Affected by many factors such as equipment and weather, there is an inevitable error in the measurement data of the power grid. Before the actual application, the appropriate estimation method should be used for data adjustment. In order to reduce the influence of bad data on estimation accuracy, this paper proposes a robust unscented Kalman filter algorithm (RUKF), which introduces a bad data detection method based on operation mode before performing unscented Kalman filtering. The measured change trend adjusts the threshold to avoid missed detection and false detection of bad data. Taking IEEE 33-bus and a real 107-node system as examples, simulations and experiments are implemented. The experimental results show that, in the presence of bad data, the data adjustment results obtained by RUKF compared with the traditional UKF have higher estimation accuracy and improve the reliability of data estimation. It shows that the RUKF algorithm proposed in this paper can provide effective theoretical support for data adjustment calculation.