Forecasting Future Climate with a Neural Network Trained on Monitored Data: An Analysis of the Energy Demand of a Detached House

Climate change is a topical issue whose main causes are related to human activity. Particularly the main causes of climate change are twofold: the first aspect is environmental; in fact, human activities such as combustion and deforestation cause greenhouse gas emission with an increase in average temperature. The second one is energy-related; in fact, the temperature increase impacts heating and cooling energy needs. Various mathematical models are widespread for forecasting future energy needs, often based on the use of a typical year. The novelty introduced by this paper is the simultaneous use of a neural network to predict the outdoor air temperature to evaluate the cooling and heating energy demand of a detached house through dynamic energy simulations. A neural network, trained on 6 years of monitored data, is used to perform air temperature prediction with a time horizon comparable to the useful life of the heating and cooling system. This prediction is used to carry out dynamic energy simulations in EnergyPlus engine for a reference building in the current state and its refurbished scenario.