In order to improve the operational reliability of wind turbines in smart wind farms and the self-confirmation of sensor status, a novel self-validation method for sensor status is proposed using the stator winding temperature sensor of wind turbines as an example. First, based on grey relational analysis theory, and utilizing sensor correlation and information fusion technology, the grey correlation degree between the abnormal stator winding temperature sensor of a specific wind field and the same type of sensor on the same machine is calculated to achieve sensor anomaly state recognition. Second, using Pearson correlation and expert system judgment, parameters with strong correlation to the stator winding temperature sensor are identified. A long short-term memory (LSTM) multi-parameter input, single-output abnormal data reconstruction model is then established and optimized using the sparrow search algorithm (SSA) to improve the model’s accuracy. To verify the model’s reconstruction accuracy, simulations of abnormal data recovery showed that the accuracy reached 99.69%. Finally, the abnormal data of the stator winding temperature sensor was recovered, and the dynamic validation uncertainty of the recovered data was calculated using a Bayesian algorithm, achieving self-validation of the sensor’s state.