Abstract: Aiming at the problem of single hot spot temperature under average state and long response time

under different load states，in order to improve the response speed，a 10 kV transformer load state detection method based on K-means algorithm was designed. The 10 kV transformer load model was established，the overlapping vector matrix was compressed，and the transformer load state parameter indexes were comprehensively measured.The hot spot temperatures under different load states were calculated，and the hot spot temperatures under different load states such as normal periodic load，long-term emergency load and short-term emergency load were discussed respectively. The load state of transformer was evaluated based on K-means algorithm，the Euclidean distance and average distance were calculated，the load state index was solved，and the load state detection of 10 kV transformer was completed. The example test results show that the maximum response time under the conditions of normal periodic load，long-term emergency load and short-term emergency load are 0.36 s，0.6 s and 0.2 s，and the detection ability for transformer load state is more stronger.

      Key words: K-means algorithm；10 kV transformer；transformer；load detection；load state detection；load

state of transformer