The thermal network is widely employed for real-time rotor temperature estimation, enhancing both control accuracy and functional safety of electric machines. However, the complex operating conditions of oil-cooled induction machines (OCIMs) in electric vehicles lead to significant variability in thermal parameters. Moreover, cost constraints often preclude using oil temperature sensors in OCIMs, resulting in a lack of critical reference temperature in the thermal network. These factors collectively pose challenges to precise rotor temperature estimation. This article first establishes a thermal network for OCIMs and determines key electromechanical variables that influence thermal parameters based on practical physical characteristics. Next, a parameter identification network (PIN) with parallel differentiated branches is proposed to learn the accurate mapping relationship between electromechanical variables and thermal parameters. Next, a closed-loop temperature observer is designed, incorporating oil temperature as an extended state and using feedback from stator temperature deviations to estimate both rotor and oil temperatures dynamically. Finally, comprehensive temperature datasets covering a wide operating range of OCIMs are constructed through bench testing. The identification results from the PIN under varying conditions are analyzed, and the effectiveness of the proposed rotor temperature estimation method is validated in the absence of reference oil temperature.

中文翻译：

---

通过具有并行差分分支的参数识别网络估算油冷感应电机的转子温度


热网络广泛用于实时转子温度估计，从而提高电机的控制精度和功能安全性。然而，电动汽车中油冷感应电机 （OCIM） 的复杂工作条件导致热参数的显著变化。此外，成本限制通常排除了在 OCIM 中使用油温度传感器的可能性，从而导致热网络中缺乏临界参考温度。这些因素共同对精确的转子温度估计构成了挑战。本文首先为 OCIM 建立了一个热网络，并根据实际物理特性确定了影响热参数的关键机电变量。接下来，提出了一个具有并行差分分支的参数识别网络 （PIN），以学习机电变量与热参数之间的精确映射关系。接下来，设计了一个闭环温度观测器，将油温作为扩展状态，并使用定子温度偏差的反馈来动态估计转子和油温。最后，通过基准测试构建涵盖广泛 OCIM 工作范围的综合温度数据集。分析了不同条件下 PIN 的识别结果，并在没有参考油温的情况下验证了所提出的转子温度估计方法的有效性。