Evaluation tool for the thermal performance of retrofitted buildings using an integrated approach of deep learning artificial neural networks and infrared thermography

Abstract

In most countries, buildings are responsible for significant energy consumption where space heating and air conditioning is responsible for the majority of this energy use. To reduce this massive consumption and decrease carbon emission, thermal insulation of buildings can play an important role. The estimation of energy savings following the improvement of a building's insulation remains a key area of research in order to calculate the cost savings and the payback period. In this paper, a case study has been presented where deep retrofitting has been introduced to an existing building to bring it closer to a Passivhaus standard with the introduction of insulation and solar photovoltaic panels. The thermal performance of the building with its improved insulation has been evaluated using infrared thermography. Artificial intelligence using deep learning neural networks is implemented to predict the thermal performance of the building and the expected energy savings. The prediction of neural networks is compared with the actual savings calculated using historical weather data. The results of the neural network show high accuracy of predicting the actual energy savings with success rate of about 82% when compared with the calculated values. The results show that this suggested approach can be used to rapidly predict energy savings from retrofitting of buildings with reasonable accuracy, hence providing a practical rapid tool for the building industry and communities to estimate energy savings. A mathematical model has been also developed which has indicated a life-long monitoring will be needed to precisely estimate the benefits of energy savings in retrofitting due to the change in weather conditions and people's behaviour.