中文摘要：

为提高永磁同步电机（Permanent magnet synchronous machine,PMSM）系统参数辨识与状态监测效率,利用图形处理器（Graphics processing unit,GPU）并行计算与人工免疫技术相结合的研究方法,建立面向永磁同步电机系统基于GPU并行动态学习型免疫进化的参数估计与状态监测模型。为提高算法的动态跟踪性能,在抗体演化进程中,通过知识学习策略来引导算法进化过程,首先将抗体群划分为B细胞群、浆细胞群以及记忆细胞群,对处于不同进化群体中的抗体分别设计免疫综合学习策略、免疫反向学习策略和高斯学习策略,以增强抗体间的信息交互;接着,应用图形处理器并行计算技术进一步加速算法求解过程;最后,将所提算法应用于永磁同步电机系统参数辨识与状态监测中,实验表明,所提方法能同时准确地对电机的定子电阻、dq轴电感和永磁磁链等系统关键参数进行估计。依据参数变化实现对系统运行状态进行在线监测与预警。计算结果表明,GPU并行技术能大幅度提高计算效率。

英文摘要：

In order to improve the efficiency of permanent magnet synchronous generator parameter estimation, a novel multi-parameters intelligent identification and condition monitoring method is proposed based on GPU parallel dynamic learnable immune evolutionary algorithm. This method named G-PDLIA combines graphics processing unit （GPU） parallel computing technology and artificial immune system. To improve the fast dynamic tracking performance of the detection algorithm, a knowledge learning scheme is used to guide the evolution process of the artificial immune algorithm. The details are as follows： firstly, the whole population is divided into B cells, plasma cells and memory cells; secondly, three learning strategies are designed for different evolutionary groups, including immune comprehensive learning strategy, immune opposition learning strategy and Gaussian dynamic learning strategy, and the interactions between antibodies are enhanced by a proposed learning operators; then, the proposed method is accelerated by GPU parallel computing technology. Finally, the proposed method is applied to PMSM parameters estimation and condition monitoring. The proposed method can effectively estimate the machine dq-axis inductances, stator winding resistance and rotor flux linkage. The task of online monitoring and early warning can be implemented for the running permanent magnet synchronous machine （PMSM） according to parameters changing. Furthermore, the computational efficiency is greatly enhanced by GPU-aecelerated parallel computing technique compared to a CPU-based serial execution.