Abstract: In order to ensure the efficient and stable operation of high-temperature submersible motors，and to

timely avoid motor failures caused by high operating temperatures，which may affect production，it is necessary to

obtain their underground temperature without temperature sensors. Based on this situation，a temperature recognition and prediction method for high-temperature submersible motors based on improved wavelet neural network （ I-WNN）was proposed. Firstly，the operating data of high-temperature submersible motors were classified. Then，an improved wavelet neural network was used to train historical data，a mapping relationship between the operating data of high-temperature submersible motors and temperature was established，and the weight parameters of the wavelet neural network were optimized to obtain the most weighted values. Finally，through experimental  simulation，the fitted temperature values and predicted temperature values of the high-temperature motor were obtained. 

      Key words: wavelet neural network（WNN）；prediction algorithm based on I-WNN；K-means clustering；

genetic algorithm（GA）