Residential building hourly operational carbon emission prediction with explainable temporal convolutional and graph convolutional network

Carbon emission prediction is the essential component of building carbon emission management. In this paper, a novel hourly operational carbon emission explainable prediction method for residential buildings is proposed, which employs correlation graph-based temporal convolutional network and graph convolutional network algorithm, and GNNExplainer. To validate the feasibility and effectiveness of the proposed method, case study was implemented with the experiment dataset from Smartline project in Cornwall, UK. The simulation results demonstrate the proposed method can accurately predict residential building hourly operational carbon emission with approximately 0.6865 average empirical correlation coefficient (Corr) value, and 0.2512 average relative squared error (RSE) value. The research results can promote residential building carbon emission management, resulting in significant benefits for more sustainable energy-efficient and low-carbon residential buildings.