基于遗传算法的径向基函数神经网络的油冷电驱动桥转子温度预测模型  

Rotor Temperature Prediction Model of the Oil-cooled Electric Drive Axle Based on Genetic Algorithm and Radial Basis Function Neural Network

作　　者：张冠楠 罗伟凡 王林玉 ZHANG Guannan;LUO Weifan;WANG Linyu(United Automotive Electronic Systems Co.,Ltd.,201206)

机构地区：[1]联合汽车电子有限公司,201206

出　　处：《电机技术》2025年第1期25-29,共5页Electrical Machinery Technology

摘　　要：开发了一种大功率油冷电驱动桥(以下简称电桥)永磁同步电机转子温度预测模型,该模型在径向基函数(radial basis function,RBF)神经网络的基础上耦合了遗传算法(genetic algorithm,GA)。基于一款200kW油冷电桥的台架实测数据对RBF-GA神经网络模型进行了训练,利用训练后的模型分别对持续工况和中国轻型汽车行驶工况(china light vehicle test cycle,CLTC)下的电桥转子温度进行预测。预测结果表明,在持续工况下(每个持续工况点运行时长为30min),电桥转子温度预测的最高误差为4℃;在CLTC下,电桥转子温度预测的最高误差为7℃,预测精度远高于现有模型。同时,这两种工况下模型的预测误差不会随着电桥的持续运行而明显增大。RBF-GA神经网络温度预测模型不仅为电桥转子温度的实时监控提供了可行的方案,还为提升电桥的性能和寿命提供了支撑。A rotor temperature prediction model for a high-power oil-cooled electric drive axle(hereinafter re-ferred to as electric drive axle)permanent magnet synchro-nous motor is developed.This model is based on the Radial Basis Function(RBF)neural network and integrated with the Genetic Algorithm(GA).This RBF-GA neural network model is trained by using bench test data from a 200 kW oil-cooled electric drive axle.Then,the trained model is used to predict the rotor temperature of the electric drive axle under continuous operating conditions and the China Light Vehicle Test Cycle(CLTC)respectively.The prediction re-sults show that under continuous operating conditions(with each operating point lasting for 30 minutes),the maximum error in rotor temperature prediction of the electric drive axle is 4°C;under CLTC conditions,the maximum error is 7°C,which is significantly higher in accuracy than existing models.Moreover,the prediction error of the model does not increase significantly with the continuous operation of the electric drive axle under these two conditions.The RBF-GA neural network temperature prediction model not only provides a feasible solution for real-time monitoring of the rotor temperature of the electric drive axle,but also supports the improvement of its performance and lifespan.

关 键 词：神经网络 遗传算法 油冷电桥 台架测试 转子温度预测 

分 类 号：TM302[电气工程—电机]