Volume 04: Proceedings of 11th International Conference on Applied Energy, Part 3, Sweden, 2019

District Heat Demand Prediction Using Artificial Neural Network With Data of Sample Buildings Si Chen, Yaxing Ren, Daniel Friedrich, Zhibin Yu


The prediction of building heat demand using engineering models will take unbearable long simulation time when they are applied to large energy networks. To deal with this problem, this paper uses the artificial neural network (ANN) to replace the engineering model in predicting the heat demand of buildings in a district heating network. To train the ANN with efficient learning process, the key data, such as the weather profiles, building fabrications, heating times, and customer heat demand, is collected from the simulation results of sample buildings in engineering models. The ANN trained with data from sample buildings is finally used to predict the heat demand of buildings in a district. In addition, different number of sample buildings for data collection and different percentage of data for training are achieved to balance prediction performance and training speed. The result shows that the ANN-based statistical model has a strong capability to accurately predict the heat demand of buildings with different types of operational functions in a district-level energy network.

Keywords building heat demand prediction, artificial neural network, statistical modelling

Copyright ©
Energy Proceedings