基于Kalman滤波的原子时算法研究
2023,31(3):294-299
摘要：守时系统的目标在于建立和保持一个稳定可靠的时间尺度，时间尺度算法正是基于此目标计算出一个频率稳定度、准确度、可靠性更高的时间尺度，时间尺度的算法本质就是综合守时系统内的原子钟，通过各原子钟与主钟的N-1组观测钟差对N台原子钟的权重和预测值进行估计；传统的加权平均算法会忽略发挥主要影响的噪声过程，更注重权重的合理分配来提高综合原子时的稳定度，缺少对噪声的关注，针对守时系统实时性的需求，对原子钟噪声模型进行了研究，在频率预测过程中研究了Kalman滤波和频率跳变检测的应用，并与传统加权平均算法进行了对比，仿真实验表明改进的算法提升了综合原子时的中长期稳定度,其中100天稳达到了〖5×10〗^(-14)数量级，既保留了AT1时间尺度连续、实时的良好特性，又避免了Kalman算法发散性的问题，经实际测试可应用于小型守时实验室的守时系统构建。
关键词：守时系统；时间尺度算法；Kalman滤波；预测值；频率跳变检测
Research on Atomic Time Algorithm Based on Kalman Filter
Abstract：The goal of the timekeeping system is to establish and maintain a stable and reliable time scale. The time scale algorithm is based on this goal to calculate a time scale with higher frequency stability, accuracy and reliability. The essence of the time scale algorithm is to integrate the primary clocks in the timekeeping system, and estimate the weight and predicted value of N atomic clocks through the N-1 group of observation clock differences between each atomic clock and the main clock. The traditional weighted average algorithm will ignore the noise process that plays a major role, and pay more attention to the reasonable distribution of weights to improve the stability of the synthetic atomic time. It lacks attention to noise. Aiming at the real-time requirements of the time-keeping system, the atomic clock noise model is studied, and the application of Kalman filter and frequency jump detection is studied in the frequency prediction process, and compared with the traditional weighted average algorithm, The simulation results show that the improved algorithm improves the stability and reliability of the synthetic atomic time in the medium and long term. It not only retains the good characteristics of continuous and real-time AT1 time scale, but also avoids the problem of divergence of Kalman algorithm. After practical testing, it can be applied to the system construction of small time-keeping laboratories.
Key words：Punctuality system; Time scale algorithm; Kalman filtering; Estimate; Frequency hopping detection
收稿日期：2023-02-06
基金项目：河北省科技重大专项（22280901Z）
