Study of Draft Tube Optimization Using a Neural Network Surrogate Model for Micro-Francis Turbines Utilized in the Water Supply System of High-Rise Buildings

• Published in: Processes 2024-06, Vol.12 (6), p.1128
• Main Authors: Xin, Qilong, Wu, Jianmin, Du, Jiyun, Ge, Zhan, Huang, Jinkuang, Yu, Wei, Yuan, Fangyang, Wang, Dongxiang, Yang, Xinjun
• Format: Article
• Language: English
• Subjects: Algorithms; Clean energy; Design parameters; Draft tubes; Energy consumption; High rise buildings; Laboratory tests; Mathematical models; Neural networks; Numerical methods; Optimization; Performance prediction; Prototypes; Quadratic programming; Radial basis function; Turbines; Water pipelines; Water supply; Water supply systems
• ISSN: 2227-9717; 2227-9717
• DOI: 10.3390/pr12061128

With the increasing popularity of clean energy, the use of micro turbines to recover surplus energy in the water supply pipelines of high-rise buildings has attracted more attention. This study adopts a predictor model based on Radial Basis Function Neural Network (RBFNN) to optimize the draft tube shape for micro-Francis turbines. The predictor model is formed on a dataset provided by numerical simulations, which are validated by lab tests. Specifically, numerical investigations are carried out in the shape of a draft tube to determine an optimal model. Additionally, the superiority of the RBFNN model in nonlinear optimization is verified by comparing it with other models under the same date sets. After that, the design parameters are optimized using RBFNN and sequential quadratic programming algorithm (SQPA). Finally, the turbine prototype is fabricated and tested on a lab test rig. The experimental results indicate that the numerical method adopted in this research is accurate enough for such a micro-Francis turbine performance prediction. Under the design conditions, the proposed micro-Francis turbine produces a power of 147 W with an efficiency of over 29%, which shows a considerable improvement compared to the initial prototype.