mohamadekabou/Tetouan-Energy-Forecasting

Power consumption forecasting for Tetouan City (Morocco) from SCADA time-series data, using XGBoost and Random Forest. # ⚡ Power Consumption Prediction — Tetouan City (Morocco) > Forecasting the electrical power consumption of Tetouan City in 10-minute windows from SCADA time-series data, using **XGBoost** and **Random Forest**. --- ## 📌 Project Overview The goal of this project is to accurately predict the electrical power consumption of **Tetouan City** in 10-minute windows. By leveraging machine learning — specifically **XGBoost** and **Random Forest** — we aim to optimize the management of the distribution network based on historical data and environmental factors. ## 🏢 Context & Data Source The dataset was collected by the **Supervisory Control and Data Acquisition System (SCADA)** of **Amendis**, the public service operator in Tetouan. The city's distribution network is powered by **3 zone stations**: 1. Quads 2. Smir 3. Boussafou The dataset covers a full year (**January 1st – December 30th, 2017**) and includes environmental metrics that influence energy demand. ## 📂 Dataset Description Observations recorded every 10 minutes. | Feature | Description | |---|---| | Date Time | Timestamp (10-min windows) | | Temperature | Weather temperature (°C) | | Humidity | Weather humidity (%) | | Wind Speed | Wind speed (km/h) | | General Diffuse Flows | General flows (air/water) | | Diffuse Flows | Specific diffuse flows | | Zone 1, 2, 3 Consumption | Power consumption (kW) for each station | ## 🛠️ Technologies & Models - **Language:** Python - **Libraries:** Pandas, scikit-learn, Matplotlib, Seaborn - **Models:** - **XGBoost Regressor** — chosen for its efficiency on structured data - **Random Forest** — used as a baseline for performance comparison ## 📊 Results - XGBoost outperformed the Random Forest baseline on the test set. - *(Add your concrete metrics here: R², RMSE, MAE per zone.)* ## 📁 Repository Structure