翼伞发电系统的GPU并行轨迹优化

GPU based trajectory parallel optimization of tethered airfoils for wind energy generator

高空风能发电是一种新型的清洁能源生产方式。针对这种非常规的带有特定目标函数优化的轨迹设计问题,采用预测控制是一条可行途径,但该方法目前需要事先离线求解,计算量极大,不具有在线自适应能力。提出了一种基于混沌的实时轨迹优化策略,以克服上述算法的不足。这是一维变量滚动次优化问题,利用均匀采样结合混沌搜索,给出了过程约束下的优化方法。通过采用数值算法的并行化,提高了在线计算效率。半实物仿真试验结果说明了该算法的有效性。

Wind energy at high altitudes is a new approach to generate clean energy. The predictive con-trol in the offline manner was previously employed to handle the problem of trajectory design with uncon-cventionally given objective function, however it is time-consuming and lacks of adaptability and flexibil-ity to varying aerodynamic parameters. A receding horizon optimization method for the tethered foil gener-ator based on an online searching strategy was presented. The nonlinear optimization problem was approx-imately reformulated to a univariate receding horizon sub-optimal issue in a short interval. By using uni-form sampling and chaotic search approaches, the sub-optimal solution, subject to the physical con-straints, was obtained. The proposed method is parallelly implemented by graphic processing unit （ GPU） to raise its online calculation efficiency. The hardware-in-the-loop simulation result demonstrates its effec-tiveness.