Modeling and Performance Analysis of a Liquid Desiccant Cooling and Dehumidification System Using ITSO-TCN-BiGRU-SA

Liquid desiccant dehumidification, an energy-efficient technology for air humidity control, has gained significant attention in recent years. In this study, the air temperature and humidity prediction models of liquid desiccant cooling and dehumidification (LDCD) system are built based on the proposed ITSO-TCN-BiGRU-SA. In the proposed model, the TCN is employed to obtain local features within sequence and improve the learning ability of temporal dependencies; BiGRU strengthens the model through global and bidirectional contextual relationships; self-attention mechanism assigns different weights to each time step. The ITSO algorithm, which combines the nonlinear adaptive weights and Levy flight strategy, is proposed to find the optimal hyperparameters of network. Accordingly, the model prediction accuracy is improved. Through comprehensive comparative analysis with other models under a series of experiment results, the superior performance of the developed models was systematically validated. Furthermore, based on the model predictions and experimental results, a comprehensive analysis was performed to systematically investigate the impact of system inlet parameters on cooling and dehumidification capacity and efficiency, which can provide valuable guidance for system control and operation.