The internal temperature rise measurement of high-frequency magnetic components is important to the design, application of magnetic components. However, due to the eddy current loss at the metal end of the thermocouple, a large additional temperature rise is caused, which makes the temperature distortion inside the high-frequency magnetic element. In order to further deduct the unpredictable influence of the high-frequency magnetic field environment on the thermocouple, this paper proposes to use the cooling process curve of the magnetic element after the power is off, combined with the thermal circuit model of the cooling process, to correct the extra temperature rise of the thermocouple caused by the eddy current loss. At the same time, it is proposed to use the neural network method to predict the temperature drop characteristic curve before the actual temperature rise of the measured point is deducted from the extra temperature rise of the thermocouple, so as to realize the accurate measurement of the temperature rise of the hot spot inside the magnetic element under the high-frequency magnetic field.