interface
Markdown output	Push to Confluence / email via SMTP / Slack webhook

---

Compiling the Desk Note (PDF)

cd desk_note
pdflatex desk_note.tex
pdflatex desk_note.tex   # second pass for references

Requires a LaTeX distribution (TeX Live / MiKTeX). Charts must be generated first.