TY - GEN
T1 - Generative Adversarial Network-Supported Permanent Magnet Temperature Estimation by Using Random Forest
AU - Zhang, Xiaotian
AU - Gong, Chao
AU - Hu, Yihua
AU - Xu, Hui
AU - Deng, Jiamei
N1 - Publisher Copyright:
© 2022, The Author(s), under exclusive license to Springer Nature Singapore Pte Ltd.
PY - 2022
Y1 - 2022
N2 - As the traditional methods are not economical and hard to directly measure permanent magnet (PM) temperature of permanent magnet synchronous motor (PMSM), currently a reasonable and more popular consideration to measure rotor temperature is prediction by artificial intelligence (AI) methods. This paper proposes a generative adversarial network (GAN)-supported AI method to solve the PM temperature measurement problem. Firstly, this paper uses CTGAN to get generated dataset and combines it with the original dataset. Secondly, a new GAN-RF method based on random forest (RF) is proposed to predict PM temperature and the performance is compared with another popular method long short-term memory (LSTM). The advantage of the GAN-RF is improving the prediction accuracy of the RF model through the size extension of datasets and getting rid of the dependence of prediction work on time series models (LSTM, etc.) through GAN. The dataset collected by the LEA department at Paderborn university verifies the effectiveness of the proposed method.
AB - As the traditional methods are not economical and hard to directly measure permanent magnet (PM) temperature of permanent magnet synchronous motor (PMSM), currently a reasonable and more popular consideration to measure rotor temperature is prediction by artificial intelligence (AI) methods. This paper proposes a generative adversarial network (GAN)-supported AI method to solve the PM temperature measurement problem. Firstly, this paper uses CTGAN to get generated dataset and combines it with the original dataset. Secondly, a new GAN-RF method based on random forest (RF) is proposed to predict PM temperature and the performance is compared with another popular method long short-term memory (LSTM). The advantage of the GAN-RF is improving the prediction accuracy of the RF model through the size extension of datasets and getting rid of the dependence of prediction work on time series models (LSTM, etc.) through GAN. The dataset collected by the LEA department at Paderborn university verifies the effectiveness of the proposed method.
KW - Artificial intelligence
KW - Permanent magnet synchronous motor
KW - Temperature prediction
UR - https://www.scopus.com/pages/publications/85141682337
U2 - 10.1007/978-981-19-3171-0_38
DO - 10.1007/978-981-19-3171-0_38
M3 - 会议稿件
AN - SCOPUS:85141682337
SN - 9789811931703
T3 - Lecture Notes in Electrical Engineering
SP - 459
EP - 472
BT - Conference Proceedings of 2021 International Joint Conference on Energy, Electrical and Power Engineering - Component Design, Optimization and Control Algorithms in Electrical and Power Engineering Systems
A2 - Cao, Wenping
A2 - Hu, Cungang
A2 - Tao, Jun
A2 - Huang, Xiaoyan
A2 - Chen, Xiangping
PB - Springer Science and Business Media Deutschland GmbH
T2 - International Joint Conference on Energy, Electrical and Power Engineering , CoEEPE 2021
Y2 - 17 September 2021 through 19 September 2021
ER -