Model-free predictive current control of FTPMV-RDM based on artificial neural network parameter optimization for ship electric drive

Aiming at the problem that the traditional finite-set model-based predictive current control (MBPCC) of fault-tolerant permanent magnet vernier rim-driven motor (FTPMV-RDM) in the field of ship electrification transportation is greatly affected by parameter mismatch, a novel model-free predictive current control (MFPCC) algorithm with strong robustness is proposed. The proposed algorithm regards the main electrical parameter items as unknown disturbances by establishing a current ultralocal model, and then uses an extended state observer (ESO) based on artificial neural network (ANN) parameter optimization to estimate it, optimize the observer performance, where ANN adopts a dual-hidden-layer structure combined with a five-fold cross-validation and model fusion strategy, ensuring strong generalization capability, and further calculate the dq-axes predicted current. Finally, all candidate voltage vectors in the complex plane are evaluated and compared by a two-norm square cost function containing the dq-axes current tracking error, and the optimal voltage vector corresponding to the minimum cost function is output to act on the six-phase H-bridge inverter to feed the FTPMV-RDM. Simulations and experiments have verified multiple groups of conventional and extreme parameter mismatch conditions. The results show that the proposed algorithm significantly reduces the impact of parameter mismatch and has strong robustness.