ProRivs/solar-forecasting-africa-minigrid-case-study

# Solar Forecasting Africa Mini-Grid Case Study This repository implements a reproducible mini-grid stress-test to evaluate how solar forecast uncertainty affects design-relevant and operational outcomes. ## System Definition - **PV system:** 100 kW AC, 120 kWp DC (DC/AC ratio = 1.2) - **Battery:** 200 kWh / 50 kW, rule-based dispatch - **Load:** synthetic peri-urban hourly demand profile ## Scenario Design Two scenarios are evaluated using the same load and battery assumptions: - **Scenario A — Perfect:** PV generation is computed from actual irradiance - **Scenario B — Forecast:** PV generation is computed from forecasted irradiance produced by the Phase 1 baseline forecasting workflow The difference between these scenarios isolates the operational penalty introduced by forecast uncertainty. ## Outputs The main outputs are mini-grid-relevant risk metrics for both scenarios and their deltas: - Unserved Energy (kWh) - Loss of Load Hours (LOLH) - Energy Curtailed (kWh) - Battery Throughput (kWh) The core Phase 2 result is the forecast penalty: - ΔUnservedEnergy - ΔLOLH - ΔCurtailment - ΔBatteryThroughput ## Objective The object is to quantify how forecast error translates into output above. This is a stress-test of design and operations under forecast uncertainty. The goal is not to build a full optimization framework, but to establish a transparent and defensible pipeline linking: 1. irradiance inputs, 2. PV power conversion, 3. peri-urban demand, 4. battery dispatch, 5. reliability-oriented risk metrics. ## Repository Structure ```text solar-forecasting-minigrid-case-study/ ├── data/ │ ├── README.md │ ├── raw/ │ └── processed/ ├── src/ │ ├── config.py │ ├── pv_model.py │ ├── load_model.py │ ├── battery_model.py │ ├── minigrid_sim.py │ └── metrics.py ├── notebooks/ │ ├── 01_build_inputs.ipynb │ ├── 02_pv_simulation.ipynb │ ├── 03_minigrid_dispatch.ipynb │ └── 04_risk_metrics.ipynb ├── reports/ │ ├── phase2_summary.md │ └── t …