Carbon-Aware Electricity Pricing for Switzerland

Simulation framework for evaluating carbon-aware dynamic electricity pricing in Switzerland. This repository accompanies the Master's thesis:

Reducing Emissions Through Carbon-Aware Electricity Pricing: A Swiss Simulation Study

Overview

This project investigates whether incorporating real-time carbon intensity signals into electricity prices can improve the temporal alignment between residential demand and low-carbon electricity availability in Switzerland.

Key Findings

• CA-Hourly reverses demand-carbon misalignment: Correlation changes from +0.24 (TOU) to -0.39 (CA-Hourly), a Δρ of -0.63
• 19.1% emission reduction through temporal demand shifting alone
• CA-Hourly outperforms CA-CPP by 73% for emission reduction (19.1% vs 11.0%)
• Results robust across 4 years (2021-2024) and 27 parameter sensitivity combinations

Repository Structure

carbon-aware-pricing-ch/
├── data/                       # Data directory
│   ├── CH_2021_hourly.csv
│   ├── CH_2022_hourly.csv
│   ├── CH_2023_hourly.csv
│   ├── CH_2024_hourly.csv
│   ├── Energy_Statistic_CH_2021.xlsx
│   ├── Energy_Statistic_CH_2022.xlsx
│   ├── Energy_Statistic_CH_2023.xlsx
│   └── Energy_Statistic_CH_2024.xlsx
├── modules/                    # Core simulation modules
│   ├── __init__.py
│   ├── parameters.py           # Model parameters and configuration
│   ├── data_pipeline.py        # Data loading and preprocessing
│   ├── eda.py                  # Exploratory data analysis
│   ├── pricing.py              # Pricing scheme implementations
│   ├── behavioral.py           # Demand response model
│   ├── hypothesis.py           # Hypothesis testing framework
│   ├── sensitivity.py          # Sensitivity analysis
│   ├── rebound.py              # Rebound effect analysis
│   ├── exporter.py             # Results export utilities
│   └── visualization.py        # Figure generation
├── results/                    # Output directory
│   ├── figures/                # Generated figures (PDF)
│   ├── #.txt                   # Summary reports
│   └── *.csv                   # Simulation results
├── run_simulation.py           # Main entry point
├── requirements.txt            # Python dependencies
├── LICENSE                     # MIT License
└── README.md                   # This file

Installation

Prerequisites

• Python 3.9 or higher
• pip package manager

Setup

1. Clone the repository:

git clone https://github.com/hymekeci/carbon-aware-pricing-ch.git
cd carbon-aware-pricing-ch

2. Create a virtual environment (recommended):

python -m venv venv
source venv/bin/activate  # On Windows: venv\Scripts\activate

3. Install dependencies:

pip install -r requirements.txt

Data Sources

Raw data files are not included in this repository.

To reproduce the analysis:

Carbon Intensity Data

• Source: Electricity Maps
• Files: CH_YYYY_hourly.csv for years 2021-2024
• Variables: Hourly carbon intensity (gCO₂eq/kWh), renewable percentage

Consumption Data

• Source: Swiss Federal Office of Energy (SFOE) via Swissgrid
• Files: Energy_Statistic_CH_YYYY.xlsx
• Variables: 15-minute national electricity consumption

Place downloaded files in the data/ directory with the exact filenames as specified.

Usage

Quick Start

Run the complete simulation for a single year:

python run_simulation.py

Command Line Options

# Single year analysis (default: 2024)
python run_simulation.py --year 2024

# Multi-year analysis (2021-2024)
python run_simulation.py --all

# Include sensitivity analysis
python run_simulation.py --sensitivity

# Skip figure generation
python run_simulation.py --no-figures

# Full analysis with all options
python run_simulation.py --all --sensitivity

Example Output

========================================================================
COMPLETE PIPELINE
Mode: MULTI-YEAR (2021-2024)
Sensitivity: True | Figures: True
========================================================================

[1/6] Loading and cleaning data...
[2/6] Running exploratory analysis...
[3/6] Applying pricing schemes...
[4/6] Modeling behavioral response...
[5/6] Testing hypotheses...
[6/6] Exporting results...

Key Results:
   Δρ = -0.635 (threshold: ≤ -0.15)
   H₁ supported: ✓ YES
   Emission reduction: 19.1%

Methodology

Pricing Schemes

The simulation compares three pricing schemes:

1. TOU (Time-of-Use) — Baseline

   • Peak: 07:00-09:00, 18:00-21:00 (1.30×)
   • Off-peak: 23:00-07:00 (0.70×)
   • Standard: all other hours (1.00×)
   • No carbon signal

2. CA-CPP (Carbon-Aware Critical Peak Pricing) — Event-Based Intervention

   P_CA-CPP(t) = P_TOU(t) × C(t)
   
   where C(t) = 3.0  if CI(t) > CI_90  (critical event)
                1.0  otherwise
   

   • Discrete signal: high prices only during top 10% carbon hours
   • Builds on TOU structure

3. CA-Hourly (Carbon-Aware Hourly Pricing) — Hourly Intervention

   P_CA-Hourly(t) = P_base × [1 + α × (CI(t) - CI_ref) / CI_ref]
   

   • α = 0.40 (carbon sensitivity)
   • Price bounds: [0.50×, 2.00×] of base price
   • Hourly signal: prices reflect full carbon intensity distribution

Behavioral Model

D_new(t) = D_baseline(t) × (P_new / P_baseline)^ε_eff

Where effective elasticity incorporates:

• Base elasticity: ε = -0.35 (Swiss-specific, Filippini 2011)
• Technology amplification: τ = 0.47 (Faruqui & Sergici 2010)
• Loss aversion: λ = 2.5 for price increases (Kahneman & Tversky 1979)
• Thermal comfort floor: 60% minimum demand (Chen et al. 2023)

Results

Hypothesis 1: Temporal Alignment

Year	ρ_TOU	ρ_CA-Hourly	Δρ	Decision
2021	+0.251	-0.503	-0.754	Reject H₀
2022	+0.096	-0.515	-0.611	Reject H₀
2023	+0.335	-0.451	-0.786	Reject H₀
2024	+0.370	-0.464	-0.834	Reject H₀

Threshold: Δρ ≤ -0.15 (derived from Hao et al. 2024 meta-analysis)

Emission Reductions

Year	TOU (ktCO₂)	CA-Hourly (ktCO₂)	CA-CPP (ktCO₂)	CA-Hourly Red.	CA-CPP Red.
2021	4,079	3,284	3,672	19.5%	10.0%
2022	5,186	4,332	4,722	16.5%	9.0%
2023	1,914	1,478	1,616	22.8%	15.5%
2024	916	693	757	24.4%	17.3%
Total	12,095	9,787	10,768	19.1%	11.0%

Key Insight

CA-Hourly outperforms CA-CPP because emission reduction requires responding to the full distribution of carbon intensity variation, not just extreme peaks. By triggering only during hours exceeding the 90th percentile, CA-CPP ignores 90% of the variation that CA-Hourly exploits.

Citation

If you use this code in your research, please cite:

@mastersthesis{mekeci2025carbon,
  author  = {Mekeci, Halil Yavuzhan},
  title   = {Reducing Emissions Through Carbon-Aware Electricity Pricing: 
             A Swiss Simulation Study},
  school  = {University of Geneva},
  year    = {2025},
  type    = {Master's Thesis}
}

License

This project is licensed under the MIT License - see the LICENSE file for details.

Acknowledgments

• Electricity Maps for carbon intensity data
• Swiss Federal Office of Energy for consumption data

This project was developed with assistance from AI tool (Claude, Anthropic) for:

• Code review and debugging
• Documentation writing
• Visualization refinement

All research design decisions, methodology, and interpretations are the author's own work.