rohitkkk/Power-Consumption-Prediction-for-Tetouan-City

This project uses machine learning techniques to predict power consumption across three distribution zones in Tetouan City, Morocco. The model employs a Random Forest regression algorithm to forecast energy consumption based on environmental data such as temperature, humidity, wind speed, and solar radiation. # Power-Consumption-Prediction-for-Tetouan-City This project uses machine learning techniques to predict power consumption across three distribution zones in Tetouan City, Morocco. The model employs a Random Forest regression algorithm to forecast energy consumption based on environmental data such as temperature, humidity, wind speed, and solar radiation. ## Dataset The dataset used in this project is sourced from the UCI Machine Learning Repository and contains hourly measurements of environmental parameters and power consumption for three zones: Source: Power Consumption of Tetouan City ### dataset attributes: * DateTime: Time of the measurement (10-minute intervals) * Temperature: Weather temperature in °C * Humidity: Relative humidity in % * Wind Speed: Wind speed in m/s * General Diffuse Flows and Diffuse Flows: Solar radiation in W/m² * Zone 1, Zone 2, Zone 3 Power Consumption: Target variables for prediction (kW) ## Tools and Libraries The following tools and libraries were used: * R: Programming language for statistical computing * Random Forest: Machine learning model for regression ### Libraries: * randomForest * ggplot2 * caret * lubridate ## Methodology ### Data Preprocessing: Handling missing values and converting the DateTime field to a suitable format. Normalizing environmental variables. Feature Engineering: Original features: Temperature, Humidity, Wind Speed, General Diffuse Flows, Diffuse Flows. Created additional features: TwoHourInterval, Weekend, DayOfWeek and Hour. ### Model Training and Evaluation: Random Forest regression was applied separately for each zone. Evaluation metrics included Root Mean Squared Error (RMSE) and R-squared. ### Improvements: Incorporating temporal features (Hour, DayOfWeek, TwoHourInterval, Weekend) led to a significant reduction in RMSE. ## Results After incorporating additional features, the model showed improved accuracy across all zones: * Before Feature Addition: Zone 1 RMSE: 5249.296, R-squa …