Large wind turbine blade design with mould sharing concept based on deep neural networks

With the maturity of design and manufacturing technology development, the drive for lower cost of wind energy continues to increase. To further reduce cost of energy, some wind turbine blade manufacturers adopt the socalled "mould case sharing" strategy, which means that blades with different size share a certain length range of baseline mould cases to improve the reuse rate of mould cases and thus control the average cost of blades. This study develops a deep learning platform for predicting blade performance based on the blade mould-sharing strategy concept. Firstly, a large number of blade samples are generated by randomly sampling geometric parameters such as chord length, twist angle and pre-bending of the blades. Subsequently, a multi-functional simulation platform computes the corresponding key parameters, such as aerodynamic efficiency, blade tip deformation and aerodynamic noise. Afterwards, a blade performance prediction model based on deep neural networks is constructed and trained to achieve better prediction accuracy. Finally, an intelligent algorithm is invoked to design a new larger blade. As compared with traditional modular design methods based on physical solvers, this data-driven approach can significantly reduce the computational cost and design cycles, it has potential application advantages to reduce the unnecessary mould cases design and production.