中文摘要：

卫星动力学模型误差是客观存在的事实，动力学模型误差传递到轨道确定算法中构成部分形式未知的模型误差，并且与测量系统自身的系统误差和随机误差耦合在一起形成定轨模型误差，严重影响轨道确定精度．详细推导了存在动力学模型误差的轨道改进方程，对模型中能准确描述的部分建立了参数化模型，对不能准确描述的误差部分，建立了非参数模型．构建了部分线性轨道改进模型，利用二阶段估计法和核函数估计法对模型误差进行拟合估计，并在轨道改进中予以补偿．根据数据深度理论，建立了非参数模型误差的深度加权核估计方法，提高了模型误差估计的抗差性．最后结合天基空间目标监视系统进行了轨道确定仿真实验．实验结果表明，模型误差是影响轨道确定精度的重要因素，核函数估计法可以有效估计定轨中的模型误差，窗宽是提高模型估计精度的重要变量，通过深度加权处理可以明显提高核函数估计的抗差性，提高轨道确定精度．

英文摘要：

It is objective that the satellite dynamic model exits error, and the dynamic model error will be transmitted to satellite orbit determination model with connotative format, commingled together with measurement system error and stochastic error, and form the new mixing model error which will badly pollute the precision of orbit determination. Orbit determination model with dynamic model error was deducted in detaile, while the parametric model for exact dynamic model was produced as well as non parametric model for mixing model error. Thus partially linear orbit determination model was proposed along with its solve method to estimate model error with two stage method and kernel function method, and the estimated model error will be compensated to improve the precision of orbit determination. Because of the non steadiness of the kernel function, a data depth weight kernel estimator for model error was produced based on the data depth theory so as to restrain the singular data. Finally, simulation experiment of SBSS was processed, the results showed that the kernel function can effectively estimate the model error, but a important parameter is the window of kernel. The precision of orbit determination can be improved by compensate the model error and the data depth weight kernel estimator can effectively restrain the singular data of observation error.