【目的】针对永磁同步电机(PMSM)多参数辨识困难、辨识精度低的问题，提出了一种改进的金枪鱼群优化(TSO)算法，同时辨识PMSM多个参数。【方法】首先，采用拉丁超立方体采样(LHS)法对金枪鱼种群进行初始化，有效地避免了由随机初始化种群造成的金枪鱼初始种群聚集现象，增强了初始种群的多样性与均匀性；其次，为了改善算法在不同迭代阶段的寻优性能，采用动态非线性权重调整策略，使算法在迭代过程早期阶段有更强的全局搜索能力以更广泛地探索搜索空间，在迭代后期阶段有更强的局部搜索能力以精确地收敛到最优解；最后，采用高斯变异策略，使算法在寻优过程中能够有效地探索新的解空间，改善了算法易陷入局部最优的问题，提高了TSO算法的收敛精度。【结果】为了验证所提方法的有效性，分别在仿真软件和电机平台上进行了仿真和试验分析。仿真和试验结果显示，基于LHS、动态非线性权重调整和高斯变异的改进TSO算法与粒子群优化算法和TSO算法相比，在辨识PMSM电阻、电感和磁链过程中有更快的收敛速度和更高的收敛精度，且所有参数辨识误差均控制在0.89%范围内。【结论】改进TSO算法能够有效地对PMSM的电阻、电感和磁链进行同步辨识，并且显示出了更高的识别精度和良好的收敛特性。

[Objective] In response to the challenges associated with the identification of multiple parameters in permanent magnet synchronous motor (PMSM), which are characterized by difficulties and low accuracy, an improved tuna swarm optimization (TSO) algorithm has been proposed. This algorithm is designed to simultaneously identify several parameters of the PMSM. [Methods] Firstly, the Latin hypercube sampling (LHS) method was employed to initialize the tuna population, effectively circumvented the issue of initial aggregation of the tuna population caused by random initialization. This approach enhanced the diversity and uniformity of the initial population. Subsequently, to enhance the algorithm’s optimization performance across different stages of iteration, a dynamic nonlinear weight adjustment strategy was adopted. This strategy equipped the algorithm with stronger global search capabilities during the early stages of iteration, allowed for a broader exploration of the search space, and stronger local search capabilities during later stages, facilitated precise convergence to the optimal solution. Finally, a Gaussian mutation strategy was implemented, enabling the algorithm to effectively explore new solution spaces during the optimization process. This strategy mitigated the tendency of the algorithm to become trapped in local optima, thereby improved the convergence accuracy of the TSO algorithm. [Results] To validate the effectiveness of the proposed method, simulations and experimental analyses were conducted on both simulation software and motor platform. The results of both simulation and experiment demonstrated that the improved TSO algorithm, which incorporated LHS, dynamic nonlinear weight adjustment, and Gaussian mutation, outperforms particle swarm optimization algorithms and the original TSO algorithm in the identification process of PMSM resistance, inductance, and magnetic flux linkage. It achieved faster convergence speeds and higher convergence accuracy, with all parameter identification errors controlled within a range of 0.89%. [Conclusion] The improved TSO algorithm is capable of effectively synchronizing the identification of resistance, inductance, and magnetic flux linkage in PMSM. It exhibits superior identification accuracy and favorable convergence characteristics.

省级重点研发项目(2022TSTD-04)