Air balancing method of multibranch ventilation systems under the condition of nonfully developed flow. (September 2022)

• Record Type: Journal Article
• Authors: Wang, Yi; Gao, Ran; Li, Angui; Gong, Zhiguo; Ni, Qichang; Yang, Yiwei; Liu, Boran; Du, Xueqing
• Is Part Of: Building and environment. Volume 223(2022)
• Publication Date: 2022-09
• Subjects: Duct systems -- Air balancing -- Multiobjective optimization -- Machine learning -- Nonfully developed flow -- Resistance calculations; Buildings -- Environmental engineering -- Periodicals; Building -- Research -- Periodicals; Constructions -- Technique de l'environnement -- Périodiques; Electronic journals; 696
• DOI: 10.1016/j.buildenv.2022.109468
• Languages: English
• ISSNs: 0360-1323

Abstract

Abstract: Air balancing is one of the cornerstones of the building environment. Research on air balancing aims to ensure the efficient operation of ventilation systems. However, there are two challenges in the air balancing process. The first issue is accurately calculating the duct resistance under the condition of nonfully developed flow. The other obstacle is calculating the damper adjustment degree when the operating flow characteristics of the dampers are affected. Based on these two challenges, this study first proposes a pressure balance model under the condition of a nonfully developed flow for the resistance calculation. The problem of parameter identification in the model is converted into a multiobjective optimization problem and solved by the elitist nondominated sorting genetic algorithm II (NSGA-II). In contrast to the traditional pressure balance model, the average error of the resistance calculation can be reduced by 78.83%–82.24%. Second, based on the resistance characteristics (degree, airflow, pressure difference) of the dampers in the duct system, XGBoost is used for machine learning to achieve accurate adjustment of the dampers. Compared with the traditional method that requires checking manuals, the average error of the damper degree calculation can be reduced by 62.27%–94.6%. Finally, the air balancing method proposed in this study is experimentally verified, which has an average relative error of airflow of only 4.11% and a maximum relative error that does not exceed 10%. Highlights: A Pressure balancing model suitable nonfully developed flow is proposed. The identification of parameter is converted into a multi-objective optimization. The machine learning method of XGBoost is used to predict the damper degree.