The effective harmonic estimation for power signals is always an important task in the field of power quality assessment. Due to the development of advanced metering infrastructure (AMI), monitoring of power quality has attracted much attention in recent years. Even though the technology of instrumentation has grown up, the advanced analysis algorithms are still not widely adopted. This is because these approaches usually suffer from the computational consumptions. To reduce the cost and to promote the implementations of power quality monitoring devices in AMI, the finite-word-length (FWL) effects of Kalman filter are taken into account. With the considered effects and related analysis of robustness, the optimal implementations of Kalman filter can be determined and redesigned for the required hardware.