The invention relates to a bilinearization-based all-distributed robust state estimation method for a multi-regional power network, and belongs to the fields of power system operation and control. The method is divided into three states: preprocessing a measurement to obtain pre-estimation of an intermediate state variable in the first stage; carrying out nonlinear conversion on the intermediate state variable in the second stage to obtain new measurements, namely a phase-angle difference between voltages at two ends of each branch and twice of the sum of logarithmic magnitudes of the voltages; and estimating the final state variable on the basis of these new measurements in the third stage. Through nonlinear conversion in the second stage, parallel solving can be carried out in various regions and communication with the other regions is not needed. State estimation in the first stage and the third stage is aimed at a linearized system; and each control region determines the local state quantity by communication of the local and a neighbor by an alternating direction multiplier method. The bilinearization-based all-distributed robust state estimation method provided by the invention is suitable for large-scale multi-regional state estimation, can well protect the data privacy of various regions and has very high agility, flexibility and robustness.