Project description:

The task is to make a 72h forecast of the heat load for a district heating network in Flensburg.

Used model are: 1. Multi Layer Perceptron (MLP) 2. Long Short-Term Memory Neural Network (LSTM) 3. Gated Recurrent Unit (GRU) 4. Convolutional Neural Network (CNN) 5. Seasonal AutoRegreessive Integrated Moving Average with eXogenous factor (SARIMAX)

The scripts can be run within the Google Colaboratory Environment using the dataset data_dummies_index for the neural network models and data_dummies for the SARIMAX model. Best to start with the MLP model as it contains the most detailed comments.