Digital watt-hour meters in intelligent substation receive IEC61850 packets, then calculate the electric power based on the data of voltage and current in those packets, so the accuracy of electric power calculation mainly depends on the accuracy of algorithm. In addition, the digital watt-hour meters also need to overcome the effects of bad points, packet loss and time delay during communications and other unfavorable factors to achieve high measuring accuracy. For that, this paper proposes a new FIR low-pass filter based on Blackman self-convolution window function and a new method for coarse data correction and packet loss compensation based on the least square method. Finally, the authors prove by experiment that the algorithm is effective and reliable on the basis of detailed information about the theory of algorithm.