Comparison of Black Box Models for Load Profile Generation of District Heating Networks

Black box modeling is a fast and efficient way of creating models for generating the heat demand of a district heating networks. A sufficient amount of high quality data has to be collected to form the basis for a valid model that can serve as training and test stand for the models. The model parameters and their influence on the heat demand are investigated and a model structure is derived. With this structure, five data mining algorithms, namely Multiple Linear Regression (LR), Support Vector Regression (SVR), Random Forest (RF), k-Nearest Neighbor (k-NN) and Artificial Neural Networks (ANN) are utilized for creating the load models for a small district heating network located in southeast of Austria. Except for LR, all algorithms showed a good performance. They are well suited for that kind of task. K-NN has the best regression score metric with an average MAPE of 13.49 %.