The invention, which belongs to the technical field of power system operation, relates to a sparse-recovery-based quick decomposing method for three-phase state estimation of a distribution network. With consideration of sparsity of bad data in a power system state estimation problem, modeling is carried out based on a sparse recovery theory, so that a state estimation value becomes accurate. Current magnitude measurement is expressed in a net loss mode, so that the state estimation model is simplified. A ratio of a resistance parameter to a reactance parameter in a distribution network is adjusted by using a complex power as a basic value, so that the three-phase power relation in the distribution network can be written in a PQ decomposition mode; calculation is carried out by using a quick decomposing method; and the efficiency of the state estimation is improved substantially because the Jacobian matrix in the quick decomposing method does not need updating in each iteration. Therefore, the sparse-recovery-based quick decomposing method is based on a state estimation model having high anti-error performance and high efficiency.