This paper presents a fine and robust least squares state estimation method for solving residual contamination problem caused by large-scale wind power integration. On the one hand, it introduces the reference value of measurement type into the weight function to distinguish different types of measurement bad data, which improves the bad data detection capability of robust state estimation; on the other hand, it uses the pre-check information of state estimation measurement to do SCADA measurement pretreatment, and then forms the bad data reference factor to eliminate bad data misjudgment caused by parameter errors, thereby improving the state estimation accuracy of large-scale wind power integration grid. In order to improve the software computing speed and the data section real-time performance of robust state estimation, parallel algorithms are used to do Givens transformation, so as to achieve the fine and rapid robust state estimation and accommodate the influence to the analysis and control class online applications caused by the large-scale wind power integration grid. Finally, the simulation tests of a regional power grid prove that the proposed method can effectively identify telemetry bad data of wind farms eliminate residual pollution caused by it, which improve the speed and accuracy of the state estimation.