A double neural network model predictive control method for series elastic actuator’s motor thermal protection
Anlong Zhang, Zhiyun Lin, Bo Wang, Zhimin Han
Hangzhou Dianzi University Southern University of Science and Technology Peng Cheng Laboratory
阅读操作
确认中在文库中上传 PDF 后可生成中文音频讲解。
摘要与影响
Traditional motor protection methods are conservative, neglecting actual thermal conditions and underutilizing the motor’s performance. This paper proposes a novel double neural network model predictive control (DNMPC)-based protection method to maximize motor performance. First, we analyze the system’s thermal properties and propose a thermal lemma to demonstrate that solving the protection problem involves determining the maximum allowable control input boundary. Second, instead of directly measuring the motor temperature, the neural network (NN) method is proposed to establish the system’s thermal model for estimating the motor temperatures. Third, leveraging the NN model, we propose a model predictive control strategy designed to determine the maximum allowable control input boundaries, which we refer to as the neural network model predictive control (NMPC) method. Consequently, this approach has led to the development of an advanced motor protection system. To further reduce the computational burden, we incorporate an additional NN to directly approximate the NMPC controller, thereby forming our proposed DNMPC approach. Finally, comparative experimental validation demonstrates the effectiveness and superiority of the proposed method. The results indicate that the NN model is highly accurate, eliminating the need for direct temperature measurements and enabling the prediction of future motor temperatures. Additionally, the DNMPC-based protection method significantly improves the joint’s peak torque output by 64.3 % compared to conservative continuous current limited protection methods. The computational efficiency is also notable, with the NMPC achieving a runtime of approximately 20 ms, while the DNMPC achieves only 0.2 ms.
逐年被引趋势
关键指标
同类平均 = 1
同领域 · 同年份 · 同类型
Google Scholar 与 OpenAlex 的被引统计范围不同,数值存在差异属正常。
AI 辅助阅读
依据:摘要
可就本文提问;依据不足时会说明。
学术脉络
学科主题
工程Advanced Control Systems Optimization
Control Systems in Engineering · Electric Motor Design and Analysis