When particle swarm optimization (PSO) is used to identify the parameters of permanent magnet synchronous motor (PMSM), the movement of particles is not selective, which makes the algorithm easy to fall into the local optimum, and the accuracy is poor. The simulated annealing particle swarm optimization (SAPSO) improves the accuracy and evolution speed, but SAPSO has redundant iteration problems. To solve these problems, a motor parameter identification method based on fast backfire double anneal particle swarm optimization (FBDAPSO) is proposed. By reducing the optimization time and quickly tempering and annealing the "misunderstood" difference, the motor adjustable model and fitness function are designed, and the number of iterations is constantly reset to achieve the effect of online identification. Under different working conditions, simulated and experimental results show that the proposed method can quickly and accurately identify the four parameters of the motor's stator, winding resistance, stator winding d-axis inductance, stator winding q-axis inductance, and permanent magnet flux linkage at the same time, compared with the traditional method of parameter identification, and it has better accuracy, rapidity, and robustness.

To further improve the steady-state performance of the conventional dual vector model predictive current control (MPCC), an improved optimal duty MPCC strategy for permanent magnet synchronous motor (PMSM) is proposed. This strategy is realized by selecting an optimal voltage vector combination and its duration from the five basic voltage vector combinations, followed by acting on the inverter. The five combinations are: the combination of the optimal voltage vector at the previous moment and basic voltage vector with an angle difference of 60°; the combination of the optimal voltage vector at the previous moment and basic voltage vector with an angle difference of 60°; the combination of the aforementioned three basic voltage vectors with the zero vector. Experimental results indicate that the method effectively reduces the stator current ripple without increasing the calculational burden. Furthermore, it improves the steady-state performance of the system without altering the dynamic performance of the system.

Aiming at the problem that extended Kalman filter (EKF) is difficult to determine the appropriate system and measurement noise matrices in parameter identification of direct‐drive permanent magnet synchronous generator (D‐PMSG), a parameter identification method of D‐PMSG based on mean particle swarm optimization with extreme disturbance (EDMPSO)‐EKF is proposed. Firstly, by analysing the principle of EKF, a dual‐thread identification model is established, then the particle swarm optimization (PSO) algorithm is improved to jump out of the local optimum by adding extremum interference and taking average extremum, and a suitable fitness function is designed. Finally, the improved PSO algorithm is applied to the adaptive optimization of EKF system noise matrix and measurement noise matrix, and the system performs the parameter identification after obtaining the optimal noise matrix. The simulation and experiment results show that the proposed method has better convergence speed and can more accurately identify the parameters of stator resistance, inductance and flux linkage, and it has better identification accuracy and generalization ability than the traditional method.

To address the unsatisfactory performance of particle swarm optimization (PSO), a novel multi-strategy self-optimizing simulated annealing particle swarm optimization (SOSAPSO) method for permanent magnet synchronous motor (PMSM) parameter identification is proposed. The full-rank mathematical model and the fitness function are developed. In SOSAPSO, the velocity term of the PSO is simplified and dynamic opposition-based learning (DOBL) is introduced in the inertia weight update process to avoid population monotonicity. Moreover, A Cauchy-Gaussian hybrid variation strategy based on similarity and density is devised to achieve self-learning in deep regions. Meanwhile, the simulated annealing (SA) with a memory and tempering mechanism is introduced into SOSAPSO, and the greedy optimization algorithm (GOA) is used to enhance local fine-exploitation capabilities when SOSAPSO evolution is stalled. The test results indicate the proposed method can effectively avoid local convergence problems and has better robustness and convergence speed.