针对数以万计的空间碎片，进行快速、精确的轨道确定与预报，以提供可靠的空间碰撞预警是当前空间态势感知的一个重要研究方向。半解析卫星/空间碎片轨道积分，可以克服轨道积分中数值法耗时、解析法低精度的不足。讨论了已初步研发成功的利用多尺度摄动原理的半解析轨道积分器及其在空间碎片轨道预报中的精度性能，并以精密数值积分结果作为“真轨道”评估半解析轨道积分器的精度。大量数据处理结果表明：面质比0.01、轨道高度300 km的碎片，预报1天，耗时不足60 ms，精度2 km左右，且随高度增加，预报精度更高，当高度超过1 000 km时，1天的预报精度仅为50 m左右，满足许多精密空间应用的要求。

For tens of thousands of Earth orbiting space debris, fast and accurate orbit determination and prediction information is essential for providing reliable collision warnings, and thus becomes an important subject in the space situational awareness(SSA). To overcome the problems of low computing efficiency using numerical integration methods and large orbit prediction errors with analytical theories, semianalytic satellite theory(SST) that could achieve the accuracy of numerical methods with the computing efficiency of analytic methods was developed. The semianalytic orbit integrator being developed using the multi-scale perturbation method at Wuhan University was discussed, together with its prediction performace. Orbits generated from the precise numerical integrator was used as the "true orbit" to compute the prediction error of the developed semianalytic integrator. Preliminary results show that, for space objects with Area-to-mass ratio 0.01 and orbit altitude 300 km, 2 km prediction accuracy for 1-day orbit prediction is achieved using the semianalytic orbit integrator, and the computation time was only 60 ms. For orbits higher than 1 000 km in altitude, the prediction errors were mostly less than 50 m, meeting the requirements for many space applications.