Siyaovo/south-africa-electricity-analysis

# South Africa Electricity System Analysis ### Demand Forecasting & Supply Optimisation **Supervised by Nicolas Maisonneuve-Bonteil, Deloitte | Université Paris 1 Panthéon-Sorbonne | M2 Sustainable Development Economics** --- ## Overview This project develops a quantitative analytical framework for South Africa's electricity system, combining **electricity demand forecasting** with **supply-side economic dispatch optimisation**. The analysis evaluates trade-offs between cost, reliability, and decarbonisation under alternative policy scenarios. The case study is motivated by South Africa's ongoing energy crisis: chronic load shedding, heavy coal dependence (~80% of generation), and a structural transition toward renewables under the Just Energy Transition Partnership (JETP). --- ## Project Structure ``` south-africa-electricity-analysis/ ├── notebooks/ │ ├── 01_demand_forecasting.ipynb # Demand forecasting (ARIMA/SARIMAX/LSTM/TFT/Hybrid) │ └── 02_supply_optimization.ipynb # Pyomo dispatch model & scenarios ├── data/ │ └── ESK17390.csv # Eskom hourly system data (2021–2025) ├── reports/ │ ├── south-africa-electricity-analysis.pdf # Full academic report (PDF) │ ├── presentation_slides.pdf # Presentation slides │ └── report_source.md # Report source (Markdown) ├── figures/ ├── requirements.txt └── README.md ``` --- ## Analytical Components ### 1. Demand Forecasting (`01_demand_forecasting.ipynb`) Forecasts daily electricity demand over a 90-day out-of-sample horizon using six modelling approaches: | Model | Type | Key feature | |-------|------|-------------| | ARIMA(1,0,1) | Statistical | Baseline univariate model | | SARIMA(1,0,1)(1,1,1,7) | Statistical | Weekly seasonality | | SARIMAX | Statistical | Exogenous regressors (outages, load shedding, pumped storage) | | LSTM | Deep learning | Nonlinear dynamics, 30-day lookback window | | Temporal Fusion Transformer (TFT) | …