Abstract: To increase the number of atmospheric sounding stations and improve the accuracy of groundbased microwave radiometers in measuring temperatures, a correction method has been developed. This method utilizes temperature products from the FY-4A meteorological satellite along with a BP neural network and genetic algorithm to refine the temperature profiles measured by microwave radiometers. The performance of this method was tested through simulation experiments on two MP-3000 ground-based microwave radiometers located at Hangzhou and Nanjing stations. The quality of the corrections was evaluated by comparing the adjusted temperature profiles with independent temperature data from radiosonde measurements and Civil Aviation AMDAR. The results show that: (1) there is a certain error between the temperature product of the microwave radiometer and the true value of the temperature. At the height of 2km, the average deviation of the two stations is the largest, and the deviation of the same station with rain is greater than that without rain. (2) The temperature measurement precision of the microwave radiometer, after being refined through BP neural network simulation, has seen a significant enhancement compared to the original temperature data. At Hangzhou station, the reductions in MAE (Mean Absolute Error), MSE (Mean Squared Error), and RMSE (Root Mean Square Error) are in the ranges of 45-55%, 65-78%, and 41-53% respectively, while at Nanjing station, these metrics have decreased by 58-66%, 83-88%, and 55-59% respectively. (3) The simulation model of the neural network, after it’s initial weights and thresholds were optimized using a genetic algorithm, has demonstrated further improved performance. There has been a significant enhancement in the rain model, with reductions in RMSE (Root Mean Square Error) ranging from 11-15%. The above-mentioned correction method for microwave radiometers can be extended to microwave radiometer stations and has practical application value.