Research data for neural network-based prediction of the dampened thermal power profile of multi-unit residential buildings using a peak-shaving strategy for buffer storage utilization

These are training data for a neural network CNN-LSTM model (including Input-profiles, a Polysun-file which includes the model used to generate the training data) and trained CNN-LSTM models with different parameters for the prediction of a dampened thermal power profile and the secondary return temperature to district heating. Prediction inputs are space heating and domestic hot water consumption profiles. The power profile at the district heating substation is dampened by application of a peak-shaving strategy utilizing the thermal buffer tank (developed in Polysun). The results (power and temperature prediction) of the best performing model are also part of this data set. The hydraulic system (suitable for a multi-unit residential building) which is modelled by the CNN-LSTM model is shown in the Polysun file.