Parameter estimation based on novel enhanced self-learning particle swarm optimization algorithm with Levy flight for PMSG

Abstract

A novel parameter estimation method is proposed for the permanent magnet synchronous generator (PMSG), which is implemented by an enhanced self-learning particle swarm optimization algorithm with Levy flight (SLPSO), and the problem of lower parameter estimation precision of standard PSO is obviated. This method injects currents of different intensities into the d-axis in a time-sharing manner to solve the problem of equation under-ranking, and the mathematical model for full-rank parameter estimation is developed. The speed term of PSO is simplified to expedite the convergence of PSO, and a strategy with chaotic decline for the inertia weight of PSO is adopted to strengthen its ability to jump out of the local optimum. Moreover, the self-learning dense fleeing strategy (SLDF) is proposed where particles perform diffusion learning based on population density information and Levy flight, the evolutionary unitary problem and human intervention in the evolutionary process is averted. Furthermore, the memory tempering annealing algorithm (MTA) and greedy algorithm (GA) is integrated into the algorithm, MTA can facilitate the exploration of potentially better regions, and GA for local optimization enhances the convergence speed and accuracy in late stage of the algorithm. Comparing the proposed method with several existing PSO algorithms through simulation and experiments, the experimental data show that the proposed method can effectively track variable parameters under different working conditions and has better robustness.