Optimizing baffle plates in the forebay of lateral inlet pump station using backpropagation neural network-genetic algorithm

Abstract

【Objective】 The baffle plates in the forebay of lateral inlet pump stations are critical components to control water flow and optimize the performance of the pump station. We proposed a method to optimize their design. 【Method】 We used computational fluid dynamics (CFD) combined with the backpropagation neural network-genetic algorithm (BPNN-GA) to optimize the design parameters of the rectification baffle plates in the forebay. The optimization process used a comprehensive evaluation index, F, which calculated the fitness of the genetic algorithm based on the uniformity of the axial flow velocity distribution and the velocity-weighted average angle. The BPNN model was fine-tuned, with the comprehensive index F as the optimization objective, leading to the identification of optimal design parameters for the rectification baffle plates. 【Result】 The optimization results showed that the rectification baffle plates designed by the BPNN-GA algorithm significantly improved the flow pattern within the inlet channel. Notable improvements included increased uniformity in axial velocity distribution and a better velocity-weighted average angle. Additionally, the improved design also significantly reduced the vortex area within the forebay, with the comprehensive evaluation index F showing a 6.31 reduction. 【Conclusion】 The proposed BPNN-GA algorithm for optimizing the design parameters of the rectification baffle plates effectively ameliorated the undesirable flow conditions in the forebay. It provides a valuable method for improving the design of similar hydraulic structures.

Keywords

• pump station
• lateral inflow
• bp neural network
• genetic algorithm
• rectification baffle plate