Abstract Efficient operation of stratum ventilation heating systems facilitates the creation of a comfortable indoor environment, improving the thermal utilization efficiency (TUE) of the indoor airflow organization. However, the system can still cause local thermal discomfort to occupants due to factors such as airflow differences and temperature gradients. To address this issue, a deep neural network and CRITIC-VIKOR-based optimization strategy for SV supply air parameters is proposed. First, data sets of TUE and PMV were constructed using computational fluid dynamics (CFD) simulations, with the inside surface temperature of the exterior envelope, supply air temperature, supply air rate, supply air angle, metabolic rate of office workers, and clothing insulation as input variables. Second, predictive models for TUE, local PMV, and overall PMV were developed using deep neural networks. Finally, by using TUE and PMV as evaluation criteria, the CRITIC-VIKOR multi-objective decision-making method was applied to optimize supply air parameters through strategy 1 (local thermal comfort as an evaluation criterion) and strategy 2 (balanced TUE, local, and overall thermal comfort as evaluation criteria). The optimal supply air parameters were analyzed under different metabolic rates and clothing insulation conditions. The results showed that strategy 2 provided more significant optimization effects when considering variations in metabolic rates and clothing insulation. This strategy led to a 27.41% increase in TUE, an improvement in average local PMV from 1.45 to 0.04, and a reduction in average overall PMV from 1.42 to 0.18, compared to the benchmark values.
Shen et al. (Tue,) studied this question.