Performance assessment of packed bed systems for humidity control in greenhouse applications: An experimental based AI modelling approach

Abstract

Optimal humidity control is essential for enhancing crop yields and ensuring favourable growth conditions in greenhouse agriculture. Packed bed devices are effective tools for regulating humidity levels; however, accurately assessing their performance, especially for temperate oceanic climates, is yet explicitly unexplored. The current paper presents a packed bed system using water as the working fluid to increase humidity during winter for greenhouse cultivation. An experimental setup is developed, and a detailed parametric study is conducted. Also, an artificial intelligence (AI) based multi-layer perceptron neural network (MLPNN) is designed to evaluate the performance of packed bed systems under varying environmental conditions with different inlet air flow rates (176 m³/hr, 286 m³/hr, 383 m³/hr, and 428 m³/hr). The results show that the system achieves a significant 50% increase in humidity ratio, transitioning from an inlet humidity ratio of 6 g/kgda to an outlet ratio of 9 g/kgda when operating with water at an average temperature of 15.7°C and a flow rate of 12.8 kg/min. The MLPNN is trained with 112 non-repeated datasets and observed that a topology of 2-10-10-1 includes 2 input neurons, 2 hidden layers with 10 neurons each, and 1 output neuron, has high prediction accuracy in estimating Δωa values for the packed bed system. The predictions closely align with experimental data, showing a maximum discrepancy within ±2.5%. This research advances the use of packed bed systems by providing a comprehensive framework for assessing and improving humidity control in greenhouse environments.