Evaluation of final energy demand for heating and cooling in buildings under different climatic conditions: A predictive tool based on Artificial Neural Networks

The object of this paper starts from a previous study done by the company Prisma Engineering. The topic was to train a neural network capable of predicting the energy needs of buildings by means of a Python code that, took some input parameters, and scaled theme, created a dataset of buildings to be run in a dynamic tool called Energyplus in order to find the energy needs for space heating and cooling to be delivered to the neural network. Starting from this point and thanks to the now days instruments our task was to improve the operations of creating the dataset in such a way to lead the Python code to be more independent and efficient, to be able to ask for cooling’s demand output, and to be run with three different weather scenarios. The operations consisted in remove the constraints from the input parameters and create more realistic buildings by means of more efficient criterion for the construction of the building’s elements and enter the commands useful for requesting the cooling outputs to the simulation in Energyplus.