Application of Artificial Neural Networks on Building Energy Estimation


  • To estimate building heating and cooling loads accurately with respect to different building envelope design and climate conditions.
  • To replace the energy analysis tool with trained ANN model or machine learning solutions for faster and accurate estimation of single design or optimization of the building envelope.


  • A neural network with alternative architectures and activation functions is constructed and the weights of the network is updated with stochastic gradient descent algorithms.
  • The performance of the proposed network is tested with 8 building envelope inputs and 2 energy load outputs. The results of the study* shows that  the proposed ANN improves the accuracy of estimation significantly compared to literature studies.
  • In the ongoing research,  the relations between  both envelope design and climate inputs with energy loads output have being examined to construct a more comprehensive and accurate model for replacing analysis tool in the design stage.