This project uses PyTorch to build a neural network model to predict residential electric vehicle (EV) charging loads based on real-world data from apartment buildings in Norway. The goal is to forecast the actual energy consumption (in kilowatt hours) during a charging session, which could be useful for predicting energy costs and planning EV charging infrastructure.
The data can be accessed from this [Link Text]: https://data.mendeley.com/datasets/jbks2rcwyj/1
Note:
The project was a challenge problem as part of the PyTorch and Neural Networks course on codecademy.
Plug-in Duration: The total time the vehicle remains plugged in.
Location Type: Whether the charging point is in a private or public setting.
Month: The month in which the charging session takes place.
Day of the Week: The specific day the charging session occurs.
Traffic Density: A measure of the traffic around the location during the session.
Charging Load (kWh): The amount of energy consumed during the charging session.
This project aims to accurately predict the energy consumption for EV charging sessions. Such a model can help in optimizing EV infrastructure planning by forecasting energy requirements, which could aid in managing energy costs effectively.
Framework: PyTorch Model Type: Feedforward Neural Network (Regression) Loss Function: Mean Squared Error (MSE) Evaluation Metrics: Root Mean Squared Error (RMSE), Mean Absolute Error (MAE)
If the model performs well, it can be applied to:
Predict energy consumption for residential EV charging. Help in cost estimation for developing charging infrastructure. Assist in energy grid planning and optimization.