yosef-getaneh/eep-hydro-ml-forecasting

Machine Learning–Based Hydropower Forecasting and Decision Support System for Ethiopian Electric Power Machine Learning-Based Hydropower Forecasting Dashboard for Ethiopian Electric Power (EEP) Overview This project develops a Machine Learning-Based Hydropower Forecasting Dashboard to support operational and planning activities at Ethiopian Electric Power (EEP). Hydropower supplies approximately 90% of Ethiopia’s electricity through major plants like Gibe, GERD, Tekeze, and others. The dashboard integrates operational data from 14 hydropower plants (2013-2025) and meteorological data to forecast key variables such as reservoir water levels, actual energy production, discharge water, and min/max unit loads. A comprehensive data preprocessing pipeline handles date standardization, missing values, outlier mitigation using IQR capping, feature scaling, and selection to reduce multicollinearity. Multiple regression models (e.g., Gradient Boosting, Random Forest, XGBoost) are trained for multi-output forecasting. Performance is evaluated using MAE, RMSE, and R² metrics. The interactive Dash-based web application provides: Dynamic plant selection and forecast horizons. Visualizations: Actual vs. predicted time-series, boxplots for outliers, and 365-day ahead forecasts (2026-2030). Anomaly detection flagging deviations >5%. Tabular outputs for decision-making. This system enhances grid stability, resource optimization, efficiency, and data-driven planning in Ethiopian hydropower operations. Keywords: Hydropower Forecasting, Ethiopian Electric Power, Machine Learning, Gradient Boosting, Random Forest, XGBoost, Multi-Output Regression, Anomaly Detection, Interactive Dashboard, Time Series Prediction. Authors Table of Contents Overview Installation Usage Data Methodology Models Dashboard Features Results Challenges and Limitations Future Work References Appendices License Contributing Contact Installation Prerequisites Python 3.8+ Virtual environment (recommended: venv or conda) Steps Clone the repository:textgit clone github.com …