Structural parameters optimization based on the neural network predicting the torque coefficient for H-type vertical-axis wind turbines

Abstract

The wind energy utilization of vertical-axis wind turbines (VAWTs) is closely related to the wind wheel geometry. Neural networks can quickly and accurately describe the nonlinear relationships between the structural parameters and aerodynamic performance of VAWTs. Hence, a novel optimization method using the neural network to predict the torque coefficient is developed for the structural design of VAWTs. The statistical parametric mapping initializes the particles and the sine algorithm, variable spiral search, and Levy flight modify the population position update for the Dung Beetle Optimization (DBO) algorithm. The improved DBO algorithm and Binary Particle Swarm Optimization algorithm optimize weight switches, weights, and thresholds of the single hidden layer back-propagation (BP) neural network, and the BP algorithm further optimizes weights and thresholds. Based on the built SMLDBO-BPNN predicting the torque coefficient, the structural parameters of the VAWT are optimized by the Particle Swarm Optimization algorithm. The pressure, vorticity, and torque coefficient of VAWTs before and after optimization are studied by the computational fluid dynamics method. The results show that after the optimization, the high- and low-pressure regions of blades in the windward region increase, the impact of shed vortexes is reduced, and the average torque coefficient increases by 15%.