Canigbobi1/nigeria-energy-forecasting

# Nigeria Electricity Consumption Forecasting > **Forecasting monthly electricity demand across Nigeria's 6 distribution zones using 8 years of NERC data and machine learning — with a best-model MAPE of ~4.2%.** --- ## Project Overview Nigeria's electricity sector is one of the most critical infrastructure challenges in Sub-Saharan Africa. This project analyses **8 years of monthly electricity consumption data (2016–2023)** across 6 Distribution Companies (DisCos) regulated by the Nigerian Electricity Regulatory Commission (NERC). The end goal: build a reliable ML forecasting model that can help grid planners anticipate demand, reduce waste, and flag high-loss distribution zones before they become crises. --- ## Objectives 1. Understand national and regional consumption trends over 8 years 2. Identify seasonal patterns and holiday effects 3. Surface distribution loss disparities by DisCo 4. Build and compare 4 ML models to forecast monthly energy demand 5. Produce a 12-month forward forecast for 2024 --- ## Key Results | Model | MAE (GWh) | MAPE (%) | R² | |-------|-----------|----------|----| | Linear Regression | — | ~8.1% | — | | Ridge Regression | — | ~7.9% | — | | Random Forest | — | ~5.1% | — | | **Gradient Boosting** | — | **~4.2%** | **~0.97** | > **Best model: Gradient Boosting** — tuned with 200 estimators, max depth 4, learning rate 0.05. --- ## Sample Visualisations ### National Consumption Trend (2016–2023) ### Actual vs Predicted (2022–2023 Holdout) ### 2024 National Demand Forecast ### Distribution Losses by DisCo --- ## Key Findings - **Lag features (1–3 months)** were the strongest predictors — recent consumption is the best signal - **December consumption drops ~8%** vs annual average due to holiday grid reduction - **Distribution losses range 18–35%** across DisCos — Kano and Enugu are the highest-loss zones - **Southern DisCos (Eko, Ikeja)** account for ~45% of national consumption --- ## Project Structure ``` nigeri …