Hybrid optimal control feedforward neural network-super twisting sliding mode algorithm applied of two series-connected multi-phase PMSM using neural SVPWM algorithm

Multiphase motors enable independent control of several motors connected in series using single inverter. Effective control of these multi-machine systems (MMS) is crucial, considering factors like torque and speed. While vector control is a common method, its reliance on conventional proportional-integral (PI) controllers makes it vulnerable to performance degradation when system parameters fluctuate, highlighting the need for accurate modeling, robust drives, and advanced control. In this context, this study proposes an advanced super-twisting sliding mode control with neural network algorithm (NSTSMC) approach for controlling two series-connected five-phase permanent magnet synchronous motors (5Ph-PMSMs) powered by a single inverter. While conventional sliding mode control (SMC) is effective for nonlinear systems, it suffers from steady-state inaccuracies and chattering. The NSTSMC design enhances system robustness and significantly reduces chattering. Additionally, a constant switching frequency is achieved using a modified neural-space vector pulse width modulation (NSVPWM) for the inverter. Digital simulations in MATLAB confirmed the proposed technique's efficacy, demonstrating its superior ability to improve system features compared to conventional and SMC-SVPWM techniques. The NSTSMC-NSVPWM technique significantly improved the performance of two 5Ph-PMSMs. For 5Ph-PMSM1, it cut response time by 45.45% compared to SMC-SVPWM technique and 89.09% over conventional methods. 5Ph-PMSM2 saw similar gains, with response time reductions of 47.27% (SMC-SVPWM technique) and 93.09% (MMS-PI Method). Additionally, the proposed technique substantially reduced torque ripple in both machines: 23.52% and 77.58% for 5Ph-PMSM1, and 21.87% and 78.26% for 5Ph-PMSM2, against conventional and SMC-SVPWM methods, respectively. This proposed control also offers robust performance against parameter variations.