Performance Improvement for Gradient-based Model Predictive Control with Optimized Switching Sequences using Adaptive Control Set Extension

The superiority of the control set extension in model predictive control (MPC) for permanent magnet synchronous motor (PMSM) leads to improved tracking performance of the control objectives in the transportation applications. From a geometric perspective, the extension of the control set can achieve smaller tracking deviations to the reference trajectory, but the improvement is accompanied by the sacrifice of increased switching action changes. To tackle this issue, an adaptive framework of extended control set based on judgment of gradient-based vector is implemented on the MPC scheme, for simultaneous improvement of current quality and switching sequence. Based on the above, a gradient-based formulation of a quadratic programming (QP) MPC problem is put forward. The judgment of relationship between optimal and reference gradient vectors in the consecutive periods is subsequently revealed. In this case, a current gradient-based MPC using consecutive optimal solutions is deployed, in which the tracking error is remarkably reduced without increment of switching changes. When the optimal vector in the former period is opposite to the direction of reference vector, the control set extension is adaptively shifted by applying two optimal vectors of current period in this circumstance. On the contrary, it is considered to avoid increased switching state changes by excluding the control set extension, as the optimal vector in the current sampling period is sufficiently close to the corresponding reference. The proposed adaptive extension framework is intended for the simultaneous enhancement of stator current and switching frequency performance for gradient-based MPC schemes. The proposed scheme is conducted on a 4.8 kW PMSM testbench for experimental verification, which confirms a concurrent optimization of current quality and switching changes in both steady and transient states.