电动旋翼动力系统参数辨识研究

Parameter Identification of Rotor Electrical Driving Systems

电动旋翼动力系统是电动飞行器的核心组成部分之一,获取其有效模型以及特性分析对飞行器控制研究有着深刻的影响。综合电机、电子调速器、螺旋桨的机制模型具有非线性特性,因此将无迹卡尔曼滤波(UKF,unscented kalman filter)引入相应的参数估计。但待估计参数过多时易导致UKF辨识中协方差矩阵非正定。为回避此问题,提出分步辨识的方案,先对一部分机制模型参数及其组合依据稳态实验数据进行拟合估计,再依据瞬态实验数据利用UKF方法辨识出各参数值,最终获得完整、精确的飞行器动力系统数学模型,为飞行器大机动飞行仿真与控制设计提供基础。实验结果表明,所采用的参数辨识方法可以不依赖经验值,估计结果能够快速收敛,并有较高的精确性和鲁棒性。

Rotor-driving system is one of the key components of a rotorcraft.The obtained valid model and analysis of the characteristics have great influence on the flight control design of the rotorcraft.The mechanical model integrating the motor,electronic speed controller and propeller presents nonlinear characteristics.Hence,the UKF(Unscented Kalman Filter)is introduced into the parameter estimation.However,the covariance matrix in the iterative identification process based on UKF is of non-positive definition when there is too many parameters to be estimated.To avoid the problem,a stepped scheme is proposed.In the first step,the partial parameters in the mechanical model and their combination are estimated according to steady-state experimental data.In the second step,a UKF-based method is used to identify the remaining parameters according to the transient experiment data.The experimental results show that the proposed method needs not to rely on empirical values and the estimation process converges quickly with high accuracy and robustness.