Aiming at some current problems of quality control in real time temperature measurements, a dynamic method based on k means clustering algorithm is proposed. The algorithm first divides the temperature sample points in the region into a number of clusters according to their similar temperatures by k means, and then for each sample point in the clusters the algorithm checks its outlier ratio and outlier speed in order to determine the final quality of the point. Compared with conventional temperature quality control methods, the algorithm uses an idea of the comparison of the single point temperature with the overall temperature, and it does not need to pre set the reference temperature value, thus it is a more real time and scientific temperature quality control method. Also, the complexity of the algorithm is low, and it is proper for the calculation of large temperature input data sets.