Abstract: A point reactor model with 6 groups delayed neutrons was adopted to simulate the typical 20 pcm and 60 pcm prompt reactivity insertion during the period of initial criticality and zero power physics test. Both Kalman filter and dynamic filter algorithms were studied. The results show that the Kalman filter approach can attenuate the impact of neutron prompt jump and gives features of self-adaptation. However, it reaches less accuracy at the earlier transient period and the protection response time is long when large reactivity is inserted. The dynamic filter approach gives high accuracy and response quickly through the optimization of the relative power change LAMMA and gain coefficient λ. However, frequent adjustment of λ takes time and increases the probability of operation mistakes. Static filter, with optimized constant gain coefficient, is accurate, timesaving and safe.